diff --git a/configure.ac b/configure.ac index 1863bed499..f2725876a2 100644 --- a/configure.ac +++ b/configure.ac @@ -1,1859 +1,1847 @@ dnl dnl autoconf for Pacemaker dnl dnl Copyright 2009-2019 the Pacemaker project contributors dnl dnl The version control history for this file may have further details. dnl dnl This source code is licensed under the GNU General Public License version 2 dnl or later (GPLv2+) WITHOUT ANY WARRANTY. dnl =============================================== dnl Bootstrap dnl =============================================== AC_PREREQ(2.64) AC_CONFIG_MACRO_DIR([m4]) AC_DEFUN([AC_DATAROOTDIR_CHECKED]) dnl Suggested structure: dnl information on the package dnl checks for programs dnl checks for libraries dnl checks for header files dnl checks for types dnl checks for structures dnl checks for compiler characteristics dnl checks for library functions dnl checks for system services m4_include([version.m4]) AC_INIT([pacemaker], VERSION_NUMBER, [users@clusterlabs.org], [pacemaker], PCMK_URL) PCMK_FEATURES="" AC_CONFIG_AUX_DIR(.) AC_CANONICAL_HOST dnl Where #defines go (e.g. `AC_CHECK_HEADERS' below) dnl dnl Internal header: include/config.h dnl - Contains ALL defines dnl - include/config.h.in is generated automatically by autoheader dnl - NOT to be included in any header files except crm_internal.h dnl (which is also not to be included in any other header files) dnl dnl External header: include/crm_config.h dnl - Contains a subset of defines checked here dnl - Manually edit include/crm_config.h.in to have configure include dnl new defines dnl - Should not include HAVE_* defines dnl - Safe to include anywhere AC_CONFIG_HEADERS([include/config.h include/crm_config.h]) AC_ARG_WITH(version, [ --with-version=version Override package version (if you are a packager needing to pretend) ], [ PACKAGE_VERSION="$withval" ]) AC_ARG_WITH(pkg-name, [ --with-pkg-name=name Override package name (if you are a packager needing to pretend) ], [ PACKAGE_NAME="$withval" ]) dnl 1.11: minimum automake version required dnl foreign: don't require GNU-standard top-level files dnl silent-rules: allow "--enable-silent-rules" (no-op in 1.13+) dnl subdir-objects: keep .o's with their .c's (no-op in 2.0+) AM_INIT_AUTOMAKE([1.11 foreign silent-rules subdir-objects]) dnl Example 2.4. Silent Custom Rule to Generate a File dnl %-bar.pc: %.pc dnl $(AM_V_GEN)$(LN_S) $(notdir $^) $@ AC_DEFINE_UNQUOTED(PACEMAKER_VERSION, "$PACKAGE_VERSION", [Current pacemaker version]) dnl Versioned attributes implementation is not yet production-ready AC_DEFINE_UNQUOTED(ENABLE_VERSIONED_ATTRS, 0, [Enable versioned attributes]) PACKAGE_SERIES=`echo $PACKAGE_VERSION | awk -F. '{ print $1"."$2 }'` AC_SUBST(PACKAGE_SERIES) AC_SUBST(PACKAGE_VERSION) CC_IN_CONFIGURE=yes export CC_IN_CONFIGURE LDD=ldd dnl ======================================================================== dnl Compiler characteristics dnl ======================================================================== AC_PROG_CC dnl Can force other with environment variable "CC". AC_PROG_CC_STDC AC_PROG_CXX dnl C++ is not needed for build, just maintainer utilities gl_EARLY # We expressly don't want to introduce OpenSSL dependency by default. gl_SET_CRYPTO_CHECK_DEFAULT([no]) gl_INIT LT_INIT([dlopen]) LTDL_INIT([convenience]) AC_TYPE_SIZE_T AC_CHECK_SIZEOF(char) AC_CHECK_SIZEOF(short) AC_CHECK_SIZEOF(int) AC_CHECK_SIZEOF(long) AC_CHECK_SIZEOF(long long) dnl =============================================== dnl Helpers dnl =============================================== cc_supports_flag() { local CFLAGS="-Werror $@" AC_MSG_CHECKING(whether $CC supports "$@") AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[ ]], [[ ]])], [RC=0; AC_MSG_RESULT(yes)], [RC=1; AC_MSG_RESULT(no)]) return $RC } # Some tests need to use their own CFLAGS cc_temp_flags() { ac_save_CFLAGS="$CFLAGS" CFLAGS="$*" } cc_restore_flags() { CFLAGS=$ac_save_CFLAGS } dnl =============================================== dnl Configure Options dnl =============================================== dnl Some systems, like Solaris require a custom package name AC_ARG_WITH(pkgname, [ --with-pkgname=name name for pkg (typically for Solaris) ], [ PKGNAME="$withval" ], [ PKGNAME="LXHAhb" ], ) AC_SUBST(PKGNAME) AC_ARG_ENABLE([ansi], [ --enable-ansi Force GCC to compile to ANSI standard for older compilers. @<:@no@:>@]) AC_ARG_ENABLE([fatal-warnings], [ --enable-fatal-warnings Enable pedantic and fatal warnings for gcc @<:@yes@:>@]) AC_ARG_ENABLE([quiet], [ --enable-quiet Suppress make output unless there is an error @<:@no@:>@]) AC_ARG_ENABLE([no-stack], [ --enable-no-stack Build only the scheduler and its requirements @<:@no@:>@]) AC_ARG_ENABLE([upstart], [ --enable-upstart Enable support for managing resources via Upstart @<:@try@:>@ ], [], [enable_upstart=try], ) AC_ARG_ENABLE([systemd], [ --enable-systemd Enable support for managing resources via systemd @<:@try@:>@], [], [enable_systemd=try], ) AC_ARG_ENABLE(hardening, [ --enable-hardening Harden the resulting executables/libraries @<:@try@:>@], [ HARDENING="${enableval}" ], [ HARDENING=try ], ) # By default, we add symlinks at the pre-2.0.0 daemon name locations, so that: # (1) tools that directly invoke those names for metadata etc. will still work # (2) this installation can be used in a bundle container image used with # cluster hosts running Pacemaker 1.1.17+ # If you know your target systems will not have any need for it, you can # disable this option. Once the above use cases are no longer in wide use, we # can disable this option by default, and once we no longer want to support # them at all, we can drop the option altogether. AC_ARG_ENABLE(legacy-links, [ --enable-legacy-links Add symlinks for old daemon names @<:@yes@:>@], [ LEGACY_LINKS="${enableval}" ], [ LEGACY_LINKS=yes ], ) AM_CONDITIONAL(BUILD_LEGACY_LINKS, test "x${LEGACY_LINKS}" = "xyes") -AC_ARG_ENABLE(hash-affected-tests, - [ --disable-hash-affected-tests Skip hash-sensitive scheduler tests (try ~ autodetect)], - [ HASH_AFFECTED="${enableval}" ], - [ HASH_AFFECTED=yes ], -) - AC_ARG_WITH(corosync, [ --with-corosync Support the Corosync messaging and membership layer ], [ SUPPORT_CS=$withval ], [ SUPPORT_CS=try ], ) AC_ARG_WITH(nagios, [ --with-nagios Support nagios remote monitoring ], [ SUPPORT_NAGIOS=$withval ], [ SUPPORT_NAGIOS=try ], ) AC_ARG_WITH(nagios-plugin-dir, [ --with-nagios-plugin-dir=DIR Directory for nagios plugins @<:@LIBEXECDIR/nagios/plugins@:>@], [ NAGIOS_PLUGIN_DIR="$withval" ] ) AC_ARG_WITH(nagios-metadata-dir, [ --with-nagios-metadata-dir=DIR Directory for nagios plugins metadata @<:@DATADIR/nagios/plugins-metadata@:>@], [ NAGIOS_METADATA_DIR="$withval" ] ) AC_ARG_WITH(acl, [ --with-acl Support CIB ACL ], [ SUPPORT_ACL=$withval ], [ SUPPORT_ACL=yes ], ) AC_ARG_WITH(cibsecrets, [ --with-cibsecrets Support separate file for CIB secrets ], [ SUPPORT_CIBSECRETS=$withval ], [ SUPPORT_CIBSECRETS=no ], ) PCMK_GNUTLS_PRIORITIES="NORMAL" AC_ARG_WITH(gnutls-priorities, [ --with-gnutls-priorities GnuTLS cipher priorities @<:@NORMAL@:>@ ], [ test x"$withval" = x"no" || PCMK_GNUTLS_PRIORITIES="$withval" ]) INITDIR="" AC_ARG_WITH(initdir, [ --with-initdir=DIR Directory for init (rc) scripts], [ INITDIR="$withval" ]) SUPPORT_PROFILING=0 AC_ARG_WITH(profiling, [ --with-profiling Disable optimizations for effective profiling ], [ SUPPORT_PROFILING=$withval ]) AC_ARG_WITH(coverage, [ --with-coverage Disable optimizations for effective profiling ], [ SUPPORT_COVERAGE=$withval ]) PUBLICAN_BRAND="common" AC_ARG_WITH(brand, [ --with-brand=brand Brand to use for generated documentation (set empty for no docs) @<:@common@:>@], [ test x"$withval" = x"no" || PUBLICAN_BRAND="$withval" ]) AC_SUBST(PUBLICAN_BRAND) BUG_URL="" AC_ARG_WITH(bug-url, [ --with-bug-url=DIR Address where users should submit bug reports @<:@https://bugs.clusterlabs.org/enter_bug.cgi?product=Pacemaker@:>@], [ BUG_URL="$withval" ] ) CONFIGDIR="" AC_ARG_WITH(configdir, [ --with-configdir=DIR Directory for Pacemaker configuration file @<:@SYSCONFDIR/sysconfig@:>@], [ CONFIGDIR="$withval" ] ) CRM_LOG_DIR="" AC_ARG_WITH(logdir, [ --with-logdir=DIR Directory for Pacemaker log file @<:@LOCALSTATEDIR/log/pacemaker@:>@ ], [ CRM_LOG_DIR="$withval" ] ) CRM_BUNDLE_DIR="" AC_ARG_WITH(bundledir, [ --with-bundledir=DIR Directory for Pacemaker bundle logs @<:@LOCALSTATEDIR/log/pacemaker/bundles@:>@ ], [ CRM_BUNDLE_DIR="$withval" ] ) dnl =============================================== dnl General Processing dnl =============================================== AC_PROG_LN_S AC_PROG_MKDIR_P if cc_supports_flag -Werror; then WERROR="-Werror" else WERROR="" fi # Normalize enable_fatal_warnings (defaulting to yes, when compiler supports it) if test "x${enable_fatal_warnings}" != "xno" ; then if test "$GCC" = "yes" && test "x${WERROR}" != "x" ; then enable_fatal_warnings=yes else AC_MSG_NOTICE(Compiler does not support fatal warnings) enable_fatal_warnings=no fi fi INIT_EXT="" echo Our Host OS: $host_os/$host AC_MSG_NOTICE(Sanitizing prefix: ${prefix}) case $prefix in NONE) prefix=/usr dnl Fix default variables - "prefix" variable if not specified if test "$localstatedir" = "\${prefix}/var"; then localstatedir="/var" fi if test "$sysconfdir" = "\${prefix}/etc"; then sysconfdir="/etc" fi ;; esac AC_MSG_NOTICE(Sanitizing exec_prefix: ${exec_prefix}) case $exec_prefix in prefix|NONE) exec_prefix=$prefix ;; esac AC_MSG_NOTICE(Sanitizing INITDIR: ${INITDIR}) case $INITDIR in prefix) INITDIR=$prefix;; "") AC_MSG_CHECKING(which init (rc) directory to use) for initdir in /etc/init.d /etc/rc.d/init.d /sbin/init.d \ /usr/local/etc/rc.d /etc/rc.d do if test -d $initdir then INITDIR=$initdir break fi done AC_MSG_RESULT($INITDIR) ;; esac AC_SUBST(INITDIR) AC_MSG_NOTICE(Sanitizing libdir: ${libdir}) case $libdir in prefix|NONE) AC_MSG_CHECKING(which lib directory to use) for aDir in lib64 lib do trydir="${exec_prefix}/${aDir}" if test -d ${trydir} then libdir=${trydir} break fi done AC_MSG_RESULT($libdir); ;; esac dnl Expand autoconf variables so that we don't end up with '${prefix}' dnl in #defines and python scripts dnl NOTE: Autoconf deliberately leaves them unexpanded to allow dnl make exec_prefix=/foo install dnl No longer being able to do this seems like no great loss to me... eval prefix="`eval echo ${prefix}`" eval exec_prefix="`eval echo ${exec_prefix}`" eval bindir="`eval echo ${bindir}`" eval sbindir="`eval echo ${sbindir}`" eval libexecdir="`eval echo ${libexecdir}`" eval datadir="`eval echo ${datadir}`" eval sysconfdir="`eval echo ${sysconfdir}`" eval sharedstatedir="`eval echo ${sharedstatedir}`" eval localstatedir="`eval echo ${localstatedir}`" eval libdir="`eval echo ${libdir}`" eval includedir="`eval echo ${includedir}`" eval oldincludedir="`eval echo ${oldincludedir}`" eval infodir="`eval echo ${infodir}`" eval mandir="`eval echo ${mandir}`" dnl Home-grown variables eval INITDIR="${INITDIR}" eval docdir="`eval echo ${docdir}`" if test x"${docdir}" = x""; then docdir=${datadir}/doc/${PACKAGE}-${VERSION} fi AC_SUBST(docdir) if test x"${CONFIGDIR}" = x""; then CONFIGDIR="${sysconfdir}/sysconfig" fi AC_SUBST(CONFIGDIR) if test x"${CRM_LOG_DIR}" = x""; then CRM_LOG_DIR="${localstatedir}/log/pacemaker" fi AC_DEFINE_UNQUOTED(CRM_LOG_DIR,"$CRM_LOG_DIR", Location for Pacemaker log file) AC_SUBST(CRM_LOG_DIR) if test x"${CRM_BUNDLE_DIR}" = x""; then CRM_BUNDLE_DIR="${localstatedir}/log/pacemaker/bundles" fi AC_DEFINE_UNQUOTED(CRM_BUNDLE_DIR,"$CRM_BUNDLE_DIR", Location for Pacemaker bundle logs) AC_SUBST(CRM_BUNDLE_DIR) if test x"${PCMK_GNUTLS_PRIORITIES}" = x""; then AC_MSG_ERROR([Empty string not applicable with --with-gnutls-priorities]) fi AC_DEFINE_UNQUOTED([PCMK_GNUTLS_PRIORITIES], ["$PCMK_GNUTLS_PRIORITIES"], [GnuTLS cipher priorities]) if test x"${BUG_URL}" = x""; then BUG_URL="https://bugs.clusterlabs.org/enter_bug.cgi?product=Pacemaker" fi AC_SUBST(BUG_URL) for j in prefix exec_prefix bindir sbindir libexecdir datadir sysconfdir \ sharedstatedir localstatedir libdir includedir oldincludedir infodir \ mandir INITDIR docdir CONFIGDIR do dirname=`eval echo '${'${j}'}'` if test ! -d "$dirname" then AC_MSG_WARN([$j directory ($dirname) does not exist!]) fi done dnl This OS-based decision-making is poor autotools practice; dnl feature-based mechanisms are strongly preferred. dnl dnl So keep this section to a bare minimum; regard as a "necessary evil". case "$host_os" in *bsd*) AC_DEFINE_UNQUOTED(ON_BSD, 1, Compiling for BSD platform) LIBS="-L/usr/local/lib" CPPFLAGS="$CPPFLAGS -I/usr/local/include" INIT_EXT=".sh" ;; *solaris*) AC_DEFINE_UNQUOTED(ON_SOLARIS, 1, Compiling for Solaris platform) ;; *linux*) AC_DEFINE_UNQUOTED(ON_LINUX, 1, Compiling for Linux platform) ;; darwin*) AC_DEFINE_UNQUOTED(ON_DARWIN, 1, Compiling for Darwin platform) LIBS="$LIBS -L${prefix}/lib" CFLAGS="$CFLAGS -I${prefix}/include" ;; esac AC_SUBST(INIT_EXT) AC_MSG_NOTICE(Host CPU: $host_cpu) case "$host_cpu" in ppc64|powerpc64) case $CFLAGS in *powerpc64*) ;; *) if test "$GCC" = yes; then CFLAGS="$CFLAGS -m64" fi ;; esac ;; esac # C99 doesn't guarantee uint64_t type and related format specifiers, but # prerequisites, corosync + libqb, use that widely, so the target platforms # are already pre-constrained to those "64bit-clean" (doesn't imply native # bit width) and hence we deliberately refrain from artificial surrogates # (sans manipulation through cached values). AC_CACHE_VAL( [pcmk_cv_decl_inttypes], [ AC_CHECK_DECLS( [PRIu64, PRIu32, PRIx32, SCNu64], [pcmk_cv_decl_inttypes="PRIu64 PRIu32 PRIx32 SCNu64"], [ # test shall only react on "no" cached result & error out respectively if test "x$ac_cv_have_decl_PRIu64" = xno; then AC_MSG_ERROR([lack of inttypes.h based specifier serving uint64_t (PRIu64)]) elif test "x$ac_cv_have_decl_PRIu32" = xno; then AC_MSG_ERROR([lack of inttypes.h based specifier serving uint32_t (PRIu32)]) elif test "x$ac_cv_have_decl_PRIx32" = xno; then AC_MSG_ERROR([lack of inttypes.h based hexa specifier serving uint32_t (PRIx32)]) elif test "x$ac_cv_have_decl_SCNu64" = xno; then AC_MSG_ERROR([lack of inttypes.h based specifier gathering uint64_t (SCNu64)]) fi ], [[#include ]] ) ] ) ( set $pcmk_cv_decl_inttypes AC_DEFINE_UNQUOTED([U64T], [$1], [Correct format specifier for U64T]) AC_DEFINE_UNQUOTED([U32T], [$2], [Correct format specifier for U32T]) AC_DEFINE_UNQUOTED([X32T], [$3], [Correct format specifier for X32T]) AC_DEFINE_UNQUOTED([U64TS], [$4], [Correct format specifier for U64TS]) ) dnl =============================================== dnl Program Paths dnl =============================================== PATH="$PATH:/sbin:/usr/sbin:/usr/local/sbin:/usr/local/bin" export PATH dnl Replacing AC_PROG_LIBTOOL with AC_CHECK_PROG because LIBTOOL dnl was NOT being expanded all the time thus causing things to fail. AC_CHECK_PROGS(LIBTOOL, glibtool libtool libtool15 libtool13) dnl Pacemaker's executable python scripts will invoke the python specified by dnl configure's PYTHON variable. If not specified, AM_PATH_PYTHON will check a dnl built-in list with (unversioned) "python" having precedence. To configure dnl Pacemaker to use a specific python interpreter version, define PYTHON dnl when calling configure, for example: ./configure PYTHON=/usr/bin/python3.6 dnl Ensure PYTHON is an absolute path AC_PATH_PROG([PYTHON], [$PYTHON]) case "x$PYTHON" in x*python3*|x*platform-python*) dnl When used with Python 3, Pacemaker requires a minimum of 3.2 AM_PATH_PYTHON([3.2]) ;; *) dnl Otherwise, Pacemaker requires a minimum of 2.7 AM_PATH_PYTHON([2.7]) ;; esac AC_PATH_PROGS([ASCIIDOC_CONV], [asciidoc asciidoctor]) AC_PATH_PROG([HELP2MAN], [help2man]) AC_PATH_PROG([PUBLICAN], [publican]) AC_PATH_PROG([INKSCAPE], [inkscape]) AC_PATH_PROG([XSLTPROC], [xsltproc]) AC_PATH_PROG([XMLCATALOG], [xmlcatalog]) dnl BASH is already an environment variable, so use something else AC_PATH_PROG([BASH_PATH], [bash]) PKG_PROG_PKG_CONFIG AC_PATH_PROGS(VALGRIND_BIN, valgrind, /usr/bin/valgrind) AC_DEFINE_UNQUOTED(VALGRIND_BIN, "$VALGRIND_BIN", Valgrind command) if test x"${LIBTOOL}" = x""; then AC_MSG_ERROR(You need (g)libtool installed in order to build ${PACKAGE}) fi dnl Bash is needed for building man pages and running regression tests if test x"${BASH_PATH}" = x""; then AC_MSG_ERROR(bash must be installed in order to build ${PACKAGE}) fi AM_CONDITIONAL(BUILD_HELP, test x"${HELP2MAN}" != x"") if test x"${HELP2MAN}" != x""; then PCMK_FEATURES="$PCMK_FEATURES generated-manpages" fi MANPAGE_XSLT="" if test x"${XSLTPROC}" != x""; then AC_MSG_CHECKING(docbook to manpage transform) # first try to figure out correct template using xmlcatalog query, # resort to extensive (semi-deterministic) file search if that fails DOCBOOK_XSL_URI='http://docbook.sourceforge.net/release/xsl/current' DOCBOOK_XSL_PATH='manpages/docbook.xsl' MANPAGE_XSLT=$(${XMLCATALOG} "" ${DOCBOOK_XSL_URI}/${DOCBOOK_XSL_PATH} \ | sed -n 's|^file://||p;q') if test x"${MANPAGE_XSLT}" = x""; then DIRS=$(find "${datadir}" -name $(basename $(dirname ${DOCBOOK_XSL_PATH})) \ -type d | LC_ALL=C sort) XSLT=$(basename ${DOCBOOK_XSL_PATH}) for d in ${DIRS}; do if test -f "${d}/${XSLT}"; then MANPAGE_XSLT="${d}/${XSLT}" break fi done fi fi AC_MSG_RESULT($MANPAGE_XSLT) AC_SUBST(MANPAGE_XSLT) AM_CONDITIONAL(BUILD_XML_HELP, test x"${MANPAGE_XSLT}" != x"") if test x"${MANPAGE_XSLT}" != x""; then PCMK_FEATURES="$PCMK_FEATURES agent-manpages" fi AM_CONDITIONAL([IS_ASCIIDOC], [echo "${ASCIIDOC_CONV}" | grep -Eq 'asciidoc$']) AM_CONDITIONAL([BUILD_ASCIIDOC], [test "x${ASCIIDOC_CONV}" != x]) if test "x${ASCIIDOC_CONV}" != x; then PCMK_FEATURES="$PCMK_FEATURES ascii-docs" fi publican_intree_brand=no if test x"${PUBLICAN_BRAND}" != x"" \ && test x"${PUBLICAN}" != x"" \ && test x"${INKSCAPE}" != x""; then dnl special handling for clusterlabs brand (possibly in-tree version used) test "${PUBLICAN_BRAND}" != "clusterlabs" \ || test -d /usr/share/publican/Common_Content/clusterlabs if test $? -ne 0; then dnl Unknown option: brand_dir vs. Option brand_dir requires an argument if ${PUBLICAN} build --brand_dir 2>&1 | grep -Eq 'brand_dir$'; then AC_MSG_WARN([Cannot use in-tree clusterlabs brand, resorting to common]) PUBLICAN_BRAND=common else publican_intree_brand=yes fi fi AC_MSG_NOTICE([Enabling Publican-generated documentation using ${PUBLICAN_BRAND} brand]) PCMK_FEATURES="$PCMK_FEATURES publican-docs" fi AM_CONDITIONAL([BUILD_DOCBOOK], [test x"${PUBLICAN_BRAND}" != x"" \ && test x"${PUBLICAN}" != x"" \ && test x"${INKSCAPE}" != x""]) AM_CONDITIONAL([PUBLICAN_INTREE_BRAND], [test x"${publican_intree_brand}" = x"yes"]) dnl Pacemaker's shell scripts (and thus man page builders) rely on GNU getopt AC_MSG_CHECKING([for GNU-compatible getopt]) IFS_orig=$IFS IFS=: for PATH_DIR in $PATH; do IFS=$IFS_orig GETOPT_PATH="${PATH_DIR}/getopt" if test -f "$GETOPT_PATH" && test -x "$GETOPT_PATH" ; then $GETOPT_PATH -T >/dev/null 2>/dev/null if test $? -eq 4; then break fi fi GETOPT_PATH="" done IFS=$IFS_orig if test -n "$GETOPT_PATH"; then AC_MSG_RESULT([$GETOPT_PATH]) else AC_MSG_RESULT([no]) AC_MSG_ERROR(Pacemaker build requires a GNU-compatible getopt) fi AC_SUBST([GETOPT_PATH]) dnl ======================================================================== dnl checks for library functions to replace them dnl dnl NoSuchFunctionName: dnl is a dummy function which no system supplies. It is here to make dnl the system compile semi-correctly on OpenBSD which doesn't know dnl how to create an empty archive dnl dnl scandir: Only on BSD. dnl System-V systems may have it, but hidden and/or deprecated. dnl A replacement function is supplied for it. dnl dnl setenv: is some bsdish function that should also be avoided (use dnl putenv instead) dnl On the other hand, putenv doesn't provide the right API for the dnl code and has memory leaks designed in (sigh...) Fortunately this dnl A replacement function is supplied for it. dnl dnl strerror: returns a string that corresponds to an errno. dnl A replacement function is supplied for it. dnl dnl strnlen: is a gnu function similar to strlen, but safer. dnl We wrote a tolearably-fast replacement function for it. dnl dnl strndup: is a gnu function similar to strdup, but safer. dnl We wrote a tolearably-fast replacement function for it. AC_REPLACE_FUNCS(alphasort NoSuchFunctionName scandir setenv strerror strchrnul unsetenv strnlen strndup) dnl =============================================== dnl Libraries dnl =============================================== AC_CHECK_LIB(socket, socket) dnl -lsocket AC_CHECK_LIB(c, dlopen) dnl if dlopen is in libc... AC_CHECK_LIB(dl, dlopen) dnl -ldl (for Linux) AC_CHECK_LIB(rt, sched_getscheduler) dnl -lrt (for Tru64) AC_CHECK_LIB(gnugetopt, getopt_long) dnl -lgnugetopt ( if available ) AC_CHECK_LIB(pam, pam_start) dnl -lpam (if available) AC_CHECK_FUNCS([sched_setscheduler]) AC_CHECK_LIB(uuid, uuid_parse) dnl load the library if necessary AC_CHECK_FUNCS(uuid_unparse) dnl OSX ships uuid_* as standard functions AC_CHECK_HEADERS(uuid/uuid.h) if test "x$ac_cv_func_uuid_unparse" != xyes; then AC_MSG_ERROR(You do not have the libuuid development package installed) fi if test x"${PKG_CONFIG}" = x""; then AC_MSG_ERROR(You need pkgconfig installed in order to build ${PACKAGE}) fi # Require glib 2.16.0 (2008-03) or later for g_hash_table_iter_init() etc. PKG_CHECK_MODULES([GLIB], [glib-2.0 >= 2.16.0], [CPPFLAGS="${CPPFLAGS} ${GLIB_CFLAGS}" LIBS="${LIBS} ${GLIB_LIBS}"]) -# cf. https://bugs.clusterlabs.org/show_bug.cgi?id=5378 -PKG_CHECK_MODULES([GLIB], [glib-2.0 >= 2.59.0], - [test "x$HASH_AFFECTED" != xtry || HASH_AFFECTED=no], - [test "x$HASH_AFFECTED" != xtry || HASH_AFFECTED=yes]) -test "x$HASH_AFFECTED" = xno && HASH_AFFECTED='#' || HASH_AFFECTED= -AC_SUBST([HASH_AFFECTED]) # # Where is dlopen? # if test "$ac_cv_lib_c_dlopen" = yes; then LIBADD_DL="" elif test "$ac_cv_lib_dl_dlopen" = yes; then LIBADD_DL=-ldl else LIBADD_DL=${lt_cv_dlopen_libs} fi dnl FreeBSD needs -lcompat for ftime() used by lrmd.c AC_CHECK_LIB([compat], [ftime], [COMPAT_LIBS='-lcompat']) AC_SUBST(COMPAT_LIBS) dnl ======================================================================== dnl Headers dnl ======================================================================== dnl Some distributions insert #warnings into deprecated headers such as dnl timeb.h. If we will enable fatal warnings for the build, then enable dnl them for the header checks as well, otherwise the build could fail dnl even though the header check succeeds. (We should probably be doing dnl this in more places.) if test "x${enable_fatal_warnings}" = xyes ; then cc_temp_flags "$CFLAGS $WERROR" fi AC_CHECK_HEADERS(arpa/inet.h) AC_CHECK_HEADERS(ctype.h) AC_CHECK_HEADERS(dirent.h) AC_CHECK_HEADERS(errno.h) AC_CHECK_HEADERS(getopt.h) AC_CHECK_HEADERS(glib.h) AC_CHECK_HEADERS(grp.h) AC_CHECK_HEADERS(limits.h) AC_CHECK_HEADERS(linux/swab.h) AC_CHECK_HEADERS(malloc.h) AC_CHECK_HEADERS(netdb.h) AC_CHECK_HEADERS(netinet/in.h) AC_CHECK_HEADERS(netinet/ip.h) AC_CHECK_HEADERS(pwd.h) AC_CHECK_HEADERS(sgtty.h) AC_CHECK_HEADERS(signal.h) AC_CHECK_HEADERS(stdarg.h) AC_CHECK_HEADERS(stddef.h) AC_CHECK_HEADERS(stdio.h) AC_CHECK_HEADERS(stdlib.h) AC_CHECK_HEADERS(string.h) AC_CHECK_HEADERS(strings.h) AC_CHECK_HEADERS(sys/dir.h) AC_CHECK_HEADERS(sys/ioctl.h) AC_CHECK_HEADERS(sys/param.h) AC_CHECK_HEADERS(sys/reboot.h) AC_CHECK_HEADERS(sys/resource.h) AC_CHECK_HEADERS(sys/socket.h) AC_CHECK_HEADERS(sys/signalfd.h) AC_CHECK_HEADERS(sys/sockio.h) AC_CHECK_HEADERS(sys/stat.h) AC_CHECK_HEADERS(sys/time.h) AC_CHECK_HEADERS(sys/timeb.h) AC_CHECK_HEADERS(sys/types.h) AC_CHECK_HEADERS(sys/utsname.h) AC_CHECK_HEADERS(sys/wait.h) AC_CHECK_HEADERS(time.h) AC_CHECK_HEADERS(unistd.h) if test "x${enable_fatal_warnings}" = xyes ; then cc_restore_flags fi dnl These headers need prerequisites before the tests will pass dnl AC_CHECK_HEADERS(net/if.h) PKG_CHECK_MODULES(LIBXML2, [libxml-2.0], [CPPFLAGS="${CPPFLAGS} ${LIBXML2_CFLAGS}" LIBS="${LIBS} ${LIBXML2_LIBS}"]) AC_CHECK_HEADERS(libxml/xpath.h) if test "$ac_cv_header_libxml_xpath_h" != "yes"; then AC_MSG_ERROR(libxml development headers not found) fi AC_CHECK_LIB(xslt, xsltApplyStylesheet, [], AC_MSG_ERROR(Unsupported libxslt library version)) AC_CHECK_HEADERS(libxslt/xslt.h) if test "$ac_cv_header_libxslt_xslt_h" != "yes"; then AC_MSG_ERROR(libxslt development headers not found) fi AC_CACHE_CHECK(whether __progname and __progname_full are available, pf_cv_var_progname, AC_TRY_LINK([extern char *__progname, *__progname_full;], [__progname = "foo"; __progname_full = "foo bar";], pf_cv_var_progname="yes", pf_cv_var_progname="no")) if test "$pf_cv_var_progname" = "yes"; then AC_DEFINE(HAVE___PROGNAME,1,[ ]) fi dnl ======================================================================== dnl Structures dnl ======================================================================== AC_CHECK_MEMBERS([struct tm.tm_gmtoff],,,[[#include ]]) AC_CHECK_MEMBERS([lrm_op_t.rsc_deleted],,,[[#include ]]) AC_CHECK_MEMBER([struct dirent.d_type], AC_DEFINE(HAVE_STRUCT_DIRENT_D_TYPE,1,[Define this if struct dirent has d_type]),, [#include ]) dnl ======================================================================== dnl Functions dnl ======================================================================== AC_CHECK_FUNCS(getopt, AC_DEFINE(HAVE_DECL_GETOPT, 1, [Have getopt function])) AC_CHECK_FUNCS(nanosleep, AC_DEFINE(HAVE_DECL_NANOSLEEP, 1, [Have nanosleep function])) AC_CACHE_CHECK(whether sscanf supports %m, pf_cv_var_sscanf, AC_RUN_IFELSE([AC_LANG_SOURCE([[ #include const char *s = "some-command-line-arg"; int main(int argc, char **argv) { char *name = NULL; int n = sscanf(s, "%ms", &name); return n == 1 ? 0 : 1; } ]])], pf_cv_var_sscanf="yes", pf_cv_var_sscanf="no", pf_cv_var_sscanf="no")) if test "$pf_cv_var_sscanf" = "yes"; then AC_DEFINE(SSCANF_HAS_M, 1, [ ]) fi dnl ======================================================================== dnl bzip2 dnl ======================================================================== AC_CHECK_HEADERS(bzlib.h) AC_CHECK_LIB(bz2, BZ2_bzBuffToBuffCompress) if test x$ac_cv_lib_bz2_BZ2_bzBuffToBuffCompress != xyes ; then AC_MSG_ERROR(BZ2 libraries not found) fi if test x$ac_cv_header_bzlib_h != xyes; then AC_MSG_ERROR(BZ2 Development headers not found) fi dnl ======================================================================== dnl sighandler_t is missing from Illumos, Solaris11 systems dnl ======================================================================== AC_MSG_CHECKING([for sighandler_t]) AC_TRY_COMPILE([#include ],[sighandler_t *f;], has_sighandler_t=yes,has_sighandler_t=no) AC_MSG_RESULT($has_sighandler_t) if test "$has_sighandler_t" = "yes" ; then AC_DEFINE( HAVE_SIGHANDLER_T, 1, [Define if sighandler_t available] ) fi dnl ======================================================================== dnl ncurses dnl ======================================================================== dnl dnl A few OSes (e.g. Linux) deliver a default "ncurses" alongside "curses". dnl Many non-Linux deliver "curses"; sites may add "ncurses". dnl dnl However, the source-code recommendation for both is to #include "curses.h" dnl (i.e. "ncurses" still wants the include to be simple, no-'n', "curses.h"). dnl dnl ncurse takes precedence. dnl AC_CHECK_HEADERS(curses.h) AC_CHECK_HEADERS(curses/curses.h) AC_CHECK_HEADERS(ncurses.h) AC_CHECK_HEADERS(ncurses/ncurses.h) dnl Although n-library is preferred, only look for it if the n-header was found. CURSESLIBS='' if test "$ac_cv_header_ncurses_h" = "yes"; then AC_CHECK_LIB(ncurses, printw, [AC_DEFINE(HAVE_LIBNCURSES,1, have ncurses library)]) CURSESLIBS=`$PKG_CONFIG --libs ncurses` || CURSESLIBS='-lncurses' fi if test "$ac_cv_header_ncurses_ncurses_h" = "yes"; then AC_CHECK_LIB(ncurses, printw, [AC_DEFINE(HAVE_LIBNCURSES,1, have ncurses library)]) CURSESLIBS=`$PKG_CONFIG --libs ncurses` || CURSESLIBS='-lncurses' fi dnl Only look for non-n-library if there was no n-library. if test X"$CURSESLIBS" = X"" -a "$ac_cv_header_curses_h" = "yes"; then AC_CHECK_LIB(curses, printw, [CURSESLIBS='-lcurses'; AC_DEFINE(HAVE_LIBCURSES,1, have curses library)]) fi dnl Only look for non-n-library if there was no n-library. if test X"$CURSESLIBS" = X"" -a "$ac_cv_header_curses_curses_h" = "yes"; then AC_CHECK_LIB(curses, printw, [CURSESLIBS='-lcurses'; AC_DEFINE(HAVE_LIBCURSES,1, have curses library)]) fi if test "x$CURSESLIBS" != "x"; then PCMK_FEATURES="$PCMK_FEATURES ncurses" fi dnl Check for printw() prototype compatibility if test X"$CURSESLIBS" != X"" && cc_supports_flag -Wcast-qual; then ac_save_LIBS=$LIBS LIBS="$CURSESLIBS" cc_temp_flags "-Wcast-qual $WERROR" # avoid broken test because of hardened build environment in Fedora 23+ # - https://fedoraproject.org/wiki/Changes/Harden_All_Packages # - https://bugzilla.redhat.com/1297985 if cc_supports_flag -fPIC; then CFLAGS="$CFLAGS -fPIC" fi AC_MSG_CHECKING(whether printw() requires argument of "const char *") AC_LINK_IFELSE( [AC_LANG_PROGRAM([ #if defined(HAVE_NCURSES_H) # include #elif defined(HAVE_NCURSES_NCURSES_H) # include #elif defined(HAVE_CURSES_H) # include #endif ], [printw((const char *)"Test");] )], [pcmk_cv_compatible_printw=yes], [pcmk_cv_compatible_printw=no] ) LIBS=$ac_save_LIBS cc_restore_flags AC_MSG_RESULT([$pcmk_cv_compatible_printw]) if test "$pcmk_cv_compatible_printw" = no; then AC_MSG_WARN([The printw() function of your ncurses or curses library is old, we will disable usage of the library. If you want to use this library anyway, please update to newer version of the library, ncurses 5.4 or later is recommended. You can get the library from http://www.gnu.org/software/ncurses/.]) AC_MSG_NOTICE([Disabling curses]) AC_DEFINE(HAVE_INCOMPATIBLE_PRINTW, 1, [Do we have incompatible printw() in curses library?]) fi fi AC_SUBST(CURSESLIBS) dnl ======================================================================== dnl Profiling and GProf dnl ======================================================================== AC_MSG_NOTICE(Old CFLAGS: $CFLAGS) case $SUPPORT_COVERAGE in 1|yes|true) SUPPORT_PROFILING=1 PCMK_FEATURES="$PCMK_FEATURES coverage" CFLAGS="$CFLAGS -fprofile-arcs -ftest-coverage" dnl During linking, make sure to specify -lgcov or -coverage ;; esac case $SUPPORT_PROFILING in 1|yes|true) SUPPORT_PROFILING=1 dnl Disable various compiler optimizations CFLAGS="$CFLAGS -fno-omit-frame-pointer -fno-inline -fno-builtin " dnl CFLAGS="$CFLAGS -fno-inline-functions -fno-default-inline -fno-inline-functions-called-once -fno-optimize-sibling-calls" dnl Turn off optimization so tools can get accurate line numbers CFLAGS=`echo $CFLAGS | sed -e 's/-O.\ //g' -e 's/-Wp,-D_FORTIFY_SOURCE=.\ //g' -e 's/-D_FORTIFY_SOURCE=.\ //g'` CFLAGS="$CFLAGS -O0 -g3 -gdwarf-2" dnl Update features PCMK_FEATURES="$PCMK_FEATURES profile" ;; *) SUPPORT_PROFILING=0 ;; esac AC_MSG_NOTICE(New CFLAGS: $CFLAGS) AC_DEFINE_UNQUOTED(SUPPORT_PROFILING, $SUPPORT_PROFILING, Support for profiling) dnl ======================================================================== dnl Cluster infrastructure - LibQB dnl ======================================================================== if test x${enable_no_stack} = xyes; then SUPPORT_CS=no fi PKG_CHECK_MODULES(libqb, libqb >= 0.13) CPPFLAGS="$libqb_CFLAGS $CPPFLAGS" LIBS="$libqb_LIBS $LIBS" dnl libqb 0.14.0+ (2012-06) AC_CHECK_LIB(qb, qb_ipcs_connection_auth_set) PCMK_FEATURES="$PCMK_FEATURES libqb-logging libqb-ipc" dnl libqb 0.17.0+ (2014-02) AC_CHECK_FUNCS(qb_ipcs_connection_get_buffer_size, AC_DEFINE(HAVE_IPCS_GET_BUFFER_SIZE, 1, [Have qb_ipcc_get_buffer_size function])) dnl libqb not yet released (as of 2018-05) CHECK_ENUM_VALUE([qb/qblog.h],[qb_log_conf],[QB_LOG_CONF_MAX_LINE_LEN]) CHECK_ENUM_VALUE([qb/qblog.h],[qb_log_conf],[QB_LOG_CONF_ELLIPSIS]) dnl Support Linux-HA fence agents if available if test "$cross_compiling" != "yes"; then CPPFLAGS="$CPPFLAGS -I${prefix}/include/heartbeat" fi AC_CHECK_HEADERS(stonith/stonith.h) if test "$ac_cv_header_stonith_stonith_h" = "yes"; then dnl On Debian, AC_CHECK_LIBS fail if a library has any unresolved symbols dnl So check for all the dependencies (so they're added to LIBS) before checking for -lplumb AC_CHECK_LIB(pils, PILLoadPlugin) AC_CHECK_LIB(plumb, G_main_add_IPC_Channel) PCMK_FEATURES="$PCMK_FEATURES lha-fencing" fi AM_CONDITIONAL([BUILD_LHA_SUPPORT], [test "$ac_cv_header_stonith_stonith_h" = "yes"]) dnl =============================================== dnl Variables needed for substitution dnl =============================================== CRM_SCHEMA_DIRECTORY="${datadir}/pacemaker" AC_DEFINE_UNQUOTED(CRM_SCHEMA_DIRECTORY,"$CRM_SCHEMA_DIRECTORY", Location for the Pacemaker Relax-NG Schema) AC_SUBST(CRM_SCHEMA_DIRECTORY) CRM_CORE_DIR="${localstatedir}/lib/pacemaker/cores" AC_DEFINE_UNQUOTED(CRM_CORE_DIR,"$CRM_CORE_DIR", Location to store core files produced by Pacemaker daemons) AC_SUBST(CRM_CORE_DIR) CRM_DAEMON_USER="hacluster" AC_DEFINE_UNQUOTED(CRM_DAEMON_USER,"$CRM_DAEMON_USER", User to run Pacemaker daemons as) AC_SUBST(CRM_DAEMON_USER) CRM_DAEMON_GROUP="haclient" AC_DEFINE_UNQUOTED(CRM_DAEMON_GROUP,"$CRM_DAEMON_GROUP", Group to run Pacemaker daemons as) AC_SUBST(CRM_DAEMON_GROUP) CRM_STATE_DIR=${localstatedir}/run/crm AC_DEFINE_UNQUOTED(CRM_STATE_DIR,"$CRM_STATE_DIR", Where to keep state files and sockets) AC_SUBST(CRM_STATE_DIR) CRM_PACEMAKER_DIR=${localstatedir}/lib/pacemaker AC_DEFINE_UNQUOTED(CRM_PACEMAKER_DIR,"$CRM_PACEMAKER_DIR", Location to store directory produced by Pacemaker daemons) AC_SUBST(CRM_PACEMAKER_DIR) CRM_BLACKBOX_DIR=${localstatedir}/lib/pacemaker/blackbox AC_DEFINE_UNQUOTED(CRM_BLACKBOX_DIR,"$CRM_BLACKBOX_DIR", Where to keep blackbox dumps) AC_SUBST(CRM_BLACKBOX_DIR) PE_STATE_DIR="${localstatedir}/lib/pacemaker/pengine" AC_DEFINE_UNQUOTED(PE_STATE_DIR,"$PE_STATE_DIR", Where to keep scheduler outputs) AC_SUBST(PE_STATE_DIR) CRM_CONFIG_DIR="${localstatedir}/lib/pacemaker/cib" AC_DEFINE_UNQUOTED(CRM_CONFIG_DIR,"$CRM_CONFIG_DIR", Where to keep configuration files) AC_SUBST(CRM_CONFIG_DIR) CRM_CONFIG_CTS="${localstatedir}/lib/pacemaker/cts" AC_DEFINE_UNQUOTED(CRM_CONFIG_CTS,"$CRM_CONFIG_CTS", Where to keep cts stateful data) AC_SUBST(CRM_CONFIG_CTS) CRM_DAEMON_DIR="${libexecdir}/pacemaker" AC_DEFINE_UNQUOTED(CRM_DAEMON_DIR,"$CRM_DAEMON_DIR", Location for Pacemaker daemons) AC_SUBST(CRM_DAEMON_DIR) HA_STATE_DIR="${localstatedir}/run" AC_DEFINE_UNQUOTED(HA_STATE_DIR,"$HA_STATE_DIR", Where sbd keeps its PID file) AC_SUBST(HA_STATE_DIR) CRM_RSCTMP_DIR="${localstatedir}/run/resource-agents" AC_DEFINE_UNQUOTED(CRM_RSCTMP_DIR,"$CRM_RSCTMP_DIR", Where resource agents should keep state files) AC_SUBST(CRM_RSCTMP_DIR) PACEMAKER_CONFIG_DIR="${sysconfdir}/pacemaker" AC_DEFINE_UNQUOTED(PACEMAKER_CONFIG_DIR,"$PACEMAKER_CONFIG_DIR", Where to keep configuration files like authkey) AC_SUBST(PACEMAKER_CONFIG_DIR) OCF_ROOT_DIR="/usr/lib/ocf" if test "X$OCF_ROOT_DIR" = X; then AC_MSG_ERROR(Could not locate OCF directory) fi AC_SUBST(OCF_ROOT_DIR) OCF_RA_DIR="$OCF_ROOT_DIR/resource.d" AC_DEFINE_UNQUOTED(OCF_RA_DIR,"$OCF_RA_DIR", Location for OCF RAs) AC_SUBST(OCF_RA_DIR) RH_STONITH_DIR="$sbindir" AC_DEFINE_UNQUOTED(RH_STONITH_DIR,"$RH_STONITH_DIR", Location for Red Hat Stonith agents) AC_DEFINE_UNQUOTED(SBIN_DIR,"$sbindir", Location for system binaries) RH_STONITH_PREFIX="fence_" AC_DEFINE_UNQUOTED(RH_STONITH_PREFIX,"$RH_STONITH_PREFIX", Prefix for Red Hat Stonith agents) AC_PATH_PROGS(GIT, git false) AC_MSG_CHECKING(build version) BUILD_VERSION=$Format:%h$ if test $BUILD_VERSION != ":%h$"; then AC_MSG_RESULT(archive hash: $BUILD_VERSION) elif test -x $GIT -a -d .git; then BUILD_VERSION=`$GIT log --pretty="format:%h" -n 1` AC_MSG_RESULT(git hash: $BUILD_VERSION) else # The current directory name make a reasonable default # Most generated archives will include the hash or tag BASE=`basename $PWD` BUILD_VERSION=`echo $BASE | sed s:.*[[Pp]]acemaker-::` AC_MSG_RESULT(directory based hash: $BUILD_VERSION) fi AC_DEFINE_UNQUOTED(BUILD_VERSION, "$BUILD_VERSION", Build version) AC_SUBST(BUILD_VERSION) HAVE_dbus=1 PKG_CHECK_MODULES([DBUS], [dbus-1], [CPPFLAGS="${CPPFLAGS} ${DBUS_CFLAGS}"], [HAVE_dbus=0]) AC_DEFINE_UNQUOTED(SUPPORT_DBUS, $HAVE_dbus, Support dbus) AM_CONDITIONAL(BUILD_DBUS, test $HAVE_dbus = 1) AC_CHECK_TYPES([DBusBasicValue],,,[[#include ]]) if test "x${enable_systemd}" != xno; then if test $HAVE_dbus = 0; then if test "x${enable_systemd}" = xyes; then AC_MSG_FAILURE([cannot enable systemd without DBus]) else enable_systemd=no fi fi if test "x${enable_systemd}" = xtry; then AC_MSG_CHECKING([for systemd version query result via dbus-send]) ret=$({ dbus-send --system --print-reply \ --dest=org.freedesktop.systemd1 \ /org/freedesktop/systemd1 \ org.freedesktop.DBus.Properties.Get \ string:org.freedesktop.systemd1.Manager \ string:Version 2>/dev/null \ || echo "this borked"; } | tail -n1) # sanitize output a bit (interested just in value, not type), # ret is intentionally unenquoted so as to normalize whitespace ret=$(echo ${ret} | cut -d' ' -f2-) AC_MSG_RESULT([${ret}]) if test "x${ret}" != xborked \ || systemctl --version 2>/dev/null | grep -q systemd; then enable_systemd=yes else enable_systemd=no fi fi fi AC_MSG_CHECKING([whether to enable support for managing resources via systemd]) AC_MSG_RESULT([${enable_systemd}]) HAVE_systemd=0 if test "x${enable_systemd}" = xyes; then HAVE_systemd=1 PCMK_FEATURES="$PCMK_FEATURES systemd" AC_MSG_CHECKING([for systemd path for system unit files]) systemdunitdir="${systemdunitdir-}" PKG_CHECK_VAR([systemdunitdir], [systemd], [systemdsystemunitdir], [], [systemdunitdir=no]) AC_MSG_RESULT([${systemdunitdir}]) if test "x${systemdunitdir}" = xno; then AC_MSG_FAILURE([cannot enable systemd when systemdunitdir unresolved]) fi fi AC_SUBST(systemdunitdir) AC_DEFINE_UNQUOTED(SUPPORT_SYSTEMD, $HAVE_systemd, Support systemd based system services) AM_CONDITIONAL(BUILD_SYSTEMD, test $HAVE_systemd = 1) AC_SUBST(SUPPORT_SYSTEMD) if test "x${enable_upstart}" != xno; then if test $HAVE_dbus = 0; then if test "x${enable_upstart}" = xyes; then AC_MSG_FAILURE([cannot enable Upstart without DBus]) else enable_upstart=no fi fi if test "x${enable_upstart}" = xtry; then AC_MSG_CHECKING([for Upstart version query result via dbus-send]) ret=$({ dbus-send --system --print-reply --dest=com.ubuntu.Upstart \ /com/ubuntu/Upstart org.freedesktop.DBus.Properties.Get \ string:com.ubuntu.Upstart0_6 string:version 2>/dev/null \ || echo "this borked"; } | tail -n1) # sanitize output a bit (interested just in value, not type), # ret is intentionally unenquoted so as to normalize whitespace ret=$(echo ${ret} | cut -d' ' -f2-) AC_MSG_RESULT([${ret}]) if test "x${ret}" != xborked \ || initctl --version 2>/dev/null | grep -q upstart; then enable_upstart=yes else enable_upstart=no fi fi fi AC_MSG_CHECKING([whether to enable support for managing resources via Upstart]) AC_MSG_RESULT([${enable_upstart}]) HAVE_upstart=0 if test "x${enable_upstart}" = xyes; then HAVE_upstart=1 PCMK_FEATURES="$PCMK_FEATURES upstart" fi AC_DEFINE_UNQUOTED(SUPPORT_UPSTART, $HAVE_upstart, Support upstart based system services) AM_CONDITIONAL(BUILD_UPSTART, test $HAVE_upstart = 1) AC_SUBST(SUPPORT_UPSTART) case $SUPPORT_NAGIOS in 1|yes|true|try) SUPPORT_NAGIOS=1 ;; *) SUPPORT_NAGIOS=0 ;; esac if test $SUPPORT_NAGIOS = 1; then PCMK_FEATURES="$PCMK_FEATURES nagios" fi AC_DEFINE_UNQUOTED(SUPPORT_NAGIOS, $SUPPORT_NAGIOS, Support nagios plugins) AM_CONDITIONAL(BUILD_NAGIOS, test $SUPPORT_NAGIOS = 1) if test x"$NAGIOS_PLUGIN_DIR" = x""; then NAGIOS_PLUGIN_DIR="${libexecdir}/nagios/plugins" fi AC_DEFINE_UNQUOTED(NAGIOS_PLUGIN_DIR, "$NAGIOS_PLUGIN_DIR", Directory for nagios plugins) AC_SUBST(NAGIOS_PLUGIN_DIR) if test x"$NAGIOS_METADATA_DIR" = x""; then NAGIOS_METADATA_DIR="${datadir}/nagios/plugins-metadata" fi AC_DEFINE_UNQUOTED(NAGIOS_METADATA_DIR, "$NAGIOS_METADATA_DIR", Directory for nagios plugins metadata) AC_SUBST(NAGIOS_METADATA_DIR) STACKS="" CLUSTERLIBS="" dnl ======================================================================== dnl Cluster stack - Corosync dnl ======================================================================== dnl Normalize the values case $SUPPORT_CS in 1|yes|true) SUPPORT_CS=yes missingisfatal=1 ;; try) missingisfatal=0 ;; *) SUPPORT_CS=no ;; esac AC_MSG_CHECKING(for native corosync) COROSYNC_LIBS="" if test $SUPPORT_CS = no; then AC_MSG_RESULT(no (disabled)) SUPPORT_CS=0 else AC_MSG_RESULT($SUPPORT_CS) SUPPORT_CS=1 PKG_CHECK_MODULES(cpg, libcpg) dnl Fatal PKG_CHECK_MODULES(cfg, libcfg) dnl Fatal PKG_CHECK_MODULES(cmap, libcmap) dnl Fatal PKG_CHECK_MODULES(quorum, libquorum) dnl Fatal PKG_CHECK_MODULES(libcorosync_common, libcorosync_common) dnl Fatal CFLAGS="$CFLAGS $libqb_FLAGS $cpg_FLAGS $cfg_FLAGS $cmap_CFLAGS $quorum_CFLAGS $libcorosync_common_CFLAGS" COROSYNC_LIBS="$COROSYNC_LIBS $libqb_LIBS $cpg_LIBS $cfg_LIBS $cmap_LIBS $quorum_LIBS $libcorosync_common_LIBS" CLUSTERLIBS="$CLUSTERLIBS $COROSYNC_LIBS" STACKS="$STACKS corosync-native" fi AC_DEFINE_UNQUOTED(SUPPORT_COROSYNC, $SUPPORT_CS, Support the Corosync messaging and membership layer) AM_CONDITIONAL(BUILD_CS_SUPPORT, test $SUPPORT_CS = 1) AC_SUBST(SUPPORT_COROSYNC) dnl dnl Cluster stack - Sanity dnl if test x${enable_no_stack} = xyes; then AC_MSG_NOTICE(No cluster stack supported, building only the scheduler) PCMK_FEATURES="$PCMK_FEATURES no-cluster-stack" else AC_MSG_CHECKING(for supported stacks) if test x"$STACKS" = x; then AC_MSG_FAILURE(You must support at least one cluster stack) fi AC_MSG_RESULT($STACKS) PCMK_FEATURES="$PCMK_FEATURES $STACKS" fi PCMK_FEATURES="$PCMK_FEATURES atomic-attrd" AC_SUBST(CLUSTERLIBS) dnl ======================================================================== dnl ACL dnl ======================================================================== case $SUPPORT_ACL in 1|yes|true) missingisfatal=1 ;; try) missingisfatal=0 ;; *) SUPPORT_ACL=no ;; esac AC_MSG_CHECKING(for acl support) if test $SUPPORT_ACL = no; then AC_MSG_RESULT(no (disabled)) SUPPORT_ACL=0 else AC_MSG_RESULT($SUPPORT_ACL) SUPPORT_ACL=1 AC_CHECK_LIB(qb, qb_ipcs_connection_auth_set) if test $ac_cv_lib_qb_qb_ipcs_connection_auth_set != yes; then SUPPORT_ACL=0 fi if test $SUPPORT_ACL = 0; then if test $missingisfatal = 0; then AC_MSG_WARN(Unable to support ACL. You need to use libqb > 0.13.0) else AC_MSG_FAILURE(Unable to support ACL. You need to use libqb > 0.13.0) fi fi fi if test $SUPPORT_ACL = 1; then PCMK_FEATURES="$PCMK_FEATURES acls" fi AM_CONDITIONAL(ENABLE_ACL, test "$SUPPORT_ACL" = "1") AC_DEFINE_UNQUOTED(ENABLE_ACL, $SUPPORT_ACL, Build in support for CIB ACL) dnl ======================================================================== dnl CIB secrets dnl ======================================================================== case $SUPPORT_CIBSECRETS in 1|yes|true|try) SUPPORT_CIBSECRETS=1 ;; *) SUPPORT_CIBSECRETS=0 ;; esac AC_DEFINE_UNQUOTED(SUPPORT_CIBSECRETS, $SUPPORT_CIBSECRETS, Support CIB secrets) AM_CONDITIONAL(BUILD_CIBSECRETS, test $SUPPORT_CIBSECRETS = 1) if test $SUPPORT_CIBSECRETS = 1; then PCMK_FEATURES="$PCMK_FEATURES cibsecrets" LRM_CIBSECRETS_DIR="${localstatedir}/lib/pacemaker/lrm/secrets" AC_DEFINE_UNQUOTED(LRM_CIBSECRETS_DIR,"$LRM_CIBSECRETS_DIR", Location for CIB secrets) AC_SUBST(LRM_CIBSECRETS_DIR) fi dnl ======================================================================== dnl GnuTLS dnl ======================================================================== dnl gnutls_priority_set_direct available since 2.1.7 (released 2007-11-29) AC_CHECK_LIB(gnutls, gnutls_priority_set_direct) if test "$ac_cv_lib_gnutls_gnutls_priority_set_direct" != ""; then AC_CHECK_HEADERS(gnutls/gnutls.h) AC_CHECK_FUNCS([gnutls_sec_param_to_pk_bits]) dnl since 2.12.0 (2011-03-24) fi dnl ======================================================================== dnl PAM dnl ======================================================================== AC_CHECK_HEADERS(security/pam_appl.h pam/pam_appl.h) dnl ======================================================================== dnl System Health dnl ======================================================================== dnl Check if servicelog development package is installed SERVICELOG=servicelog-1 SERVICELOG_EXISTS="no" AC_MSG_CHECKING(for $SERVICELOG packages) if $PKG_CONFIG --exists $SERVICELOG then PKG_CHECK_MODULES([SERVICELOG], [servicelog-1]) SERVICELOG_EXISTS="yes" fi AC_MSG_RESULT($SERVICELOG_EXISTS) AM_CONDITIONAL(BUILD_SERVICELOG, test "$SERVICELOG_EXISTS" = "yes") dnl Check if OpenIMPI packages and servicelog are installed OPENIPMI="OpenIPMI OpenIPMIposix" OPENIPMI_SERVICELOG_EXISTS="no" AC_MSG_CHECKING(for $SERVICELOG $OPENIPMI packages) if $PKG_CONFIG --exists $OPENIPMI $SERVICELOG then PKG_CHECK_MODULES([OPENIPMI_SERVICELOG],[OpenIPMI OpenIPMIposix]) OPENIPMI_SERVICELOG_EXISTS="yes" fi AC_MSG_RESULT($OPENIPMI_SERVICELOG_EXISTS) AM_CONDITIONAL(BUILD_OPENIPMI_SERVICELOG, test "$OPENIPMI_SERVICELOG_EXISTS" = "yes") dnl ======================================================================== dnl Compiler flags dnl ======================================================================== dnl Make sure that CFLAGS is not exported. If the user did dnl not have CFLAGS in their environment then this should have dnl no effect. However if CFLAGS was exported from the user's dnl environment, then the new CFLAGS will also be exported dnl to sub processes. if export | fgrep " CFLAGS=" > /dev/null; then SAVED_CFLAGS="$CFLAGS" unset CFLAGS CFLAGS="$SAVED_CFLAGS" unset SAVED_CFLAGS fi AC_ARG_VAR([CFLAGS_HARDENED_LIB], [extra C compiler flags for hardened libraries]) AC_ARG_VAR([LDFLAGS_HARDENED_LIB], [extra linker flags for hardened libraries]) AC_ARG_VAR([CFLAGS_HARDENED_EXE], [extra C compiler flags for hardened executables]) AC_ARG_VAR([LDFLAGS_HARDENED_EXE], [extra linker flags for hardened executables]) CC_EXTRAS="" if test "$GCC" != yes; then CFLAGS="$CFLAGS -g" else CFLAGS="$CFLAGS -ggdb" dnl When we don't have diagnostic push / pull, we can't explicitly disable dnl checking for nonliteral formats in the places where they occur on purpose dnl thus we disable nonliteral format checking globally as we are aborting dnl on warnings. dnl what makes the things really ugly is that nonliteral format checking is dnl obviously available as an extra switch in very modern gcc but for older dnl gcc this is part of -Wformat=2 dnl so if we have push/pull we can enable -Wformat=2 -Wformat-nonliteral dnl if we don't have push/pull but -Wformat-nonliteral we can enable -Wformat=2 dnl otherwise none of both gcc_diagnostic_push_pull=no cc_temp_flags "$CFLAGS $WERROR" AC_MSG_CHECKING([for gcc diagnostic push / pull]) AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[ #pragma GCC diagnostic push #pragma GCC diagnostic pop ]])], [ AC_MSG_RESULT([yes]) gcc_diagnostic_push_pull=yes ], AC_MSG_RESULT([no])) cc_restore_flags if cc_supports_flag "-Wformat-nonliteral"; then gcc_format_nonliteral=yes else gcc_format_nonliteral=no fi # We had to eliminate -Wnested-externs because of libtool changes # Make sure to order options so that the former stand for prerequisites # of the latter (e.g., -Wformat-nonliteral requires -Wformat). EXTRA_FLAGS="-fgnu89-inline -Wall -Waggregate-return -Wbad-function-cast -Wcast-align -Wdeclaration-after-statement -Wendif-labels -Wfloat-equal -Wformat-security -Wmissing-prototypes -Wmissing-declarations -Wnested-externs -Wno-long-long -Wno-strict-aliasing -Wpointer-arith -Wstrict-prototypes -Wwrite-strings -Wunused-but-set-variable -Wunsigned-char" if test "x$gcc_diagnostic_push_pull" = "xyes"; then AC_DEFINE([GCC_FORMAT_NONLITERAL_CHECKING_ENABLED], [], [gcc can complain about nonliterals in format]) EXTRA_FLAGS="$EXTRA_FLAGS -Wformat=2 -Wformat-nonliteral" else if test "x$gcc_format_nonliteral" = "xyes"; then EXTRA_FLAGS="$EXTRA_FLAGS -Wformat=2" fi fi # Additional warnings it might be nice to enable one day # -Wshadow # -Wunreachable-code for j in $EXTRA_FLAGS do if cc_supports_flag $CC_EXTRAS $j then CC_EXTRAS="$CC_EXTRAS $j" fi done if test "x${enable_ansi}" = xyes && cc_supports_flag -std=iso9899:199409 ; then AC_MSG_NOTICE(Enabling ANSI Compatibility) CC_EXTRAS="$CC_EXTRAS -ansi -D_GNU_SOURCE -DANSI_ONLY" fi AC_MSG_NOTICE(Activated additional gcc flags: ${CC_EXTRAS}) fi dnl dnl Hardening flags dnl dnl The prime control of whether to apply (targeted) hardening build flags and dnl which ones is --{enable,disable}-hardening option passed to ./configure: dnl dnl --enable-hardening=try (default): dnl depending on whether any of CFLAGS_HARDENED_EXE, LDFLAGS_HARDENED_EXE, dnl CFLAGS_HARDENED_LIB or LDFLAGS_HARDENED_LIB environment variables dnl (see below) is set and non-null, all these custom flags (even if not dnl set) are used as are, otherwise the best effort is made to offer dnl reasonably strong hardening in several categories (RELRO, PIE, dnl "bind now", stack protector) according to what the selected toolchain dnl can offer dnl dnl --enable-hardening: dnl same effect as --enable-hardening=try when the environment variables dnl in question are suppressed dnl dnl --disable-hardening: dnl do not apply any targeted hardening measures at all dnl dnl The user-injected environment variables that regulate the hardening in dnl default case are as follows: dnl dnl * CFLAGS_HARDENED_EXE, LDFLAGS_HARDENED_EXE dnl compiler and linker flags (respectively) for daemon programs dnl (pacemakerd, pacemaker-attrd, pacemaker-controld, pacemaker-execd, dnl cib, stonithd, pacemaker-remoted, pacemaker-schedulerd) dnl dnl * CFLAGS_HARDENED_LIB, LDFLAGS_HARDENED_LIB dnl compiler and linker flags (respectively) for libraries linked dnl with the daemon programs dnl dnl Note that these are purposedly targeted variables (addressing particular dnl targets all over the scattered Makefiles) and have no effect outside of dnl the predestined scope (e.g., CLI utilities). For a global reach, dnl use CFLAGS, LDFLAGS, etc. as usual. dnl dnl For guidance on the suitable flags consult, for instance: dnl https://fedoraproject.org/wiki/Changes/Harden_All_Packages#Detailed_Harden_Flags_Description dnl https://owasp.org/index.php/C-Based_Toolchain_Hardening#GCC.2FBinutils dnl if test "x${HARDENING}" != "xtry"; then unset CFLAGS_HARDENED_EXE unset CFLAGS_HARDENED_LIB unset LDFLAGS_HARDENED_EXE unset LDFLAGS_HARDENED_LIB fi if test "x${HARDENING}" = "xno"; then AC_MSG_NOTICE([Hardening: explicitly disabled]) elif test "x${HARDENING}" = "xyes" \ || test "$(env | grep -Ec '^(C|LD)FLAGS_HARDENED_(EXE|LIB)=.')" = 0; then dnl We'll figure out on our own... CFLAGS_HARDENED_EXE= CFLAGS_HARDENED_LIB= LDFLAGS_HARDENED_EXE= LDFLAGS_HARDENED_LIB= relro=0 pie=0 bindnow=0 # daemons incl. libs: partial RELRO flag="-Wl,-z,relro" CC_CHECK_LDFLAGS(["${flag}"], [LDFLAGS_HARDENED_EXE="${LDFLAGS_HARDENED_EXE} ${flag}"; LDFLAGS_HARDENED_LIB="${LDFLAGS_HARDENED_LIB} ${flag}"; relro=1]) # daemons: PIE for both CFLAGS and LDFLAGS if cc_supports_flag -fPIE; then flag="-pie" CC_CHECK_LDFLAGS(["${flag}"], [CFLAGS_HARDENED_EXE="${CFLAGS_HARDENED_EXE} -fPIE"; LDFLAGS_HARDENED_EXE="${LDFLAGS_HARDENED_EXE} ${flag}"; pie=1]) fi # daemons incl. libs: full RELRO if sensible + as-needed linking # so as to possibly mitigate startup performance # hit caused by excessive linking with unneeded # libraries if test "${relro}" = 1 && test "${pie}" = 1; then flag="-Wl,-z,now" CC_CHECK_LDFLAGS(["${flag}"], [LDFLAGS_HARDENED_EXE="${LDFLAGS_HARDENED_EXE} ${flag}"; LDFLAGS_HARDENED_LIB="${LDFLAGS_HARDENED_LIB} ${flag}"; bindnow=1]) fi if test "${bindnow}" = 1; then flag="-Wl,--as-needed" CC_CHECK_LDFLAGS(["${flag}"], [LDFLAGS_HARDENED_EXE="${LDFLAGS_HARDENED_EXE} ${flag}"; LDFLAGS_HARDENED_LIB="${LDFLAGS_HARDENED_LIB} ${flag}"]) fi # universal: prefer strong > all > default stack protector if possible flag= if cc_supports_flag -fstack-protector-strong; then flag="-fstack-protector-strong" elif cc_supports_flag -fstack-protector-all; then flag="-fstack-protector-all" elif cc_supports_flag -fstack-protector; then flag="-fstack-protector" fi if test -n "${flag}"; then CC_EXTRAS="${CC_EXTRAS} ${flag}" stackprot=1 fi if test "${relro}" = 1 \ || test "${pie}" = 1 \ || test "${stackprot}" = 1; then AC_MSG_NOTICE([Hardening: relro=${relro} pie=${pie} bindnow=${bindnow} stackprot=${flag}]) else AC_MSG_WARN([Hardening: no suitable features in the toolchain detected]) fi else AC_MSG_NOTICE([Hardening: using custom flags]) fi CFLAGS="$CFLAGS $CC_EXTRAS" NON_FATAL_CFLAGS="$CFLAGS" AC_SUBST(NON_FATAL_CFLAGS) dnl dnl We reset CFLAGS to include our warnings *after* all function dnl checking goes on, so that our warning flags don't keep the dnl AC_*FUNCS() calls above from working. In particular, -Werror will dnl *always* cause us troubles if we set it before here. dnl dnl if test "x${enable_fatal_warnings}" = xyes ; then AC_MSG_NOTICE(Enabling Fatal Warnings) CFLAGS="$CFLAGS $WERROR" fi AC_SUBST(CFLAGS) dnl This is useful for use in Makefiles that need to remove one specific flag CFLAGS_COPY="$CFLAGS" AC_SUBST(CFLAGS_COPY) AC_SUBST(LIBADD_DL) dnl extra flags for dynamic linking libraries AC_SUBST(LIBADD_INTL) dnl extra flags for GNU gettext stuff... AC_SUBST(LOCALE) dnl Options for cleaning up the compiler output QUIET_LIBTOOL_OPTS="" QUIET_MAKE_OPTS="" if test "x${enable_quiet}" = "xyes"; then QUIET_LIBTOOL_OPTS="--silent" QUIET_MAKE_OPTS="-s" # POSIX compliant fi AC_MSG_RESULT(Suppress make details: ${enable_quiet}) dnl Put the above variables to use LIBTOOL="${LIBTOOL} --tag=CC \$(QUIET_LIBTOOL_OPTS)" MAKEFLAGS="${MAKEFLAGS} ${QUIET_MAKE_OPTS}" AC_SUBST(CC) AC_SUBST(MAKEFLAGS) AC_SUBST(LIBTOOL) AC_SUBST(QUIET_LIBTOOL_OPTS) AC_DEFINE_UNQUOTED(CRM_FEATURES, "$PCMK_FEATURES", Set of enabled features) AC_SUBST(PCMK_FEATURES) dnl Files we output that need to be executable AC_CONFIG_FILES([cts/CTSlab.py], [chmod +x cts/CTSlab.py]) AC_CONFIG_FILES([cts/LSBDummy], [chmod +x cts/LSBDummy]) AC_CONFIG_FILES([cts/OCFIPraTest.py], [chmod +x cts/OCFIPraTest.py]) AC_CONFIG_FILES([cts/cluster_test], [chmod +x cts/cluster_test]) AC_CONFIG_FILES([cts/cts], [chmod +x cts/cts]) AC_CONFIG_FILES([cts/cts-cli], [chmod +x cts/cts-cli]) AC_CONFIG_FILES([cts/cts-coverage], [chmod +x cts/cts-coverage]) AC_CONFIG_FILES([cts/cts-exec], [chmod +x cts/cts-exec]) AC_CONFIG_FILES([cts/cts-fencing], [chmod +x cts/cts-fencing]) AC_CONFIG_FILES([cts/cts-log-watcher], [chmod +x cts/cts-log-watcher]) AC_CONFIG_FILES([cts/cts-regression], [chmod +x cts/cts-regression]) AC_CONFIG_FILES([cts/cts-scheduler], [chmod +x cts/cts-scheduler]) AC_CONFIG_FILES([cts/cts-support], [chmod +x cts/cts-support]) AC_CONFIG_FILES([cts/lxc_autogen.sh], [chmod +x cts/lxc_autogen.sh]) AC_CONFIG_FILES([cts/benchmark/clubench], [chmod +x cts/benchmark/clubench]) AC_CONFIG_FILES([cts/fence_dummy], [chmod +x cts/fence_dummy]) AC_CONFIG_FILES([cts/pacemaker-cts-dummyd], [chmod +x cts/pacemaker-cts-dummyd]) AC_CONFIG_FILES([daemons/fenced/fence_legacy], [chmod +x daemons/fenced/fence_legacy]) AC_CONFIG_FILES([doc/abi-check], [chmod +x doc/abi-check]) AC_CONFIG_FILES([extra/resources/ClusterMon], [chmod +x extra/resources/ClusterMon]) AC_CONFIG_FILES([extra/resources/HealthSMART], [chmod +x extra/resources/HealthSMART]) AC_CONFIG_FILES([extra/resources/SysInfo], [chmod +x extra/resources/SysInfo]) AC_CONFIG_FILES([extra/resources/ifspeed], [chmod +x extra/resources/ifspeed]) AC_CONFIG_FILES([extra/resources/o2cb], [chmod +x extra/resources/o2cb]) AC_CONFIG_FILES([tools/crm_failcount], [chmod +x tools/crm_failcount]) AC_CONFIG_FILES([tools/crm_master], [chmod +x tools/crm_master]) AC_CONFIG_FILES([tools/crm_report], [chmod +x tools/crm_report]) AC_CONFIG_FILES([tools/crm_standby], [chmod +x tools/crm_standby]) AC_CONFIG_FILES([tools/cibsecret], [chmod +x tools/cibsecret]) dnl Other files we output AC_CONFIG_FILES(Makefile \ cts/Makefile \ cts/CTS.py \ cts/CTSvars.py \ cts/benchmark/Makefile \ cts/pacemaker-cts-dummyd@.service \ daemons/Makefile \ daemons/attrd/Makefile \ daemons/based/Makefile \ daemons/controld/Makefile \ daemons/execd/Makefile \ daemons/execd/pacemaker_remote \ daemons/execd/pacemaker_remote.service \ daemons/fenced/Makefile \ daemons/pacemakerd/Makefile \ daemons/pacemakerd/pacemaker \ daemons/pacemakerd/pacemaker.service \ daemons/pacemakerd/pacemaker.upstart \ daemons/pacemakerd/pacemaker.combined.upstart \ daemons/schedulerd/Makefile \ doc/Doxyfile \ doc/Makefile \ doc/Clusters_from_Scratch/publican.cfg \ doc/Pacemaker_Administration/publican.cfg \ doc/Pacemaker_Development/publican.cfg \ doc/Pacemaker_Explained/publican.cfg \ doc/Pacemaker_Remote/publican.cfg \ extra/Makefile \ extra/alerts/Makefile \ extra/resources/Makefile \ extra/logrotate/Makefile \ extra/logrotate/pacemaker \ include/Makefile \ include/crm/Makefile \ include/crm/cib/Makefile \ include/crm/common/Makefile \ include/crm/cluster/Makefile \ include/crm/fencing/Makefile \ include/crm/pengine/Makefile \ include/pcmki/Makefile \ replace/Makefile \ lib/Makefile \ lib/libpacemaker.pc \ lib/pacemaker.pc \ lib/pacemaker-cib.pc \ lib/pacemaker-lrmd.pc \ lib/pacemaker-service.pc \ lib/pacemaker-pe_rules.pc \ lib/pacemaker-pe_status.pc \ lib/pacemaker-fencing.pc \ lib/pacemaker-cluster.pc \ lib/common/Makefile \ lib/cluster/Makefile \ lib/cib/Makefile \ lib/gnu/Makefile \ lib/pacemaker/Makefile \ lib/pengine/Makefile \ lib/fencing/Makefile \ lib/lrmd/Makefile \ lib/services/Makefile \ maint/Makefile \ tools/Makefile \ tools/report.collector \ tools/report.common \ tools/crm_mon.service \ tools/crm_mon.upstart \ xml/Makefile \ ) dnl Now process the entire list of files added by previous dnl calls to AC_CONFIG_FILES() AC_OUTPUT() dnl ***************** dnl Configure summary dnl ***************** AC_MSG_RESULT([]) AC_MSG_RESULT([$PACKAGE configuration:]) AC_MSG_RESULT([ Version = ${VERSION} (Build: $BUILD_VERSION)]) AC_MSG_RESULT([ Features =${PCMK_FEATURES}]) AC_MSG_RESULT([]) AC_MSG_RESULT([ Prefix = ${prefix}]) AC_MSG_RESULT([ Executables = ${sbindir}]) AC_MSG_RESULT([ Man pages = ${mandir}]) AC_MSG_RESULT([ Libraries = ${libdir}]) AC_MSG_RESULT([ Header files = ${includedir}]) AC_MSG_RESULT([ Arch-independent files = ${datadir}]) AC_MSG_RESULT([ State information = ${localstatedir}]) AC_MSG_RESULT([ System configuration = ${sysconfdir}]) AC_MSG_RESULT([]) AC_MSG_RESULT([ HA group name = ${CRM_DAEMON_GROUP}]) AC_MSG_RESULT([ HA user name = ${CRM_DAEMON_USER}]) AC_MSG_RESULT([]) AC_MSG_RESULT([ CFLAGS = ${CFLAGS}]) AC_MSG_RESULT([ CFLAGS_HARDENED_EXE = ${CFLAGS_HARDENED_EXE}]) AC_MSG_RESULT([ CFLAGS_HARDENED_LIB = ${CFLAGS_HARDENED_LIB}]) AC_MSG_RESULT([ LDFLAGS_HARDENED_EXE = ${LDFLAGS_HARDENED_EXE}]) AC_MSG_RESULT([ LDFLAGS_HARDENED_LIB = ${LDFLAGS_HARDENED_LIB}]) AC_MSG_RESULT([ Libraries = ${LIBS}]) AC_MSG_RESULT([ Stack Libraries = ${CLUSTERLIBS}]) diff --git a/cts/cts-scheduler.in b/cts/cts-scheduler.in index fbfc004476..635085bfd9 100644 --- a/cts/cts-scheduler.in +++ b/cts/cts-scheduler.in @@ -1,1301 +1,1301 @@ #!@BASH_PATH@ # # Copyright 2004-2019 the Pacemaker project contributors # # The version control history for this file may have further details. # # This source code is licensed under the GNU General Public License version 2 # or later (GPLv2+) WITHOUT ANY WARRANTY. # USAGE_TEXT="Usage: cts-scheduler [] Options: --help Display this text, then exit -V, --verbose Display any differences from expected output --run TEST Run only single specified test --update Update expected results with actual results -b, --binary PATH Specify path to crm_simulate -i, --io-dir PATH Specify path to regression test data directory -v, --valgrind Run all commands under valgrind --valgrind-dhat Run all commands under valgrind with heap analyzer --valgrind-skip-output If running under valgrind, don't display output --testcmd-options Additional options for command under test" SBINDIR="@sbindir@" BUILDDIR="@abs_top_builddir@" CRM_SCHEMA_DIRECTORY="@CRM_SCHEMA_DIRECTORY@" # If readlink supports -e (i.e. GNU), use it readlink -e / >/dev/null 2>/dev/null if [ $? -eq 0 ]; then test_home="$(dirname "$(readlink -e "$0")")" else test_home="$(dirname "$0")" fi io_dir="$test_home/scheduler" failed="$test_home/.regression.failed.diff" test_binary= testcmd_options= single_test= verbose=0 num_failed=0 num_tests=0 VALGRIND_CMD="" VALGRIND_OPTS="-q --gen-suppressions=all --log-file=%q{valgrind_output} --time-stamp=yes --trace-children=no --show-reachable=no --leak-check=full --num-callers=20 --suppressions=$test_home/valgrind-pcmk.suppressions" VALGRIND_DHAT_OPTS="--tool=exp-dhat --log-file=%q{valgrind_output} --time-stamp=yes --trace-children=no --show-top-n=100 --num-callers=4" diff_opts="--ignore-all-space --ignore-blank-lines -u -N" # These constants must track crm_exit_t values CRM_EX_OK=0 CRM_EX_ERROR=1 CRM_EX_NOT_INSTALLED=5 CRM_EX_USAGE=64 CRM_EX_NOINPUT=66 EXITCODE=$CRM_EX_OK function info() { printf "$*\n" } function error() { printf " * ERROR: $*\n" } function failed() { printf " * FAILED: $*\n" } function show_test() { name=$1; shift printf " Test %-25s $*\n" "$name:" } # Normalize scheduler output for comparison normalize() { for NORMALIZE_FILE in "$@"; do # sed -i is not portable :-( sed -e 's/crm_feature_set="[^"]*"//' \ -e 's/batch-limit="[0-9]*"//' \ "$NORMALIZE_FILE" > "${NORMALIZE_FILE}.$$" mv -- "${NORMALIZE_FILE}.$$" "$NORMALIZE_FILE" done } info "Test home is:\t$test_home" create_mode="false" while [ $# -gt 0 ] ; do case "$1" in -V|--verbose) verbose=1 shift ;; -v|--valgrind) export G_SLICE=always-malloc VALGRIND_CMD="valgrind $VALGRIND_OPTS" shift ;; --valgrind-dhat) VALGRIND_CMD="valgrind $VALGRIND_DHAT_OPTS" shift ;; --valgrind-skip-output) VALGRIND_SKIP_OUTPUT=1 shift ;; --update) create_mode="true" shift ;; --run) single_test=$(basename "$2" ".xml") shift 2 break # any remaining arguments will be passed to test command ;; -b|--binary) test_binary="$2" shift 2 ;; -i|--io-dir) io_dir="$2" shift 2 ;; --help) echo "$USAGE_TEXT" exit $CRM_EX_OK ;; --testcmd-options) testcmd_options=$2 shift 2 ;; *) error "unknown option: $1" exit $CRM_EX_USAGE ;; esac done if [ -z "$PCMK_schema_directory" ]; then if [ -d "$BUILDDIR/xml" ]; then export PCMK_schema_directory="$BUILDDIR/xml" elif [ -d "$CRM_SCHEMA_DIRECTORY" ]; then export PCMK_schema_directory="$CRM_SCHEMA_DIRECTORY" fi fi if [ -z "$test_binary" ]; then if [ -x "$BUILDDIR/tools/crm_simulate" ]; then test_binary="$BUILDDIR/tools/crm_simulate" elif [ -x "$SBINDIR/crm_simulate" ]; then test_binary="$SBINDIR/crm_simulate" fi fi if [ ! -x "$test_binary" ]; then error "Test binary $test_binary not found" exit $CRM_EX_NOT_INSTALLED fi info "Test binary is:\t$test_binary" if [ -n "$PCMK_schema_directory" ]; then info "Schema home is:\t$PCMK_schema_directory" fi if [ "x$VALGRIND_CMD" != "x" ]; then info "Activating memory testing with valgrind"; fi info " " test_cmd="$VALGRIND_CMD $test_binary $testcmd_options" #echo $test_cmd if [ "$(whoami)" != "root" ]; then declare -x CIB_shadow_dir=/tmp fi do_test() { did_fail=0 expected_rc=0 num_tests=$(( $num_tests + 1 )) base=$1; shift name=$1; shift input=$io_dir/${base}.xml output=$io_dir/${base}.out expected=$io_dir/${base}.exp dot_expected=$io_dir/${base}.dot dot_output=$io_dir/${base}.pe.dot scores=$io_dir/${base}.scores score_output=$io_dir/${base}.scores.pe stderr_expected=$io_dir/${base}.stderr stderr_output=$io_dir/${base}.stderr.pe summary=$io_dir/${base}.summary summary_output=$io_dir/${base}.summary.pe valgrind_output=$io_dir/${base}.valgrind export valgrind_output if [ "x$1" = "x--rc" ]; then expected_rc=$2 shift; shift; fi show_test "$base" "$name" if [ ! -f $input ]; then error "No input"; did_fail=1 num_failed=$(( $num_failed + 1 )) return $CRM_EX_NOINPUT; fi if [ "$create_mode" != "true" ] && [ ! -f "$expected" ]; then error "no stored output"; return $CRM_EX_NOINPUT; fi # ../admin/crm_verify -X $input if [ ! -z "$single_test" ]; then echo "CIB_shadow_dir=\"$io_dir\" $test_cmd -x \"$input\" -D \"$dot_output\" -G \"$output\" -S" "$@" CIB_shadow_dir="$io_dir" $test_cmd -x "$input" -D "$dot_output" \ -G "$output" -S "$@" 2>&1 | tee "$summary_output" else CIB_shadow_dir="$io_dir" $test_cmd -x "$input" -S &> "$summary_output" fi CIB_shadow_dir="$io_dir" $test_cmd -x "$input" -D "$dot_output" \ -G "$output" -SQ -s "$@" 2> "$stderr_output" > "$score_output" rc=$? if [ $rc -ne $expected_rc ]; then failed "Test returned: $rc"; did_fail=1 echo "CIB_shadow_dir=\"$io_dir\" $test_cmd -x \"$input\" -D \"$dot_output\" -G \"$output\" -SQ -s" "$@" fi if [ -z "$VALGRIND_SKIP_OUTPUT" ]; then if [ -s "${valgrind_output}" ]; then error "Valgrind reported errors"; did_fail=1 cat ${valgrind_output} fi rm -f ${valgrind_output} fi if [ -s core ]; then error "Core-file detected: core.${base}"; did_fail=1 rm -f $test_home/core.$base mv core $test_home/core.$base fi if [ -e "$stderr_expected" ]; then diff $diff_opts $stderr_expected $stderr_output >/dev/null rc2=$? if [ $rc2 -ne 0 ]; then failed "stderr changed"; diff $diff_opts $stderr_expected $stderr_output 2>/dev/null >> $failed echo "" >> $failed did_fail=1 fi elif [ -s "$stderr_output" ]; then error "Output was written to stderr" did_fail=1 cat $stderr_output fi rm -f $stderr_output if [ ! -s $output ]; then error "No graph produced"; did_fail=1 num_failed=$(( $num_failed + 1 )) rm -f $output return $CRM_EX_ERROR; fi if [ ! -s $dot_output ]; then error "No dot-file summary produced"; did_fail=1 num_failed=$(( $num_failed + 1 )) rm -f $output return $CRM_EX_ERROR; else echo "digraph \"g\" {" > $dot_output.sort LC_ALL=POSIX sort -u $dot_output | grep -v -e '^}$' -e digraph >> $dot_output.sort echo "}" >> $dot_output.sort mv -f $dot_output.sort $dot_output fi if [ ! -s $score_output ]; then error "No allocation scores produced"; did_fail=1 num_failed=$(( $num_failed + 1 )) rm $output return $CRM_EX_ERROR; else LC_ALL=POSIX sort $score_output > $score_output.sorted mv -f $score_output.sorted $score_output fi if [ "$create_mode" = "true" ]; then cp "$output" "$expected" cp "$dot_output" "$dot_expected" cp "$score_output" "$scores" cp "$summary_output" "$summary" info " Updated expected outputs" fi diff $diff_opts $summary $summary_output >/dev/null rc2=$? if [ $rc2 -ne 0 ]; then failed "summary changed"; diff $diff_opts $summary $summary_output 2>/dev/null >> $failed echo "" >> $failed did_fail=1 fi diff $diff_opts $dot_expected $dot_output >/dev/null rc=$? if [ $rc -ne 0 ]; then failed "dot-file summary changed"; diff $diff_opts $dot_expected $dot_output 2>/dev/null >> $failed echo "" >> $failed did_fail=1 else rm -f $dot_output fi normalize "$expected" "$output" diff $diff_opts $expected $output >/dev/null rc2=$? if [ $rc2 -ne 0 ]; then failed "xml-file changed"; diff $diff_opts $expected $output 2>/dev/null >> $failed echo "" >> $failed did_fail=1 fi diff $diff_opts $scores $score_output >/dev/null rc=$? if [ $rc -ne 0 ]; then failed "scores-file changed"; diff $diff_opts $scores $score_output 2>/dev/null >> $failed echo "" >> $failed did_fail=1 fi rm -f $output $score_output $summary_output if [ $did_fail -eq 1 ]; then num_failed=$(( $num_failed + 1 )) return $CRM_EX_ERROR fi return $CRM_EX_OK } function test_results { if [ $num_failed -ne 0 ]; then if [ -s "$failed" ]; then if [ $verbose -eq 1 ]; then error "Results of $num_failed failed tests (out of $num_tests):" cat $failed else error "Results of $num_failed failed tests (out of $num_tests) are in $failed" error "Use -V to display them after running the tests" fi else error "$num_failed (of $num_tests) tests failed (no diff results)" rm $failed fi EXITCODE=$CRM_EX_ERROR fi } # zero out the error log true > $failed if [ -n "$single_test" ]; then do_test "$single_test" "Single shot" "$@" TEST_RC=$? cat "$failed" exit $TEST_RC fi DO_VERSIONED_TESTS=0 info Performing the following tests from $io_dir echo "" do_test simple1 "Offline " do_test simple2 "Start " do_test simple3 "Start 2 " do_test simple4 "Start Failed" do_test simple6 "Stop Start " do_test simple7 "Shutdown " #do_test simple8 "Stonith " #do_test simple9 "Lower version" #do_test simple10 "Higher version" do_test simple11 "Priority (ne)" do_test simple12 "Priority (eq)" do_test simple8 "Stickiness" echo "" do_test group1 "Group " do_test group2 "Group + Native " do_test group3 "Group + Group " do_test group4 "Group + Native (nothing)" do_test group5 "Group + Native (move) " do_test group6 "Group + Group (move) " do_test group7 "Group colocation" do_test group13 "Group colocation (cant run)" do_test group8 "Group anti-colocation" do_test group9 "Group recovery" do_test group10 "Group partial recovery" do_test group11 "Group target_role" do_test group14 "Group stop (graph terminated)" do_test group15 "Negative group colocation" do_test bug-1573 "Partial stop of a group with two children" do_test bug-1718 "Mandatory group ordering - Stop group_FUN" -@HASH_AFFECTED@do_test bug-lf-2613 "Move group on failure" -@HASH_AFFECTED@do_test bug-lf-2619 "Move group on clone failure" +do_test bug-lf-2613 "Move group on failure" +do_test bug-lf-2619 "Move group on clone failure" do_test group-fail "Ensure stop order is preserved for partially active groups" do_test group-unmanaged "No need to restart r115 because r114 is unmanaged" do_test group-unmanaged-stopped "Make sure r115 is stopped when r114 fails" do_test group-dependents "Account for the location preferences of things colocated with a group" do_test group-stop-ordering "Ensure blocked group member stop does not force other member stops" echo "" do_test rsc_dep1 "Must not " do_test rsc_dep3 "Must " do_test rsc_dep5 "Must not 3 " do_test rsc_dep7 "Must 3 " do_test rsc_dep10 "Must (but cant)" do_test rsc_dep2 "Must (running) " do_test rsc_dep8 "Must (running : alt) " do_test rsc_dep4 "Must (running + move)" do_test asymmetric "Asymmetric - require explicit location constraints" echo "" do_test orphan-0 "Orphan ignore" do_test orphan-1 "Orphan stop" do_test orphan-2 "Orphan stop, remove failcount" echo "" do_test params-0 "Params: No change" do_test params-1 "Params: Changed" do_test params-2 "Params: Resource definition" do_test params-4 "Params: Reload" do_test params-5 "Params: Restart based on probe digest" do_test novell-251689 "Resource definition change + target_role=stopped" do_test bug-lf-2106 "Restart all anonymous clone instances after config change" do_test params-6 "Params: Detect reload in previously migrated resource" do_test nvpair-id-ref "Support id-ref in nvpair with optional name" do_test not-reschedule-unneeded-monitor "Do not reschedule unneeded monitors while resource definitions have changed" do_test reload-becomes-restart "Cancel reload if restart becomes required" echo "" do_test target-0 "Target Role : baseline" do_test target-1 "Target Role : master" do_test target-2 "Target Role : invalid" echo "" do_test base-score "Set a node's default score for all nodes" echo "" do_test date-1 "Dates" -t "2005-020" do_test date-2 "Date Spec - Pass" -t "2005-020T12:30" do_test date-3 "Date Spec - Fail" -t "2005-020T11:30" do_test origin "Timing of recurring operations" -t "2014-05-07 00:28:00" do_test probe-0 "Probe (anon clone)" do_test probe-1 "Pending Probe" -@HASH_AFFECTED@do_test probe-2 "Correctly re-probe cloned groups" +do_test probe-2 "Correctly re-probe cloned groups" do_test probe-3 "Probe (pending node)" do_test probe-4 "Probe (pending node + stopped resource)" do_test standby "Standby" do_test comments "Comments" echo "" do_test one-or-more-0 "Everything starts" do_test one-or-more-1 "Nothing starts because of A" do_test one-or-more-2 "D can start because of C" do_test one-or-more-3 "D cannot start because of B and C" do_test one-or-more-4 "D cannot start because of target-role" do_test one-or-more-5 "Start A and F even though C and D are stopped" do_test one-or-more-6 "Leave A running even though B is stopped" do_test one-or-more-7 "Leave A running even though C is stopped" do_test bug-5140-require-all-false "Allow basegrp:0 to stop" do_test clone-require-all-1 "clone B starts node 3 and 4" do_test clone-require-all-2 "clone B remains stopped everywhere" do_test clone-require-all-3 "clone B stops everywhere because A stops everywhere" do_test clone-require-all-4 "clone B remains on node 3 and 4 with only one instance of A remaining." do_test clone-require-all-5 "clone B starts on node 1 3 and 4" do_test clone-require-all-6 "clone B remains active after shutting down instances of A" do_test clone-require-all-7 "clone A and B both start at the same time. all instances of A start before B." do_test clone-require-all-no-interleave-1 "C starts everywhere after A and B" do_test clone-require-all-no-interleave-2 "C starts on nodes 1, 2, and 4 with only one active instance of B" do_test clone-require-all-no-interleave-3 "C remains active when instance of B is stopped on one node and started on another." -@HASH_AFFECTED@do_test one-or-more-unrunnable-instances "Avoid dependencies on instances that won't ever be started" +do_test one-or-more-unrunnable-instances "Avoid dependencies on instances that won't ever be started" echo "" do_test order1 "Order start 1 " do_test order2 "Order start 2 " do_test order3 "Order stop " do_test order4 "Order (multiple) " do_test order5 "Order (move) " do_test order6 "Order (move w/ restart) " do_test order7 "Order (mandatory) " do_test order-optional "Order (score=0) " do_test order-required "Order (score=INFINITY) " do_test bug-lf-2171 "Prevent group start when clone is stopped" do_test order-clone "Clone ordering should be able to prevent startup of dependent clones" do_test order-sets "Ordering for resource sets" do_test order-serialize "Serialize resources without inhibiting migration" do_test order-serialize-set "Serialize a set of resources without inhibiting migration" do_test clone-order-primitive "Order clone start after a primitive" do_test clone-order-16instances "Verify ordering of 16 cloned resources" do_test order-optional-keyword "Order (optional keyword)" do_test order-mandatory "Order (mandatory keyword)" do_test bug-lf-2493 "Don't imply colocation requirements when applying ordering constraints with clones" do_test ordered-set-basic-startup "Constraint set with default order settings." do_test ordered-set-natural "Allow natural set ordering" do_test order-wrong-kind "Order (error)" echo "" do_test coloc-loop "Colocation - loop" do_test coloc-many-one "Colocation - many-to-one" do_test coloc-list "Colocation - many-to-one with list" do_test coloc-group "Colocation - groups" do_test coloc-slave-anti "Anti-colocation with slave shouldn't prevent master colocation" do_test coloc-attr "Colocation based on node attributes" do_test coloc-negative-group "Negative colocation with a group" do_test coloc-intra-set "Intra-set colocation" do_test bug-lf-2435 "Colocation sets with a negative score" do_test coloc-clone-stays-active "Ensure clones don't get stopped/demoted because a dependent must stop" do_test coloc_fp_logic "Verify floating point calculations in colocation are working" do_test colo_master_w_native "cl#5070 - Verify promotion order is affected when colocating master to native rsc." do_test colo_slave_w_native "cl#5070 - Verify promotion order is affected when colocating slave to native rsc." do_test anti-colocation-order "cl#5187 - Prevent resources in an anti-colocation from even temporarily running on a same node" do_test anti-colocation-master "Organize order of actions for master resources in anti-colocations" do_test anti-colocation-slave "Organize order of actions for slave resources in anti-colocations" do_test enforce-colo1 "Always enforce B with A INFINITY." do_test complex_enforce_colo "Always enforce B with A INFINITY. (make sure heat-engine stops)" echo "" do_test rsc-sets-seq-true "Resource Sets - sequential=false" do_test rsc-sets-seq-false "Resource Sets - sequential=true" do_test rsc-sets-clone "Resource Sets - Clone" do_test rsc-sets-master "Resource Sets - Master" do_test rsc-sets-clone-1 "Resource Sets - Clone (lf#2404)" #echo "" #do_test agent1 "version: lt (empty)" #do_test agent2 "version: eq " #do_test agent3 "version: gt " echo "" do_test attrs1 "string: eq (and) " do_test attrs2 "string: lt / gt (and)" do_test attrs3 "string: ne (or) " do_test attrs4 "string: exists " do_test attrs5 "string: not_exists " do_test attrs6 "is_dc: true " do_test attrs7 "is_dc: false " do_test attrs8 "score_attribute " do_test per-node-attrs "Per node resource parameters" echo "" do_test mon-rsc-1 "Schedule Monitor - start" do_test mon-rsc-2 "Schedule Monitor - move " do_test mon-rsc-3 "Schedule Monitor - pending start " do_test mon-rsc-4 "Schedule Monitor - move/pending start" echo "" do_test rec-rsc-0 "Resource Recover - no start " do_test rec-rsc-1 "Resource Recover - start " do_test rec-rsc-2 "Resource Recover - monitor " do_test rec-rsc-3 "Resource Recover - stop - ignore" do_test rec-rsc-4 "Resource Recover - stop - block " do_test rec-rsc-5 "Resource Recover - stop - fence " do_test rec-rsc-6 "Resource Recover - multiple - restart" do_test rec-rsc-7 "Resource Recover - multiple - stop " do_test rec-rsc-8 "Resource Recover - multiple - block " do_test rec-rsc-9 "Resource Recover - group/group" do_test monitor-recovery "on-fail=block + resource recovery detected by recurring monitor" do_test stop-failure-no-quorum "Stop failure without quorum" do_test stop-failure-no-fencing "Stop failure without fencing available" do_test stop-failure-with-fencing "Stop failure with fencing available" do_test multiple-active-block-group "Support of multiple-active=block for resource groups" do_test multiple-monitor-one-failed "Consider resource failed if any of the configured monitor operations failed" echo "" do_test quorum-1 "No quorum - ignore" do_test quorum-2 "No quorum - freeze" do_test quorum-3 "No quorum - stop " do_test quorum-4 "No quorum - start anyway" do_test quorum-5 "No quorum - start anyway (group)" do_test quorum-6 "No quorum - start anyway (clone)" -@HASH_AFFECTED@do_test bug-cl-5212 "No promotion with no-quorum-policy=freeze" +do_test bug-cl-5212 "No promotion with no-quorum-policy=freeze" do_test suicide-needed-inquorate "no-quorum-policy=suicide: suicide necessary" do_test suicide-not-needed-initial-quorum "no-quorum-policy=suicide: suicide not necessary at initial quorum" do_test suicide-not-needed-never-quorate "no-quorum-policy=suicide: suicide not necessary if never quorate" do_test suicide-not-needed-quorate "no-quorum-policy=suicide: suicide necessary if quorate" echo "" do_test rec-node-1 "Node Recover - Startup - no fence" do_test rec-node-2 "Node Recover - Startup - fence " do_test rec-node-3 "Node Recover - HA down - no fence" do_test rec-node-4 "Node Recover - HA down - fence " do_test rec-node-5 "Node Recover - CRM down - no fence" do_test rec-node-6 "Node Recover - CRM down - fence " do_test rec-node-7 "Node Recover - no quorum - ignore " do_test rec-node-8 "Node Recover - no quorum - freeze " do_test rec-node-9 "Node Recover - no quorum - stop " do_test rec-node-10 "Node Recover - no quorum - stop w/fence" do_test rec-node-11 "Node Recover - CRM down w/ group - fence " do_test rec-node-12 "Node Recover - nothing active - fence " do_test rec-node-13 "Node Recover - failed resource + shutdown - fence " do_test rec-node-15 "Node Recover - unknown lrm section" do_test rec-node-14 "Serialize all stonith's" echo "" do_test multi1 "Multiple Active (stop/start)" echo "" do_test migrate-begin "Normal migration" do_test migrate-success "Completed migration" do_test migrate-partial-1 "Completed migration, missing stop on source" do_test migrate-partial-2 "Successful migrate_to only" do_test migrate-partial-3 "Successful migrate_to only, target down" -@HASH_AFFECTED@do_test migrate-partial-4 "Migrate from the correct host after migrate_to+migrate_from" +do_test migrate-partial-4 "Migrate from the correct host after migrate_to+migrate_from" do_test bug-5186-partial-migrate "Handle partial migration when src node loses membership" do_test migrate-fail-2 "Failed migrate_from" do_test migrate-fail-3 "Failed migrate_from + stop on source" do_test migrate-fail-4 "Failed migrate_from + stop on target - ideally we wouldn't need to re-stop on target" do_test migrate-fail-5 "Failed migrate_from + stop on source and target" do_test migrate-fail-6 "Failed migrate_to" do_test migrate-fail-7 "Failed migrate_to + stop on source" do_test migrate-fail-8 "Failed migrate_to + stop on target - ideally we wouldn't need to re-stop on target" do_test migrate-fail-9 "Failed migrate_to + stop on source and target" do_test migrate-stop "Migration in a stopping stack" do_test migrate-start "Migration in a starting stack" do_test migrate-stop_start "Migration in a restarting stack" do_test migrate-stop-complex "Migration in a complex stopping stack" do_test migrate-start-complex "Migration in a complex starting stack" do_test migrate-stop-start-complex "Migration in a complex moving stack" do_test migrate-shutdown "Order the post-migration 'stop' before node shutdown" do_test migrate-1 "Migrate (migrate)" do_test migrate-2 "Migrate (stable)" do_test migrate-3 "Migrate (failed migrate_to)" do_test migrate-4 "Migrate (failed migrate_from)" do_test novell-252693 "Migration in a stopping stack" -@HASH_AFFECTED@do_test novell-252693-2 "Migration in a starting stack" -@HASH_AFFECTED@do_test novell-252693-3 "Non-Migration in a starting and stopping stack" +do_test novell-252693-2 "Migration in a starting stack" +do_test novell-252693-3 "Non-Migration in a starting and stopping stack" do_test bug-1820 "Migration in a group" do_test bug-1820-1 "Non-migration in a group" do_test migrate-5 "Primitive migration with a clone" do_test migrate-fencing "Migration after Fencing" do_test migrate-both-vms "Migrate two VMs that have no colocation" do_test migration-behind-migrating-remote "Migrate resource behind migrating remote connection" do_test 1-a-then-bm-move-b "Advanced migrate logic. A then B. migrate B." do_test 2-am-then-b-move-a "Advanced migrate logic, A then B, migrate A without stopping B" do_test 3-am-then-bm-both-migrate "Advanced migrate logic. A then B. migrate both" do_test 4-am-then-bm-b-not-migratable "Advanced migrate logic, A then B, B not migratable" do_test 5-am-then-bm-a-not-migratable "Advanced migrate logic. A then B. move both, a not migratable" do_test 6-migrate-group "Advanced migrate logic, migrate a group" do_test 7-migrate-group-one-unmigratable "Advanced migrate logic, migrate group mixed with allow-migrate true/false" do_test 8-am-then-bm-a-migrating-b-stopping "Advanced migrate logic, A then B, A migrating, B stopping" do_test 9-am-then-bm-b-migrating-a-stopping "Advanced migrate logic, A then B, B migrate, A stopping" do_test 10-a-then-bm-b-move-a-clone "Advanced migrate logic, A clone then B, migrate B while stopping A" do_test 11-a-then-bm-b-move-a-clone-starting "Advanced migrate logic, A clone then B, B moving while A is start/stopping" do_test a-promote-then-b-migrate "A promote then B start. migrate B" do_test a-demote-then-b-migrate "A demote then B stop. migrate B" if [ $DO_VERSIONED_TESTS -eq 1 ]; then do_test migrate-versioned "Disable migration for versioned resources" fi #echo "" #do_test complex1 "Complex " do_test bug-lf-2422 "Dependency on partially active group - stop ocfs:*" echo "" do_test clone-anon-probe-1 "Probe the correct (anonymous) clone instance for each node" do_test clone-anon-probe-2 "Avoid needless re-probing of anonymous clones" do_test clone-anon-failcount "Merge failcounts for anonymous clones" do_test force-anon-clone-max "Update clone-max properly when forcing a clone to be anonymous" do_test anon-instance-pending "Assign anonymous clone instance numbers properly when action pending" do_test inc0 "Incarnation start" do_test inc1 "Incarnation start order" do_test inc2 "Incarnation silent restart, stop, move" do_test inc3 "Inter-incarnation ordering, silent restart, stop, move" do_test inc4 "Inter-incarnation ordering, silent restart, stop, move (ordered)" do_test inc5 "Inter-incarnation ordering, silent restart, stop, move (restart 1)" do_test inc6 "Inter-incarnation ordering, silent restart, stop, move (restart 2)" do_test inc7 "Clone colocation" do_test inc8 "Clone anti-colocation" do_test inc9 "Non-unique clone" -@HASH_AFFECTED@do_test inc10 "Non-unique clone (stop)" +do_test inc10 "Non-unique clone (stop)" do_test inc11 "Primitive colocation with clones" do_test inc12 "Clone shutdown" do_test cloned-group "Make sure only the correct number of cloned groups are started" do_test cloned-group-stop "Ensure stopping qpidd also stops glance and cinder" do_test clone-no-shuffle "Don't prioritize allocation of instances that must be moved" do_test clone-max-zero "Orphan processing with clone-max=0" do_test clone-anon-dup "Bug LF#2087 - Correctly parse the state of anonymous clones that are active more than once per node" do_test bug-lf-2160 "Don't shuffle clones due to colocation" do_test bug-lf-2213 "clone-node-max enforcement for cloned groups" -@HASH_AFFECTED@do_test bug-lf-2153 "Clone ordering constraints" -@HASH_AFFECTED@do_test bug-lf-2361 "Ensure clones observe mandatory ordering constraints if the LHS is unrunnable" -@HASH_AFFECTED@do_test bug-lf-2317 "Avoid needless restart of primitive depending on a clone" +do_test bug-lf-2153 "Clone ordering constraints" +do_test bug-lf-2361 "Ensure clones observe mandatory ordering constraints if the LHS is unrunnable" +do_test bug-lf-2317 "Avoid needless restart of primitive depending on a clone" do_test clone-colocate-instance-1 "Colocation with a specific clone instance (negative example)" do_test clone-colocate-instance-2 "Colocation with a specific clone instance" do_test clone-order-instance "Ordering with specific clone instances" do_test bug-lf-2453 "Enforce mandatory clone ordering without colocation" do_test bug-lf-2508 "Correctly reconstruct the status of anonymous cloned groups" do_test bug-lf-2544 "Balanced clone placement" do_test bug-lf-2445 "Redistribute clones with node-max > 1 and stickiness = 0" do_test bug-lf-2574 "Avoid clone shuffle" do_test bug-lf-2581 "Avoid group restart due to unrelated clone (re)start" -@HASH_AFFECTED@do_test bug-cl-5168 "Don't shuffle clones" +do_test bug-cl-5168 "Don't shuffle clones" do_test bug-cl-5170 "Prevent clone from starting with on-fail=block" do_test clone-fail-block-colocation "Move colocated group when failed clone has on-fail=block" do_test clone-interleave-1 "Clone-3 cannot start on pcmk-1 due to interleaved ordering (no colocation)" do_test clone-interleave-2 "Clone-3 must stop on pcmk-1 due to interleaved ordering (no colocation)" do_test clone-interleave-3 "Clone-3 must be recovered on pcmk-1 due to interleaved ordering (no colocation)" do_test rebalance-unique-clones "Rebalance unique clone instances with no stickiness" do_test clone-requires-quorum-recovery "Clone with requires=quorum on failed node needing recovery" do_test clone-requires-quorum "Clone with requires=quorum with presumed-inactive instance on failed node" echo "" do_test cloned_start_one "order first clone then clone... first clone_min=2" do_test cloned_start_two "order first clone then clone... first clone_min=2" do_test cloned_stop_one "order first clone then clone... first clone_min=2" do_test cloned_stop_two "order first clone then clone... first clone_min=2" do_test clone_min_interleave_start_one "order first clone then clone... first clone_min=2 and then has interleave=true" do_test clone_min_interleave_start_two "order first clone then clone... first clone_min=2 and then has interleave=true" do_test clone_min_interleave_stop_one "order first clone then clone... first clone_min=2 and then has interleave=true" do_test clone_min_interleave_stop_two "order first clone then clone... first clone_min=2 and then has interleave=true" do_test clone_min_start_one "order first clone then primitive... first clone_min=2" do_test clone_min_start_two "order first clone then primitive... first clone_min=2" do_test clone_min_stop_all "order first clone then primitive... first clone_min=2" do_test clone_min_stop_one "order first clone then primitive... first clone_min=2" do_test clone_min_stop_two "order first clone then primitive... first clone_min=2" echo "" -@HASH_AFFECTED@do_test unfence-startup "Clean unfencing" -@HASH_AFFECTED@do_test unfence-definition "Unfencing when the agent changes" -@HASH_AFFECTED@do_test unfence-parameters "Unfencing when the agent parameters changes" +do_test unfence-startup "Clean unfencing" +do_test unfence-definition "Unfencing when the agent changes" +do_test unfence-parameters "Unfencing when the agent parameters changes" do_test unfence-device "Unfencing when a cluster has only fence devices" echo "" do_test master-0 "Stopped -> Slave" do_test master-1 "Stopped -> Promote" do_test master-2 "Stopped -> Promote : notify" do_test master-3 "Stopped -> Promote : master location" do_test master-4 "Started -> Promote : master location" do_test master-5 "Promoted -> Promoted" do_test master-6 "Promoted -> Promoted (2)" do_test master-7 "Promoted -> Fenced" do_test master-8 "Promoted -> Fenced -> Moved" do_test master-9 "Stopped + Promotable + No quorum" do_test master-10 "Stopped -> Promotable : notify with monitor" do_test master-11 "Stopped -> Promote : colocation" -@HASH_AFFECTED@do_test novell-239082 "Demote/Promote ordering" +do_test novell-239082 "Demote/Promote ordering" do_test novell-239087 "Stable master placement" do_test master-12 "Promotion based solely on rsc_location constraints" do_test master-13 "Include preferences of colocated resources when placing master" -@HASH_AFFECTED@do_test master-demote "Ordering when actions depends on demoting a slave resource" -@HASH_AFFECTED@do_test master-ordering "Prevent resources from starting that need a master" +do_test master-demote "Ordering when actions depends on demoting a slave resource" +do_test master-ordering "Prevent resources from starting that need a master" do_test bug-1765 "Master-Master Colocation (dont stop the slaves)" do_test master-group "Promotion of cloned groups" -@HASH_AFFECTED@do_test bug-lf-1852 "Don't shuffle master/slave instances unnecessarily" +do_test bug-lf-1852 "Don't shuffle master/slave instances unnecessarily" do_test master-failed-demote "Don't retry failed demote actions" do_test master-failed-demote-2 "Don't retry failed demote actions (notify=false)" do_test master-depend "Ensure resources that depend on the master don't get allocated until the master does" do_test master-reattach "Re-attach to a running master" do_test master-allow-start "Don't include master score if it would prevent allocation" do_test master-colocation "Allow master instances placemaker to be influenced by colocation constraints" do_test master-pseudo "Make sure promote/demote pseudo actions are created correctly" do_test master-role "Prevent target-role from promoting more than master-max instances" -@HASH_AFFECTED@do_test bug-lf-2358 "Master-Master anti-colocation" +do_test bug-lf-2358 "Master-Master anti-colocation" do_test master-promotion-constraint "Mandatory master colocation constraints" do_test unmanaged-master "Ensure role is preserved for unmanaged resources" do_test master-unmanaged-monitor "Start the correct monitor operation for unmanaged masters" do_test master-demote-2 "Demote does not clear past failure" -@HASH_AFFECTED@do_test master-move "Move master based on failure of colocated group" +do_test master-move "Move master based on failure of colocated group" do_test master-probed-score "Observe the promotion score of probed resources" -@HASH_AFFECTED@do_test colocation_constraint_stops_master "cl#5054 - Ensure master is demoted when stopped by colocation constraint" -@HASH_AFFECTED@do_test colocation_constraint_stops_slave "cl#5054 - Ensure slave is not demoted when stopped by colocation constraint" -@HASH_AFFECTED@do_test order_constraint_stops_master "cl#5054 - Ensure master is demoted when stopped by order constraint" -@HASH_AFFECTED@do_test order_constraint_stops_slave "cl#5054 - Ensure slave is not demoted when stopped by order constraint" +do_test colocation_constraint_stops_master "cl#5054 - Ensure master is demoted when stopped by colocation constraint" +do_test colocation_constraint_stops_slave "cl#5054 - Ensure slave is not demoted when stopped by colocation constraint" +do_test order_constraint_stops_master "cl#5054 - Ensure master is demoted when stopped by order constraint" +do_test order_constraint_stops_slave "cl#5054 - Ensure slave is not demoted when stopped by order constraint" do_test master_monitor_restart "cl#5072 - Ensure master monitor operation will start after promotion." do_test bug-rh-880249 "Handle replacement of an m/s resource with a primitive" -@HASH_AFFECTED@do_test bug-5143-ms-shuffle "Prevent master shuffling due to promotion score" +do_test bug-5143-ms-shuffle "Prevent master shuffling due to promotion score" do_test master-demote-block "Block promotion if demote fails with on-fail=block" -@HASH_AFFECTED@do_test master-dependent-ban "Don't stop instances from being active because a dependent is banned from that host" +do_test master-dependent-ban "Don't stop instances from being active because a dependent is banned from that host" do_test master-stop "Stop instances due to location constraint with role=Started" do_test master-partially-demoted-group "Allow partially demoted group to finish demoting" do_test bug-cl-5213 "Ensure role colocation with -INFINITY is enforced" do_test bug-cl-5219 "Allow unrelated resources with a common colocation target to remain promoted" do_test master-asymmetrical-order "Fix the behaviors of multi-state resources with asymmetrical ordering" -@HASH_AFFECTED@do_test master-notify "Master promotion with notifies" +do_test master-notify "Master promotion with notifies" do_test master-score-startup "Use permanent master scores without LRM history" do_test failed-demote-recovery "Recover resource in slave role after demote fails" do_test failed-demote-recovery-master "Recover resource in master role after demote fails" echo "" do_test history-1 "Correctly parse stateful-1 resource state" echo "" do_test managed-0 "Managed (reference)" do_test managed-1 "Not managed - down " do_test managed-2 "Not managed - up " do_test bug-5028 "Shutdown should block if anything depends on an unmanaged resource" do_test bug-5028-detach "Ensure detach still works" do_test bug-5028-bottom "Ensure shutdown still blocks if the blocked resource is at the bottom of the stack" do_test unmanaged-stop-1 "cl#5155 - Block the stop of resources if any depending resource is unmanaged " do_test unmanaged-stop-2 "cl#5155 - Block the stop of resources if the first resource in a mandatory stop order is unmanaged " do_test unmanaged-stop-3 "cl#5155 - Block the stop of resources if any depending resource in a group is unmanaged " do_test unmanaged-stop-4 "cl#5155 - Block the stop of resources if any depending resource in the middle of a group is unmanaged " do_test unmanaged-block-restart "Block restart of resources if any dependent resource in a group is unmanaged" echo "" do_test interleave-0 "Interleave (reference)" do_test interleave-1 "coloc - not interleaved" do_test interleave-2 "coloc - interleaved " do_test interleave-3 "coloc - interleaved (2)" do_test interleave-pseudo-stop "Interleaved clone during stonith" do_test interleave-stop "Interleaved clone during stop" do_test interleave-restart "Interleaved clone during dependency restart" echo "" do_test notify-0 "Notify reference" do_test notify-1 "Notify simple" do_test notify-2 "Notify simple, confirm" do_test notify-3 "Notify move, confirm" -@HASH_AFFECTED@do_test novell-239079 "Notification priority" +do_test novell-239079 "Notification priority" #do_test notify-2 "Notify - 764" do_test notifs-for-unrunnable "Don't schedule notifications for an unrunnable action" -@HASH_AFFECTED@do_test route-remote-notify "Route remote notify actions through correct cluster node" -@HASH_AFFECTED@do_test notify-behind-stopping-remote "Don't schedule notifications behind stopped remote" +do_test route-remote-notify "Route remote notify actions through correct cluster node" +do_test notify-behind-stopping-remote "Don't schedule notifications behind stopped remote" echo "" do_test 594 "OSDL #594 - Unrunnable actions scheduled in transition" do_test 662 "OSDL #662 - Two resources start on one node when incarnation_node_max = 1" do_test 696 "OSDL #696 - CRM starts stonith RA without monitor" do_test 726 "OSDL #726 - Attempting to schedule rsc_posic041_monitor_5000 _after_ a stop" do_test 735 "OSDL #735 - Correctly detect that rsc_hadev1 is stopped on hadev3" do_test 764 "OSDL #764 - Missing monitor op for DoFencing:child_DoFencing:1" do_test 797 "OSDL #797 - Assert triggered: task_id_i > max_call_id" do_test 829 "OSDL #829" do_test 994 "OSDL #994 - Stopping the last resource in a resource group causes the entire group to be restarted" do_test 994-2 "OSDL #994 - with a dependent resource" do_test 1360 "OSDL #1360 - Clone stickiness" do_test 1484 "OSDL #1484 - on_fail=stop" do_test 1494 "OSDL #1494 - Clone stability" do_test unrunnable-1 "Unrunnable" do_test unrunnable-2 "Unrunnable 2" do_test stonith-0 "Stonith loop - 1" do_test stonith-1 "Stonith loop - 2" do_test stonith-2 "Stonith loop - 3" do_test stonith-3 "Stonith startup" do_test stonith-4 "Stonith node state" do_test dc-fence-ordering "DC needs fencing while other nodes are shutting down" do_test bug-1572-1 "Recovery of groups depending on master/slave" do_test bug-1572-2 "Recovery of groups depending on master/slave when the master is never re-promoted" -@HASH_AFFECTED@do_test bug-1685 "Depends-on-master ordering" +do_test bug-1685 "Depends-on-master ordering" do_test bug-1822 "Don't promote partially active groups" do_test bug-pm-11 "New resource added to a m/s group" do_test bug-pm-12 "Recover only the failed portion of a cloned group" do_test bug-n-387749 "Don't shuffle clone instances" do_test bug-n-385265 "Don't ignore the failure stickiness of group children - resource_idvscommon should stay stopped" do_test bug-n-385265-2 "Ensure groups are migrated instead of remaining partially active on the current node" do_test bug-lf-1920 "Correctly handle probes that find active resources" do_test bnc-515172 "Location constraint with multiple expressions" do_test colocate-primitive-with-clone "Optional colocation with a clone" do_test use-after-free-merge "Use-after-free in native_merge_weights" -@HASH_AFFECTED@do_test bug-lf-2551 "STONITH ordering for stop" +do_test bug-lf-2551 "STONITH ordering for stop" do_test bug-lf-2606 "Stonith implies demote" do_test bug-lf-2474 "Ensure resource op timeout takes precedence over op_defaults" -@HASH_AFFECTED@do_test bug-suse-707150 "Prevent vm-01 from starting due to colocation/ordering" +do_test bug-suse-707150 "Prevent vm-01 from starting due to colocation/ordering" do_test bug-5014-A-start-B-start "Verify when A starts B starts using symmetrical=false" do_test bug-5014-A-stop-B-started "Verify when A stops B does not stop if it has already started using symmetric=false" do_test bug-5014-A-stopped-B-stopped "Verify when A is stopped and B has not started, B does not start before A using symmetric=false" do_test bug-5014-CthenAthenB-C-stopped "Verify when C then A is symmetrical=true, A then B is symmetric=false, and C is stopped that nothing starts." do_test bug-5014-CLONE-A-start-B-start "Verify when A starts B starts using clone resources with symmetric=false" do_test bug-5014-CLONE-A-stop-B-started "Verify when A stops B does not stop if it has already started using clone resources with symmetric=false." do_test bug-5014-GROUP-A-start-B-start "Verify when A starts B starts when using group resources with symmetric=false." do_test bug-5014-GROUP-A-stopped-B-started "Verify when A stops B does not stop if it has already started using group resources with symmetric=false." do_test bug-5014-GROUP-A-stopped-B-stopped "Verify when A is stopped and B has not started, B does not start before A using group resources with symmetric=false." do_test bug-5014-ordered-set-symmetrical-false "Verify ordered sets work with symmetrical=false" do_test bug-5014-ordered-set-symmetrical-true "Verify ordered sets work with symmetrical=true" do_test bug-5007-masterslave_colocation "Verify use of colocation scores other than INFINITY and -INFINITY work on multi-state resources." do_test bug-5038 "Prevent restart of anonymous clones when clone-max decreases" do_test bug-5025-1 "Automatically clean up failcount after resource config change with reload" do_test bug-5025-2 "Make sure clear failcount action isn't set when config does not change." do_test bug-5025-3 "Automatically clean up failcount after resource config change with restart" do_test bug-5025-4 "Clear failcount when last failure is a start op and rsc attributes changed." do_test failcount "Ensure failcounts are correctly expired" do_test failcount-block "Ensure failcounts are not expired when on-fail=block is present" do_test per-op-failcount "Ensure per-operation failcount is handled and not passed to fence agent" do_test on-fail-ignore "Ensure on-fail=ignore works even beyond migration-threshold" do_test monitor-onfail-restart "bug-5058 - Monitor failure with on-fail set to restart" do_test monitor-onfail-stop "bug-5058 - Monitor failure wiht on-fail set to stop" do_test bug-5059 "No need to restart p_stateful1:*" do_test bug-5069-op-enabled "Test on-fail=ignore with failure when monitor is enabled." do_test bug-5069-op-disabled "Test on-fail-ignore with failure when monitor is disabled." do_test obsolete-lrm-resource "cl#5115 - Do not use obsolete lrm_resource sections" do_test expire-non-blocked-failure "Ignore failure-timeout only if the failed operation has on-fail=block" do_test asymmetrical-order-move "Respect asymmetrical ordering when trying to move resources" do_test asymmetrical-order-restart "Respect asymmetrical ordering when restarting dependent resource" do_test start-then-stop-with-unfence "Avoid graph loop with start-then-stop constraint plus unfencing" do_test order-expired-failure "Order failcount cleanup after remote fencing" do_test ignore_stonith_rsc_order1 "cl#5056- Ignore order constraint between stonith and non-stonith rsc." do_test ignore_stonith_rsc_order2 "cl#5056- Ignore order constraint with group rsc containing mixed stonith and non-stonith." do_test ignore_stonith_rsc_order3 "cl#5056- Ignore order constraint, stonith clone and mixed group" do_test ignore_stonith_rsc_order4 "cl#5056- Ignore order constraint, stonith clone and clone with nested mixed group" do_test honor_stonith_rsc_order1 "cl#5056- Honor order constraint, stonith clone and pure stonith group(single rsc)." do_test honor_stonith_rsc_order2 "cl#5056- Honor order constraint, stonith clone and pure stonith group(multiple rsc)" do_test honor_stonith_rsc_order3 "cl#5056- Honor order constraint, stonith clones with nested pure stonith group." do_test honor_stonith_rsc_order4 "cl#5056- Honor order constraint, between two native stonith rscs." do_test multiply-active-stonith do_test probe-timeout "cl#5099 - Default probe timeout" do_test order-first-probes "cl#5301 - respect order constraints when relevant resources are being probed" do_test concurrent-fencing "Allow performing fencing operations in parallel" echo "" do_test systemhealth1 "System Health () #1" do_test systemhealth2 "System Health () #2" do_test systemhealth3 "System Health () #3" do_test systemhealthn1 "System Health (None) #1" do_test systemhealthn2 "System Health (None) #2" do_test systemhealthn3 "System Health (None) #3" do_test systemhealthm1 "System Health (Migrate On Red) #1" do_test systemhealthm2 "System Health (Migrate On Red) #2" do_test systemhealthm3 "System Health (Migrate On Red) #3" do_test systemhealtho1 "System Health (Only Green) #1" do_test systemhealtho2 "System Health (Only Green) #2" do_test systemhealtho3 "System Health (Only Green) #3" do_test systemhealthp1 "System Health (Progessive) #1" do_test systemhealthp2 "System Health (Progessive) #2" do_test systemhealthp3 "System Health (Progessive) #3" echo "" -@HASH_AFFECTED@do_test utilization "Placement Strategy - utilization" -@HASH_AFFECTED@do_test minimal "Placement Strategy - minimal" -@HASH_AFFECTED@do_test balanced "Placement Strategy - balanced" +do_test utilization "Placement Strategy - utilization" +do_test minimal "Placement Strategy - minimal" +do_test balanced "Placement Strategy - balanced" echo "" do_test placement-stickiness "Optimized Placement Strategy - stickiness" do_test placement-priority "Optimized Placement Strategy - priority" do_test placement-location "Optimized Placement Strategy - location" do_test placement-capacity "Optimized Placement Strategy - capacity" echo "" do_test utilization-order1 "Utilization Order - Simple" do_test utilization-order2 "Utilization Order - Complex" do_test utilization-order3 "Utilization Order - Migrate" do_test utilization-order4 "Utilization Order - Live Migration (bnc#695440)" -@HASH_AFFECTED@do_test utilization-shuffle "Don't displace prmExPostgreSQLDB2 on act2, Start prmExPostgreSQLDB1 on act3" +do_test utilization-shuffle "Don't displace prmExPostgreSQLDB2 on act2, Start prmExPostgreSQLDB1 on act3" do_test load-stopped-loop "Avoid transition loop due to load_stopped (cl#5044)" -@HASH_AFFECTED@do_test load-stopped-loop-2 "cl#5235 - Prevent graph loops that can be introduced by load_stopped -> migrate_to ordering" +do_test load-stopped-loop-2 "cl#5235 - Prevent graph loops that can be introduced by load_stopped -> migrate_to ordering" echo "" do_test colocated-utilization-primitive-1 "Colocated Utilization - Primitive" do_test colocated-utilization-primitive-2 "Colocated Utilization - Choose the most capable node" do_test colocated-utilization-group "Colocated Utilization - Group" -@HASH_AFFECTED@do_test colocated-utilization-clone "Colocated Utilization - Clone" +do_test colocated-utilization-clone "Colocated Utilization - Clone" do_test utilization-check-allowed-nodes "Only check the capacities of the nodes that can run the resource" echo "" do_test reprobe-target_rc "Ensure correct target_rc for reprobe of inactive resources" do_test node-maintenance-1 "cl#5128 - Node maintenance" do_test node-maintenance-2 "cl#5128 - Node maintenance (coming out of maintenance mode)" do_test shutdown-maintenance-node "Do not fence a maintenance node if it shuts down cleanly" do_test rsc-maintenance "Per-resource maintenance" echo "" do_test not-installed-agent "The resource agent is missing" do_test not-installed-tools "Something the resource agent needs is missing" echo "" do_test stopped-monitor-00 "Stopped Monitor - initial start" do_test stopped-monitor-01 "Stopped Monitor - failed started" do_test stopped-monitor-02 "Stopped Monitor - started multi-up" do_test stopped-monitor-03 "Stopped Monitor - stop started" do_test stopped-monitor-04 "Stopped Monitor - failed stop" do_test stopped-monitor-05 "Stopped Monitor - start unmanaged" do_test stopped-monitor-06 "Stopped Monitor - unmanaged multi-up" do_test stopped-monitor-07 "Stopped Monitor - start unmanaged multi-up" do_test stopped-monitor-08 "Stopped Monitor - migrate" do_test stopped-monitor-09 "Stopped Monitor - unmanage started" do_test stopped-monitor-10 "Stopped Monitor - unmanaged started multi-up" do_test stopped-monitor-11 "Stopped Monitor - stop unmanaged started" do_test stopped-monitor-12 "Stopped Monitor - unmanaged started multi-up (target-role=Stopped)" do_test stopped-monitor-20 "Stopped Monitor - initial stop" do_test stopped-monitor-21 "Stopped Monitor - stopped single-up" do_test stopped-monitor-22 "Stopped Monitor - stopped multi-up" do_test stopped-monitor-23 "Stopped Monitor - start stopped" do_test stopped-monitor-24 "Stopped Monitor - unmanage stopped" do_test stopped-monitor-25 "Stopped Monitor - unmanaged stopped multi-up" do_test stopped-monitor-26 "Stopped Monitor - start unmanaged stopped" do_test stopped-monitor-27 "Stopped Monitor - unmanaged stopped multi-up (target-role=Started)" do_test stopped-monitor-30 "Stopped Monitor - new node started" do_test stopped-monitor-31 "Stopped Monitor - new node stopped" echo "" # This is a combo test to check: # - probe timeout defaults to the minimum-interval monitor's # - duplicate recurring operations are ignored # - if timeout spec is bad, the default timeout is used # - failure is blocked with on-fail=block even if ISO8601 interval is specified # - started/stopped role monitors are started/stopped on right nodes do_test intervals "Recurring monitor interval handling" echo"" do_test ticket-primitive-1 "Ticket - Primitive (loss-policy=stop, initial)" do_test ticket-primitive-2 "Ticket - Primitive (loss-policy=stop, granted)" do_test ticket-primitive-3 "Ticket - Primitive (loss-policy-stop, revoked)" do_test ticket-primitive-4 "Ticket - Primitive (loss-policy=demote, initial)" do_test ticket-primitive-5 "Ticket - Primitive (loss-policy=demote, granted)" do_test ticket-primitive-6 "Ticket - Primitive (loss-policy=demote, revoked)" do_test ticket-primitive-7 "Ticket - Primitive (loss-policy=fence, initial)" do_test ticket-primitive-8 "Ticket - Primitive (loss-policy=fence, granted)" do_test ticket-primitive-9 "Ticket - Primitive (loss-policy=fence, revoked)" do_test ticket-primitive-10 "Ticket - Primitive (loss-policy=freeze, initial)" do_test ticket-primitive-11 "Ticket - Primitive (loss-policy=freeze, granted)" do_test ticket-primitive-12 "Ticket - Primitive (loss-policy=freeze, revoked)" do_test ticket-primitive-13 "Ticket - Primitive (loss-policy=stop, standby, granted)" do_test ticket-primitive-14 "Ticket - Primitive (loss-policy=stop, granted, standby)" do_test ticket-primitive-15 "Ticket - Primitive (loss-policy=stop, standby, revoked)" do_test ticket-primitive-16 "Ticket - Primitive (loss-policy=demote, standby, granted)" do_test ticket-primitive-17 "Ticket - Primitive (loss-policy=demote, granted, standby)" do_test ticket-primitive-18 "Ticket - Primitive (loss-policy=demote, standby, revoked)" do_test ticket-primitive-19 "Ticket - Primitive (loss-policy=fence, standby, granted)" do_test ticket-primitive-20 "Ticket - Primitive (loss-policy=fence, granted, standby)" do_test ticket-primitive-21 "Ticket - Primitive (loss-policy=fence, standby, revoked)" do_test ticket-primitive-22 "Ticket - Primitive (loss-policy=freeze, standby, granted)" do_test ticket-primitive-23 "Ticket - Primitive (loss-policy=freeze, granted, standby)" do_test ticket-primitive-24 "Ticket - Primitive (loss-policy=freeze, standby, revoked)" echo"" do_test ticket-group-1 "Ticket - Group (loss-policy=stop, initial)" do_test ticket-group-2 "Ticket - Group (loss-policy=stop, granted)" do_test ticket-group-3 "Ticket - Group (loss-policy-stop, revoked)" do_test ticket-group-4 "Ticket - Group (loss-policy=demote, initial)" do_test ticket-group-5 "Ticket - Group (loss-policy=demote, granted)" do_test ticket-group-6 "Ticket - Group (loss-policy=demote, revoked)" do_test ticket-group-7 "Ticket - Group (loss-policy=fence, initial)" do_test ticket-group-8 "Ticket - Group (loss-policy=fence, granted)" do_test ticket-group-9 "Ticket - Group (loss-policy=fence, revoked)" do_test ticket-group-10 "Ticket - Group (loss-policy=freeze, initial)" do_test ticket-group-11 "Ticket - Group (loss-policy=freeze, granted)" do_test ticket-group-12 "Ticket - Group (loss-policy=freeze, revoked)" do_test ticket-group-13 "Ticket - Group (loss-policy=stop, standby, granted)" do_test ticket-group-14 "Ticket - Group (loss-policy=stop, granted, standby)" do_test ticket-group-15 "Ticket - Group (loss-policy=stop, standby, revoked)" do_test ticket-group-16 "Ticket - Group (loss-policy=demote, standby, granted)" do_test ticket-group-17 "Ticket - Group (loss-policy=demote, granted, standby)" do_test ticket-group-18 "Ticket - Group (loss-policy=demote, standby, revoked)" do_test ticket-group-19 "Ticket - Group (loss-policy=fence, standby, granted)" do_test ticket-group-20 "Ticket - Group (loss-policy=fence, granted, standby)" do_test ticket-group-21 "Ticket - Group (loss-policy=fence, standby, revoked)" do_test ticket-group-22 "Ticket - Group (loss-policy=freeze, standby, granted)" do_test ticket-group-23 "Ticket - Group (loss-policy=freeze, granted, standby)" do_test ticket-group-24 "Ticket - Group (loss-policy=freeze, standby, revoked)" echo"" do_test ticket-clone-1 "Ticket - Clone (loss-policy=stop, initial)" do_test ticket-clone-2 "Ticket - Clone (loss-policy=stop, granted)" do_test ticket-clone-3 "Ticket - Clone (loss-policy-stop, revoked)" do_test ticket-clone-4 "Ticket - Clone (loss-policy=demote, initial)" do_test ticket-clone-5 "Ticket - Clone (loss-policy=demote, granted)" do_test ticket-clone-6 "Ticket - Clone (loss-policy=demote, revoked)" do_test ticket-clone-7 "Ticket - Clone (loss-policy=fence, initial)" do_test ticket-clone-8 "Ticket - Clone (loss-policy=fence, granted)" do_test ticket-clone-9 "Ticket - Clone (loss-policy=fence, revoked)" do_test ticket-clone-10 "Ticket - Clone (loss-policy=freeze, initial)" do_test ticket-clone-11 "Ticket - Clone (loss-policy=freeze, granted)" do_test ticket-clone-12 "Ticket - Clone (loss-policy=freeze, revoked)" do_test ticket-clone-13 "Ticket - Clone (loss-policy=stop, standby, granted)" do_test ticket-clone-14 "Ticket - Clone (loss-policy=stop, granted, standby)" do_test ticket-clone-15 "Ticket - Clone (loss-policy=stop, standby, revoked)" do_test ticket-clone-16 "Ticket - Clone (loss-policy=demote, standby, granted)" do_test ticket-clone-17 "Ticket - Clone (loss-policy=demote, granted, standby)" do_test ticket-clone-18 "Ticket - Clone (loss-policy=demote, standby, revoked)" do_test ticket-clone-19 "Ticket - Clone (loss-policy=fence, standby, granted)" do_test ticket-clone-20 "Ticket - Clone (loss-policy=fence, granted, standby)" do_test ticket-clone-21 "Ticket - Clone (loss-policy=fence, standby, revoked)" do_test ticket-clone-22 "Ticket - Clone (loss-policy=freeze, standby, granted)" do_test ticket-clone-23 "Ticket - Clone (loss-policy=freeze, granted, standby)" do_test ticket-clone-24 "Ticket - Clone (loss-policy=freeze, standby, revoked)" echo"" do_test ticket-master-1 "Ticket - Master (loss-policy=stop, initial)" do_test ticket-master-2 "Ticket - Master (loss-policy=stop, granted)" do_test ticket-master-3 "Ticket - Master (loss-policy-stop, revoked)" do_test ticket-master-4 "Ticket - Master (loss-policy=demote, initial)" do_test ticket-master-5 "Ticket - Master (loss-policy=demote, granted)" do_test ticket-master-6 "Ticket - Master (loss-policy=demote, revoked)" do_test ticket-master-7 "Ticket - Master (loss-policy=fence, initial)" do_test ticket-master-8 "Ticket - Master (loss-policy=fence, granted)" do_test ticket-master-9 "Ticket - Master (loss-policy=fence, revoked)" do_test ticket-master-10 "Ticket - Master (loss-policy=freeze, initial)" do_test ticket-master-11 "Ticket - Master (loss-policy=freeze, granted)" do_test ticket-master-12 "Ticket - Master (loss-policy=freeze, revoked)" do_test ticket-master-13 "Ticket - Master (loss-policy=stop, standby, granted)" do_test ticket-master-14 "Ticket - Master (loss-policy=stop, granted, standby)" do_test ticket-master-15 "Ticket - Master (loss-policy=stop, standby, revoked)" do_test ticket-master-16 "Ticket - Master (loss-policy=demote, standby, granted)" do_test ticket-master-17 "Ticket - Master (loss-policy=demote, granted, standby)" do_test ticket-master-18 "Ticket - Master (loss-policy=demote, standby, revoked)" do_test ticket-master-19 "Ticket - Master (loss-policy=fence, standby, granted)" do_test ticket-master-20 "Ticket - Master (loss-policy=fence, granted, standby)" do_test ticket-master-21 "Ticket - Master (loss-policy=fence, standby, revoked)" do_test ticket-master-22 "Ticket - Master (loss-policy=freeze, standby, granted)" do_test ticket-master-23 "Ticket - Master (loss-policy=freeze, granted, standby)" do_test ticket-master-24 "Ticket - Master (loss-policy=freeze, standby, revoked)" echo "" do_test ticket-rsc-sets-1 "Ticket - Resource sets (1 ticket, initial)" do_test ticket-rsc-sets-2 "Ticket - Resource sets (1 ticket, granted)" do_test ticket-rsc-sets-3 "Ticket - Resource sets (1 ticket, revoked)" do_test ticket-rsc-sets-4 "Ticket - Resource sets (2 tickets, initial)" do_test ticket-rsc-sets-5 "Ticket - Resource sets (2 tickets, granted)" do_test ticket-rsc-sets-6 "Ticket - Resource sets (2 tickets, granted)" do_test ticket-rsc-sets-7 "Ticket - Resource sets (2 tickets, revoked)" do_test ticket-rsc-sets-8 "Ticket - Resource sets (1 ticket, standby, granted)" do_test ticket-rsc-sets-9 "Ticket - Resource sets (1 ticket, granted, standby)" do_test ticket-rsc-sets-10 "Ticket - Resource sets (1 ticket, standby, revoked)" do_test ticket-rsc-sets-11 "Ticket - Resource sets (2 tickets, standby, granted)" do_test ticket-rsc-sets-12 "Ticket - Resource sets (2 tickets, standby, granted)" do_test ticket-rsc-sets-13 "Ticket - Resource sets (2 tickets, granted, standby)" do_test ticket-rsc-sets-14 "Ticket - Resource sets (2 tickets, standby, revoked)" do_test cluster-specific-params "Cluster-specific instance attributes based on rules" do_test site-specific-params "Site-specific instance attributes based on rules" echo "" do_test template-1 "Template - 1" do_test template-2 "Template - 2" do_test template-3 "Template - 3 (merge operations)" do_test template-coloc-1 "Template - Colocation 1" do_test template-coloc-2 "Template - Colocation 2" do_test template-coloc-3 "Template - Colocation 3" do_test template-order-1 "Template - Order 1" do_test template-order-2 "Template - Order 2" do_test template-order-3 "Template - Order 3" do_test template-ticket "Template - Ticket" do_test template-rsc-sets-1 "Template - Resource Sets 1" do_test template-rsc-sets-2 "Template - Resource Sets 2" do_test template-rsc-sets-3 "Template - Resource Sets 3" do_test template-rsc-sets-4 "Template - Resource Sets 4" do_test template-clone-primitive "Cloned primitive from template" do_test template-clone-group "Cloned group from template" do_test location-sets-templates "Resource sets and templates - Location" do_test tags-coloc-order-1 "Tags - Colocation and Order (Simple)" do_test tags-coloc-order-2 "Tags - Colocation and Order (Resource Sets with Templates)" do_test tags-location "Tags - Location" do_test tags-ticket "Tags - Ticket" echo "" do_test container-1 "Container - initial" do_test container-2 "Container - monitor failed" do_test container-3 "Container - stop failed" do_test container-4 "Container - reached migration-threshold" do_test container-group-1 "Container in group - initial" do_test container-group-2 "Container in group - monitor failed" do_test container-group-3 "Container in group - stop failed" do_test container-group-4 "Container in group - reached migration-threshold" do_test container-is-remote-node "Place resource within container when container is remote-node" do_test bug-rh-1097457 "Kill user defined container/contents ordering" -@HASH_AFFECTED@do_test bug-cl-5247 "Graph loop when recovering m/s resource in a container" +do_test bug-cl-5247 "Graph loop when recovering m/s resource in a container" do_test bundle-order-startup "Bundle startup ordering" do_test bundle-order-partial-start "Bundle startup ordering when some dependancies are already running" do_test bundle-order-partial-start-2 "Bundle startup ordering when some dependancies and the container are already running" do_test bundle-order-stop "Bundle stop ordering" do_test bundle-order-partial-stop "Bundle startup ordering when some dependancies are already stopped" -@HASH_AFFECTED@do_test bundle-order-stop-on-remote "Stop nested resource after bringing up the connection" +do_test bundle-order-stop-on-remote "Stop nested resource after bringing up the connection" -@HASH_AFFECTED@do_test bundle-order-startup-clone "Prevent startup because bundle isn't promoted" -@HASH_AFFECTED@do_test bundle-order-startup-clone-2 "Bundle startup with clones" +do_test bundle-order-startup-clone "Prevent startup because bundle isn't promoted" +do_test bundle-order-startup-clone-2 "Bundle startup with clones" do_test bundle-order-stop-clone "Stop bundle because clone is stopping" -@HASH_AFFECTED@do_test bundle-nested-colocation "Colocation of nested connection resources" +do_test bundle-nested-colocation "Colocation of nested connection resources" -@HASH_AFFECTED@do_test bundle-order-fencing "Order pseudo bundle fencing after parent node fencing if both are happening" +do_test bundle-order-fencing "Order pseudo bundle fencing after parent node fencing if both are happening" do_test bundle-probe-order-1 "order 1" do_test bundle-probe-order-2 "order 2" do_test bundle-probe-order-3 "order 3" do_test bundle-probe-remotes "Ensure remotes get probed too" do_test bundle-replicas-change "Change bundle from 1 replica to multiple" -@HASH_AFFECTED@do_test nested-remote-recovery "Recover bundle's container hosted on remote node" +do_test nested-remote-recovery "Recover bundle's container hosted on remote node" echo "" do_test whitebox-fail1 "Fail whitebox container rsc." do_test whitebox-fail2 "Fail cluster connection to guest node" do_test whitebox-fail3 "Failed containers should not run nested on remote nodes." do_test whitebox-start "Start whitebox container with resources assigned to it" do_test whitebox-stop "Stop whitebox container with resources assigned to it" do_test whitebox-move "Move whitebox container with resources assigned to it" do_test whitebox-asymmetric "Verify connection rsc opts-in based on container resource" do_test whitebox-ms-ordering "Verify promote/demote can not occur before connection is established" do_test whitebox-ms-ordering-move "Stop/Start cycle within a moving container" do_test whitebox-orphaned "Properly shutdown orphaned whitebox container" do_test whitebox-orphan-ms "Properly tear down orphan ms resources on remote-nodes" do_test whitebox-unexpectedly-running "Recover container nodes the cluster did not start." do_test whitebox-migrate1 "Migrate both container and connection resource" do_test whitebox-imply-stop-on-fence "imply stop action on container node rsc when host node is fenced" do_test whitebox-nested-group "Verify guest remote-node works nested in a group" do_test guest-node-host-dies "Verify guest node is recovered if host goes away" do_test guest-node-cleanup "Order guest node connection recovery after container probe" echo "" do_test remote-startup-probes "Baremetal remote-node startup probes" do_test remote-startup "Startup a newly discovered remote-nodes with no status." do_test remote-fence-unclean "Fence unclean baremetal remote-node" do_test remote-fence-unclean2 "Fence baremetal remote-node after cluster node fails and connection can not be recovered" do_test remote-fence-unclean-3 "Probe failed remote nodes (triggers fencing)" do_test remote-move "Move remote-node connection resource" do_test remote-disable "Disable a baremetal remote-node" do_test remote-probe-disable "Probe then stop a baremetal remote-node" do_test remote-orphaned "Properly shutdown orphaned connection resource" do_test remote-orphaned2 "verify we can handle orphaned remote connections with active resources on the remote" do_test remote-recover "Recover connection resource after cluster-node fails." do_test remote-stale-node-entry "Make sure we properly handle leftover remote-node entries in the node section" do_test remote-partial-migrate "Make sure partial migrations are handled before ops on the remote node." do_test remote-partial-migrate2 "Make sure partial migration target is prefered for remote connection." do_test remote-recover-fail "Make sure start failure causes fencing if rsc are active on remote." do_test remote-start-fail "Make sure a start failure does not result in fencing if no active resources are on remote." do_test remote-unclean2 "Make monitor failure always results in fencing, even if no rsc are active on remote." do_test remote-fence-before-reconnect "Fence before clearing recurring monitor failure" -@HASH_AFFECTED@do_test remote-recovery "Recover remote connections before attempting demotion" -@HASH_AFFECTED@do_test remote-recover-connection "Optimistically recovery of only the connection" -@HASH_AFFECTED@do_test remote-recover-all "Fencing when the connection has no home" -@HASH_AFFECTED@do_test remote-recover-no-resources "Fencing when the connection has no home and no active resources" -@HASH_AFFECTED@do_test remote-recover-unknown "Fencing when the connection has no home and the remote has no operation history" +do_test remote-recovery "Recover remote connections before attempting demotion" +do_test remote-recover-connection "Optimistically recovery of only the connection" +do_test remote-recover-all "Fencing when the connection has no home" +do_test remote-recover-no-resources "Fencing when the connection has no home and no active resources" +do_test remote-recover-unknown "Fencing when the connection has no home and the remote has no operation history" do_test remote-reconnect-delay "Waiting for remote reconnect interval to expire" do_test remote-connection-unrecoverable "Remote connection host must be fenced, with connection unrecoverable" echo "" do_test resource-discovery "Exercises resource-discovery location constraint option." do_test rsc-discovery-per-node "Disable resource discovery per node" if [ $DO_VERSIONED_TESTS -eq 1 ]; then echo "" do_test versioned-resources "Start resources with #ra-version rules" do_test restart-versioned "Restart resources on #ra-version change" do_test reload-versioned "Reload resources on #ra-version change" echo "" do_test versioned-operations-1 "Use #ra-version to configure operations of native resources" do_test versioned-operations-2 "Use #ra-version to configure operations of stonith resources" do_test versioned-operations-3 "Use #ra-version to configure operations of master/slave resources" do_test versioned-operations-4 "Use #ra-version to configure operations of groups of the resources" fi echo "" test_results exit $EXITCODE diff --git a/daemons/controld/controld_execd.c b/daemons/controld/controld_execd.c index 9fd1ba9c7e..976fed1392 100644 --- a/daemons/controld/controld_execd.c +++ b/daemons/controld/controld_execd.c @@ -1,2738 +1,2733 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #define START_DELAY_THRESHOLD 5 * 60 * 1000 #define MAX_LRM_REG_FAILS 30 #define s_if_plural(i) (((i) == 1)? "" : "s") struct delete_event_s { int rc; const char *rsc; lrm_state_t *lrm_state; }; static gboolean is_rsc_active(lrm_state_t * lrm_state, const char *rsc_id); static gboolean build_active_RAs(lrm_state_t * lrm_state, xmlNode * rsc_list); static gboolean stop_recurring_actions(gpointer key, gpointer value, gpointer user_data); static int delete_rsc_status(lrm_state_t * lrm_state, const char *rsc_id, int call_options, const char *user_name); static lrmd_event_data_t *construct_op(lrm_state_t * lrm_state, xmlNode * rsc_op, const char *rsc_id, const char *operation); static void do_lrm_rsc_op(lrm_state_t * lrm_state, lrmd_rsc_info_t * rsc, const char *operation, xmlNode * msg, xmlNode * request); void send_direct_ack(const char *to_host, const char *to_sys, lrmd_rsc_info_t * rsc, lrmd_event_data_t * op, const char *rsc_id); static gboolean lrm_state_verify_stopped(lrm_state_t * lrm_state, enum crmd_fsa_state cur_state, int log_level); static int do_update_resource(const char *node_name, lrmd_rsc_info_t * rsc, lrmd_event_data_t * op); static void lrm_connection_destroy(void) { if (is_set(fsa_input_register, R_LRM_CONNECTED)) { crm_crit("Connection to executor failed"); register_fsa_input(C_FSA_INTERNAL, I_ERROR, NULL); clear_bit(fsa_input_register, R_LRM_CONNECTED); } else { crm_info("Disconnected from executor"); } } static char * make_stop_id(const char *rsc, int call_id) { return crm_strdup_printf("%s:%d", rsc, call_id); } static void copy_instance_keys(gpointer key, gpointer value, gpointer user_data) { if (strstr(key, CRM_META "_") == NULL) { g_hash_table_replace(user_data, strdup((const char *)key), strdup((const char *)value)); } } static void copy_meta_keys(gpointer key, gpointer value, gpointer user_data) { if (strstr(key, CRM_META "_") != NULL) { g_hash_table_replace(user_data, strdup((const char *)key), strdup((const char *)value)); } } /*! * \internal * \brief Remove a recurring operation from a resource's history * * \param[in,out] history Resource history to modify * \param[in] op Operation to remove * * \return TRUE if the operation was found and removed, FALSE otherwise */ static gboolean history_remove_recurring_op(rsc_history_t *history, const lrmd_event_data_t *op) { GList *iter; for (iter = history->recurring_op_list; iter != NULL; iter = iter->next) { lrmd_event_data_t *existing = iter->data; if ((op->interval_ms == existing->interval_ms) && crm_str_eq(op->rsc_id, existing->rsc_id, TRUE) && safe_str_eq(op->op_type, existing->op_type)) { history->recurring_op_list = g_list_delete_link(history->recurring_op_list, iter); lrmd_free_event(existing); return TRUE; } } return FALSE; } /*! * \internal * \brief Free all recurring operations in resource history * * \param[in,out] history Resource history to modify */ static void history_free_recurring_ops(rsc_history_t *history) { GList *iter; for (iter = history->recurring_op_list; iter != NULL; iter = iter->next) { lrmd_free_event(iter->data); } g_list_free(history->recurring_op_list); history->recurring_op_list = NULL; } /*! * \internal * \brief Free resource history * * \param[in,out] history Resource history to free */ void history_free(gpointer data) { rsc_history_t *history = (rsc_history_t*)data; if (history->stop_params) { g_hash_table_destroy(history->stop_params); } /* Don't need to free history->rsc.id because it's set to history->id */ free(history->rsc.type); free(history->rsc.standard); free(history->rsc.provider); lrmd_free_event(history->failed); lrmd_free_event(history->last); free(history->id); history_free_recurring_ops(history); free(history); } static void update_history_cache(lrm_state_t * lrm_state, lrmd_rsc_info_t * rsc, lrmd_event_data_t * op) { int target_rc = 0; rsc_history_t *entry = NULL; if (op->rsc_deleted) { crm_debug("Purged history for '%s' after %s", op->rsc_id, op->op_type); delete_rsc_status(lrm_state, op->rsc_id, cib_quorum_override, NULL); return; } if (safe_str_eq(op->op_type, RSC_NOTIFY)) { return; } crm_debug("Updating history for '%s' with %s op", op->rsc_id, op->op_type); entry = g_hash_table_lookup(lrm_state->resource_history, op->rsc_id); if (entry == NULL && rsc) { entry = calloc(1, sizeof(rsc_history_t)); entry->id = strdup(op->rsc_id); g_hash_table_insert(lrm_state->resource_history, entry->id, entry); entry->rsc.id = entry->id; entry->rsc.type = strdup(rsc->type); entry->rsc.standard = strdup(rsc->standard); if (rsc->provider) { entry->rsc.provider = strdup(rsc->provider); } else { entry->rsc.provider = NULL; } } else if (entry == NULL) { crm_info("Resource %s no longer exists, not updating cache", op->rsc_id); return; } entry->last_callid = op->call_id; target_rc = rsc_op_expected_rc(op); if (op->op_status == PCMK_LRM_OP_CANCELLED) { if (op->interval_ms > 0) { crm_trace("Removing cancelled recurring op: " CRM_OP_FMT, op->rsc_id, op->op_type, op->interval_ms); history_remove_recurring_op(entry, op); return; } else { crm_trace("Skipping " CRM_OP_FMT " rc=%d, status=%d", op->rsc_id, op->op_type, op->interval_ms, op->rc, op->op_status); } } else if (did_rsc_op_fail(op, target_rc)) { /* Store failed monitors here, otherwise the block below will cause them * to be forgotten when a stop happens. */ if (entry->failed) { lrmd_free_event(entry->failed); } entry->failed = lrmd_copy_event(op); } else if (op->interval_ms == 0) { if (entry->last) { lrmd_free_event(entry->last); } entry->last = lrmd_copy_event(op); if (op->params && (safe_str_eq(CRMD_ACTION_START, op->op_type) || safe_str_eq("reload", op->op_type) || safe_str_eq(CRMD_ACTION_STATUS, op->op_type))) { if (entry->stop_params) { g_hash_table_destroy(entry->stop_params); } entry->stop_params = crm_str_table_new(); g_hash_table_foreach(op->params, copy_instance_keys, entry->stop_params); } } if (op->interval_ms > 0) { /* Ensure there are no duplicates */ history_remove_recurring_op(entry, op); crm_trace("Adding recurring op: " CRM_OP_FMT, op->rsc_id, op->op_type, op->interval_ms); entry->recurring_op_list = g_list_prepend(entry->recurring_op_list, lrmd_copy_event(op)); } else if (entry->recurring_op_list && safe_str_eq(op->op_type, RSC_STATUS) == FALSE) { crm_trace("Dropping %d recurring ops because of: " CRM_OP_FMT, g_list_length(entry->recurring_op_list), op->rsc_id, op->op_type, op->interval_ms); history_free_recurring_ops(entry); } } /*! * \internal * \brief Send a direct OK ack for a resource task * * \param[in] lrm_state LRM connection * \param[in] input Input message being ack'ed * \param[in] rsc_id ID of affected resource * \param[in] rsc Affected resource (if available) * \param[in] task Operation task being ack'ed * \param[in] ack_host Name of host to send ack to * \param[in] ack_sys IPC system name to ack */ static void send_task_ok_ack(lrm_state_t *lrm_state, ha_msg_input_t *input, const char *rsc_id, lrmd_rsc_info_t *rsc, const char *task, const char *ack_host, const char *ack_sys) { lrmd_event_data_t *op = construct_op(lrm_state, input->xml, rsc_id, task); op->rc = PCMK_OCF_OK; op->op_status = PCMK_LRM_OP_DONE; send_direct_ack(ack_host, ack_sys, rsc, op, rsc_id); lrmd_free_event(op); } static inline const char * op_node_name(lrmd_event_data_t *op) { return op->remote_nodename? op->remote_nodename : fsa_our_uname; } void lrm_op_callback(lrmd_event_data_t * op) { CRM_CHECK(op != NULL, return); switch (op->type) { case lrmd_event_disconnect: if (op->remote_nodename == NULL) { /* If this is the local executor IPC connection, set the right * bits in the controller when the connection goes down. */ lrm_connection_destroy(); } break; case lrmd_event_exec_complete: { lrm_state_t *lrm_state = lrm_state_find(op_node_name(op)); CRM_ASSERT(lrm_state != NULL); process_lrm_event(lrm_state, op, NULL, NULL); } break; default: break; } } /* A_LRM_CONNECT */ void do_lrm_control(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { /* This only pertains to local executor connections. Remote connections are * handled as resources within the scheduler. Connecting and disconnecting * from remote executor instances is handled differently. */ lrm_state_t *lrm_state = NULL; if(fsa_our_uname == NULL) { return; /* Nothing to do */ } lrm_state = lrm_state_find_or_create(fsa_our_uname); if (lrm_state == NULL) { register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); return; } if (action & A_LRM_DISCONNECT) { if (lrm_state_verify_stopped(lrm_state, cur_state, LOG_INFO) == FALSE) { if (action == A_LRM_DISCONNECT) { crmd_fsa_stall(FALSE); return; } } clear_bit(fsa_input_register, R_LRM_CONNECTED); crm_info("Disconnecting from the executor"); lrm_state_disconnect(lrm_state); lrm_state_reset_tables(lrm_state, FALSE); crm_notice("Disconnected from the executor"); } if (action & A_LRM_CONNECT) { int ret = pcmk_ok; crm_debug("Connecting to the executor"); ret = lrm_state_ipc_connect(lrm_state); if (ret != pcmk_ok) { if (lrm_state->num_lrm_register_fails < MAX_LRM_REG_FAILS) { crm_warn("Failed to connect to the executor %d time%s (%d max)", lrm_state->num_lrm_register_fails, s_if_plural(lrm_state->num_lrm_register_fails), MAX_LRM_REG_FAILS); crm_timer_start(wait_timer); crmd_fsa_stall(FALSE); return; } } if (ret != pcmk_ok) { crm_err("Failed to connect to the executor the max allowed %d time%s", lrm_state->num_lrm_register_fails, s_if_plural(lrm_state->num_lrm_register_fails)); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); return; } set_bit(fsa_input_register, R_LRM_CONNECTED); crm_info("Connection to the executor established"); } if (action & ~(A_LRM_CONNECT | A_LRM_DISCONNECT)) { crm_err("Unexpected action %s in %s", fsa_action2string(action), __FUNCTION__); } } static gboolean lrm_state_verify_stopped(lrm_state_t * lrm_state, enum crmd_fsa_state cur_state, int log_level) { int counter = 0; gboolean rc = TRUE; const char *when = "lrm disconnect"; GHashTableIter gIter; const char *key = NULL; rsc_history_t *entry = NULL; struct recurring_op_s *pending = NULL; crm_debug("Checking for active resources before exit"); if (cur_state == S_TERMINATE) { log_level = LOG_ERR; when = "shutdown"; } else if (is_set(fsa_input_register, R_SHUTDOWN)) { when = "shutdown... waiting"; } if (lrm_state->pending_ops && lrm_state_is_connected(lrm_state) == TRUE) { guint removed = g_hash_table_foreach_remove( lrm_state->pending_ops, stop_recurring_actions, lrm_state); guint nremaining = g_hash_table_size(lrm_state->pending_ops); if (removed || nremaining) { crm_notice("Stopped %u recurring operation%s at %s (%u remaining)", removed, s_if_plural(removed), when, nremaining); } } if (lrm_state->pending_ops) { g_hash_table_iter_init(&gIter, lrm_state->pending_ops); while (g_hash_table_iter_next(&gIter, NULL, (void **)&pending)) { /* Ignore recurring actions in the shutdown calculations */ if (pending->interval_ms == 0) { counter++; } } } if (counter > 0) { do_crm_log(log_level, "%d pending executor operation%s at %s", counter, s_if_plural(counter), when); if (cur_state == S_TERMINATE || !is_set(fsa_input_register, R_SENT_RSC_STOP)) { g_hash_table_iter_init(&gIter, lrm_state->pending_ops); while (g_hash_table_iter_next(&gIter, (gpointer*)&key, (gpointer*)&pending)) { do_crm_log(log_level, "Pending action: %s (%s)", key, pending->op_key); } } else { rc = FALSE; } return rc; } if (lrm_state->resource_history == NULL) { return rc; } if (is_set(fsa_input_register, R_SHUTDOWN)) { /* At this point we're not waiting, we're just shutting down */ when = "shutdown"; } counter = 0; g_hash_table_iter_init(&gIter, lrm_state->resource_history); while (g_hash_table_iter_next(&gIter, NULL, (gpointer*)&entry)) { if (is_rsc_active(lrm_state, entry->id) == FALSE) { continue; } counter++; if (log_level == LOG_ERR) { crm_info("Found %s active at %s", entry->id, when); } else { crm_trace("Found %s active at %s", entry->id, when); } if (lrm_state->pending_ops) { GHashTableIter hIter; g_hash_table_iter_init(&hIter, lrm_state->pending_ops); while (g_hash_table_iter_next(&hIter, (gpointer*)&key, (gpointer*)&pending)) { if (crm_str_eq(entry->id, pending->rsc_id, TRUE)) { crm_notice("%sction %s (%s) incomplete at %s", pending->interval_ms == 0 ? "A" : "Recurring a", key, pending->op_key, when); } } } } if (counter) { crm_err("%d resource%s active at %s", counter, (counter == 1)? " was" : "s were", when); } return rc; } static char * build_parameter_list(const lrmd_event_data_t *op, const struct ra_metadata_s *metadata, xmlNode *result, enum ra_param_flags_e param_type, bool invert_for_xml) { int len = 0; int max = 0; char *list = NULL; GList *iter = NULL; /* Newer resource agents support the "private" parameter attribute to * indicate sensitive parameters. For backward compatibility with older * agents, this list is used if the agent doesn't specify any as "private". */ const char *secure_terms[] = { "password", "passwd", "user", }; if (is_not_set(metadata->ra_flags, ra_uses_private) && (param_type == ra_param_private)) { max = DIMOF(secure_terms); } for (iter = metadata->ra_params; iter != NULL; iter = iter->next) { struct ra_param_s *param = (struct ra_param_s *) iter->data; bool accept = FALSE; if (is_set(param->rap_flags, param_type)) { accept = TRUE; } else if (max) { for (int lpc = 0; lpc < max; lpc++) { if (safe_str_eq(secure_terms[lpc], param->rap_name)) { accept = TRUE; break; } } } if (accept) { int start = len; crm_trace("Attr %s is %s", param->rap_name, ra_param_flag2text(param_type)); len += strlen(param->rap_name) + 2; // include spaces around list = realloc_safe(list, len + 1); // include null terminator // spaces before and after make parsing simpler sprintf(list + start, " %s ", param->rap_name); } else { crm_trace("Rejecting %s for %s", param->rap_name, ra_param_flag2text(param_type)); } if (result && (invert_for_xml? !accept : accept)) { const char *v = g_hash_table_lookup(op->params, param->rap_name); if (v != NULL) { crm_trace("Adding attr %s=%s to the xml result", param->rap_name, v); crm_xml_add(result, param->rap_name, v); } } } return list; } static void append_restart_list(lrmd_event_data_t *op, struct ra_metadata_s *metadata, xmlNode *update, const char *version) { char *list = NULL; char *digest = NULL; xmlNode *restart = NULL; CRM_LOG_ASSERT(op->params != NULL); if (op->interval_ms > 0) { /* monitors are not reloadable */ return; } if (is_set(metadata->ra_flags, ra_supports_reload)) { restart = create_xml_node(NULL, XML_TAG_PARAMS); /* Add any parameters with unique="1" to the "op-force-restart" list. * * (Currently, we abuse "unique=0" to indicate reloadability. This is * nonstandard and should eventually be replaced once the OCF standard * is updated with something better.) */ list = build_parameter_list(op, metadata, restart, ra_param_unique, FALSE); } else { /* Resource does not support reloads */ return; } digest = calculate_operation_digest(restart, version); /* Add "op-force-restart" and "op-restart-digest" to indicate the resource supports reload, * no matter if it actually supports any parameters with unique="1"). */ crm_xml_add(update, XML_LRM_ATTR_OP_RESTART, list? list: ""); crm_xml_add(update, XML_LRM_ATTR_RESTART_DIGEST, digest); crm_trace("%s: %s, %s", op->rsc_id, digest, list); crm_log_xml_trace(restart, "restart digest source"); free_xml(restart); free(digest); free(list); } static void append_secure_list(lrmd_event_data_t *op, struct ra_metadata_s *metadata, xmlNode *update, const char *version) { char *list = NULL; char *digest = NULL; xmlNode *secure = NULL; CRM_LOG_ASSERT(op->params != NULL); /* * To keep XML_LRM_ATTR_OP_SECURE short, we want it to contain the * secure parameters but XML_LRM_ATTR_SECURE_DIGEST to be based on * the insecure ones */ secure = create_xml_node(NULL, XML_TAG_PARAMS); list = build_parameter_list(op, metadata, secure, ra_param_private, TRUE); if (list != NULL) { digest = calculate_operation_digest(secure, version); crm_xml_add(update, XML_LRM_ATTR_OP_SECURE, list); crm_xml_add(update, XML_LRM_ATTR_SECURE_DIGEST, digest); crm_trace("%s: %s, %s", op->rsc_id, digest, list); crm_log_xml_trace(secure, "secure digest source"); } else { crm_trace("%s: no secure parameters", op->rsc_id); } free_xml(secure); free(digest); free(list); } static gboolean build_operation_update(xmlNode * parent, lrmd_rsc_info_t * rsc, lrmd_event_data_t * op, const char *node_name, const char *src) { int target_rc = 0; xmlNode *xml_op = NULL; struct ra_metadata_s *metadata = NULL; const char *caller_version = NULL; lrm_state_t *lrm_state = NULL; if (op == NULL) { return FALSE; } target_rc = rsc_op_expected_rc(op); /* there is a small risk in formerly mixed clusters that it will * be sub-optimal. * * however with our upgrade policy, the update we send should * still be completely supported anyway */ caller_version = g_hash_table_lookup(op->params, XML_ATTR_CRM_VERSION); CRM_LOG_ASSERT(caller_version != NULL); if(caller_version == NULL) { caller_version = CRM_FEATURE_SET; } crm_trace("Building %s operation update with originator version: %s", op->rsc_id, caller_version); xml_op = create_operation_update(parent, op, caller_version, target_rc, fsa_our_uname, src, LOG_DEBUG); if (xml_op == NULL) { return TRUE; } if ((rsc == NULL) || (op->params == NULL) || !crm_op_needs_metadata(rsc->standard, op->op_type)) { crm_trace("No digests needed for %s action on %s (params=%p rsc=%p)", op->op_type, op->rsc_id, op->params, rsc); return TRUE; } lrm_state = lrm_state_find(node_name); if (lrm_state == NULL) { crm_warn("Cannot calculate digests for operation " CRM_OP_FMT " because we have no connection to executor for %s", op->rsc_id, op->op_type, op->interval_ms, node_name); return TRUE; } metadata = metadata_cache_get(lrm_state->metadata_cache, rsc); if (metadata == NULL) { /* For now, we always collect resource agent meta-data via a local, * synchronous, direct execution of the agent. This has multiple issues: * the executor should execute agents, not the controller; meta-data for * Pacemaker Remote nodes should be collected on those nodes, not * locally; and the meta-data call shouldn't eat into the timeout of the * real action being performed. * * These issues are planned to be addressed by having the scheduler * schedule a meta-data cache check at the beginning of each transition. * Once that is working, this block will only be a fallback in case the * initial collection fails. */ char *metadata_str = NULL; int rc = lrm_state_get_metadata(lrm_state, rsc->standard, rsc->provider, rsc->type, &metadata_str, 0); if (rc != pcmk_ok) { crm_warn("Failed to get metadata for %s (%s:%s:%s)", rsc->id, rsc->standard, rsc->provider, rsc->type); return TRUE; } metadata = metadata_cache_update(lrm_state->metadata_cache, rsc, metadata_str); free(metadata_str); if (metadata == NULL) { crm_warn("Failed to update metadata for %s (%s:%s:%s)", rsc->id, rsc->standard, rsc->provider, rsc->type); return TRUE; } } #if ENABLE_VERSIONED_ATTRS crm_xml_add(xml_op, XML_ATTR_RA_VERSION, metadata->ra_version); #endif crm_trace("Including additional digests for %s::%s:%s", rsc->standard, rsc->provider, rsc->type); append_restart_list(op, metadata, xml_op, caller_version); append_secure_list(op, metadata, xml_op, caller_version); return TRUE; } static gboolean is_rsc_active(lrm_state_t * lrm_state, const char *rsc_id) { rsc_history_t *entry = NULL; entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); if (entry == NULL || entry->last == NULL) { return FALSE; } crm_trace("Processing %s: %s.%d=%d", rsc_id, entry->last->op_type, entry->last->interval_ms, entry->last->rc); if (entry->last->rc == PCMK_OCF_OK && safe_str_eq(entry->last->op_type, CRMD_ACTION_STOP)) { return FALSE; } else if (entry->last->rc == PCMK_OCF_OK && safe_str_eq(entry->last->op_type, CRMD_ACTION_MIGRATE)) { /* a stricter check is too complex... * leave that to the PE */ return FALSE; } else if (entry->last->rc == PCMK_OCF_NOT_RUNNING) { return FALSE; } else if ((entry->last->interval_ms == 0) && (entry->last->rc == PCMK_OCF_NOT_CONFIGURED)) { /* Badly configured resources can't be reliably stopped */ return FALSE; } return TRUE; } static gboolean build_active_RAs(lrm_state_t * lrm_state, xmlNode * rsc_list) { GHashTableIter iter; rsc_history_t *entry = NULL; g_hash_table_iter_init(&iter, lrm_state->resource_history); while (g_hash_table_iter_next(&iter, NULL, (void **)&entry)) { GList *gIter = NULL; xmlNode *xml_rsc = create_xml_node(rsc_list, XML_LRM_TAG_RESOURCE); crm_xml_add(xml_rsc, XML_ATTR_ID, entry->id); crm_xml_add(xml_rsc, XML_ATTR_TYPE, entry->rsc.type); crm_xml_add(xml_rsc, XML_AGENT_ATTR_CLASS, entry->rsc.standard); crm_xml_add(xml_rsc, XML_AGENT_ATTR_PROVIDER, entry->rsc.provider); if (entry->last && entry->last->params) { const char *container = g_hash_table_lookup(entry->last->params, CRM_META"_"XML_RSC_ATTR_CONTAINER); if (container) { crm_trace("Resource %s is a part of container resource %s", entry->id, container); crm_xml_add(xml_rsc, XML_RSC_ATTR_CONTAINER, container); } } build_operation_update(xml_rsc, &(entry->rsc), entry->failed, lrm_state->node_name, __FUNCTION__); build_operation_update(xml_rsc, &(entry->rsc), entry->last, lrm_state->node_name, __FUNCTION__); for (gIter = entry->recurring_op_list; gIter != NULL; gIter = gIter->next) { build_operation_update(xml_rsc, &(entry->rsc), gIter->data, lrm_state->node_name, __FUNCTION__); } } return FALSE; } static xmlNode * do_lrm_query_internal(lrm_state_t *lrm_state, int update_flags) { xmlNode *xml_state = NULL; xmlNode *xml_data = NULL; xmlNode *rsc_list = NULL; crm_node_t *peer = NULL; peer = crm_get_peer_full(0, lrm_state->node_name, CRM_GET_PEER_ANY); CRM_CHECK(peer != NULL, return NULL); xml_state = create_node_state_update(peer, update_flags, NULL, __FUNCTION__); if (xml_state == NULL) { return NULL; } xml_data = create_xml_node(xml_state, XML_CIB_TAG_LRM); crm_xml_add(xml_data, XML_ATTR_ID, peer->uuid); rsc_list = create_xml_node(xml_data, XML_LRM_TAG_RESOURCES); /* Build a list of active (not always running) resources */ build_active_RAs(lrm_state, rsc_list); crm_log_xml_trace(xml_state, "Current executor state"); return xml_state; } xmlNode * do_lrm_query(gboolean is_replace, const char *node_name) { lrm_state_t *lrm_state = lrm_state_find(node_name); if (!lrm_state) { crm_err("Could not find executor state for node %s", node_name); return NULL; } return do_lrm_query_internal(lrm_state, node_update_cluster|node_update_peer); } static void notify_deleted(lrm_state_t * lrm_state, ha_msg_input_t * input, const char *rsc_id, int rc) { lrmd_event_data_t *op = NULL; const char *from_sys = crm_element_value(input->msg, F_CRM_SYS_FROM); const char *from_host = crm_element_value(input->msg, F_CRM_HOST_FROM); crm_info("Notifying %s on %s that %s was%s deleted", from_sys, (from_host? from_host : "localhost"), rsc_id, ((rc == pcmk_ok)? "" : " not")); op = construct_op(lrm_state, input->xml, rsc_id, CRMD_ACTION_DELETE); if (rc == pcmk_ok) { op->op_status = PCMK_LRM_OP_DONE; op->rc = PCMK_OCF_OK; } else { op->op_status = PCMK_LRM_OP_ERROR; op->rc = PCMK_OCF_UNKNOWN_ERROR; } send_direct_ack(from_host, from_sys, NULL, op, rsc_id); lrmd_free_event(op); if (safe_str_neq(from_sys, CRM_SYSTEM_TENGINE)) { /* this isn't expected - trigger a new transition */ time_t now = time(NULL); char *now_s = crm_itoa(now); crm_debug("Triggering a refresh after %s deleted %s from the executor", from_sys, rsc_id); update_attr_delegate(fsa_cib_conn, cib_none, XML_CIB_TAG_CRMCONFIG, NULL, NULL, NULL, NULL, "last-lrm-refresh", now_s, FALSE, NULL, NULL); free(now_s); } } static gboolean lrm_remove_deleted_rsc(gpointer key, gpointer value, gpointer user_data) { struct delete_event_s *event = user_data; struct pending_deletion_op_s *op = value; if (crm_str_eq(event->rsc, op->rsc, TRUE)) { notify_deleted(event->lrm_state, op->input, event->rsc, event->rc); return TRUE; } return FALSE; } static gboolean lrm_remove_deleted_op(gpointer key, gpointer value, gpointer user_data) { const char *rsc = user_data; struct recurring_op_s *pending = value; if (crm_str_eq(rsc, pending->rsc_id, TRUE)) { crm_info("Removing op %s:%d for deleted resource %s", pending->op_key, pending->call_id, rsc); return TRUE; } return FALSE; } /* * Remove the rsc from the CIB * * Avoids refreshing the entire LRM section of this host */ #define RSC_TEMPLATE "//"XML_CIB_TAG_STATE"[@uname='%s']//"XML_LRM_TAG_RESOURCE"[@id='%s']" static int delete_rsc_status(lrm_state_t * lrm_state, const char *rsc_id, int call_options, const char *user_name) { char *rsc_xpath = NULL; int rc = pcmk_ok; CRM_CHECK(rsc_id != NULL, return -ENXIO); rsc_xpath = crm_strdup_printf(RSC_TEMPLATE, lrm_state->node_name, rsc_id); rc = cib_internal_op(fsa_cib_conn, CIB_OP_DELETE, NULL, rsc_xpath, NULL, NULL, call_options | cib_xpath, user_name); free(rsc_xpath); return rc; } static void delete_rsc_entry(lrm_state_t * lrm_state, ha_msg_input_t * input, const char *rsc_id, GHashTableIter * rsc_gIter, int rc, const char *user_name) { struct delete_event_s event; CRM_CHECK(rsc_id != NULL, return); if (rc == pcmk_ok) { char *rsc_id_copy = strdup(rsc_id); if (rsc_gIter) g_hash_table_iter_remove(rsc_gIter); else g_hash_table_remove(lrm_state->resource_history, rsc_id_copy); crm_debug("sync: Sending delete op for %s", rsc_id_copy); delete_rsc_status(lrm_state, rsc_id_copy, cib_quorum_override, user_name); g_hash_table_foreach_remove(lrm_state->pending_ops, lrm_remove_deleted_op, rsc_id_copy); free(rsc_id_copy); } if (input) { notify_deleted(lrm_state, input, rsc_id, rc); } event.rc = rc; event.rsc = rsc_id; event.lrm_state = lrm_state; g_hash_table_foreach_remove(lrm_state->deletion_ops, lrm_remove_deleted_rsc, &event); } /*! * \internal * \brief Erase an LRM history entry from the CIB, given the operation data * * \param[in] lrm_state LRM state of the desired node * \param[in] op Operation whose history should be deleted */ static void erase_lrm_history_by_op(lrm_state_t *lrm_state, lrmd_event_data_t *op) { xmlNode *xml_top = NULL; CRM_CHECK(op != NULL, return); xml_top = create_xml_node(NULL, XML_LRM_TAG_RSC_OP); crm_xml_add_int(xml_top, XML_LRM_ATTR_CALLID, op->call_id); crm_xml_add(xml_top, XML_ATTR_TRANSITION_KEY, op->user_data); if (op->interval_ms > 0) { char *op_id = generate_op_key(op->rsc_id, op->op_type, op->interval_ms); /* Avoid deleting last_failure too (if it was a result of this recurring op failing) */ crm_xml_add(xml_top, XML_ATTR_ID, op_id); free(op_id); } crm_debug("Erasing resource operation history for " CRM_OP_FMT " (call=%d)", op->rsc_id, op->op_type, op->interval_ms, op->call_id); fsa_cib_conn->cmds->remove(fsa_cib_conn, XML_CIB_TAG_STATUS, xml_top, cib_quorum_override); crm_log_xml_trace(xml_top, "op:cancel"); free_xml(xml_top); } /* Define xpath to find LRM resource history entry by node and resource */ #define XPATH_HISTORY \ "/" XML_TAG_CIB "/" XML_CIB_TAG_STATUS \ "/" XML_CIB_TAG_STATE "[@" XML_ATTR_UNAME "='%s']" \ "/" XML_CIB_TAG_LRM "/" XML_LRM_TAG_RESOURCES \ "/" XML_LRM_TAG_RESOURCE "[@" XML_ATTR_ID "='%s']" \ "/" XML_LRM_TAG_RSC_OP /* ... and also by operation key */ #define XPATH_HISTORY_ID XPATH_HISTORY \ "[@" XML_ATTR_ID "='%s']" /* ... and also by operation key and operation call ID */ #define XPATH_HISTORY_CALL XPATH_HISTORY \ "[@" XML_ATTR_ID "='%s' and @" XML_LRM_ATTR_CALLID "='%d']" /* ... and also by operation key and original operation key */ #define XPATH_HISTORY_ORIG XPATH_HISTORY \ "[@" XML_ATTR_ID "='%s' and @" XML_LRM_ATTR_TASK_KEY "='%s']" /*! * \internal * \brief Erase an LRM history entry from the CIB, given operation identifiers * * \param[in] lrm_state LRM state of the node to clear history for * \param[in] rsc_id Name of resource to clear history for * \param[in] key Operation key of operation to clear history for * \param[in] orig_op If specified, delete only if it has this original op * \param[in] call_id If specified, delete entry only if it has this call ID */ static void erase_lrm_history_by_id(lrm_state_t *lrm_state, const char *rsc_id, const char *key, const char *orig_op, int call_id) { char *op_xpath = NULL; CRM_CHECK((rsc_id != NULL) && (key != NULL), return); if (call_id > 0) { op_xpath = crm_strdup_printf(XPATH_HISTORY_CALL, lrm_state->node_name, rsc_id, key, call_id); } else if (orig_op) { op_xpath = crm_strdup_printf(XPATH_HISTORY_ORIG, lrm_state->node_name, rsc_id, key, orig_op); } else { op_xpath = crm_strdup_printf(XPATH_HISTORY_ID, lrm_state->node_name, rsc_id, key); } crm_debug("Erasing resource operation history for %s on %s (call=%d)", key, rsc_id, call_id); fsa_cib_conn->cmds->remove(fsa_cib_conn, op_xpath, NULL, cib_quorum_override | cib_xpath); free(op_xpath); } static inline gboolean last_failed_matches_op(rsc_history_t *entry, const char *op, guint interval_ms) { if (entry == NULL) { return FALSE; } if (op == NULL) { return TRUE; } return (safe_str_eq(op, entry->failed->op_type) && (interval_ms == entry->failed->interval_ms)); } /*! * \internal * \brief Clear a resource's last failure * * Erase a resource's last failure on a particular node from both the * LRM resource history in the CIB, and the resource history remembered * for the LRM state. * * \param[in] rsc_id Resource name * \param[in] node_name Node name * \param[in] operation If specified, only clear if matching this operation * \param[in] interval_ms If operation is specified, it has this interval */ void lrm_clear_last_failure(const char *rsc_id, const char *node_name, const char *operation, guint interval_ms) { char *op_key = NULL; char *orig_op_key = NULL; lrm_state_t *lrm_state = NULL; lrm_state = lrm_state_find(node_name); if (lrm_state == NULL) { return; } /* Erase from CIB */ op_key = generate_op_key(rsc_id, "last_failure", 0); if (operation) { orig_op_key = generate_op_key(rsc_id, operation, interval_ms); } erase_lrm_history_by_id(lrm_state, rsc_id, op_key, orig_op_key, 0); free(op_key); free(orig_op_key); /* Remove from memory */ if (lrm_state->resource_history) { rsc_history_t *entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); if (last_failed_matches_op(entry, operation, interval_ms)) { lrmd_free_event(entry->failed); entry->failed = NULL; } } } /* Returns: gboolean - cancellation is in progress */ static gboolean cancel_op(lrm_state_t * lrm_state, const char *rsc_id, const char *key, int op, gboolean remove) { int rc = pcmk_ok; char *local_key = NULL; struct recurring_op_s *pending = NULL; CRM_CHECK(op != 0, return FALSE); CRM_CHECK(rsc_id != NULL, return FALSE); if (key == NULL) { local_key = make_stop_id(rsc_id, op); key = local_key; } pending = g_hash_table_lookup(lrm_state->pending_ops, key); if (pending) { if (remove && pending->remove == FALSE) { pending->remove = TRUE; crm_debug("Scheduling %s for removal", key); } if (pending->cancelled) { crm_debug("Operation %s already cancelled", key); free(local_key); return FALSE; } pending->cancelled = TRUE; } else { crm_info("No pending op found for %s", key); free(local_key); return FALSE; } crm_debug("Cancelling op %d for %s (%s)", op, rsc_id, key); rc = lrm_state_cancel(lrm_state, pending->rsc_id, pending->op_type, pending->interval_ms); if (rc == pcmk_ok) { crm_debug("Op %d for %s (%s): cancelled", op, rsc_id, key); free(local_key); return TRUE; } crm_debug("Op %d for %s (%s): Nothing to cancel", op, rsc_id, key); /* The caller needs to make sure the entry is * removed from the pending_ops list * * Usually by returning TRUE inside the worker function * supplied to g_hash_table_foreach_remove() * * Not removing the entry from pending_ops will block * the node from shutting down */ free(local_key); return FALSE; } struct cancel_data { gboolean done; gboolean remove; const char *key; lrmd_rsc_info_t *rsc; lrm_state_t *lrm_state; }; static gboolean cancel_action_by_key(gpointer key, gpointer value, gpointer user_data) { gboolean remove = FALSE; struct cancel_data *data = user_data; struct recurring_op_s *op = (struct recurring_op_s *)value; if (crm_str_eq(op->op_key, data->key, TRUE)) { data->done = TRUE; remove = !cancel_op(data->lrm_state, data->rsc->id, key, op->call_id, data->remove); } return remove; } static gboolean cancel_op_key(lrm_state_t * lrm_state, lrmd_rsc_info_t * rsc, const char *key, gboolean remove) { guint removed = 0; struct cancel_data data; CRM_CHECK(rsc != NULL, return FALSE); CRM_CHECK(key != NULL, return FALSE); data.key = key; data.rsc = rsc; data.done = FALSE; data.remove = remove; data.lrm_state = lrm_state; removed = g_hash_table_foreach_remove(lrm_state->pending_ops, cancel_action_by_key, &data); crm_trace("Removed %u op cache entries, new size: %u", removed, g_hash_table_size(lrm_state->pending_ops)); return data.done; } /*! * \internal * \brief Retrieve resource information from LRM * * \param[in] lrm_state LRM connection to use * \param[in] rsc_xml XML containing resource configuration * \param[in] do_create If true, register resource with LRM if not already * \param[out] rsc_info Where to store resource information obtained from LRM * * \retval pcmk_ok Success (and rsc_info holds newly allocated result) * \retval -EINVAL Required information is missing from arguments * \retval -ENOTCONN No active connection to LRM * \retval -ENODEV Resource not found * \retval -errno Error communicating with executor when registering resource * * \note Caller is responsible for freeing result on success. */ static int get_lrm_resource(lrm_state_t *lrm_state, xmlNode *rsc_xml, gboolean do_create, lrmd_rsc_info_t **rsc_info) { const char *id = ID(rsc_xml); CRM_CHECK(lrm_state && rsc_xml && rsc_info, return -EINVAL); CRM_CHECK(id, return -EINVAL); if (lrm_state_is_connected(lrm_state) == FALSE) { return -ENOTCONN; } crm_trace("Retrieving resource information for %s from the executor", id); *rsc_info = lrm_state_get_rsc_info(lrm_state, id, 0); // If resource isn't known by ID, try clone name, if provided if (!*rsc_info) { const char *long_id = crm_element_value(rsc_xml, XML_ATTR_ID_LONG); if (long_id) { *rsc_info = lrm_state_get_rsc_info(lrm_state, long_id, 0); } } if ((*rsc_info == NULL) && do_create) { const char *class = crm_element_value(rsc_xml, XML_AGENT_ATTR_CLASS); const char *provider = crm_element_value(rsc_xml, XML_AGENT_ATTR_PROVIDER); const char *type = crm_element_value(rsc_xml, XML_ATTR_TYPE); int rc; crm_trace("Registering resource %s with the executor", id); rc = lrm_state_register_rsc(lrm_state, id, class, provider, type, lrmd_opt_drop_recurring); if (rc != pcmk_ok) { fsa_data_t *msg_data = NULL; crm_err("Could not register resource %s with the executor on %s: %s " CRM_XS " rc=%d", id, lrm_state->node_name, pcmk_strerror(rc), rc); /* Register this as an internal error if this involves the local * executor. Otherwise, we're likely dealing with an unresponsive * remote node, which is not an FSA failure. */ if (lrm_state_is_local(lrm_state) == TRUE) { register_fsa_error(C_FSA_INTERNAL, I_FAIL, NULL); } return rc; } *rsc_info = lrm_state_get_rsc_info(lrm_state, id, 0); } return *rsc_info? pcmk_ok : -ENODEV; } static void delete_resource(lrm_state_t * lrm_state, const char *id, lrmd_rsc_info_t * rsc, GHashTableIter * gIter, const char *sys, const char *host, const char *user, ha_msg_input_t * request, gboolean unregister) { int rc = pcmk_ok; crm_info("Removing resource %s for %s (%s) on %s", id, sys, user ? user : "internal", host); if (rsc && unregister) { rc = lrm_state_unregister_rsc(lrm_state, id, 0); } if (rc == pcmk_ok) { crm_trace("Resource '%s' deleted", id); } else if (rc == -EINPROGRESS) { crm_info("Deletion of resource '%s' pending", id); if (request) { struct pending_deletion_op_s *op = NULL; char *ref = crm_element_value_copy(request->msg, XML_ATTR_REFERENCE); op = calloc(1, sizeof(struct pending_deletion_op_s)); op->rsc = strdup(rsc->id); op->input = copy_ha_msg_input(request); g_hash_table_insert(lrm_state->deletion_ops, ref, op); } return; } else { crm_warn("Deletion of resource '%s' for %s (%s) on %s failed: %d", id, sys, user ? user : "internal", host, rc); } delete_rsc_entry(lrm_state, request, id, gIter, rc, user); } static int get_fake_call_id(lrm_state_t *lrm_state, const char *rsc_id) { int call_id = 999999999; rsc_history_t *entry = NULL; if(lrm_state) { entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); } /* Make sure the call id is greater than the last successful operation, * otherwise the failure will not result in a possible recovery of the resource * as it could appear the failure occurred before the successful start */ if (entry) { call_id = entry->last_callid + 1; } if (call_id < 0) { call_id = 1; } return call_id; } static void fake_op_status(lrm_state_t *lrm_state, lrmd_event_data_t *op, int op_status, enum ocf_exitcode op_exitcode) { op->call_id = get_fake_call_id(lrm_state, op->rsc_id); op->t_run = time(NULL); op->t_rcchange = op->t_run; op->op_status = op_status; op->rc = op_exitcode; } static void force_reprobe(lrm_state_t *lrm_state, const char *from_sys, const char *from_host, const char *user_name, gboolean is_remote_node) { GHashTableIter gIter; rsc_history_t *entry = NULL; crm_info("Clearing resource history on node %s", lrm_state->node_name); g_hash_table_iter_init(&gIter, lrm_state->resource_history); while (g_hash_table_iter_next(&gIter, NULL, (void **)&entry)) { /* only unregister the resource during a reprobe if it is not a remote connection * resource. otherwise unregistering the connection will terminate remote-node * membership */ gboolean unregister = TRUE; if (is_remote_lrmd_ra(NULL, NULL, entry->id)) { lrm_state_t *remote_lrm_state = lrm_state_find(entry->id); if (remote_lrm_state) { /* when forcing a reprobe, make sure to clear remote node before * clearing the remote node's connection resource */ force_reprobe(remote_lrm_state, from_sys, from_host, user_name, TRUE); } unregister = FALSE; } delete_resource(lrm_state, entry->id, &entry->rsc, &gIter, from_sys, from_host, user_name, NULL, unregister); } /* Now delete the copy in the CIB */ erase_status_tag(lrm_state->node_name, XML_CIB_TAG_LRM, cib_scope_local); /* Finally, _delete_ the value in pacemaker-attrd -- setting it to FALSE * would result in the scheduler sending us back here again */ update_attrd(lrm_state->node_name, CRM_OP_PROBED, NULL, user_name, is_remote_node); } static void synthesize_lrmd_failure(lrm_state_t *lrm_state, xmlNode *action, int rc) { lrmd_event_data_t *op = NULL; const char *operation = crm_element_value(action, XML_LRM_ATTR_TASK); const char *target_node = crm_element_value(action, XML_LRM_ATTR_TARGET); xmlNode *xml_rsc = find_xml_node(action, XML_CIB_TAG_RESOURCE, TRUE); if ((xml_rsc == NULL) || (ID(xml_rsc) == NULL)) { /* @TODO Should we do something else, like direct ack? */ crm_info("Can't fake %s failure (%d) on %s without resource configuration", crm_element_value(action, XML_LRM_ATTR_TASK_KEY), rc, target_node); return; } else if(operation == NULL) { /* This probably came from crm_resource -C, nothing to do */ crm_info("Can't fake %s failure (%d) on %s without operation", ID(xml_rsc), rc, target_node); return; } op = construct_op(lrm_state, action, ID(xml_rsc), operation); if (safe_str_eq(operation, RSC_NOTIFY)) { // Notifications can't fail fake_op_status(lrm_state, op, PCMK_LRM_OP_DONE, PCMK_OCF_OK); } else { fake_op_status(lrm_state, op, PCMK_LRM_OP_ERROR, rc); } crm_info("Faking " CRM_OP_FMT " result (%d) on %s", op->rsc_id, op->op_type, op->interval_ms, op->rc, target_node); // Process the result as if it came from the LRM process_lrm_event(lrm_state, op, NULL, action); lrmd_free_event(op); } /*! * \internal * \brief Get target of an LRM operation * * \param[in] xml LRM operation data XML * * \return LRM operation target node name (local node or Pacemaker Remote node) */ static const char * lrm_op_target(xmlNode *xml) { const char *target = NULL; if (xml) { target = crm_element_value(xml, XML_LRM_ATTR_TARGET); } if (target == NULL) { target = fsa_our_uname; } return target; } static void fail_lrm_resource(xmlNode *xml, lrm_state_t *lrm_state, const char *user_name, const char *from_host, const char *from_sys) { lrmd_event_data_t *op = NULL; lrmd_rsc_info_t *rsc = NULL; xmlNode *xml_rsc = find_xml_node(xml, XML_CIB_TAG_RESOURCE, TRUE); CRM_CHECK(xml_rsc != NULL, return); /* The executor simply executes operations and reports the results, without * any concept of success or failure, so to fail a resource, we must fake * what a failure looks like. * * To do this, we create a fake executor operation event for the resource, * and pass that event to the executor client callback so it will be * processed as if it came from the executor. */ op = construct_op(lrm_state, xml, ID(xml_rsc), "asyncmon"); fake_op_status(lrm_state, op, PCMK_LRM_OP_DONE, PCMK_OCF_UNKNOWN_ERROR); free((char*) op->user_data); op->user_data = NULL; op->interval_ms = 0; #if ENABLE_ACL if (user_name && is_privileged(user_name) == FALSE) { crm_err("%s does not have permission to fail %s", user_name, ID(xml_rsc)); send_direct_ack(from_host, from_sys, NULL, op, ID(xml_rsc)); lrmd_free_event(op); return; } #endif if (get_lrm_resource(lrm_state, xml_rsc, TRUE, &rsc) == pcmk_ok) { crm_info("Failing resource %s...", rsc->id); op->exit_reason = strdup("Simulated failure"); process_lrm_event(lrm_state, op, NULL, xml); op->op_status = PCMK_LRM_OP_DONE; op->rc = PCMK_OCF_OK; lrmd_free_rsc_info(rsc); } else { crm_info("Cannot find/create resource in order to fail it..."); crm_log_xml_warn(xml, "bad input"); } send_direct_ack(from_host, from_sys, NULL, op, ID(xml_rsc)); lrmd_free_event(op); } static void handle_refresh_op(lrm_state_t *lrm_state, const char *user_name, const char *from_host, const char *from_sys) { int rc = pcmk_ok; xmlNode *fragment = do_lrm_query_internal(lrm_state, node_update_all); fsa_cib_update(XML_CIB_TAG_STATUS, fragment, cib_quorum_override, rc, user_name); crm_info("Forced a local resource history refresh: call=%d", rc); if (safe_str_neq(CRM_SYSTEM_CRMD, from_sys)) { xmlNode *reply = create_request(CRM_OP_INVOKE_LRM, fragment, from_host, from_sys, CRM_SYSTEM_LRMD, fsa_our_uuid); crm_debug("ACK'ing refresh from %s (%s)", from_sys, from_host); if (relay_message(reply, TRUE) == FALSE) { crm_log_xml_err(reply, "Unable to route reply"); } free_xml(reply); } free_xml(fragment); } static void handle_query_op(xmlNode *msg, lrm_state_t *lrm_state) { xmlNode *data = do_lrm_query_internal(lrm_state, node_update_all); xmlNode *reply = create_reply(msg, data); if (relay_message(reply, TRUE) == FALSE) { crm_err("Unable to route reply"); crm_log_xml_err(reply, "reply"); } free_xml(reply); free_xml(data); } static void handle_reprobe_op(lrm_state_t *lrm_state, const char *from_sys, const char *from_host, const char *user_name, gboolean is_remote_node) { crm_notice("Forcing the status of all resources to be redetected"); force_reprobe(lrm_state, from_sys, from_host, user_name, is_remote_node); if (safe_str_neq(CRM_SYSTEM_PENGINE, from_sys) && safe_str_neq(CRM_SYSTEM_TENGINE, from_sys)) { xmlNode *reply = create_request(CRM_OP_INVOKE_LRM, NULL, from_host, from_sys, CRM_SYSTEM_LRMD, fsa_our_uuid); crm_debug("ACK'ing re-probe from %s (%s)", from_sys, from_host); if (relay_message(reply, TRUE) == FALSE) { crm_log_xml_err(reply, "Unable to route reply"); } free_xml(reply); } } static bool do_lrm_cancel(ha_msg_input_t *input, lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, const char *from_host, const char *from_sys) { char *op_key = NULL; char *meta_key = NULL; int call = 0; const char *call_id = NULL; const char *op_task = NULL; const char *interval_ms_s = NULL; gboolean in_progress = FALSE; xmlNode *params = find_xml_node(input->xml, XML_TAG_ATTRS, TRUE); CRM_CHECK(params != NULL, return FALSE); meta_key = crm_meta_name(XML_LRM_ATTR_INTERVAL_MS); interval_ms_s = crm_element_value(params, meta_key); free(meta_key); CRM_CHECK(interval_ms_s != NULL, return FALSE); meta_key = crm_meta_name(XML_LRM_ATTR_TASK); op_task = crm_element_value(params, meta_key); free(meta_key); CRM_CHECK(op_task != NULL, return FALSE); meta_key = crm_meta_name(XML_LRM_ATTR_CALLID); call_id = crm_element_value(params, meta_key); free(meta_key); op_key = generate_op_key(rsc->id, op_task, crm_parse_ms(interval_ms_s)); crm_debug("Scheduler requested op %s (call=%s) be cancelled", op_key, (call_id? call_id : "NA")); call = crm_parse_int(call_id, "0"); if (call == 0) { // Normal case when the scheduler cancels a recurring op in_progress = cancel_op_key(lrm_state, rsc, op_key, TRUE); } else { // Normal case when the scheduler cancels an orphan op in_progress = cancel_op(lrm_state, rsc->id, NULL, call, TRUE); } // Acknowledge cancellation operation if for a remote connection resource if (!in_progress || is_remote_lrmd_ra(NULL, NULL, rsc->id)) { char *op_id = make_stop_id(rsc->id, call); if (is_remote_lrmd_ra(NULL, NULL, rsc->id) == FALSE) { crm_info("Nothing known about operation %d for %s", call, op_key); } erase_lrm_history_by_id(lrm_state, rsc->id, op_key, NULL, call); send_task_ok_ack(lrm_state, input, rsc->id, rsc, op_task, from_host, from_sys); /* needed at least for cancellation of a remote operation */ g_hash_table_remove(lrm_state->pending_ops, op_id); free(op_id); } else { /* No ack is needed since abcdaa8, but peers with older versions * in a rolling upgrade need one. We didn't bump the feature set * at that commit, so we can only compare against the previous * CRM version (3.0.8). If any peers have feature set 3.0.9 but * not abcdaa8, they will time out waiting for the ack (no * released versions of Pacemaker are affected). */ const char *peer_version = crm_element_value(params, XML_ATTR_CRM_VERSION); if (compare_version(peer_version, "3.0.8") <= 0) { crm_info("Sending compatibility ack for %s cancellation to %s (CRM version %s)", op_key, from_host, peer_version); send_task_ok_ack(lrm_state, input, rsc->id, rsc, op_task, from_host, from_sys); } } free(op_key); return TRUE; } static void do_lrm_delete(ha_msg_input_t *input, lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, const char *from_sys, const char *from_host, bool crm_rsc_delete, const char *user_name) { gboolean unregister = TRUE; #if ENABLE_ACL int cib_rc = delete_rsc_status(lrm_state, rsc->id, cib_dryrun|cib_sync_call, user_name); if (cib_rc != pcmk_ok) { lrmd_event_data_t *op = NULL; crm_err("Could not delete resource status of %s for %s (user %s) on %s: %s" CRM_XS " rc=%d", rsc->id, from_sys, (user_name? user_name : "unknown"), from_host, pcmk_strerror(cib_rc), cib_rc); op = construct_op(lrm_state, input->xml, rsc->id, CRMD_ACTION_DELETE); op->op_status = PCMK_LRM_OP_ERROR; if (cib_rc == -EACCES) { op->rc = PCMK_OCF_INSUFFICIENT_PRIV; } else { op->rc = PCMK_OCF_UNKNOWN_ERROR; } send_direct_ack(from_host, from_sys, NULL, op, rsc->id); lrmd_free_event(op); return; } #endif if (crm_rsc_delete && is_remote_lrmd_ra(NULL, NULL, rsc->id)) { unregister = FALSE; } delete_resource(lrm_state, rsc->id, rsc, NULL, from_sys, from_host, user_name, input, unregister); } /* A_LRM_INVOKE */ void do_lrm_invoke(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { lrm_state_t *lrm_state = NULL; const char *crm_op = NULL; const char *from_sys = NULL; const char *from_host = NULL; const char *operation = NULL; ha_msg_input_t *input = fsa_typed_data(fsa_dt_ha_msg); const char *user_name = NULL; const char *target_node = NULL; gboolean is_remote_node = FALSE; bool crm_rsc_delete = FALSE; target_node = lrm_op_target(input->xml); is_remote_node = safe_str_neq(target_node, fsa_our_uname); lrm_state = lrm_state_find(target_node); if ((lrm_state == NULL) && is_remote_node) { crm_err("Failing action because local node has never had connection to remote node %s", target_node); synthesize_lrmd_failure(NULL, input->xml, PCMK_OCF_CONNECTION_DIED); return; } CRM_ASSERT(lrm_state != NULL); #if ENABLE_ACL user_name = crm_acl_get_set_user(input->msg, F_CRM_USER, NULL); crm_trace("Executor command from user '%s'", user_name); #endif crm_op = crm_element_value(input->msg, F_CRM_TASK); from_sys = crm_element_value(input->msg, F_CRM_SYS_FROM); if (safe_str_neq(from_sys, CRM_SYSTEM_TENGINE)) { from_host = crm_element_value(input->msg, F_CRM_HOST_FROM); } crm_trace("Executor %s command from %s", crm_op, from_sys); if (safe_str_eq(crm_op, CRM_OP_LRM_DELETE)) { crm_rsc_delete = TRUE; // Only crm_resource uses this op operation = CRMD_ACTION_DELETE; } else if (safe_str_eq(crm_op, CRM_OP_LRM_FAIL)) { fail_lrm_resource(input->xml, lrm_state, user_name, from_host, from_sys); return; } else if (input->xml != NULL) { operation = crm_element_value(input->xml, XML_LRM_ATTR_TASK); } if (safe_str_eq(crm_op, CRM_OP_LRM_REFRESH)) { handle_refresh_op(lrm_state, user_name, from_host, from_sys); } else if (safe_str_eq(crm_op, CRM_OP_LRM_QUERY)) { handle_query_op(input->msg, lrm_state); } else if (safe_str_eq(operation, CRM_OP_PROBED)) { update_attrd(lrm_state->node_name, CRM_OP_PROBED, XML_BOOLEAN_TRUE, user_name, is_remote_node); } else if (safe_str_eq(operation, CRM_OP_REPROBE) || safe_str_eq(crm_op, CRM_OP_REPROBE)) { handle_reprobe_op(lrm_state, from_sys, from_host, user_name, is_remote_node); } else if (operation != NULL) { lrmd_rsc_info_t *rsc = NULL; xmlNode *xml_rsc = find_xml_node(input->xml, XML_CIB_TAG_RESOURCE, TRUE); gboolean create_rsc = safe_str_neq(operation, CRMD_ACTION_DELETE); int rc; // We can't return anything meaningful without a resource ID CRM_CHECK(xml_rsc && ID(xml_rsc), return); rc = get_lrm_resource(lrm_state, xml_rsc, create_rsc, &rsc); if (rc == -ENOTCONN) { synthesize_lrmd_failure(lrm_state, input->xml, PCMK_OCF_CONNECTION_DIED); return; } else if ((rc < 0) && !create_rsc) { /* Delete of malformed or nonexistent resource * (deleting something that does not exist is a success) */ crm_notice("Not registering resource '%s' for a %s event " CRM_XS " get-rc=%d (%s) transition-key=%s", ID(xml_rsc), operation, rc, pcmk_strerror(rc), ID(input->xml)); delete_rsc_entry(lrm_state, input, ID(xml_rsc), NULL, pcmk_ok, user_name); send_task_ok_ack(lrm_state, input, ID(xml_rsc), NULL, operation, from_host, from_sys); return; } else if (rc == -EINVAL) { // Resource operation on malformed resource crm_err("Invalid resource definition for %s", ID(xml_rsc)); crm_log_xml_warn(input->msg, "invalid resource"); synthesize_lrmd_failure(lrm_state, input->xml, PCMK_OCF_NOT_CONFIGURED); // fatal error return; } else if (rc < 0) { // Error communicating with the executor crm_err("Could not register resource '%s' with executor: %s " CRM_XS " rc=%d", ID(xml_rsc), pcmk_strerror(rc), rc); crm_log_xml_warn(input->msg, "failed registration"); synthesize_lrmd_failure(lrm_state, input->xml, PCMK_OCF_INVALID_PARAM); // hard error return; } if (safe_str_eq(operation, CRMD_ACTION_CANCEL)) { if (!do_lrm_cancel(input, lrm_state, rsc, from_host, from_sys)) { crm_log_xml_warn(input->xml, "Bad command"); } } else if (safe_str_eq(operation, CRMD_ACTION_DELETE)) { do_lrm_delete(input, lrm_state, rsc, from_sys, from_host, crm_rsc_delete, user_name); } else { do_lrm_rsc_op(lrm_state, rsc, operation, input->xml, input->msg); } lrmd_free_rsc_info(rsc); } else { crm_err("Cannot perform operation %s of unknown type", crm_str(crm_op)); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } } static lrmd_event_data_t * construct_op(lrm_state_t * lrm_state, xmlNode * rsc_op, const char *rsc_id, const char *operation) { lrmd_event_data_t *op = NULL; const char *op_delay = NULL; const char *op_timeout = NULL; const char *interval_ms_s = NULL; GHashTable *params = NULL; const char *transition = NULL; CRM_ASSERT(rsc_id && operation); op = calloc(1, sizeof(lrmd_event_data_t)); CRM_ASSERT(op != NULL); op->type = lrmd_event_exec_complete; op->op_type = strdup(operation); op->op_status = PCMK_LRM_OP_PENDING; op->rc = -1; op->rsc_id = strdup(rsc_id); op->interval_ms = 0; op->timeout = 0; op->start_delay = 0; if (rsc_op == NULL) { CRM_LOG_ASSERT(safe_str_eq(CRMD_ACTION_STOP, operation)); op->user_data = NULL; /* the stop_all_resources() case * by definition there is no DC (or they'd be shutting * us down). * So we should put our version here. */ op->params = crm_str_table_new(); g_hash_table_insert(op->params, strdup(XML_ATTR_CRM_VERSION), strdup(CRM_FEATURE_SET)); crm_trace("Constructed %s op for %s", operation, rsc_id); return op; } params = xml2list(rsc_op); g_hash_table_remove(params, CRM_META "_op_target_rc"); op_delay = crm_meta_value(params, XML_OP_ATTR_START_DELAY); op_timeout = crm_meta_value(params, XML_ATTR_TIMEOUT); interval_ms_s = crm_meta_value(params, XML_LRM_ATTR_INTERVAL_MS); op->interval_ms = crm_parse_ms(interval_ms_s); op->timeout = crm_parse_int(op_timeout, "0"); op->start_delay = crm_parse_int(op_delay, "0"); #if ENABLE_VERSIONED_ATTRS // Resolve any versioned parameters if (lrm_state && safe_str_neq(op->op_type, RSC_METADATA) && safe_str_neq(op->op_type, CRMD_ACTION_DELETE) && !is_remote_lrmd_ra(NULL, NULL, rsc_id)) { // Resource info *should* already be cached, so we don't get executor call lrmd_rsc_info_t *rsc = lrm_state_get_rsc_info(lrm_state, rsc_id, 0); struct ra_metadata_s *metadata; metadata = metadata_cache_get(lrm_state->metadata_cache, rsc); if (metadata) { xmlNode *versioned_attrs = NULL; GHashTable *hash = NULL; char *key = NULL; char *value = NULL; GHashTableIter iter; versioned_attrs = first_named_child(rsc_op, XML_TAG_OP_VER_ATTRS); hash = pe_unpack_versioned_parameters(versioned_attrs, metadata->ra_version); g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) { g_hash_table_iter_steal(&iter); g_hash_table_replace(params, key, value); } g_hash_table_destroy(hash); versioned_attrs = first_named_child(rsc_op, XML_TAG_OP_VER_META); hash = pe_unpack_versioned_parameters(versioned_attrs, metadata->ra_version); g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) { g_hash_table_replace(params, crm_meta_name(key), strdup(value)); if (safe_str_eq(key, XML_ATTR_TIMEOUT)) { op->timeout = crm_parse_int(value, "0"); } else if (safe_str_eq(key, XML_OP_ATTR_START_DELAY)) { op->start_delay = crm_parse_int(value, "0"); } } g_hash_table_destroy(hash); versioned_attrs = first_named_child(rsc_op, XML_TAG_RSC_VER_ATTRS); hash = pe_unpack_versioned_parameters(versioned_attrs, metadata->ra_version); g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) { g_hash_table_iter_steal(&iter); g_hash_table_replace(params, key, value); } g_hash_table_destroy(hash); } lrmd_free_rsc_info(rsc); } #endif if (safe_str_neq(operation, RSC_STOP)) { op->params = params; } else { rsc_history_t *entry = NULL; if (lrm_state) { entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); } /* If we do not have stop parameters cached, use * whatever we are given */ if (!entry || !entry->stop_params) { op->params = params; } else { /* Copy the cached parameter list so that we stop the resource * with the old attributes, not the new ones */ op->params = crm_str_table_new(); g_hash_table_foreach(params, copy_meta_keys, op->params); g_hash_table_foreach(entry->stop_params, copy_instance_keys, op->params); g_hash_table_destroy(params); params = NULL; } } /* sanity */ if (op->timeout <= 0) { op->timeout = op->interval_ms; } if (op->start_delay < 0) { op->start_delay = 0; } transition = crm_element_value(rsc_op, XML_ATTR_TRANSITION_KEY); CRM_CHECK(transition != NULL, return op); op->user_data = strdup(transition); if (op->interval_ms != 0) { if (safe_str_eq(operation, CRMD_ACTION_START) || safe_str_eq(operation, CRMD_ACTION_STOP)) { crm_err("Start and Stop actions cannot have an interval: %u", op->interval_ms); op->interval_ms = 0; } } crm_trace("Constructed %s op for %s: interval=%u", operation, rsc_id, op->interval_ms); return op; } void send_direct_ack(const char *to_host, const char *to_sys, lrmd_rsc_info_t * rsc, lrmd_event_data_t * op, const char *rsc_id) { xmlNode *reply = NULL; xmlNode *update, *iter; crm_node_t *peer = NULL; CRM_CHECK(op != NULL, return); if (op->rsc_id == NULL) { CRM_ASSERT(rsc_id != NULL); op->rsc_id = strdup(rsc_id); } if (to_sys == NULL) { to_sys = CRM_SYSTEM_TENGINE; } peer = crm_get_peer(0, fsa_our_uname); update = create_node_state_update(peer, node_update_none, NULL, __FUNCTION__); iter = create_xml_node(update, XML_CIB_TAG_LRM); crm_xml_add(iter, XML_ATTR_ID, fsa_our_uuid); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCES); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCE); crm_xml_add(iter, XML_ATTR_ID, op->rsc_id); build_operation_update(iter, rsc, op, fsa_our_uname, __FUNCTION__); reply = create_request(CRM_OP_INVOKE_LRM, update, to_host, to_sys, CRM_SYSTEM_LRMD, NULL); crm_log_xml_trace(update, "ACK Update"); crm_debug("ACK'ing resource op " CRM_OP_FMT " from %s: %s", op->rsc_id, op->op_type, op->interval_ms, op->user_data, crm_element_value(reply, XML_ATTR_REFERENCE)); if (relay_message(reply, TRUE) == FALSE) { crm_log_xml_err(reply, "Unable to route reply"); } free_xml(update); free_xml(reply); } gboolean verify_stopped(enum crmd_fsa_state cur_state, int log_level) { gboolean res = TRUE; GList *lrm_state_list = lrm_state_get_list(); GList *state_entry; for (state_entry = lrm_state_list; state_entry != NULL; state_entry = state_entry->next) { lrm_state_t *lrm_state = state_entry->data; if (!lrm_state_verify_stopped(lrm_state, cur_state, log_level)) { /* keep iterating through all even when false is returned */ res = FALSE; } } set_bit(fsa_input_register, R_SENT_RSC_STOP); g_list_free(lrm_state_list); lrm_state_list = NULL; return res; } struct stop_recurring_action_s { lrmd_rsc_info_t *rsc; lrm_state_t *lrm_state; }; static gboolean stop_recurring_action_by_rsc(gpointer key, gpointer value, gpointer user_data) { gboolean remove = FALSE; struct stop_recurring_action_s *event = user_data; struct recurring_op_s *op = (struct recurring_op_s *)value; if ((op->interval_ms != 0) && crm_str_eq(op->rsc_id, event->rsc->id, TRUE)) { crm_debug("Cancelling op %d for %s (%s)", op->call_id, op->rsc_id, (char*)key); remove = !cancel_op(event->lrm_state, event->rsc->id, key, op->call_id, FALSE); } return remove; } static gboolean stop_recurring_actions(gpointer key, gpointer value, gpointer user_data) { gboolean remove = FALSE; lrm_state_t *lrm_state = user_data; struct recurring_op_s *op = (struct recurring_op_s *)value; if (op->interval_ms != 0) { crm_info("Cancelling op %d for %s (%s)", op->call_id, op->rsc_id, (const char *) key); remove = !cancel_op(lrm_state, op->rsc_id, key, op->call_id, FALSE); } return remove; } static void record_pending_op(const char *node_name, lrmd_rsc_info_t *rsc, lrmd_event_data_t *op) { const char *record_pending = NULL; CRM_CHECK(node_name != NULL, return); CRM_CHECK(rsc != NULL, return); CRM_CHECK(op != NULL, return); // Never record certain operation types as pending if ((op->op_type == NULL) || (op->params == NULL) || !controld_action_is_recordable(op->op_type)) { return; } // defaults to true record_pending = crm_meta_value(op->params, XML_OP_ATTR_PENDING); if (record_pending && !crm_is_true(record_pending)) { return; } op->call_id = -1; op->op_status = PCMK_LRM_OP_PENDING; op->rc = PCMK_OCF_UNKNOWN; op->t_run = time(NULL); op->t_rcchange = op->t_run; /* write a "pending" entry to the CIB, inhibit notification */ crm_debug("Recording pending op " CRM_OP_FMT " on %s in the CIB", op->rsc_id, op->op_type, op->interval_ms, node_name); do_update_resource(node_name, rsc, op); } static void do_lrm_rsc_op(lrm_state_t * lrm_state, lrmd_rsc_info_t * rsc, const char *operation, xmlNode * msg, xmlNode * request) { int call_id = 0; char *op_id = NULL; lrmd_event_data_t *op = NULL; lrmd_key_value_t *params = NULL; fsa_data_t *msg_data = NULL; const char *transition = NULL; gboolean stop_recurring = FALSE; bool send_nack = FALSE; CRM_CHECK(rsc != NULL, return); CRM_CHECK(operation != NULL, return); if (msg != NULL) { transition = crm_element_value(msg, XML_ATTR_TRANSITION_KEY); if (transition == NULL) { crm_log_xml_err(msg, "Missing transition number"); } } op = construct_op(lrm_state, msg, rsc->id, operation); CRM_CHECK(op != NULL, return); if (is_remote_lrmd_ra(NULL, NULL, rsc->id) && (op->interval_ms == 0) && strcmp(operation, CRMD_ACTION_MIGRATE) == 0) { /* pcmk remote connections are a special use case. * We never ever want to stop monitoring a connection resource until * the entire migration has completed. If the connection is unexpectedly * severed, even during a migration, this is an event we must detect.*/ stop_recurring = FALSE; } else if ((op->interval_ms == 0) && strcmp(operation, CRMD_ACTION_STATUS) != 0 && strcmp(operation, CRMD_ACTION_NOTIFY) != 0) { /* stop any previous monitor operations before changing the resource state */ stop_recurring = TRUE; } if (stop_recurring == TRUE) { guint removed = 0; struct stop_recurring_action_s data; data.rsc = rsc; data.lrm_state = lrm_state; removed = g_hash_table_foreach_remove( lrm_state->pending_ops, stop_recurring_action_by_rsc, &data); if (removed) { crm_debug("Stopped %u recurring operation%s in preparation for " CRM_OP_FMT, removed, s_if_plural(removed), rsc->id, operation, op->interval_ms); } } /* now do the op */ crm_info("Performing key=%s op=" CRM_OP_FMT, transition, rsc->id, operation, op->interval_ms); if (is_set(fsa_input_register, R_SHUTDOWN) && safe_str_eq(operation, RSC_START)) { register_fsa_input(C_SHUTDOWN, I_SHUTDOWN, NULL); send_nack = TRUE; } else if (fsa_state != S_NOT_DC && fsa_state != S_POLICY_ENGINE /* Recalculating */ && fsa_state != S_TRANSITION_ENGINE && safe_str_neq(operation, CRMD_ACTION_STOP)) { send_nack = TRUE; } if(send_nack) { crm_notice("Discarding attempt to perform action %s on %s in state %s (shutdown=%s)", operation, rsc->id, fsa_state2string(fsa_state), is_set(fsa_input_register, R_SHUTDOWN)?"true":"false"); op->rc = CRM_DIRECT_NACK_RC; op->op_status = PCMK_LRM_OP_ERROR; send_direct_ack(NULL, NULL, rsc, op, rsc->id); lrmd_free_event(op); free(op_id); return; } record_pending_op(lrm_state->node_name, rsc, op); op_id = generate_op_key(rsc->id, op->op_type, op->interval_ms); if (op->interval_ms > 0) { /* cancel it so we can then restart it without conflict */ cancel_op_key(lrm_state, rsc, op_id, FALSE); } if (op->params) { char *key = NULL; char *value = NULL; GHashTableIter iter; g_hash_table_iter_init(&iter, op->params); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { params = lrmd_key_value_add(params, key, value); } } call_id = lrm_state_exec(lrm_state, rsc->id, op->op_type, op->user_data, op->interval_ms, op->timeout, op->start_delay, params); if (call_id <= 0 && lrm_state_is_local(lrm_state)) { crm_err("Operation %s on %s failed: %d", operation, rsc->id, call_id); register_fsa_error(C_FSA_INTERNAL, I_FAIL, NULL); } else if (call_id <= 0) { crm_err("Operation %s on resource %s failed to execute on remote node %s: %d", operation, rsc->id, lrm_state->node_name, call_id); fake_op_status(lrm_state, op, PCMK_LRM_OP_DONE, PCMK_OCF_UNKNOWN_ERROR); process_lrm_event(lrm_state, op, NULL, NULL); } else { /* record all operations so we can wait * for them to complete during shutdown */ char *call_id_s = make_stop_id(rsc->id, call_id); struct recurring_op_s *pending = NULL; pending = calloc(1, sizeof(struct recurring_op_s)); crm_trace("Recording pending op: %d - %s %s", call_id, op_id, call_id_s); pending->call_id = call_id; pending->interval_ms = op->interval_ms; pending->op_type = strdup(operation); pending->op_key = strdup(op_id); pending->rsc_id = strdup(rsc->id); pending->start_time = time(NULL); pending->user_data = op->user_data? strdup(op->user_data) : NULL; g_hash_table_replace(lrm_state->pending_ops, call_id_s, pending); if ((op->interval_ms > 0) && (op->start_delay > START_DELAY_THRESHOLD)) { - - char *uuid = NULL; - int dummy = 0, target_rc = 0; + int target_rc = 0; crm_info("Faking confirmation of %s: execution postponed for over 5 minutes", op_id); - - decode_transition_key(op->user_data, &uuid, &dummy, &dummy, &target_rc); - free(uuid); - + decode_transition_key(op->user_data, NULL, NULL, NULL, &target_rc); op->rc = target_rc; op->op_status = PCMK_LRM_OP_DONE; send_direct_ack(NULL, NULL, rsc, op, rsc->id); } pending->params = op->params; op->params = NULL; } free(op_id); lrmd_free_event(op); return; } int last_resource_update = 0; static void cib_rsc_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { switch (rc) { case pcmk_ok: case -pcmk_err_diff_failed: case -pcmk_err_diff_resync: crm_trace("Resource update %d complete: rc=%d", call_id, rc); break; default: crm_warn("Resource update %d failed: (rc=%d) %s", call_id, rc, pcmk_strerror(rc)); } if (call_id == last_resource_update) { last_resource_update = 0; trigger_fsa(fsa_source); } } static int do_update_resource(const char *node_name, lrmd_rsc_info_t * rsc, lrmd_event_data_t * op) { /* */ int rc = pcmk_ok; xmlNode *update, *iter = NULL; int call_opt = crmd_cib_smart_opt(); const char *uuid = NULL; CRM_CHECK(op != NULL, return 0); iter = create_xml_node(iter, XML_CIB_TAG_STATUS); update = iter; iter = create_xml_node(iter, XML_CIB_TAG_STATE); if (safe_str_eq(node_name, fsa_our_uname)) { uuid = fsa_our_uuid; } else { /* remote nodes uuid and uname are equal */ uuid = node_name; crm_xml_add(iter, XML_NODE_IS_REMOTE, "true"); } CRM_LOG_ASSERT(uuid != NULL); if(uuid == NULL) { rc = -EINVAL; goto done; } crm_xml_add(iter, XML_ATTR_UUID, uuid); crm_xml_add(iter, XML_ATTR_UNAME, node_name); crm_xml_add(iter, XML_ATTR_ORIGIN, __FUNCTION__); iter = create_xml_node(iter, XML_CIB_TAG_LRM); crm_xml_add(iter, XML_ATTR_ID, uuid); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCES); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCE); crm_xml_add(iter, XML_ATTR_ID, op->rsc_id); build_operation_update(iter, rsc, op, node_name, __FUNCTION__); if (rsc) { const char *container = NULL; crm_xml_add(iter, XML_ATTR_TYPE, rsc->type); crm_xml_add(iter, XML_AGENT_ATTR_CLASS, rsc->standard); crm_xml_add(iter, XML_AGENT_ATTR_PROVIDER, rsc->provider); if (op->params) { container = g_hash_table_lookup(op->params, CRM_META"_"XML_RSC_ATTR_CONTAINER); } if (container) { crm_trace("Resource %s is a part of container resource %s", op->rsc_id, container); crm_xml_add(iter, XML_RSC_ATTR_CONTAINER, container); } } else { crm_warn("Resource %s no longer exists in the executor", op->rsc_id); send_direct_ack(NULL, NULL, rsc, op, op->rsc_id); goto cleanup; } crm_log_xml_trace(update, __FUNCTION__); /* make it an asynchronous call and be done with it * * Best case: * the resource state will be discovered during * the next signup or election. * * Bad case: * we are shutting down and there is no DC at the time, * but then why were we shutting down then anyway? * (probably because of an internal error) * * Worst case: * we get shot for having resources "running" that really weren't * * the alternative however means blocking here for too long, which * isn't acceptable */ fsa_cib_update(XML_CIB_TAG_STATUS, update, call_opt, rc, NULL); if (rc > 0) { last_resource_update = rc; } done: /* the return code is a call number, not an error code */ crm_trace("Sent resource state update message: %d for %s=%u on %s", rc, op->op_type, op->interval_ms, op->rsc_id); fsa_register_cib_callback(rc, FALSE, NULL, cib_rsc_callback); cleanup: free_xml(update); return rc; } void do_lrm_event(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input cur_input, fsa_data_t * msg_data) { CRM_CHECK(FALSE, return); } static char * unescape_newlines(const char *string) { char *pch = NULL; char *ret = NULL; static const char *escaped_newline = "\\n"; if (!string) { return NULL; } ret = strdup(string); pch = strstr(ret, escaped_newline); while (pch != NULL) { /* 2 chars for 2 chars, null-termination irrelevant */ memcpy(pch, "\n ", 2 * sizeof(char)); pch = strstr(pch, escaped_newline); } return ret; } void process_lrm_event(lrm_state_t *lrm_state, lrmd_event_data_t *op, struct recurring_op_s *pending, xmlNode *action_xml) { char *op_id = NULL; char *op_key = NULL; int update_id = 0; gboolean remove = FALSE; gboolean removed = FALSE; bool need_direct_ack = FALSE; lrmd_rsc_info_t *rsc = NULL; const char *node_name = NULL; CRM_CHECK(op != NULL, return); CRM_CHECK(op->rsc_id != NULL, return); op_id = make_stop_id(op->rsc_id, op->call_id); op_key = generate_op_key(op->rsc_id, op->op_type, op->interval_ms); // Get resource info if available (from executor state or action XML) if (lrm_state) { rsc = lrm_state_get_rsc_info(lrm_state, op->rsc_id, 0); } if ((rsc == NULL) && action_xml) { xmlNode *xml = find_xml_node(action_xml, XML_CIB_TAG_RESOURCE, TRUE); const char *standard = crm_element_value(xml, XML_AGENT_ATTR_CLASS); const char *provider = crm_element_value(xml, XML_AGENT_ATTR_PROVIDER); const char *type = crm_element_value(xml, XML_ATTR_TYPE); if (standard && type) { crm_info("%s agent information not cached, using %s%s%s:%s from action XML", op->rsc_id, standard, (provider? ":" : ""), (provider? provider : ""), type); rsc = lrmd_new_rsc_info(op->rsc_id, standard, provider, type); } else { crm_err("Can't process %s result because %s agent information not cached or in XML", op_key, op->rsc_id); } } // Get node name if available (from executor state or action XML) if (lrm_state) { node_name = lrm_state->node_name; } else if (action_xml) { node_name = crm_element_value(action_xml, XML_LRM_ATTR_TARGET); } if(pending == NULL) { remove = TRUE; if (lrm_state) { pending = g_hash_table_lookup(lrm_state->pending_ops, op_id); } } if (op->op_status == PCMK_LRM_OP_ERROR) { switch(op->rc) { case PCMK_OCF_NOT_RUNNING: case PCMK_OCF_RUNNING_MASTER: case PCMK_OCF_DEGRADED: case PCMK_OCF_DEGRADED_MASTER: // Leave it to the TE/scheduler to decide if this is an error op->op_status = PCMK_LRM_OP_DONE; break; default: /* Nothing to do */ break; } } if (op->op_status != PCMK_LRM_OP_CANCELLED) { /* We might not record the result, so directly acknowledge it to the * originator instead, so it doesn't time out waiting for the result * (especially important if part of a transition). */ need_direct_ack = TRUE; if (controld_action_is_recordable(op->op_type)) { if (node_name && rsc) { // We should record the result, and happily, we can update_id = do_update_resource(node_name, rsc, op); need_direct_ack = FALSE; } else if (op->rsc_deleted) { /* We shouldn't record the result (likely the resource was * refreshed, cleaned, or removed while this operation was * in flight). */ crm_notice("Not recording %s result in CIB because " "resource information was removed since it was initiated", op_key); } else { /* This shouldn't be possible; the executor didn't consider the * resource deleted, but we couldn't find resource or node * information. */ crm_err("Unable to record %s result in CIB: %s", op_key, (node_name? "No resource information" : "No node name")); } } } else if (op->interval_ms == 0) { /* A non-recurring operation was cancelled. Most likely, the * never-initiated action was removed from the executor's pending * operations list upon resource removal. */ need_direct_ack = TRUE; } else if (pending == NULL) { /* This recurring operation was cancelled, but was not pending. No * transition actions are waiting on it, nothing needs to be done. */ } else if (op->user_data == NULL) { /* This recurring operation was cancelled and pending, but we don't * have a transition key. This should never happen. */ crm_err("Recurring operation %s was cancelled without transition information", op_key); } else if (pending->remove) { /* This recurring operation was cancelled (by us) and pending, and we * have been waiting for it to finish. */ if (lrm_state) { erase_lrm_history_by_op(lrm_state, op); } } else if (op->rsc_deleted) { /* This recurring operation was cancelled (but not by us, and the * executor does not have resource information, likely due to resource * cleanup, refresh, or removal) and pending. */ crm_debug("Recurring op %s was cancelled due to resource deletion", op_key); need_direct_ack = TRUE; } else { /* This recurring operation was cancelled (but not by us, likely by the * executor before stopping the resource) and pending. We don't need to * do anything special. */ } if (need_direct_ack) { send_direct_ack(NULL, NULL, NULL, op, op->rsc_id); } if(remove == FALSE) { /* The caller will do this afterwards, but keep the logging consistent */ removed = TRUE; } else if (lrm_state && ((op->interval_ms == 0) || (op->op_status == PCMK_LRM_OP_CANCELLED))) { gboolean found = g_hash_table_remove(lrm_state->pending_ops, op_id); if (op->interval_ms != 0) { removed = TRUE; } else if (found) { removed = TRUE; crm_trace("Op %s (call=%d, stop-id=%s, remaining=%u): Confirmed", op_key, op->call_id, op_id, g_hash_table_size(lrm_state->pending_ops)); } } if (node_name == NULL) { node_name = "unknown node"; // for logging } switch (op->op_status) { case PCMK_LRM_OP_CANCELLED: crm_info("Result of %s operation for %s on %s: %s " CRM_XS " call=%d key=%s confirmed=%s", crm_action_str(op->op_type, op->interval_ms), op->rsc_id, node_name, services_lrm_status_str(op->op_status), op->call_id, op_key, (removed? "true" : "false")); break; case PCMK_LRM_OP_DONE: do_crm_log((op->interval_ms? LOG_INFO : LOG_NOTICE), "Result of %s operation for %s on %s: %d (%s) " CRM_XS " call=%d key=%s confirmed=%s cib-update=%d", crm_action_str(op->op_type, op->interval_ms), op->rsc_id, node_name, op->rc, services_ocf_exitcode_str(op->rc), op->call_id, op_key, (removed? "true" : "false"), update_id); break; case PCMK_LRM_OP_TIMEOUT: crm_err("Result of %s operation for %s on %s: %s " CRM_XS " call=%d key=%s timeout=%dms", crm_action_str(op->op_type, op->interval_ms), op->rsc_id, node_name, services_lrm_status_str(op->op_status), op->call_id, op_key, op->timeout); break; default: crm_err("Result of %s operation for %s on %s: %s " CRM_XS " call=%d key=%s confirmed=%s status=%d cib-update=%d", crm_action_str(op->op_type, op->interval_ms), op->rsc_id, node_name, services_lrm_status_str(op->op_status), op->call_id, op_key, (removed? "true" : "false"), op->op_status, update_id); } if (op->output) { char *prefix = crm_strdup_printf("%s-" CRM_OP_FMT ":%d", node_name, op->rsc_id, op->op_type, op->interval_ms, op->call_id); if (op->rc) { crm_log_output(LOG_NOTICE, prefix, op->output); } else { crm_log_output(LOG_DEBUG, prefix, op->output); } free(prefix); } if (lrm_state) { if (safe_str_neq(op->op_type, RSC_METADATA)) { crmd_alert_resource_op(lrm_state->node_name, op); } else if (rsc && (op->rc == PCMK_OCF_OK)) { char *metadata = unescape_newlines(op->output); metadata_cache_update(lrm_state->metadata_cache, rsc, metadata); free(metadata); } } if (op->rsc_deleted) { crm_info("Deletion of resource '%s' complete after %s", op->rsc_id, op_key); if (lrm_state) { delete_rsc_entry(lrm_state, NULL, op->rsc_id, NULL, pcmk_ok, NULL); } } /* If a shutdown was escalated while operations were pending, * then the FSA will be stalled right now... allow it to continue */ mainloop_set_trigger(fsa_source); if (lrm_state && rsc) { update_history_cache(lrm_state, rsc, op); } lrmd_free_rsc_info(rsc); free(op_key); free(op_id); } diff --git a/daemons/controld/controld_te_callbacks.c b/daemons/controld/controld_te_callbacks.c index 02ab103f4c..51d908e90b 100644 --- a/daemons/controld/controld_te_callbacks.c +++ b/daemons/controld/controld_te_callbacks.c @@ -1,962 +1,959 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include /* For ONLINESTATUS etc */ void te_update_confirm(const char *event, xmlNode * msg); extern char *te_uuid; gboolean shuttingdown = FALSE; crm_graph_t *transition_graph; crm_trigger_t *transition_trigger = NULL; static unsigned long int stonith_max_attempts = 10; /* #define RSC_OP_TEMPLATE "//"XML_TAG_DIFF_ADDED"//"XML_TAG_CIB"//"XML_CIB_TAG_STATE"[@uname='%s']"//"XML_LRM_TAG_RSC_OP"[@id='%s]" */ #define RSC_OP_TEMPLATE "//"XML_TAG_DIFF_ADDED"//"XML_TAG_CIB"//"XML_LRM_TAG_RSC_OP"[@id='%s']" static const char * get_node_id(xmlNode * rsc_op) { xmlNode *node = rsc_op; while (node != NULL && safe_str_neq(XML_CIB_TAG_STATE, TYPE(node))) { node = node->parent; } CRM_CHECK(node != NULL, return NULL); return ID(node); } void update_stonith_max_attempts(const char* value) { if (safe_str_eq(value, CRM_INFINITY_S)) { stonith_max_attempts = CRM_SCORE_INFINITY; } else { stonith_max_attempts = crm_int_helper(value, NULL); } } static void te_update_diff_v1(const char *event, xmlNode *diff) { int lpc, max; xmlXPathObject *xpathObj = NULL; CRM_CHECK(diff != NULL, return); xml_log_patchset(LOG_TRACE, __FUNCTION__, diff); if (cib_config_changed(NULL, NULL, &diff)) { abort_transition(INFINITY, tg_restart, "Non-status change", diff); goto bail; /* configuration changed */ } /* Tickets Attributes - Added/Updated */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_CIB_TAG_TICKETS); if (numXpathResults(xpathObj) > 0) { xmlNode *aborted = getXpathResult(xpathObj, 0); abort_transition(INFINITY, tg_restart, "Ticket attribute: update", aborted); goto bail; } freeXpathObject(xpathObj); /* Tickets Attributes - Removed */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_CIB_TAG_TICKETS); if (numXpathResults(xpathObj) > 0) { xmlNode *aborted = getXpathResult(xpathObj, 0); abort_transition(INFINITY, tg_restart, "Ticket attribute: removal", aborted); goto bail; } freeXpathObject(xpathObj); /* Transient Attributes - Added/Updated */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_TAG_TRANSIENT_NODEATTRS "//" XML_CIB_TAG_NVPAIR); max = numXpathResults(xpathObj); for (lpc = 0; lpc < max; lpc++) { xmlNode *attr = getXpathResult(xpathObj, lpc); const char *name = crm_element_value(attr, XML_NVPAIR_ATTR_NAME); const char *value = NULL; if (safe_str_eq(CRM_OP_PROBED, name)) { value = crm_element_value(attr, XML_NVPAIR_ATTR_VALUE); } if (crm_is_true(value) == FALSE) { abort_transition(INFINITY, tg_restart, "Transient attribute: update", attr); crm_log_xml_trace(attr, "Abort"); goto bail; } } freeXpathObject(xpathObj); /* Transient Attributes - Removed */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_TAG_TRANSIENT_NODEATTRS); if (numXpathResults(xpathObj) > 0) { xmlNode *aborted = getXpathResult(xpathObj, 0); abort_transition(INFINITY, tg_restart, "Transient attribute: removal", aborted); goto bail; } freeXpathObject(xpathObj); /* * Updates by, or in response to, TE actions will never contain updates * for more than one resource at a time, so such updates indicate an * LRM refresh. * * In that case, start a new transition rather than check each result * individually, which can result in _huge_ speedups in large clusters. * * Unfortunately, we can only do so when there are no pending actions. * Otherwise, we could mistakenly throw away those results here, and * the cluster will stall waiting for them and time out the operation. */ if (transition_graph->pending == 0) { xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_LRM_TAG_RESOURCE); max = numXpathResults(xpathObj); if (max > 1) { crm_debug("Ignoring resource operation updates due to history refresh of %d resources", max); crm_log_xml_trace(diff, "lrm-refresh"); abort_transition(INFINITY, tg_restart, "History refresh", NULL); goto bail; } freeXpathObject(xpathObj); } /* Process operation updates */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_LRM_TAG_RSC_OP); max = numXpathResults(xpathObj); if (max > 0) { int lpc = 0; for (lpc = 0; lpc < max; lpc++) { xmlNode *rsc_op = getXpathResult(xpathObj, lpc); const char *node = get_node_id(rsc_op); process_graph_event(rsc_op, node); } } freeXpathObject(xpathObj); /* Detect deleted (as opposed to replaced or added) actions - eg. crm_resource -C */ xpathObj = xpath_search(diff, "//" XML_TAG_DIFF_REMOVED "//" XML_LRM_TAG_RSC_OP); max = numXpathResults(xpathObj); for (lpc = 0; lpc < max; lpc++) { int path_max = 0; const char *op_id = NULL; char *rsc_op_xpath = NULL; xmlXPathObject *op_match = NULL; xmlNode *match = getXpathResult(xpathObj, lpc); CRM_LOG_ASSERT(match != NULL); if(match == NULL) { continue; }; op_id = ID(match); path_max = strlen(RSC_OP_TEMPLATE) + strlen(op_id) + 1; rsc_op_xpath = calloc(1, path_max); snprintf(rsc_op_xpath, path_max, RSC_OP_TEMPLATE, op_id); op_match = xpath_search(diff, rsc_op_xpath); if (numXpathResults(op_match) == 0) { /* Prevent false positives by matching cancelations too */ const char *node = get_node_id(match); crm_action_t *cancelled = get_cancel_action(op_id, node); if (cancelled == NULL) { crm_debug("No match for deleted action %s (%s on %s)", rsc_op_xpath, op_id, node); abort_transition(INFINITY, tg_restart, "Resource op removal", match); freeXpathObject(op_match); free(rsc_op_xpath); goto bail; } else { crm_debug("Deleted lrm_rsc_op %s on %s was for graph event %d", op_id, node, cancelled->id); } } freeXpathObject(op_match); free(rsc_op_xpath); } bail: freeXpathObject(xpathObj); } static void process_lrm_resource_diff(xmlNode *lrm_resource, const char *node) { for (xmlNode *rsc_op = __xml_first_child(lrm_resource); rsc_op != NULL; rsc_op = __xml_next(rsc_op)) { process_graph_event(rsc_op, node); } } static void process_resource_updates(const char *node, xmlNode *xml, xmlNode *change, const char *op, const char *xpath) { xmlNode *rsc = NULL; if (xml == NULL) { return; } else if (strcmp((const char*)xml->name, XML_CIB_TAG_LRM) == 0) { xml = first_named_child(xml, XML_LRM_TAG_RESOURCES); crm_trace("Got %p in %s", xml, XML_CIB_TAG_LRM); } CRM_ASSERT(strcmp((const char*)xml->name, XML_LRM_TAG_RESOURCES) == 0); /* * Updates by, or in response to, TE actions will never contain updates * for more than one resource at a time, so such updates indicate an * LRM refresh. * * In that case, start a new transition rather than check each result * individually, which can result in _huge_ speedups in large clusters. * * Unfortunately, we can only do so when there are no pending actions. * Otherwise, we could mistakenly throw away those results here, and * the cluster will stall waiting for them and time out the operation. */ if ((transition_graph->pending == 0) && xml->children && xml->children->next) { crm_log_xml_trace(change, "lrm-refresh"); abort_transition(INFINITY, tg_restart, "History refresh", NULL); return; } for (rsc = __xml_first_child(xml); rsc != NULL; rsc = __xml_next(rsc)) { crm_trace("Processing %s", ID(rsc)); process_lrm_resource_diff(rsc, node); } } #define NODE_PATT "/lrm[@id=" static char *get_node_from_xpath(const char *xpath) { char *nodeid = NULL; char *tmp = strstr(xpath, NODE_PATT); if(tmp) { tmp += strlen(NODE_PATT); tmp += 1; nodeid = strdup(tmp); tmp = strstr(nodeid, "\'"); CRM_ASSERT(tmp); tmp[0] = 0; } return nodeid; } static char *extract_node_uuid(const char *xpath) { char *mutable_path = strdup(xpath); char *node_uuid = NULL; char *search = NULL; char *match = NULL; match = strstr(mutable_path, "node_state[@id=\'"); if (match == NULL) { free(mutable_path); return NULL; } match += strlen("node_state[@id=\'"); search = strchr(match, '\''); if (search == NULL) { free(mutable_path); return NULL; } search[0] = 0; node_uuid = strdup(match); free(mutable_path); return node_uuid; } static void abort_unless_down(const char *xpath, const char *op, xmlNode *change, const char *reason) { char *node_uuid = NULL; crm_action_t *down = NULL; if(safe_str_neq(op, "delete")) { abort_transition(INFINITY, tg_restart, reason, change); return; } node_uuid = extract_node_uuid(xpath); if(node_uuid == NULL) { crm_err("Could not extract node ID from %s", xpath); abort_transition(INFINITY, tg_restart, reason, change); return; } down = match_down_event(node_uuid); if (down == NULL) { crm_trace("Not expecting %s to be down (%s)", node_uuid, xpath); abort_transition(INFINITY, tg_restart, reason, change); } else { crm_trace("Expecting changes to %s (%s)", node_uuid, xpath); } free(node_uuid); } static void process_op_deletion(const char *xpath, xmlNode *change) { char *mutable_key = strdup(xpath); char *key; char *node_uuid; crm_action_t *cancel = NULL; // Extract the part of xpath between last pair of single quotes key = strrchr(mutable_key, '\''); if (key != NULL) { *key = '\0'; key = strrchr(mutable_key, '\''); } if (key == NULL) { crm_warn("Ignoring malformed CIB update (resource deletion of %s)", xpath); free(mutable_key); return; } ++key; node_uuid = extract_node_uuid(xpath); cancel = get_cancel_action(key, node_uuid); if (cancel) { crm_info("Cancellation of %s on %s confirmed (%d)", key, node_uuid, cancel->id); stop_te_timer(cancel->timer); te_action_confirmed(cancel); update_graph(transition_graph, cancel); trigger_graph(); } else { abort_transition(INFINITY, tg_restart, "Resource operation removal", change); } free(mutable_key); free(node_uuid); } static void process_delete_diff(const char *xpath, const char *op, xmlNode *change) { if (strstr(xpath, "/" XML_LRM_TAG_RSC_OP "[")) { process_op_deletion(xpath, change); } else if (strstr(xpath, "/" XML_CIB_TAG_LRM "[")) { abort_unless_down(xpath, op, change, "Resource state removal"); } else if (strstr(xpath, "/" XML_CIB_TAG_STATE "[")) { abort_unless_down(xpath, op, change, "Node state removal"); } else { crm_trace("Ignoring delete of %s", xpath); } } static void process_node_state_diff(xmlNode *state, xmlNode *change, const char *op, const char *xpath) { xmlNode *lrm = first_named_child(state, XML_CIB_TAG_LRM); process_resource_updates(ID(state), lrm, change, op, xpath); } static void process_status_diff(xmlNode *status, xmlNode *change, const char *op, const char *xpath) { for (xmlNode *state = __xml_first_child(status); state != NULL; state = __xml_next(state)) { process_node_state_diff(state, change, op, xpath); } } static void process_cib_diff(xmlNode *cib, xmlNode *change, const char *op, const char *xpath) { xmlNode *status = first_named_child(cib, XML_CIB_TAG_STATUS); xmlNode *config = first_named_child(cib, XML_CIB_TAG_CONFIGURATION); if (status) { process_status_diff(status, change, op, xpath); } if (config) { abort_transition(INFINITY, tg_restart, "Non-status-only change", change); } } static void te_update_diff_v2(xmlNode *diff) { crm_log_xml_trace(diff, "Patch:Raw"); for (xmlNode *change = __xml_first_child(diff); change != NULL; change = __xml_next(change)) { xmlNode *match = NULL; const char *name = NULL; const char *xpath = crm_element_value(change, XML_DIFF_PATH); // Possible ops: create, modify, delete, move const char *op = crm_element_value(change, XML_DIFF_OP); // Ignore uninteresting updates if (op == NULL) { continue; } else if (xpath == NULL) { crm_trace("Ignoring %s change for version field", op); continue; } else if (strcmp(op, "move") == 0) { crm_trace("Ignoring move change at %s", xpath); continue; } // Find the result of create/modify ops if (strcmp(op, "create") == 0) { match = change->children; } else if (strcmp(op, "modify") == 0) { match = first_named_child(change, XML_DIFF_RESULT); if(match) { match = match->children; } } else if (strcmp(op, "delete") != 0) { crm_warn("Ignoring malformed CIB update (%s operation on %s is unrecognized)", op, xpath); continue; } if (match) { if (match->type == XML_COMMENT_NODE) { crm_trace("Ignoring %s operation for comment at %s", op, xpath); continue; } name = (const char *)match->name; } crm_trace("Handling %s operation for %s%s%s", op, (xpath? xpath : "CIB"), (name? " matched by " : ""), (name? name : "")); if (strstr(xpath, "/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION)) { abort_transition(INFINITY, tg_restart, "Configuration change", change); break; // Won't be packaged with operation results we may be waiting for } else if (strstr(xpath, "/" XML_CIB_TAG_TICKETS) || safe_str_eq(name, XML_CIB_TAG_TICKETS)) { abort_transition(INFINITY, tg_restart, "Ticket attribute change", change); break; // Won't be packaged with operation results we may be waiting for } else if (strstr(xpath, "/" XML_TAG_TRANSIENT_NODEATTRS "[") || safe_str_eq(name, XML_TAG_TRANSIENT_NODEATTRS)) { abort_unless_down(xpath, op, change, "Transient attribute change"); break; // Won't be packaged with operation results we may be waiting for } else if (strcmp(op, "delete") == 0) { process_delete_diff(xpath, op, change); } else if (name == NULL) { crm_warn("Ignoring malformed CIB update (%s at %s has no result)", op, xpath); } else if (strcmp(name, XML_TAG_CIB) == 0) { process_cib_diff(match, change, op, xpath); } else if (strcmp(name, XML_CIB_TAG_STATUS) == 0) { process_status_diff(match, change, op, xpath); } else if (strcmp(name, XML_CIB_TAG_STATE) == 0) { process_node_state_diff(match, change, op, xpath); } else if (strcmp(name, XML_CIB_TAG_LRM) == 0) { process_resource_updates(ID(match), match, change, op, xpath); } else if (strcmp(name, XML_LRM_TAG_RESOURCES) == 0) { char *local_node = get_node_from_xpath(xpath); process_resource_updates(local_node, match, change, op, xpath); free(local_node); } else if (strcmp(name, XML_LRM_TAG_RESOURCE) == 0) { char *local_node = get_node_from_xpath(xpath); process_lrm_resource_diff(match, local_node); free(local_node); } else if (strcmp(name, XML_LRM_TAG_RSC_OP) == 0) { char *local_node = get_node_from_xpath(xpath); process_graph_event(match, local_node); free(local_node); } else { crm_warn("Ignoring malformed CIB update (%s at %s has unrecognized result %s)", op, xpath, name); } } } void te_update_diff(const char *event, xmlNode * msg) { xmlNode *diff = NULL; const char *op = NULL; int rc = -EINVAL; int format = 1; int p_add[] = { 0, 0, 0 }; int p_del[] = { 0, 0, 0 }; CRM_CHECK(msg != NULL, return); crm_element_value_int(msg, F_CIB_RC, &rc); if (transition_graph == NULL) { crm_trace("No graph"); return; } else if (rc < pcmk_ok) { crm_trace("Filter rc=%d (%s)", rc, pcmk_strerror(rc)); return; } else if (transition_graph->complete && fsa_state != S_IDLE && fsa_state != S_TRANSITION_ENGINE && fsa_state != S_POLICY_ENGINE) { crm_trace("Filter state=%s, complete=%d", fsa_state2string(fsa_state), transition_graph->complete); return; } op = crm_element_value(msg, F_CIB_OPERATION); diff = get_message_xml(msg, F_CIB_UPDATE_RESULT); xml_patch_versions(diff, p_add, p_del); crm_debug("Processing (%s) diff: %d.%d.%d -> %d.%d.%d (%s)", op, p_del[0], p_del[1], p_del[2], p_add[0], p_add[1], p_add[2], fsa_state2string(fsa_state)); crm_element_value_int(diff, "format", &format); switch (format) { case 1: te_update_diff_v1(event, diff); break; case 2: te_update_diff_v2(diff); break; default: crm_warn("Ignoring malformed CIB update (unknown patch format %d)", format); } } gboolean process_te_message(xmlNode * msg, xmlNode * xml_data) { const char *from = crm_element_value(msg, F_ORIG); const char *sys_to = crm_element_value(msg, F_CRM_SYS_TO); const char *sys_from = crm_element_value(msg, F_CRM_SYS_FROM); const char *ref = crm_element_value(msg, F_CRM_REFERENCE); const char *op = crm_element_value(msg, F_CRM_TASK); const char *type = crm_element_value(msg, F_CRM_MSG_TYPE); crm_trace("Processing %s (%s) message", op, ref); crm_log_xml_trace(msg, "ipc"); if (op == NULL) { /* error */ } else if (sys_to == NULL || strcasecmp(sys_to, CRM_SYSTEM_TENGINE) != 0) { crm_trace("Bad sys-to %s", crm_str(sys_to)); return FALSE; } else if (safe_str_eq(op, CRM_OP_INVOKE_LRM) && safe_str_eq(sys_from, CRM_SYSTEM_LRMD) /* && safe_str_eq(type, XML_ATTR_RESPONSE) */ ) { xmlXPathObject *xpathObj = NULL; crm_log_xml_trace(msg, "Processing (N)ACK"); crm_debug("Processing (N)ACK %s from %s", crm_element_value(msg, F_CRM_REFERENCE), from); xpathObj = xpath_search(xml_data, "//" XML_LRM_TAG_RSC_OP); if (numXpathResults(xpathObj)) { int lpc = 0, max = numXpathResults(xpathObj); for (lpc = 0; lpc < max; lpc++) { xmlNode *rsc_op = getXpathResult(xpathObj, lpc); const char *node = get_node_id(rsc_op); process_graph_event(rsc_op, node); } freeXpathObject(xpathObj); } else { crm_log_xml_err(msg, "Invalid (N)ACK"); freeXpathObject(xpathObj); return FALSE; } } else { crm_err("Unknown command: %s::%s from %s", type, op, sys_from); } crm_trace("finished processing message"); return TRUE; } GHashTable *stonith_failures = NULL; struct st_fail_rec { int count; }; static gboolean too_many_st_failures(const char *target) { GHashTableIter iter; const char *key = NULL; struct st_fail_rec *value = NULL; if (stonith_failures == NULL) { return FALSE; } if (target == NULL) { g_hash_table_iter_init(&iter, stonith_failures); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { if (value->count >= stonith_max_attempts) { target = (const char*)key; goto too_many; } } } else { value = g_hash_table_lookup(stonith_failures, target); if ((value != NULL) && (value->count >= stonith_max_attempts)) { goto too_many; } } return FALSE; too_many: crm_warn("Too many failures (%d) to fence %s, giving up", value->count, target); return TRUE; } /*! * \internal * \brief Reset a stonith fail count * * \param[in] target Name of node to reset, or NULL for all */ void st_fail_count_reset(const char *target) { if (stonith_failures == NULL) { return; } if (target) { struct st_fail_rec *rec = NULL; rec = g_hash_table_lookup(stonith_failures, target); if (rec) { rec->count = 0; } } else { GHashTableIter iter; const char *key = NULL; struct st_fail_rec *rec = NULL; g_hash_table_iter_init(&iter, stonith_failures); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &rec)) { rec->count = 0; } } } void st_fail_count_increment(const char *target) { struct st_fail_rec *rec = NULL; if (stonith_failures == NULL) { stonith_failures = crm_str_table_new(); } rec = g_hash_table_lookup(stonith_failures, target); if (rec) { rec->count++; } else { rec = malloc(sizeof(struct st_fail_rec)); if(rec == NULL) { return; } rec->count = 1; g_hash_table_insert(stonith_failures, strdup(target), rec); } } /*! * \internal * \brief Abort transition due to stonith failure * * \param[in] abort_action Whether to restart or stop transition * \param[in] target Don't restart if this (NULL for any) has too many failures * \param[in] reason Log this stonith action XML as abort reason (or NULL) */ void abort_for_stonith_failure(enum transition_action abort_action, const char *target, xmlNode *reason) { /* If stonith repeatedly fails, we eventually give up on starting a new * transition for that reason. */ if ((abort_action != tg_stop) && too_many_st_failures(target)) { abort_action = tg_stop; } abort_transition(INFINITY, abort_action, "Stonith failed", reason); } void tengine_stonith_callback(stonith_t * stonith, stonith_callback_data_t * data) { char *uuid = NULL; - int target_rc = -1; int stonith_id = -1; int transition_id = -1; crm_action_t *action = NULL; int call_id = data->call_id; int rc = data->rc; char *userdata = data->userdata; CRM_CHECK(userdata != NULL, return); crm_notice("Stonith operation %d/%s: %s (%d)", call_id, (char *)userdata, pcmk_strerror(rc), rc); if (AM_I_DC == FALSE) { return; } /* crm_info("call=%d, optype=%d, node_name=%s, result=%d, node_list=%s, action=%s", */ /* op->call_id, op->optype, op->node_name, op->op_result, */ /* (char *)op->node_list, op->private_data); */ /* filter out old STONITH actions */ - CRM_CHECK(decode_transition_key(userdata, &uuid, &transition_id, &stonith_id, &target_rc), - crm_err("Invalid event detected"); - goto bail; - ); + CRM_CHECK(decode_transition_key(userdata, &uuid, &transition_id, &stonith_id, NULL), + goto bail); if (transition_graph->complete || stonith_id < 0 || safe_str_neq(uuid, te_uuid) || transition_graph->id != transition_id) { crm_info("Ignoring STONITH action initiated outside of the current transition"); goto bail; } action = get_action(stonith_id, FALSE); if (action == NULL) { crm_err("Stonith action not matched"); goto bail; } stop_te_timer(action->timer); if (rc == pcmk_ok) { const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); const char *uuid = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID); const char *op = crm_meta_value(action->params, "stonith_action"); crm_info("Stonith operation %d for %s passed", call_id, target); if (action->confirmed == FALSE) { te_action_confirmed(action); if (safe_str_eq("on", op)) { const char *value = NULL; char *now = crm_itoa(time(NULL)); update_attrd(target, CRM_ATTR_UNFENCED, now, NULL, FALSE); free(now); value = crm_meta_value(action->params, XML_OP_ATTR_DIGESTS_ALL); update_attrd(target, CRM_ATTR_DIGESTS_ALL, value, NULL, FALSE); value = crm_meta_value(action->params, XML_OP_ATTR_DIGESTS_SECURE); update_attrd(target, CRM_ATTR_DIGESTS_SECURE, value, NULL, FALSE); } else if (action->sent_update == FALSE) { send_stonith_update(action, target, uuid); action->sent_update = TRUE; } } st_fail_count_reset(target); } else { const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); enum transition_action abort_action = tg_restart; action->failed = TRUE; crm_notice("Stonith operation %d for %s failed (%s): aborting transition.", call_id, target, pcmk_strerror(rc)); /* If no fence devices were available, there's no use in immediately * checking again, so don't start a new transition in that case. */ if (rc == -ENODEV) { crm_warn("No devices found in cluster to fence %s, giving up", target); abort_action = tg_stop; } /* Increment the fail count now, so abort_for_stonith_failure() can * check it. Non-DC nodes will increment it in tengine_stonith_notify(). */ st_fail_count_increment(target); abort_for_stonith_failure(abort_action, target, NULL); } update_graph(transition_graph, action); trigger_graph(); bail: free(userdata); free(uuid); return; } void cib_fencing_updated(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { if (rc < pcmk_ok) { crm_err("Fencing update %d for %s: failed - %s (%d)", call_id, (char *)user_data, pcmk_strerror(rc), rc); crm_log_xml_warn(msg, "Failed update"); abort_transition(INFINITY, tg_shutdown, "CIB update failed", NULL); } else { crm_info("Fencing update %d for %s: complete", call_id, (char *)user_data); } } void cib_action_updated(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { if (rc < pcmk_ok) { crm_err("Update %d FAILED: %s", call_id, pcmk_strerror(rc)); } } /*! * \brief Handle a timeout in node-to-node communication * * \param[in] data Pointer to action timer * * \return FALSE (indicating that source should be not be re-added) */ gboolean action_timer_callback(gpointer data) { crm_action_timer_t *timer = NULL; const char *task = NULL; const char *on_node = NULL; const char *via_node = NULL; CRM_CHECK(data != NULL, return FALSE); timer = (crm_action_timer_t *) data; stop_te_timer(timer); CRM_CHECK(timer->action != NULL, return FALSE); task = crm_element_value(timer->action->xml, XML_LRM_ATTR_TASK); on_node = crm_element_value(timer->action->xml, XML_LRM_ATTR_TARGET); via_node = crm_element_value(timer->action->xml, XML_LRM_ATTR_ROUTER_NODE); if (transition_graph->complete) { crm_notice("Node %s did not send %s result (via %s) within %dms " "(ignoring because transition not in progress)", (on_node? on_node : ""), (task? task : "unknown action"), (via_node? via_node : "controller"), timer->timeout); } else { /* fail the action */ crm_err("Node %s did not send %s result (via %s) within %dms " "(action timeout plus cluster-delay)", (on_node? on_node : ""), (task? task : "unknown action"), (via_node? via_node : "controller"), timer->timeout); print_action(LOG_ERR, "Aborting transition, action lost: ", timer->action); timer->action->failed = TRUE; te_action_confirmed(timer->action); abort_transition(INFINITY, tg_restart, "Action lost", NULL); update_graph(transition_graph, timer->action); trigger_graph(); // Record timeout in the CIB if appropriate if ((timer->action->type == action_type_rsc) && controld_action_is_recordable(task)) { controld_record_action_timeout(timer->action); } } return FALSE; } diff --git a/daemons/execd/remoted_proxy.c b/daemons/execd/remoted_proxy.c index 8ec8425821..f202b00c13 100644 --- a/daemons/execd/remoted_proxy.c +++ b/daemons/execd/remoted_proxy.c @@ -1,440 +1,455 @@ /* - * Copyright 2012-2018 David Vossel + * Copyright 2012-2019 the Pacemaker project contributors + * + * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include "pacemaker-execd.h" #include #include #include #include #include #include #include #include static qb_ipcs_service_t *cib_ro = NULL; static qb_ipcs_service_t *cib_rw = NULL; static qb_ipcs_service_t *cib_shm = NULL; static qb_ipcs_service_t *attrd_ipcs = NULL; static qb_ipcs_service_t *crmd_ipcs = NULL; static qb_ipcs_service_t *stonith_ipcs = NULL; // An IPC provider is a cluster node controller connecting as a client static GList *ipc_providers = NULL; /* ipc clients == things like cibadmin, crm_resource, connecting locally */ static GHashTable *ipc_clients = NULL; /*! * \internal * \brief Get an IPC proxy provider * * \return Pointer to a provider if one exists, NULL otherwise * * \note Grab the first provider, which is the most recent connection. That way, * if we haven't yet timed out an old, failed connection, we don't try to * use it. */ crm_client_t * ipc_proxy_get_provider() { return ipc_providers? (crm_client_t*) (ipc_providers->data) : NULL; } +/*! + * \internal + * \brief Accept a client connection on a proxy IPC server + * + * \param[in] c Client's IPC connection + * \param[in] uid Client's user ID + * \param[in] gid Client's group ID + * \param[in] ipc_channel Name of IPC server to proxy + * + * \return pcmk_ok on success, -errno on error + */ static int32_t ipc_proxy_accept(qb_ipcs_connection_t * c, uid_t uid, gid_t gid, const char *ipc_channel) { crm_client_t *client; crm_client_t *ipc_proxy = ipc_proxy_get_provider(); xmlNode *msg; - crm_trace("Connection %p on channel %s", c, ipc_channel); - if (ipc_proxy == NULL) { - crm_err("No ipc providers available for uid %d gid %d", uid, gid); + crm_warn("Cannot proxy request from uid %d gid %d " + "because not connected to cluster", uid, gid); return -EREMOTEIO; } - /* this new client is a local ipc client on a remote - * guest wanting to access the ipc on any available cluster nodes */ + /* This new client is a local IPC client on a Pacemaker Remote controlled + * node, needing to access cluster node IPC services. + */ client = crm_client_new(c, uid, gid); if (client == NULL) { return -EREMOTEIO; } /* This ipc client is bound to a single ipc provider. If the * provider goes away, this client is disconnected */ client->userdata = strdup(ipc_proxy->id); client->name = crm_strdup_printf("proxy-%s-%d-%.8s", ipc_channel, client->pid, client->id); g_hash_table_insert(ipc_clients, client->id, client); msg = create_xml_node(NULL, T_LRMD_IPC_PROXY); crm_xml_add(msg, F_LRMD_IPC_OP, LRMD_IPC_OP_NEW); crm_xml_add(msg, F_LRMD_IPC_IPC_SERVER, ipc_channel); crm_xml_add(msg, F_LRMD_IPC_SESSION, client->id); lrmd_server_send_notify(ipc_proxy, msg); free_xml(msg); - crm_debug("created new ipc proxy with session id %s", client->id); + crm_debug("Accepted IPC proxy connection (session ID %s) " + "from uid %d gid %d on channel %s", + client->id, uid, gid, ipc_channel); return 0; } static int32_t crmd_proxy_accept(qb_ipcs_connection_t * c, uid_t uid, gid_t gid) { return ipc_proxy_accept(c, uid, gid, CRM_SYSTEM_CRMD); } static int32_t attrd_proxy_accept(qb_ipcs_connection_t * c, uid_t uid, gid_t gid) { return ipc_proxy_accept(c, uid, gid, T_ATTRD); } static int32_t stonith_proxy_accept(qb_ipcs_connection_t * c, uid_t uid, gid_t gid) { return ipc_proxy_accept(c, uid, gid, "stonith-ng"); } static int32_t cib_proxy_accept_rw(qb_ipcs_connection_t * c, uid_t uid, gid_t gid) { return ipc_proxy_accept(c, uid, gid, CIB_CHANNEL_RW); } static int32_t cib_proxy_accept_ro(qb_ipcs_connection_t * c, uid_t uid, gid_t gid) { return ipc_proxy_accept(c, uid, gid, CIB_CHANNEL_RO); } static void ipc_proxy_created(qb_ipcs_connection_t * c) { crm_trace("Connection %p", c); } void ipc_proxy_forward_client(crm_client_t *ipc_proxy, xmlNode *xml) { const char *session = crm_element_value(xml, F_LRMD_IPC_SESSION); const char *msg_type = crm_element_value(xml, F_LRMD_IPC_OP); xmlNode *msg = get_message_xml(xml, F_LRMD_IPC_MSG); crm_client_t *ipc_client; int rc = 0; /* If the IPC provider is acknowledging our shutdown request, * defuse the short exit timer to give the cluster time to * stop any resources we're running. */ if (safe_str_eq(msg_type, LRMD_IPC_OP_SHUTDOWN_ACK)) { handle_shutdown_ack(); return; } if (safe_str_eq(msg_type, LRMD_IPC_OP_SHUTDOWN_NACK)) { handle_shutdown_nack(); return; } ipc_client = crm_client_get_by_id(session); if (ipc_client == NULL) { xmlNode *msg = create_xml_node(NULL, T_LRMD_IPC_PROXY); crm_xml_add(msg, F_LRMD_IPC_OP, LRMD_IPC_OP_DESTROY); crm_xml_add(msg, F_LRMD_IPC_SESSION, session); lrmd_server_send_notify(ipc_proxy, msg); free_xml(msg); return; } /* This is an event or response from the ipc provider * going to the local ipc client. * * Looking at the chain of events. * * -----remote node----------------|---- cluster node ------ * ipc_client <--1--> this code * <--2--> pacemaker-controld:remote_proxy_cb/remote_proxy_relay_event() * <--3--> ipc server * * This function is receiving a msg from connection 2 * and forwarding it to connection 1. */ if (safe_str_eq(msg_type, LRMD_IPC_OP_EVENT)) { crm_trace("Sending event to %s", ipc_client->id); rc = crm_ipcs_send(ipc_client, 0, msg, crm_ipc_server_event); } else if (safe_str_eq(msg_type, LRMD_IPC_OP_RESPONSE)) { int msg_id = 0; crm_element_value_int(xml, F_LRMD_IPC_MSG_ID, &msg_id); crm_trace("Sending response to %d - %s", ipc_client->request_id, ipc_client->id); rc = crm_ipcs_send(ipc_client, msg_id, msg, FALSE); CRM_LOG_ASSERT(msg_id == ipc_client->request_id); ipc_client->request_id = 0; } else if (safe_str_eq(msg_type, LRMD_IPC_OP_DESTROY)) { qb_ipcs_disconnect(ipc_client->ipcs); } else { crm_err("Unknown ipc proxy msg type %s" , msg_type); } if (rc < 0) { crm_warn("IPC Proxy send to ipc client %s failed, rc = %d", ipc_client->id, rc); } } static int32_t ipc_proxy_dispatch(qb_ipcs_connection_t * c, void *data, size_t size) { uint32_t id = 0; uint32_t flags = 0; crm_client_t *client = crm_client_get(c); crm_client_t *ipc_proxy = crm_client_get_by_id(client->userdata); xmlNode *request = NULL; xmlNode *msg = NULL; if (!ipc_proxy) { qb_ipcs_disconnect(client->ipcs); return 0; } /* This is a request from the local ipc client going * to the ipc provider. * * Looking at the chain of events. * * -----remote node----------------|---- cluster node ------ * ipc_client <--1--> this code * <--2--> pacemaker-controld:remote_proxy_dispatch_internal() * <--3--> ipc server * * This function is receiving a request from connection * 1 and forwarding it to connection 2. */ request = crm_ipcs_recv(client, data, size, &id, &flags); if (!request) { return 0; } CRM_CHECK(client != NULL, crm_err("Invalid client"); free_xml(request); return FALSE); CRM_CHECK(client->id != NULL, crm_err("Invalid client: %p", client); free_xml(request); return FALSE); /* This ensures that synced request/responses happen over the event channel * in the controller, allowing the controller to process the messages async. */ set_bit(flags, crm_ipc_proxied); client->request_id = id; msg = create_xml_node(NULL, T_LRMD_IPC_PROXY); crm_xml_add(msg, F_LRMD_IPC_OP, LRMD_IPC_OP_REQUEST); crm_xml_add(msg, F_LRMD_IPC_SESSION, client->id); crm_xml_add(msg, F_LRMD_IPC_CLIENT, crm_client_name(client)); crm_xml_add(msg, F_LRMD_IPC_USER, client->user); crm_xml_add_int(msg, F_LRMD_IPC_MSG_ID, id); crm_xml_add_int(msg, F_LRMD_IPC_MSG_FLAGS, flags); add_message_xml(msg, F_LRMD_IPC_MSG, request); lrmd_server_send_notify(ipc_proxy, msg); free_xml(request); free_xml(msg); return 0; } /*! * \internal * \brief Notify a proxy provider that we wish to shut down * * \return 0 on success, -1 on error */ int ipc_proxy_shutdown_req(crm_client_t *ipc_proxy) { xmlNode *msg = create_xml_node(NULL, T_LRMD_IPC_PROXY); int rc; crm_xml_add(msg, F_LRMD_IPC_OP, LRMD_IPC_OP_SHUTDOWN_REQ); /* We don't really have a session, but the controller needs this attribute * to recognize this as proxy communication. */ crm_xml_add(msg, F_LRMD_IPC_SESSION, "0"); rc = (lrmd_server_send_notify(ipc_proxy, msg) < 0)? -1 : 0; free_xml(msg); return rc; } static int32_t ipc_proxy_closed(qb_ipcs_connection_t * c) { crm_client_t *client = crm_client_get(c); crm_client_t *ipc_proxy; if (client == NULL) { return 0; } ipc_proxy = crm_client_get_by_id(client->userdata); crm_trace("Connection %p", c); if (ipc_proxy) { xmlNode *msg = create_xml_node(NULL, T_LRMD_IPC_PROXY); crm_xml_add(msg, F_LRMD_IPC_OP, LRMD_IPC_OP_DESTROY); crm_xml_add(msg, F_LRMD_IPC_SESSION, client->id); lrmd_server_send_notify(ipc_proxy, msg); free_xml(msg); } g_hash_table_remove(ipc_clients, client->id); free(client->userdata); client->userdata = NULL; crm_client_destroy(client); return 0; } static void ipc_proxy_destroy(qb_ipcs_connection_t * c) { crm_trace("Connection %p", c); ipc_proxy_closed(c); } static struct qb_ipcs_service_handlers crmd_proxy_callbacks = { .connection_accept = crmd_proxy_accept, .connection_created = ipc_proxy_created, .msg_process = ipc_proxy_dispatch, .connection_closed = ipc_proxy_closed, .connection_destroyed = ipc_proxy_destroy }; static struct qb_ipcs_service_handlers attrd_proxy_callbacks = { .connection_accept = attrd_proxy_accept, .connection_created = ipc_proxy_created, .msg_process = ipc_proxy_dispatch, .connection_closed = ipc_proxy_closed, .connection_destroyed = ipc_proxy_destroy }; static struct qb_ipcs_service_handlers stonith_proxy_callbacks = { .connection_accept = stonith_proxy_accept, .connection_created = ipc_proxy_created, .msg_process = ipc_proxy_dispatch, .connection_closed = ipc_proxy_closed, .connection_destroyed = ipc_proxy_destroy }; static struct qb_ipcs_service_handlers cib_proxy_callbacks_ro = { .connection_accept = cib_proxy_accept_ro, .connection_created = ipc_proxy_created, .msg_process = ipc_proxy_dispatch, .connection_closed = ipc_proxy_closed, .connection_destroyed = ipc_proxy_destroy }; static struct qb_ipcs_service_handlers cib_proxy_callbacks_rw = { .connection_accept = cib_proxy_accept_rw, .connection_created = ipc_proxy_created, .msg_process = ipc_proxy_dispatch, .connection_closed = ipc_proxy_closed, .connection_destroyed = ipc_proxy_destroy }; void ipc_proxy_add_provider(crm_client_t *ipc_proxy) { // Prepending ensures the most recent connection is always first ipc_providers = g_list_prepend(ipc_providers, ipc_proxy); } void ipc_proxy_remove_provider(crm_client_t *ipc_proxy) { GHashTableIter iter; crm_client_t *ipc_client = NULL; char *key = NULL; GList *remove_these = NULL; GListPtr gIter = NULL; ipc_providers = g_list_remove(ipc_providers, ipc_proxy); g_hash_table_iter_init(&iter, ipc_clients); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & ipc_client)) { const char *proxy_id = ipc_client->userdata; if (safe_str_eq(proxy_id, ipc_proxy->id)) { crm_info("ipc proxy connection for client %s pid %d destroyed because cluster node disconnected.", ipc_client->id, ipc_client->pid); /* we can't remove during the iteration, so copy items * to a list we can destroy later */ remove_these = g_list_append(remove_these, ipc_client); } } for (gIter = remove_these; gIter != NULL; gIter = gIter->next) { ipc_client = gIter->data; // Disconnection callback will free the client here qb_ipcs_disconnect(ipc_client->ipcs); } /* just frees the list, not the elements in the list */ g_list_free(remove_these); } void ipc_proxy_init(void) { ipc_clients = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, NULL); cib_ipc_servers_init(&cib_ro, &cib_rw, &cib_shm, &cib_proxy_callbacks_ro, &cib_proxy_callbacks_rw); attrd_ipc_server_init(&attrd_ipcs, &attrd_proxy_callbacks); stonith_ipc_server_init(&stonith_ipcs, &stonith_proxy_callbacks); crmd_ipcs = crmd_ipc_server_init(&crmd_proxy_callbacks); if (crmd_ipcs == NULL) { crm_err("Failed to create controller: exiting and inhibiting respawn"); crm_warn("Verify pacemaker and pacemaker_remote are not both enabled"); crm_exit(CRM_EX_FATAL); } } void ipc_proxy_cleanup(void) { if (ipc_providers) { g_list_free(ipc_providers); ipc_providers = NULL; } if (ipc_clients) { g_hash_table_destroy(ipc_clients); ipc_clients = NULL; } cib_ipc_servers_destroy(cib_ro, cib_rw, cib_shm); qb_ipcs_destroy(attrd_ipcs); qb_ipcs_destroy(stonith_ipcs); qb_ipcs_destroy(crmd_ipcs); cib_ro = NULL; cib_rw = NULL; cib_shm = NULL; } diff --git a/daemons/fenced/pacemaker-fenced.c b/daemons/fenced/pacemaker-fenced.c index b7d6072375..105012b6df 100644 --- a/daemons/fenced/pacemaker-fenced.c +++ b/daemons/fenced/pacemaker-fenced.c @@ -1,1503 +1,1501 @@ /* * Copyright 2009-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include /* U32T ~ PRIu32, X32T ~ PRIx32 */ #include #include #include #include #include #include #include #include #include #include #include #include #include char *stonith_our_uname = NULL; char *stonith_our_uuid = NULL; long stonith_watchdog_timeout_ms = 0; static GMainLoop *mainloop = NULL; gboolean stand_alone = FALSE; static gboolean no_cib_connect = FALSE; static gboolean stonith_shutdown_flag = FALSE; static qb_ipcs_service_t *ipcs = NULL; static xmlNode *local_cib = NULL; static pe_working_set_t *fenced_data_set = NULL; static cib_t *cib_api = NULL; static void *cib_library = NULL; static void stonith_shutdown(int nsig); static void stonith_cleanup(void); static int32_t st_ipc_accept(qb_ipcs_connection_t * c, uid_t uid, gid_t gid) { if (stonith_shutdown_flag) { crm_info("Ignoring new client [%d] during shutdown", crm_ipcs_client_pid(c)); return -EPERM; } if (crm_client_new(c, uid, gid) == NULL) { return -EIO; } return 0; } static void st_ipc_created(qb_ipcs_connection_t * c) { crm_trace("Connection created for %p", c); } /* Exit code means? */ static int32_t st_ipc_dispatch(qb_ipcs_connection_t * qbc, void *data, size_t size) { uint32_t id = 0; uint32_t flags = 0; int call_options = 0; xmlNode *request = NULL; crm_client_t *c = crm_client_get(qbc); const char *op = NULL; if (c == NULL) { crm_info("Invalid client: %p", qbc); return 0; } request = crm_ipcs_recv(c, data, size, &id, &flags); if (request == NULL) { crm_ipcs_send_ack(c, id, flags, "nack", __FUNCTION__, __LINE__); return 0; } op = crm_element_value(request, F_CRM_TASK); if(safe_str_eq(op, CRM_OP_RM_NODE_CACHE)) { crm_xml_add(request, F_TYPE, T_STONITH_NG); crm_xml_add(request, F_STONITH_OPERATION, op); crm_xml_add(request, F_STONITH_CLIENTID, c->id); crm_xml_add(request, F_STONITH_CLIENTNAME, crm_client_name(c)); crm_xml_add(request, F_STONITH_CLIENTNODE, stonith_our_uname); send_cluster_message(NULL, crm_msg_stonith_ng, request, FALSE); free_xml(request); return 0; } if (c->name == NULL) { const char *value = crm_element_value(request, F_STONITH_CLIENTNAME); if (value == NULL) { value = "unknown"; } c->name = crm_strdup_printf("%s.%u", value, c->pid); } crm_element_value_int(request, F_STONITH_CALLOPTS, &call_options); crm_trace("Flags %" X32T "/%u for command %" U32T " from %s", flags, call_options, id, crm_client_name(c)); if (is_set(call_options, st_opt_sync_call)) { CRM_ASSERT(flags & crm_ipc_client_response); CRM_LOG_ASSERT(c->request_id == 0); /* This means the client has two synchronous events in-flight */ c->request_id = id; /* Reply only to the last one */ } crm_xml_add(request, F_STONITH_CLIENTID, c->id); crm_xml_add(request, F_STONITH_CLIENTNAME, crm_client_name(c)); crm_xml_add(request, F_STONITH_CLIENTNODE, stonith_our_uname); crm_log_xml_trace(request, "Client[inbound]"); stonith_command(c, id, flags, request, NULL); free_xml(request); return 0; } /* Error code means? */ static int32_t st_ipc_closed(qb_ipcs_connection_t * c) { crm_client_t *client = crm_client_get(c); if (client == NULL) { return 0; } crm_trace("Connection %p closed", c); crm_client_destroy(client); /* 0 means: yes, go ahead and destroy the connection */ return 0; } static void st_ipc_destroy(qb_ipcs_connection_t * c) { crm_trace("Connection %p destroyed", c); st_ipc_closed(c); } static void stonith_peer_callback(xmlNode * msg, void *private_data) { const char *remote_peer = crm_element_value(msg, F_ORIG); const char *op = crm_element_value(msg, F_STONITH_OPERATION); if (crm_str_eq(op, "poke", TRUE)) { return; } crm_log_xml_trace(msg, "Peer[inbound]"); stonith_command(NULL, 0, 0, msg, remote_peer); } #if SUPPORT_COROSYNC static void stonith_peer_ais_callback(cpg_handle_t handle, const struct cpg_name *groupName, uint32_t nodeid, uint32_t pid, void *msg, size_t msg_len) { uint32_t kind = 0; xmlNode *xml = NULL; const char *from = NULL; char *data = pcmk_message_common_cs(handle, nodeid, pid, msg, &kind, &from); if(data == NULL) { return; } if (kind == crm_class_cluster) { xml = string2xml(data); if (xml == NULL) { crm_err("Invalid XML: '%.120s'", data); free(data); return; } crm_xml_add(xml, F_ORIG, from); /* crm_xml_add_int(xml, F_SEQ, wrapper->id); */ stonith_peer_callback(xml, NULL); } free_xml(xml); free(data); return; } static void stonith_peer_cs_destroy(gpointer user_data) { crm_crit("Lost connection to cluster layer, shutting down"); stonith_shutdown(0); } #endif void do_local_reply(xmlNode * notify_src, const char *client_id, gboolean sync_reply, gboolean from_peer) { /* send callback to originating child */ crm_client_t *client_obj = NULL; int local_rc = pcmk_ok; crm_trace("Sending response"); client_obj = crm_client_get_by_id(client_id); crm_trace("Sending callback to request originator"); if (client_obj == NULL) { local_rc = -1; crm_trace("No client to sent the response to. F_STONITH_CLIENTID not set."); } else { int rid = 0; if (sync_reply) { CRM_LOG_ASSERT(client_obj->request_id); rid = client_obj->request_id; client_obj->request_id = 0; crm_trace("Sending response %d to %s %s", rid, client_obj->name, from_peer ? "(originator of delegated request)" : ""); } else { crm_trace("Sending an event to %s %s", client_obj->name, from_peer ? "(originator of delegated request)" : ""); } local_rc = crm_ipcs_send(client_obj, rid, notify_src, sync_reply?crm_ipc_flags_none:crm_ipc_server_event); } if (local_rc < pcmk_ok && client_obj != NULL) { crm_warn("%sSync reply to %s failed: %s", sync_reply ? "" : "A-", client_obj ? client_obj->name : "", pcmk_strerror(local_rc)); } } long long get_stonith_flag(const char *name) { if (safe_str_eq(name, T_STONITH_NOTIFY_FENCE)) { return st_callback_notify_fence; } else if (safe_str_eq(name, STONITH_OP_DEVICE_ADD)) { return st_callback_device_add; } else if (safe_str_eq(name, STONITH_OP_DEVICE_DEL)) { return st_callback_device_del; } else if (safe_str_eq(name, T_STONITH_NOTIFY_HISTORY)) { return st_callback_notify_history; } return st_callback_unknown; } static void stonith_notify_client(gpointer key, gpointer value, gpointer user_data) { xmlNode *update_msg = user_data; crm_client_t *client = value; const char *type = NULL; CRM_CHECK(client != NULL, return); CRM_CHECK(update_msg != NULL, return); type = crm_element_value(update_msg, F_SUBTYPE); CRM_CHECK(type != NULL, crm_log_xml_err(update_msg, "notify"); return); if (client->ipcs == NULL) { crm_trace("Skipping client with NULL channel"); return; } if (client->options & get_stonith_flag(type)) { int rc = crm_ipcs_send(client, 0, update_msg, crm_ipc_server_event | crm_ipc_server_error); if (rc <= 0) { crm_warn("%s notification of client %s.%.6s failed: %s (%d)", type, crm_client_name(client), client->id, pcmk_strerror(rc), rc); } else { crm_trace("Sent %s notification to client %s.%.6s", type, crm_client_name(client), client->id); } } } void do_stonith_async_timeout_update(const char *client_id, const char *call_id, int timeout) { crm_client_t *client = NULL; xmlNode *notify_data = NULL; if (!timeout || !call_id || !client_id) { return; } client = crm_client_get_by_id(client_id); if (!client) { return; } notify_data = create_xml_node(NULL, T_STONITH_TIMEOUT_VALUE); crm_xml_add(notify_data, F_TYPE, T_STONITH_TIMEOUT_VALUE); crm_xml_add(notify_data, F_STONITH_CALLID, call_id); crm_xml_add_int(notify_data, F_STONITH_TIMEOUT, timeout); crm_trace("timeout update is %d for client %s and call id %s", timeout, client_id, call_id); if (client) { crm_ipcs_send(client, 0, notify_data, crm_ipc_server_event); } free_xml(notify_data); } void do_stonith_notify(int options, const char *type, int result, xmlNode * data) { /* TODO: Standardize the contents of data */ xmlNode *update_msg = create_xml_node(NULL, "notify"); CRM_CHECK(type != NULL,;); crm_xml_add(update_msg, F_TYPE, T_STONITH_NOTIFY); crm_xml_add(update_msg, F_SUBTYPE, type); crm_xml_add(update_msg, F_STONITH_OPERATION, type); crm_xml_add_int(update_msg, F_STONITH_RC, result); if (data != NULL) { add_message_xml(update_msg, F_STONITH_CALLDATA, data); } crm_trace("Notifying clients"); g_hash_table_foreach(client_connections, stonith_notify_client, update_msg); free_xml(update_msg); crm_trace("Notify complete"); } static void do_stonith_notify_config(int options, const char *op, int rc, const char *desc, int active) { xmlNode *notify_data = create_xml_node(NULL, op); CRM_CHECK(notify_data != NULL, return); crm_xml_add(notify_data, F_STONITH_DEVICE, desc); crm_xml_add_int(notify_data, F_STONITH_ACTIVE, active); do_stonith_notify(options, op, rc, notify_data); free_xml(notify_data); } void do_stonith_notify_device(int options, const char *op, int rc, const char *desc) { do_stonith_notify_config(options, op, rc, desc, g_hash_table_size(device_list)); } void do_stonith_notify_level(int options, const char *op, int rc, const char *desc) { do_stonith_notify_config(options, op, rc, desc, g_hash_table_size(topology)); } static void topology_remove_helper(const char *node, int level) { int rc; char *desc = NULL; xmlNode *data = create_xml_node(NULL, XML_TAG_FENCING_LEVEL); crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__); crm_xml_add_int(data, XML_ATTR_STONITH_INDEX, level); crm_xml_add(data, XML_ATTR_STONITH_TARGET, node); rc = stonith_level_remove(data, &desc); do_stonith_notify_level(0, STONITH_OP_LEVEL_DEL, rc, desc); free_xml(data); free(desc); } static void remove_cib_device(xmlXPathObjectPtr xpathObj) { int max = numXpathResults(xpathObj), lpc = 0; for (lpc = 0; lpc < max; lpc++) { const char *rsc_id = NULL; const char *standard = NULL; xmlNode *match = getXpathResult(xpathObj, lpc); CRM_LOG_ASSERT(match != NULL); if(match != NULL) { standard = crm_element_value(match, XML_AGENT_ATTR_CLASS); } if (safe_str_neq(standard, PCMK_RESOURCE_CLASS_STONITH)) { continue; } rsc_id = crm_element_value(match, XML_ATTR_ID); stonith_device_remove(rsc_id, TRUE); } } static void handle_topology_change(xmlNode *match, bool remove) { int rc; char *desc = NULL; CRM_CHECK(match != NULL, return); crm_trace("Updating %s", ID(match)); if(remove) { int index = 0; char *key = stonith_level_key(match, -1); crm_element_value_int(match, XML_ATTR_STONITH_INDEX, &index); topology_remove_helper(key, index); free(key); } rc = stonith_level_register(match, &desc); do_stonith_notify_level(0, STONITH_OP_LEVEL_ADD, rc, desc); free(desc); } static void remove_fencing_topology(xmlXPathObjectPtr xpathObj) { int max = numXpathResults(xpathObj), lpc = 0; for (lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpathObj, lpc); CRM_LOG_ASSERT(match != NULL); if (match && crm_element_value(match, XML_DIFF_MARKER)) { /* Deletion */ int index = 0; char *target = stonith_level_key(match, -1); crm_element_value_int(match, XML_ATTR_STONITH_INDEX, &index); if (target == NULL) { crm_err("Invalid fencing target in element %s", ID(match)); } else if (index <= 0) { crm_err("Invalid level for %s in element %s", target, ID(match)); } else { topology_remove_helper(target, index); } /* } else { Deal with modifications during the 'addition' stage */ } } } static void register_fencing_topology(xmlXPathObjectPtr xpathObj) { int max = numXpathResults(xpathObj), lpc = 0; for (lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpathObj, lpc); handle_topology_change(match, TRUE); } } /* Fencing */ static void fencing_topology_init() { xmlXPathObjectPtr xpathObj = NULL; const char *xpath = "//" XML_TAG_FENCING_LEVEL; crm_trace("Full topology refresh"); free_topology_list(); init_topology_list(); /* Grab everything */ xpathObj = xpath_search(local_cib, xpath); register_fencing_topology(xpathObj); freeXpathObject(xpathObj); } #define rsc_name(x) x->clone_name?x->clone_name:x->id /*! * \internal * \brief Check whether our uname is in a resource's allowed node list * * \param[in] rsc Resource to check * * \return Pointer to node object if found, NULL otherwise */ static node_t * our_node_allowed_for(resource_t *rsc) { GHashTableIter iter; node_t *node = NULL; if (rsc && stonith_our_uname) { g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if (node && strcmp(node->details->uname, stonith_our_uname) == 0) { break; } node = NULL; } } return node; } /*! * \internal * \brief If a resource or any of its children are STONITH devices, update their * definitions given a cluster working set. * * \param[in] rsc Resource to check * \param[in] data_set Cluster working set with device information */ static void cib_device_update(resource_t *rsc, pe_working_set_t *data_set) { node_t *node = NULL; const char *value = NULL; const char *rclass = NULL; node_t *parent = NULL; gboolean remove = TRUE; /* If this is a complex resource, check children rather than this resource itself. * TODO: Mark each installed device and remove if untouched when this process finishes. */ if(rsc->children) { GListPtr gIter = NULL; for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) { cib_device_update(gIter->data, data_set); if(pe_rsc_is_clone(rsc)) { crm_trace("Only processing one copy of the clone %s", rsc->id); break; } } return; } /* We only care about STONITH resources. */ rclass = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS); if (safe_str_neq(rclass, PCMK_RESOURCE_CLASS_STONITH)) { return; } /* If this STONITH resource is disabled, just remove it. */ value = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE); if (safe_str_eq(value, RSC_STOPPED)) { crm_info("Device %s has been disabled", rsc->id); goto update_done; } /* Check whether our node is allowed for this resource (and its parent if in a group) */ node = our_node_allowed_for(rsc); if (rsc->parent && (rsc->parent->variant == pe_group)) { parent = our_node_allowed_for(rsc->parent); } if(node == NULL) { /* Our node is disallowed, so remove the device */ GHashTableIter iter; crm_info("Device %s has been disabled on %s: unknown", rsc->id, stonith_our_uname); g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { crm_trace("Available: %s = %d", node->details->uname, node->weight); } goto update_done; } else if(node->weight < 0 || (parent && parent->weight < 0)) { /* Our node (or its group) is disallowed by score, so remove the device */ char *score = score2char((node->weight < 0) ? node->weight : parent->weight); crm_info("Device %s has been disabled on %s: score=%s", rsc->id, stonith_our_uname, score); free(score); goto update_done; } else { /* Our node is allowed, so update the device information */ int rc; xmlNode *data; GHashTableIter gIter; stonith_key_value_t *params = NULL; const char *name = NULL; const char *agent = crm_element_value(rsc->xml, XML_EXPR_ATTR_TYPE); const char *rsc_provides = NULL; crm_debug("Device %s is allowed on %s: score=%d", rsc->id, stonith_our_uname, node->weight); get_rsc_attributes(rsc->parameters, rsc, node, data_set); get_meta_attributes(rsc->meta, rsc, node, data_set); rsc_provides = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_PROVIDES); g_hash_table_iter_init(&gIter, rsc->parameters); while (g_hash_table_iter_next(&gIter, (gpointer *) & name, (gpointer *) & value)) { if (!name || !value) { continue; } params = stonith_key_value_add(params, name, value); crm_trace(" %s=%s", name, value); } remove = FALSE; data = create_device_registration_xml(rsc_name(rsc), st_namespace_any, agent, params, rsc_provides); stonith_key_value_freeall(params, 1, 1); rc = stonith_device_register(data, NULL, TRUE); CRM_ASSERT(rc == pcmk_ok); free_xml(data); } update_done: if(remove && g_hash_table_lookup(device_list, rsc_name(rsc))) { stonith_device_remove(rsc_name(rsc), TRUE); } } -extern xmlNode *do_calculations(pe_working_set_t * data_set, xmlNode * xml_input, crm_time_t * now); - /*! * \internal * \brief Update all STONITH device definitions based on current CIB */ static void cib_devices_update(void) { GListPtr gIter = NULL; crm_info("Updating devices to version %s.%s.%s", crm_element_value(local_cib, XML_ATTR_GENERATION_ADMIN), crm_element_value(local_cib, XML_ATTR_GENERATION), crm_element_value(local_cib, XML_ATTR_NUMUPDATES)); CRM_ASSERT(fenced_data_set != NULL); fenced_data_set->input = local_cib; fenced_data_set->now = crm_time_new(NULL); fenced_data_set->flags |= pe_flag_quick_location; fenced_data_set->localhost = stonith_our_uname; cluster_status(fenced_data_set); - do_calculations(fenced_data_set, NULL, NULL); + pcmk__schedule_actions(fenced_data_set, NULL, NULL); for (gIter = fenced_data_set->resources; gIter != NULL; gIter = gIter->next) { cib_device_update(gIter->data, fenced_data_set); } fenced_data_set->input = NULL; // Wasn't a copy, so don't let API free it pe_reset_working_set(fenced_data_set); } static void update_cib_stonith_devices_v2(const char *event, xmlNode * msg) { xmlNode *change = NULL; char *reason = NULL; bool needs_update = FALSE; xmlNode *patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT); for (change = __xml_first_child(patchset); change != NULL; change = __xml_next(change)) { const char *op = crm_element_value(change, XML_DIFF_OP); const char *xpath = crm_element_value(change, XML_DIFF_PATH); const char *shortpath = NULL; if ((op == NULL) || (strcmp(op, "move") == 0) || strstr(xpath, "/"XML_CIB_TAG_STATUS)) { continue; } else if (safe_str_eq(op, "delete") && strstr(xpath, "/"XML_CIB_TAG_RESOURCE)) { const char *rsc_id = NULL; char *search = NULL; char *mutable = NULL; if (strstr(xpath, XML_TAG_ATTR_SETS) || strstr(xpath, XML_TAG_META_SETS)) { needs_update = TRUE; reason = strdup("(meta) attribute deleted from resource"); break; } mutable = strdup(xpath); rsc_id = strstr(mutable, "primitive[@id=\'"); if (rsc_id != NULL) { rsc_id += strlen("primitive[@id=\'"); search = strchr(rsc_id, '\''); } if (search != NULL) { *search = 0; stonith_device_remove(rsc_id, TRUE); } else { crm_warn("Ignoring malformed CIB update (resource deletion)"); } free(mutable); } else if (strstr(xpath, "/"XML_CIB_TAG_RESOURCES) || strstr(xpath, "/"XML_CIB_TAG_CONSTRAINTS) || strstr(xpath, "/"XML_CIB_TAG_RSCCONFIG)) { shortpath = strrchr(xpath, '/'); CRM_ASSERT(shortpath); reason = crm_strdup_printf("%s %s", op, shortpath+1); needs_update = TRUE; break; } } if(needs_update) { crm_info("Updating device list from the cib: %s", reason); cib_devices_update(); } else { crm_trace("No updates for device list found in cib"); } free(reason); } static void update_cib_stonith_devices_v1(const char *event, xmlNode * msg) { const char *reason = "none"; gboolean needs_update = FALSE; xmlXPathObjectPtr xpath_obj = NULL; /* process new constraints */ xpath_obj = xpath_search(msg, "//" F_CIB_UPDATE_RESULT "//" XML_CONS_TAG_RSC_LOCATION); if (numXpathResults(xpath_obj) > 0) { int max = numXpathResults(xpath_obj), lpc = 0; /* Safest and simplest to always recompute */ needs_update = TRUE; reason = "new location constraint"; for (lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpath_obj, lpc); crm_log_xml_trace(match, "new constraint"); } } freeXpathObject(xpath_obj); /* process deletions */ xpath_obj = xpath_search(msg, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_CIB_TAG_RESOURCE); if (numXpathResults(xpath_obj) > 0) { remove_cib_device(xpath_obj); } freeXpathObject(xpath_obj); /* process additions */ xpath_obj = xpath_search(msg, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_CIB_TAG_RESOURCE); if (numXpathResults(xpath_obj) > 0) { int max = numXpathResults(xpath_obj), lpc = 0; for (lpc = 0; lpc < max; lpc++) { const char *rsc_id = NULL; const char *standard = NULL; xmlNode *match = getXpathResult(xpath_obj, lpc); rsc_id = crm_element_value(match, XML_ATTR_ID); standard = crm_element_value(match, XML_AGENT_ATTR_CLASS); if (safe_str_neq(standard, PCMK_RESOURCE_CLASS_STONITH)) { continue; } crm_trace("Fencing resource %s was added or modified", rsc_id); reason = "new resource"; needs_update = TRUE; } } freeXpathObject(xpath_obj); if(needs_update) { crm_info("Updating device list from the cib: %s", reason); cib_devices_update(); } } static void update_cib_stonith_devices(const char *event, xmlNode * msg) { int format = 1; xmlNode *patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT); CRM_ASSERT(patchset); crm_element_value_int(patchset, "format", &format); switch(format) { case 1: update_cib_stonith_devices_v1(event, msg); break; case 2: update_cib_stonith_devices_v2(event, msg); break; default: crm_warn("Unknown patch format: %d", format); } } /* Needs to hold node name + attribute name + attribute value + 75 */ #define XPATH_MAX 512 /*! * \internal * \brief Check whether a node has a specific attribute name/value * * \param[in] node Name of node to check * \param[in] name Name of an attribute to look for * \param[in] value The value the named attribute needs to be set to in order to be considered a match * * \return TRUE if the locally cached CIB has the specified node attribute */ gboolean node_has_attr(const char *node, const char *name, const char *value) { char xpath[XPATH_MAX]; xmlNode *match; int n; CRM_CHECK(local_cib != NULL, return FALSE); /* Search for the node's attributes in the CIB. While the schema allows * multiple sets of instance attributes, and allows instance attributes to * use id-ref to reference values elsewhere, that is intended for resources, * so we ignore that here. */ n = snprintf(xpath, XPATH_MAX, "//" XML_CIB_TAG_NODES "/" XML_CIB_TAG_NODE "[@uname='%s']/" XML_TAG_ATTR_SETS "/" XML_CIB_TAG_NVPAIR "[@name='%s' and @value='%s']", node, name, value); match = get_xpath_object(xpath, local_cib, LOG_TRACE); CRM_CHECK(n < XPATH_MAX, return FALSE); return (match != NULL); } static void update_fencing_topology(const char *event, xmlNode * msg) { int format = 1; const char *xpath; xmlXPathObjectPtr xpathObj = NULL; xmlNode *patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT); CRM_ASSERT(patchset); crm_element_value_int(patchset, "format", &format); if(format == 1) { /* Process deletions (only) */ xpath = "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_TAG_FENCING_LEVEL; xpathObj = xpath_search(msg, xpath); remove_fencing_topology(xpathObj); freeXpathObject(xpathObj); /* Process additions and changes */ xpath = "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_TAG_FENCING_LEVEL; xpathObj = xpath_search(msg, xpath); register_fencing_topology(xpathObj); freeXpathObject(xpathObj); } else if(format == 2) { xmlNode *change = NULL; int add[] = { 0, 0, 0 }; int del[] = { 0, 0, 0 }; xml_patch_versions(patchset, add, del); for (change = __xml_first_child(patchset); change != NULL; change = __xml_next(change)) { const char *op = crm_element_value(change, XML_DIFF_OP); const char *xpath = crm_element_value(change, XML_DIFF_PATH); if(op == NULL) { continue; } else if(strstr(xpath, "/" XML_TAG_FENCING_LEVEL) != NULL) { /* Change to a specific entry */ crm_trace("Handling %s operation %d.%d.%d for %s", op, add[0], add[1], add[2], xpath); if(strcmp(op, "move") == 0) { continue; } else if(strcmp(op, "create") == 0) { handle_topology_change(change->children, FALSE); } else if(strcmp(op, "modify") == 0) { xmlNode *match = first_named_child(change, XML_DIFF_RESULT); if(match) { handle_topology_change(match->children, TRUE); } } else if(strcmp(op, "delete") == 0) { /* Nuclear option, all we have is the path and an id... not enough to remove a specific entry */ crm_info("Re-initializing fencing topology after %s operation %d.%d.%d for %s", op, add[0], add[1], add[2], xpath); fencing_topology_init(); return; } } else if (strstr(xpath, "/" XML_TAG_FENCING_TOPOLOGY) != NULL) { /* Change to the topology in general */ crm_info("Re-initializing fencing topology after top-level %s operation %d.%d.%d for %s", op, add[0], add[1], add[2], xpath); fencing_topology_init(); return; } else if (strstr(xpath, "/" XML_CIB_TAG_CONFIGURATION)) { /* Changes to the whole config section, possibly including the topology as a whild */ if(first_named_child(change, XML_TAG_FENCING_TOPOLOGY) == NULL) { crm_trace("Nothing for us in %s operation %d.%d.%d for %s.", op, add[0], add[1], add[2], xpath); } else if(strcmp(op, "delete") == 0 || strcmp(op, "create") == 0) { crm_info("Re-initializing fencing topology after top-level %s operation %d.%d.%d for %s.", op, add[0], add[1], add[2], xpath); fencing_topology_init(); return; } } else { crm_trace("Nothing for us in %s operation %d.%d.%d for %s", op, add[0], add[1], add[2], xpath); } } } else { crm_warn("Unknown patch format: %d", format); } } static bool have_cib_devices = FALSE; static void update_cib_cache_cb(const char *event, xmlNode * msg) { int rc = pcmk_ok; xmlNode *stonith_enabled_xml = NULL; xmlNode *stonith_watchdog_xml = NULL; const char *stonith_enabled_s = NULL; static gboolean stonith_enabled_saved = TRUE; if(!have_cib_devices) { crm_trace("Skipping updates until we get a full dump"); return; } else if(msg == NULL) { crm_trace("Missing %s update", event); return; } /* Maintain a local copy of the CIB so that we have full access * to device definitions, location constraints, and node attributes */ if (local_cib != NULL) { int rc = pcmk_ok; xmlNode *patchset = NULL; crm_element_value_int(msg, F_CIB_RC, &rc); if (rc != pcmk_ok) { return; } patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT); xml_log_patchset(LOG_TRACE, "Config update", patchset); rc = xml_apply_patchset(local_cib, patchset, TRUE); switch (rc) { case pcmk_ok: case -pcmk_err_old_data: break; case -pcmk_err_diff_resync: case -pcmk_err_diff_failed: crm_notice("[%s] Patch aborted: %s (%d)", event, pcmk_strerror(rc), rc); free_xml(local_cib); local_cib = NULL; break; default: crm_warn("[%s] ABORTED: %s (%d)", event, pcmk_strerror(rc), rc); free_xml(local_cib); local_cib = NULL; } } if (local_cib == NULL) { crm_trace("Re-requesting the full cib"); rc = cib_api->cmds->query(cib_api, NULL, &local_cib, cib_scope_local | cib_sync_call); if(rc != pcmk_ok) { crm_err("Couldn't retrieve the CIB: %s (%d)", pcmk_strerror(rc), rc); return; } CRM_ASSERT(local_cib != NULL); stonith_enabled_saved = FALSE; /* Trigger a full refresh below */ } crm_peer_caches_refresh(local_cib); stonith_enabled_xml = get_xpath_object("//nvpair[@name='stonith-enabled']", local_cib, LOG_TRACE); if (stonith_enabled_xml) { stonith_enabled_s = crm_element_value(stonith_enabled_xml, XML_NVPAIR_ATTR_VALUE); } if (stonith_enabled_s == NULL || crm_is_true(stonith_enabled_s)) { long timeout_ms = 0; const char *value = NULL; stonith_watchdog_xml = get_xpath_object("//nvpair[@name='stonith-watchdog-timeout']", local_cib, LOG_TRACE); if (stonith_watchdog_xml) { value = crm_element_value(stonith_watchdog_xml, XML_NVPAIR_ATTR_VALUE); } if(value) { timeout_ms = crm_get_msec(value); } if (timeout_ms < 0) { timeout_ms = crm_auto_watchdog_timeout(); } if(timeout_ms != stonith_watchdog_timeout_ms) { crm_notice("New watchdog timeout %lds (was %lds)", timeout_ms/1000, stonith_watchdog_timeout_ms/1000); stonith_watchdog_timeout_ms = timeout_ms; } } else { stonith_watchdog_timeout_ms = 0; } if (stonith_enabled_s && crm_is_true(stonith_enabled_s) == FALSE) { crm_trace("Ignoring cib updates while stonith is disabled"); stonith_enabled_saved = FALSE; return; } else if (stonith_enabled_saved == FALSE) { crm_info("Updating stonith device and topology lists now that stonith is enabled"); stonith_enabled_saved = TRUE; fencing_topology_init(); cib_devices_update(); } else { update_fencing_topology(event, msg); update_cib_stonith_devices(event, msg); } } static void init_cib_cache_cb(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { crm_info("Updating device list from the cib: init"); have_cib_devices = TRUE; local_cib = copy_xml(output); crm_peer_caches_refresh(local_cib); fencing_topology_init(); cib_devices_update(); } static void stonith_shutdown(int nsig) { stonith_shutdown_flag = TRUE; crm_info("Terminating with %d clients", crm_hash_table_size(client_connections)); if (mainloop != NULL && g_main_loop_is_running(mainloop)) { g_main_loop_quit(mainloop); } else { stonith_cleanup(); crm_exit(CRM_EX_OK); } } static void cib_connection_destroy(gpointer user_data) { if (stonith_shutdown_flag) { crm_info("Connection to the CIB manager closed"); return; } else { crm_crit("Lost connection to the CIB manager, shutting down"); } if (cib_api) { cib_api->cmds->signoff(cib_api); } stonith_shutdown(0); } static void stonith_cleanup(void) { if (cib_api) { cib_api->cmds->signoff(cib_api); } if (ipcs) { qb_ipcs_destroy(ipcs); } crm_peer_destroy(); crm_client_cleanup(); free_stonith_remote_op_list(); free_topology_list(); free_device_list(); free_metadata_cache(); free(stonith_our_uname); stonith_our_uname = NULL; free_xml(local_cib); local_cib = NULL; } /* *INDENT-OFF* */ static struct crm_option long_options[] = { {"stand-alone", 0, 0, 's'}, {"stand-alone-w-cpg", 0, 0, 'c'}, {"logfile", 1, 0, 'l'}, {"verbose", 0, 0, 'V'}, {"version", 0, 0, '$'}, {"help", 0, 0, '?'}, {0, 0, 0, 0} }; /* *INDENT-ON* */ static void setup_cib(void) { int rc, retries = 0; static cib_t *(*cib_new_fn) (void) = NULL; if (cib_new_fn == NULL) { cib_new_fn = find_library_function(&cib_library, CIB_LIBRARY, "cib_new", TRUE); } if (cib_new_fn != NULL) { cib_api = (*cib_new_fn) (); } if (cib_api == NULL) { crm_err("No connection to the CIB manager"); return; } do { sleep(retries); rc = cib_api->cmds->signon(cib_api, CRM_SYSTEM_STONITHD, cib_command); } while (rc == -ENOTCONN && ++retries < 5); if (rc != pcmk_ok) { crm_err("Could not connect to the CIB manager: %s (%d)", pcmk_strerror(rc), rc); } else if (pcmk_ok != cib_api->cmds->add_notify_callback(cib_api, T_CIB_DIFF_NOTIFY, update_cib_cache_cb)) { crm_err("Could not set CIB notification callback"); } else { rc = cib_api->cmds->query(cib_api, NULL, NULL, cib_scope_local); cib_api->cmds->register_callback(cib_api, rc, 120, FALSE, NULL, "init_cib_cache_cb", init_cib_cache_cb); cib_api->cmds->set_connection_dnotify(cib_api, cib_connection_destroy); crm_info("Watching for stonith topology changes"); } } struct qb_ipcs_service_handlers ipc_callbacks = { .connection_accept = st_ipc_accept, .connection_created = st_ipc_created, .msg_process = st_ipc_dispatch, .connection_closed = st_ipc_closed, .connection_destroyed = st_ipc_destroy }; /*! * \internal * \brief Callback for peer status changes * * \param[in] type What changed * \param[in] node What peer had the change * \param[in] data Previous value of what changed */ static void st_peer_update_callback(enum crm_status_type type, crm_node_t * node, const void *data) { if ((type != crm_status_processes) && !is_set(node->flags, crm_remote_node)) { /* * This is a hack until we can send to a nodeid and/or we fix node name lookups * These messages are ignored in stonith_peer_callback() */ xmlNode *query = create_xml_node(NULL, "stonith_command"); crm_xml_add(query, F_XML_TAGNAME, "stonith_command"); crm_xml_add(query, F_TYPE, T_STONITH_NG); crm_xml_add(query, F_STONITH_OPERATION, "poke"); crm_debug("Broadcasting our uname because of node %u", node->id); send_cluster_message(NULL, crm_msg_stonith_ng, query, FALSE); free_xml(query); } } int main(int argc, char **argv) { int flag; int lpc = 0; int argerr = 0; int option_index = 0; crm_cluster_t cluster; const char *actions[] = { "reboot", "off", "on", "list", "monitor", "status" }; crm_log_preinit(NULL, argc, argv); crm_set_options(NULL, "mode [options]", long_options, "Provides a summary of cluster's current state." "\n\nOutputs varying levels of detail in a number of different formats.\n"); while (1) { flag = crm_get_option(argc, argv, &option_index); if (flag == -1) { break; } switch (flag) { case 'V': crm_bump_log_level(argc, argv); break; case 'l': crm_add_logfile(optarg); break; case 's': stand_alone = TRUE; break; case 'c': stand_alone = FALSE; no_cib_connect = TRUE; break; case '$': case '?': crm_help(flag, CRM_EX_OK); break; default: ++argerr; break; } } if (argc - optind == 1 && safe_str_eq("metadata", argv[optind])) { printf("\n"); printf("\n"); printf(" 1.0\n"); printf(" Instance attributes available for all \"stonith\"-class resources" " and used by Pacemaker's fence daemon, formerly known as stonithd\n"); printf(" Instance attributes available for all \"stonith\"-class resources\n"); printf(" \n"); #if 0 // priority is not implemented yet printf(" \n"); printf(" Devices that are not in a topology " "are tried in order of highest to lowest integer priority\n"); printf(" \n"); printf(" \n"); #endif printf(" \n", STONITH_ATTR_HOSTARG); printf (" Advanced use only: An alternate parameter to supply instead of 'port'\n"); printf (" Some devices do not support the standard 'port' parameter or may provide additional ones.\n" "Use this to specify an alternate, device-specific, parameter that should indicate the machine to be fenced.\n" "A value of 'none' can be used to tell the cluster not to supply any additional parameters.\n" " \n"); printf(" \n"); printf(" \n"); printf(" \n", STONITH_ATTR_HOSTMAP); printf (" A mapping of host names to ports numbers for devices that do not support host names.\n"); printf (" Eg. node1:1;node2:2,3 would tell the cluster to use port 1 for node1 and ports 2 and 3 for node2\n"); printf(" \n"); printf(" \n"); printf(" \n", STONITH_ATTR_HOSTLIST); printf (" A list of machines controlled by this device (Optional unless %s=static-list).\n", STONITH_ATTR_HOSTCHECK); printf(" \n"); printf(" \n"); printf(" \n", STONITH_ATTR_HOSTCHECK); printf (" How to determine which machines are controlled by the device.\n"); printf(" Allowed values: dynamic-list " "(query the device via the 'list' command), static-list " "(check the " STONITH_ATTR_HOSTLIST " attribute), status " "(query the device via the 'status' command), none (assume " "every device can fence every machine)\n"); printf(" \n"); printf(" \n"); printf(" \n", STONITH_ATTR_DELAY_MAX); printf (" Enable a random delay for stonith actions and specify the maximum of random delay.\n"); printf (" This prevents double fencing when using slow devices such as sbd.\n" "Use this to enable a random delay for stonith actions.\n" "The overall delay is derived from this random delay value adding a static delay so that the sum is kept below the maximum delay.\n"); printf(" \n"); printf(" \n"); printf(" \n", STONITH_ATTR_DELAY_BASE); printf (" Enable a base delay for stonith actions and specify base delay value.\n"); printf (" This prevents double fencing when different delays are configured on the nodes.\n" "Use this to enable a static delay for stonith actions.\n" "The overall delay is derived from a random delay value adding this static delay so that the sum is kept below the maximum delay.\n"); printf(" \n"); printf(" \n"); printf(" \n", STONITH_ATTR_ACTION_LIMIT); printf (" The maximum number of actions can be performed in parallel on this device\n"); printf (" Cluster property concurrent-fencing=true needs to be configured first.\n" "Then use this to specify the maximum number of actions can be performed in parallel on this device. -1 is unlimited.\n"); printf(" \n"); printf(" \n"); for (lpc = 0; lpc < DIMOF(actions); lpc++) { printf(" \n", actions[lpc]); printf (" Advanced use only: An alternate command to run instead of '%s'\n", actions[lpc]); printf (" Some devices do not support the standard commands or may provide additional ones.\n" "Use this to specify an alternate, device-specific, command that implements the '%s' action.\n", actions[lpc]); printf(" \n", actions[lpc]); printf(" \n"); printf(" \n", actions[lpc]); printf (" Advanced use only: Specify an alternate timeout to use for %s actions instead of stonith-timeout\n", actions[lpc]); printf (" Some devices need much more/less time to complete than normal.\n" "Use this to specify an alternate, device-specific, timeout for '%s' actions.\n", actions[lpc]); printf(" \n"); printf(" \n"); printf(" \n", actions[lpc]); printf (" Advanced use only: The maximum number of times to retry the '%s' command within the timeout period\n", actions[lpc]); printf(" Some devices do not support multiple connections." " Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining." " Use this option to alter the number of times Pacemaker retries '%s' actions before giving up." "\n", actions[lpc]); printf(" \n"); printf(" \n"); } printf(" \n"); printf("\n"); return CRM_EX_OK; } if (optind != argc) { ++argerr; } if (argerr) { crm_help('?', CRM_EX_USAGE); } crm_log_init(NULL, LOG_INFO, TRUE, FALSE, argc, argv, FALSE); mainloop_add_signal(SIGTERM, stonith_shutdown); crm_peer_init(); fenced_data_set = pe_new_working_set(); CRM_ASSERT(fenced_data_set != NULL); if (stand_alone == FALSE) { if (is_corosync_cluster()) { #if SUPPORT_COROSYNC cluster.destroy = stonith_peer_cs_destroy; cluster.cpg.cpg_deliver_fn = stonith_peer_ais_callback; cluster.cpg.cpg_confchg_fn = pcmk_cpg_membership; #endif } crm_set_status_callback(&st_peer_update_callback); if (crm_cluster_connect(&cluster) == FALSE) { crm_crit("Cannot sign in to the cluster... terminating"); crm_exit(CRM_EX_FATAL); } stonith_our_uname = cluster.uname; stonith_our_uuid = cluster.uuid; if (no_cib_connect == FALSE) { setup_cib(); } } else { stonith_our_uname = strdup("localhost"); } init_device_list(); init_topology_list(); if(stonith_watchdog_timeout_ms > 0) { int rc; xmlNode *xml; stonith_key_value_t *params = NULL; params = stonith_key_value_add(params, STONITH_ATTR_HOSTLIST, stonith_our_uname); xml = create_device_registration_xml("watchdog", st_namespace_internal, STONITH_WATCHDOG_AGENT, params, NULL); stonith_key_value_freeall(params, 1, 1); rc = stonith_device_register(xml, NULL, FALSE); free_xml(xml); if (rc != pcmk_ok) { crm_crit("Cannot register watchdog pseudo fence agent"); crm_exit(CRM_EX_FATAL); } } stonith_ipc_server_init(&ipcs, &ipc_callbacks); /* Create the mainloop and run it... */ mainloop = g_main_loop_new(NULL, FALSE); crm_info("Starting %s mainloop", crm_system_name); g_main_loop_run(mainloop); stonith_cleanup(); pe_free_working_set(fenced_data_set); crm_exit(CRM_EX_OK); } diff --git a/daemons/schedulerd/pacemaker-schedulerd.c b/daemons/schedulerd/pacemaker-schedulerd.c index 3b896dd895..e349bdc099 100644 --- a/daemons/schedulerd/pacemaker-schedulerd.c +++ b/daemons/schedulerd/pacemaker-schedulerd.c @@ -1,377 +1,350 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #define OPTARGS "hVc" static GMainLoop *mainloop = NULL; static qb_ipcs_service_t *ipcs = NULL; static pe_working_set_t *sched_data_set = NULL; -extern int transition_id; #define get_series() was_processing_error?1:was_processing_warning?2:3 typedef struct series_s { const char *name; const char *param; int wrap; } series_t; series_t series[] = { {"pe-unknown", "_dont_match_anything_", -1}, {"pe-error", "pe-error-series-max", -1}, {"pe-warn", "pe-warn-series-max", 200}, {"pe-input", "pe-input-series-max", 400}, }; void pengine_shutdown(int nsig); -extern xmlNode *do_calculations(pe_working_set_t * data_set, xmlNode * xml_input, crm_time_t * now); static gboolean process_pe_message(xmlNode * msg, xmlNode * xml_data, crm_client_t * sender) { static char *last_digest = NULL; static char *filename = NULL; time_t execution_date = time(NULL); const char *sys_to = crm_element_value(msg, F_CRM_SYS_TO); const char *op = crm_element_value(msg, F_CRM_TASK); const char *ref = crm_element_value(msg, F_CRM_REFERENCE); crm_trace("Processing %s op (ref=%s)...", op, ref); if (op == NULL) { /* error */ } else if (strcasecmp(op, CRM_OP_HELLO) == 0) { /* ignore */ } else if (safe_str_eq(crm_element_value(msg, F_CRM_MSG_TYPE), XML_ATTR_RESPONSE)) { /* ignore */ } else if (sys_to == NULL || strcasecmp(sys_to, CRM_SYSTEM_PENGINE) != 0) { crm_trace("Bad sys-to %s", crm_str(sys_to)); return FALSE; } else if (strcasecmp(op, CRM_OP_PECALC) == 0) { int seq = -1; int series_id = 0; int series_wrap = 0; char *digest = NULL; const char *value = NULL; xmlNode *converted = NULL; xmlNode *reply = NULL; gboolean is_repoke = FALSE; gboolean process = TRUE; crm_config_error = FALSE; crm_config_warning = FALSE; was_processing_error = FALSE; was_processing_warning = FALSE; if (sched_data_set == NULL) { sched_data_set = pe_new_working_set(); CRM_ASSERT(sched_data_set != NULL); } digest = calculate_xml_versioned_digest(xml_data, FALSE, FALSE, CRM_FEATURE_SET); converted = copy_xml(xml_data); if (cli_config_update(&converted, NULL, TRUE) == FALSE) { sched_data_set->graph = create_xml_node(NULL, XML_TAG_GRAPH); crm_xml_add_int(sched_data_set->graph, "transition_id", 0); crm_xml_add_int(sched_data_set->graph, "cluster-delay", 0); process = FALSE; free(digest); } else if (safe_str_eq(digest, last_digest)) { crm_info("Input has not changed since last time, not saving to disk"); is_repoke = TRUE; free(digest); } else { free(last_digest); last_digest = digest; } if (process) { - do_calculations(sched_data_set, converted, NULL); + pcmk__schedule_actions(sched_data_set, converted, NULL); } series_id = get_series(); series_wrap = series[series_id].wrap; value = pe_pref(sched_data_set->config_hash, series[series_id].param); if (value != NULL) { series_wrap = crm_int_helper(value, NULL); if (errno != 0) { series_wrap = series[series_id].wrap; } } else { crm_config_warn("No value specified for cluster" " preference: %s", series[series_id].param); } seq = get_last_sequence(PE_STATE_DIR, series[series_id].name); crm_trace("Series %s: wrap=%d, seq=%d, pref=%s", series[series_id].name, series_wrap, seq, value); sched_data_set->input = NULL; reply = create_reply(msg, sched_data_set->graph); CRM_ASSERT(reply != NULL); if (is_repoke == FALSE) { free(filename); filename = generate_series_filename(PE_STATE_DIR, series[series_id].name, seq, HAVE_BZLIB_H); } crm_xml_add(reply, F_CRM_TGRAPH_INPUT, filename); crm_xml_add_int(reply, "graph-errors", was_processing_error); crm_xml_add_int(reply, "graph-warnings", was_processing_warning); crm_xml_add_int(reply, "config-errors", crm_config_error); crm_xml_add_int(reply, "config-warnings", crm_config_warning); if (crm_ipcs_send(sender, 0, reply, crm_ipc_server_event) == FALSE) { int graph_file_fd = 0; char *graph_file = NULL; umask(S_IWGRP | S_IWOTH | S_IROTH); graph_file = crm_strdup_printf("%s/pengine.graph.XXXXXX", PE_STATE_DIR); graph_file_fd = mkstemp(graph_file); crm_err("Couldn't send transition graph to peer, writing to %s instead", graph_file); crm_xml_add(reply, F_CRM_TGRAPH, graph_file); write_xml_fd(sched_data_set->graph, graph_file, graph_file_fd, FALSE); free(graph_file); free_xml(first_named_child(reply, F_CRM_DATA)); CRM_ASSERT(crm_ipcs_send(sender, 0, reply, crm_ipc_server_event)); } free_xml(reply); pe_reset_working_set(sched_data_set); - - if (was_processing_error) { - crm_err("Calculated transition %d (with errors), saving inputs in %s", - transition_id, filename); - - } else if (was_processing_warning) { - crm_warn("Calculated transition %d (with warnings), saving inputs in %s", - transition_id, filename); - - } else { - crm_notice("Calculated transition %d, saving inputs in %s", - transition_id, filename); - } - - if (crm_config_error) { - crm_notice("Configuration errors found during scheduler processing," - " please run \"crm_verify -L\" to identify issues"); - } + pcmk__log_transition_summary(filename); if (is_repoke == FALSE && series_wrap != 0) { unlink(filename); crm_xml_add_int(xml_data, "execution-date", execution_date); write_xml_file(xml_data, filename, HAVE_BZLIB_H); write_last_sequence(PE_STATE_DIR, series[series_id].name, seq + 1, series_wrap); } else { crm_trace("Not writing out %s: %d & %d", filename, is_repoke, series_wrap); } free_xml(converted); } return TRUE; } -// only needed if process_pe_message() is called -static void -libpengine_fini() -{ - pe_free_working_set(sched_data_set); - sched_data_set = NULL; -} - static int32_t pe_ipc_accept(qb_ipcs_connection_t * c, uid_t uid, gid_t gid) { crm_trace("Connection %p", c); if (crm_client_new(c, uid, gid) == NULL) { return -EIO; } return 0; } static void pe_ipc_created(qb_ipcs_connection_t * c) { crm_trace("Connection %p", c); } gboolean process_pe_message(xmlNode * msg, xmlNode * xml_data, crm_client_t * sender); static int32_t pe_ipc_dispatch(qb_ipcs_connection_t * qbc, void *data, size_t size) { uint32_t id = 0; uint32_t flags = 0; crm_client_t *c = crm_client_get(qbc); xmlNode *msg = crm_ipcs_recv(c, data, size, &id, &flags); crm_ipcs_send_ack(c, id, flags, "ack", __FUNCTION__, __LINE__); if (msg != NULL) { xmlNode *data_xml = get_message_xml(msg, F_CRM_DATA); process_pe_message(msg, data_xml, c); free_xml(msg); } return 0; } /* Error code means? */ static int32_t pe_ipc_closed(qb_ipcs_connection_t * c) { crm_client_t *client = crm_client_get(c); if (client == NULL) { return 0; } crm_trace("Connection %p", c); crm_client_destroy(client); return 0; } static void pe_ipc_destroy(qb_ipcs_connection_t * c) { crm_trace("Connection %p", c); pe_ipc_closed(c); } struct qb_ipcs_service_handlers ipc_callbacks = { .connection_accept = pe_ipc_accept, .connection_created = pe_ipc_created, .msg_process = pe_ipc_dispatch, .connection_closed = pe_ipc_closed, .connection_destroyed = pe_ipc_destroy }; /* *INDENT-OFF* */ static struct crm_option long_options[] = { /* Top-level Options */ {"help", 0, 0, '?', "\tThis text"}, {"verbose", 0, 0, 'V', "\tIncrease debug output"}, {0, 0, 0, 0} }; /* *INDENT-ON* */ int main(int argc, char **argv) { int flag; int index = 0; int argerr = 0; crm_log_preinit(NULL, argc, argv); crm_set_options(NULL, "[options]", long_options, "Daemon for calculating the cluster's response to events"); mainloop_add_signal(SIGTERM, pengine_shutdown); while (1) { flag = crm_get_option(argc, argv, &index); if (flag == -1) break; switch (flag) { case 'V': crm_bump_log_level(argc, argv); break; case 'h': /* Help message */ crm_help('?', CRM_EX_OK); break; default: ++argerr; break; } } if (argc - optind == 1 && safe_str_eq("metadata", argv[optind])) { pe_metadata(); return CRM_EX_OK; } if (optind > argc) { ++argerr; } if (argerr) { crm_help('?', CRM_EX_USAGE); } crm_log_init(NULL, LOG_INFO, TRUE, FALSE, argc, argv, FALSE); if (pcmk__daemon_can_write(PE_STATE_DIR, NULL) == FALSE) { crm_err("Terminating due to bad permissions on " PE_STATE_DIR); fprintf(stderr, "ERROR: Bad permissions on " PE_STATE_DIR " (see logs for details)\n"); fflush(stderr); return CRM_EX_FATAL; } crm_debug("Init server comms"); ipcs = mainloop_add_ipc_server(CRM_SYSTEM_PENGINE, QB_IPC_SHM, &ipc_callbacks); if (ipcs == NULL) { crm_err("Failed to create IPC server: shutting down and inhibiting respawn"); crm_exit(CRM_EX_FATAL); } /* Create the mainloop and run it... */ crm_info("Starting %s", crm_system_name); mainloop = g_main_loop_new(NULL, FALSE); g_main_loop_run(mainloop); - libpengine_fini(); + pe_free_working_set(sched_data_set); crm_info("Exiting %s", crm_system_name); crm_exit(CRM_EX_OK); } void pengine_shutdown(int nsig) { mainloop_del_ipc_server(ipcs); - libpengine_fini(); + pe_free_working_set(sched_data_set); crm_exit(CRM_EX_OK); } diff --git a/doc/Pacemaker_Administration/en-US/Ch-Configuring.txt b/doc/Pacemaker_Administration/en-US/Ch-Configuring.txt index 1fb630ff04..38d77b19f4 100644 --- a/doc/Pacemaker_Administration/en-US/Ch-Configuring.txt +++ b/doc/Pacemaker_Administration/en-US/Ch-Configuring.txt @@ -1,442 +1,461 @@ :compat-mode: legacy = Configuring Pacemaker = == How Should the Configuration be Updated? == -There are three basic rules for updating the cluster configuration: +=== General Guidelines === - * Rule 1 - Never edit the +cib.xml+ file manually. Ever. I'm not making this up. - * Rule 2 - Read Rule 1 again. - * Rule 3 - The cluster will notice if you ignored rules 1 & 2 and refuse to use the configuration. +Pacemaker's configuration (the CIB) is stored in XML format. Cluster +administrators have multiple options for modifying the configuration either via +the XML, or at a more abstract (and easier for humans to understand) level. -Now that it is clear how 'not' to update the configuration, we can begin -to explain how you 'should'. +Pacemaker reacts to configuration changes as soon as they are saved. Most tools +provide the ability to batch changes together and commit them at once, rather +than make a series of small changes, which could cause avoid unnecessary +actions as Pacemaker responds to each change individually. + +Pacemaker keeps track of revisions to the configuration and will reject any +update that is older than the current revision. Thus, it is a good idea to +serialize all changes to the configuration. Avoid attempting simultaneous +changes, whether on the same node or different nodes, and whether manually or +using some automated configuration tool. + +[NOTE] +==== +It is not necessary to update the configuration on all cluster nodes. All +changes are immediately synchronized to all active members of the cluster. To +reduce bandwidth, the cluster only broadcasts the incremental updates that +result from your changes and uses MD5 checksums to ensure that each copy is +completely consistent. +==== + +=== Higher-level Tools === + +Most users will benefit from using higher-level tools that are provided by +projects separate from Pacemaker. Some of the most commonly used include the +crm shell, hawk, and pcs. footnote:[For a list, see "Configuration Tools" at +https://clusterlabs.org/components.html] + +See those projects' documentation for details on how to configure Pacemaker +using them. + +=== Pacemaker's Command-Line Tools === + +Most configuration tasks can be performed without needing any XML knowledge, +using one of the lower-level command-line tools provided with Pacemaker. + +To enable STONITH for example, one could run: + +---- +# crm_attribute --name stonith-enabled --update 1 +---- + +Or, to check whether *node1* is allowed to run resources, there is: + +---- +# crm_standby --query --node node1 +---- + +Or, to change the failure threshold of *my-test-rsc*, one can use: + +---- +# crm_resource -r my-test-rsc --set-parameter migration-threshold --parameter-value 3 --meta +---- + +Examples of using these tools for specific cases will be given throughout this +document where appropriate. See the man pages for further details. === Editing the CIB Using XML === -The most powerful tool for modifying the configuration is the +The most flexible tool for modifying the configuration is Pacemaker's +cibadmin+ command. With +cibadmin+, you can query, add, remove, update or replace any part of the configuration. All changes take effect immediately, so there is no need to perform a reload-like operation. The simplest way of using `cibadmin` is to use it to save the current configuration to a temporary file, edit that file with your favorite -text or XML editor, and then upload the revised configuration. footnote:[This -process might appear to risk overwriting changes that happen after the initial -cibadmin call, but pacemaker will reject any update that is "too old". If the -CIB is updated in some other fashion after the initial cibadmin, the second -cibadmin will be rejected because the version number will be too low.] +text or XML editor, and then upload the revised configuration. .Safely using an editor to modify the cluster configuration ====== -------- # cibadmin --query > tmp.xml # vi tmp.xml # cibadmin --replace --xml-file tmp.xml -------- ====== Some of the better XML editors can make use of a Relax NG schema to help make sure any changes you make are valid. The schema describing the configuration can be found in +pacemaker.rng+, which may be -deployed in a location such as +/usr/share/pacemaker+ or -+/usr/lib/heartbeat+ depending on your operating system and how you -installed the software. +deployed in a location such as +/usr/share/pacemaker+ depending on your +operating system distribution and how you installed the software. If you want to modify just one section of the configuration, you can query and replace just that section to avoid modifying any others. .Safely using an editor to modify only the resources section ====== -------- # cibadmin --query --scope resources > tmp.xml # vi tmp.xml # cibadmin --replace --scope resources --xml-file tmp.xml -------- ====== -=== Quickly Deleting Part of the Configuration === - -Identify the object you wish to delete by XML tag and id. For example, -you might search the CIB for all STONITH-related configuration: +To quickly delete a part of the configuration, identify the object you wish to +delete by XML tag and id. For example, you might search the CIB for all +STONITH-related configuration: .Searching for STONITH-related configuration items ====== ---- -# cibadmin -Q | grep stonith +# cibadmin --query | grep stonith ---- ====== If you wanted to delete the +primitive+ tag with id +child_DoFencing+, you would run: ---- # cibadmin --delete --xml-text '' ---- -=== Updating the Configuration Without Using XML === - -Most tasks can be performed with one of the other command-line -tools provided with pacemaker, avoiding the need to read or edit XML. - -To enable STONITH for example, one could run: - ----- -# crm_attribute --name stonith-enabled --update 1 ----- - -Or, to check whether *somenode* is allowed to run resources, there is: - ----- -# crm_standby --query --node somenode ----- - -Or, to find the current location of *my-test-rsc*, one can use: - ----- -# crm_resource --locate --resource my-test-rsc ----- +See the cibadmin man page for more options. -Examples of using these tools for specific cases will be given throughout this -document where appropriate. +[IMPORTANT] +==== +Never edit the live +cib.xml+ file directly. Pacemaker will detect such changes +and refuse to use the configuration. +==== [[s-config-sandboxes]] == Making Configuration Changes in a Sandbox == Often it is desirable to preview the effects of a series of changes before updating the configuration all at once. For this purpose, we have created `crm_shadow` which creates a "shadow" copy of the configuration and arranges for all the command line tools to use it. To begin, simply invoke `crm_shadow --create` with the name of a configuration to create footnote:[Shadow copies are identified with a name, making it possible to have more than one.], and follow the simple on-screen instructions. [WARNING] ==== Read this section and the on-screen instructions carefully; failure to do so could result in destroying the cluster's active configuration! ==== - .Creating and displaying the active sandbox ====== ---- # crm_shadow --create test Setting up shadow instance Type Ctrl-D to exit the crm_shadow shell shadow[test]: shadow[test] # crm_shadow --which test ---- ====== From this point on, all cluster commands will automatically use the shadow copy instead of talking to the cluster's active configuration. Once you have finished experimenting, you can either make the changes active via the `--commit` option, or discard them using the `--delete` option. Again, be sure to follow the on-screen instructions carefully! For a full list of `crm_shadow` options and commands, invoke it with the `--help` option. .Use sandbox to make multiple changes all at once, discard them, and verify real configuration is untouched ====== ---- shadow[test] # crm_failcount -r rsc_c001n01 -G scope=status name=fail-count-rsc_c001n01 value=0 shadow[test] # crm_standby --node c001n02 -v on shadow[test] # crm_standby --node c001n02 -G scope=nodes name=standby value=on shadow[test] # cibadmin --erase --force shadow[test] # cibadmin --query shadow[test] # crm_shadow --delete test --force Now type Ctrl-D to exit the crm_shadow shell shadow[test] # exit # crm_shadow --which No active shadow configuration defined # cibadmin -Q ---- ====== [[s-config-testing-changes]] == Testing Your Configuration Changes == We saw previously how to make a series of changes to a "shadow" copy of the configuration. Before loading the changes back into the cluster (e.g. `crm_shadow --commit mytest --force`), it is often advisable to simulate the effect of the changes with +crm_simulate+. For example: ---- # crm_simulate --live-check -VVVVV --save-graph tmp.graph --save-dotfile tmp.dot ---- This tool uses the same library as the live cluster to show what it would have done given the supplied input. Its output, in addition to a significant amount of logging, is stored in two files +tmp.graph+ and +tmp.dot+. Both files are representations of the same thing: the cluster's response to your changes. The graph file stores the complete transition from the existing cluster state to your desired new state, containing a list of all the actions, their parameters and their pre-requisites. Because the transition graph is not terribly easy to read, the tool also generates a Graphviz footnote:[Graph visualization software. See http://www.graphviz.org/ for details.] dot-file representing the same information. For information on the options supported by `crm_simulate`, use its `--help` option. .Interpreting the Graphviz output * Arrows indicate ordering dependencies * Dashed arrows indicate dependencies that are not present in the transition graph * Actions with a dashed border of any color do not form part of the transition graph * Actions with a green border form part of the transition graph * Actions with a red border are ones the cluster would like to execute but cannot run * Actions with a blue border are ones the cluster does not feel need to be executed * Actions with orange text are pseudo/pretend actions that the cluster uses to simplify the graph * Actions with black text are sent to the executor * Resource actions have text of the form pass:[rsc]_pass:[action]_pass:[interval] pass:[node] * Any action depending on an action with a red border will not be able to execute. * Loops are _really_ bad. Please report them to the development team. === Small Cluster Transition === image::images/Policy-Engine-small.png["An example transition graph as represented by Graphviz",width="16cm",height="6cm",align="center"] In the above example, it appears that a new node, *pcmk-2*, has come online and that the cluster is checking to make sure *rsc1*, *rsc2* and *rsc3* are not already running there (Indicated by the *rscN_monitor_0* entries). Once it did that, and assuming the resources were not active there, it would have liked to stop *rsc1* and *rsc2* on *pcmk-1* and move them to *pcmk-2*. However, there appears to be some problem and the cluster cannot or is not permitted to perform the stop actions which implies it also cannot perform the start actions. For some reason the cluster does not want to start *rsc3* anywhere. === Complex Cluster Transition === image::images/Policy-Engine-big.png["Another, slightly more complex, transition graph that you're not expected to be able to read",width="16cm",height="20cm",align="center"] -== Do I Need to Update the Configuration on All Cluster Nodes? == - -No. Any changes are immediately synchronized to the other active -members of the cluster. - -To reduce bandwidth, the cluster only broadcasts the incremental -updates that result from your changes and uses MD5 checksums to ensure -that each copy is completely consistent. - == Working with CIB Properties == Although these fields can be written to by the user, in most cases the cluster will overwrite any values specified by the user with the "correct" ones. To change the ones that can be specified by the user, for example +admin_epoch+, one should use: ---- # cibadmin --modify --xml-text '' ---- A complete set of CIB properties will look something like this: .Attributes set for a cib object ====== [source,XML] ------- ------- ====== == Querying and Setting Cluster Options == indexterm:[Querying,Cluster Option] indexterm:[Setting,Cluster Option] indexterm:[Cluster,Querying Options] indexterm:[Cluster,Setting Options] Cluster options can be queried and modified using the `crm_attribute` tool. To get the current value of +cluster-delay+, you can run: ---- # crm_attribute --query --name cluster-delay ---- which is more simply written as ---- # crm_attribute -G -n cluster-delay ---- If a value is found, you'll see a result like this: ---- # crm_attribute -G -n cluster-delay scope=crm_config name=cluster-delay value=60s ---- If no value is found, the tool will display an error: ---- # crm_attribute -G -n clusta-deway scope=crm_config name=clusta-deway value=(null) Error performing operation: No such device or address ---- To use a different value (for example, 30 seconds), simply run: ---- # crm_attribute --name cluster-delay --update 30s ---- To go back to the cluster's default value, you can delete the value, for example: ---- # crm_attribute --name cluster-delay --delete Deleted crm_config option: id=cib-bootstrap-options-cluster-delay name=cluster-delay ---- === When Options are Listed More Than Once === If you ever see something like the following, it means that the option you're modifying is present more than once. .Deleting an option that is listed twice ======= ------ # crm_attribute --name batch-limit --delete Multiple attributes match name=batch-limit in crm_config: Value: 50 (set=cib-bootstrap-options, id=cib-bootstrap-options-batch-limit) Value: 100 (set=custom, id=custom-batch-limit) Please choose from one of the matches above and supply the 'id' with --id ------- ======= In such cases, follow the on-screen instructions to perform the requested action. To determine which value is currently being used by the cluster, refer to the 'Rules' chapter of 'Pacemaker Explained'. [[s-remote-connection]] == Connecting from a Remote Machine == indexterm:[Cluster,Remote connection] indexterm:[Cluster,Remote administration] Provided Pacemaker is installed on a machine, it is possible to connect to the cluster even if the machine itself is not in the same cluster. To do this, one simply sets up a number of environment variables and runs the same commands as when working on a cluster node. .Environment Variables Used to Connect to Remote Instances of the CIB [width="95%",cols="1m,1,<3",options="header",align="center"] |========================================================= |Environment Variable |Default |Description |CIB_user |$USER |The user to connect as. Needs to be part of the +haclient+ group on the target host. indexterm:[Environment Variable,CIB_user] |CIB_passwd | |The user's password. Read from the command line if unset. indexterm:[Environment Variable,CIB_passwd] |CIB_server |localhost |The host to contact indexterm:[Environment Variable,CIB_server] |CIB_port | |The port on which to contact the server; required. indexterm:[Environment Variable,CIB_port] |CIB_encrypted |TRUE |Whether to encrypt network traffic indexterm:[Environment Variable,CIB_encrypted] |========================================================= So, if *c001n01* is an active cluster node and is listening on port 1234 for connections, and *someuser* is a member of the *haclient* group, then the following would prompt for *someuser*'s password and return the cluster's current configuration: ---- # export CIB_port=1234; export CIB_server=c001n01; export CIB_user=someuser; # cibadmin -Q ---- For security reasons, the cluster does not listen for remote connections by default. If you wish to allow remote access, you need to set the +remote-tls-port+ (encrypted) or +remote-clear-port+ (unencrypted) CIB properties (i.e., those kept in the +cib+ tag, like +num_updates+ and +epoch+). .Extra top-level CIB properties for remote access [width="95%",cols="1m,1,<3",options="header",align="center"] |========================================================= |Field |Default |Description |remote-tls-port |_none_ |Listen for encrypted remote connections on this port. indexterm:[remote-tls-port,Remote Connection Option] indexterm:[Remote Connection,Option,remote-tls-port] |remote-clear-port |_none_ |Listen for plaintext remote connections on this port. indexterm:[remote-clear-port,Remote Connection Option] indexterm:[Remote Connection,Option,remote-clear-port] |========================================================= [IMPORTANT] ==== The Pacemaker version on the administration host must be the same or greater than the version(s) on the cluster nodes. Otherwise, it may not have the schema files necessary to validate the CIB. ==== diff --git a/include/crm/common/mainloop.h b/include/crm/common/mainloop.h index c7b43ca532..5d423cb0d2 100644 --- a/include/crm/common/mainloop.h +++ b/include/crm/common/mainloop.h @@ -1,129 +1,134 @@ /* - * Copyright 2009-2018 the Pacemaker project contributors + * Copyright 2009-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef CRM_COMMON_MAINLOOP__H # define CRM_COMMON_MAINLOOP__H #ifdef __cplusplus extern "C" { #endif /** * \file * \brief Wrappers for and extensions to glib mainloop * \ingroup core */ +# include // sighandler_t # include enum mainloop_child_flags { /* don't kill pid group on timeout, only kill the pid */ mainloop_leave_pid_group = 0x01, }; typedef struct trigger_s crm_trigger_t; typedef struct mainloop_io_s mainloop_io_t; typedef struct mainloop_child_s mainloop_child_t; typedef struct mainloop_timer_s mainloop_timer_t; void mainloop_cleanup(void); crm_trigger_t *mainloop_add_trigger(int priority, int (*dispatch) (gpointer user_data), gpointer userdata); void mainloop_set_trigger(crm_trigger_t * source); void mainloop_trigger_complete(crm_trigger_t * trig); gboolean mainloop_destroy_trigger(crm_trigger_t * source); -gboolean crm_signal(int sig, void (*dispatch) (int sig)); +# ifndef HAVE_SIGHANDLER_T +typedef void (*sighandler_t)(int); +# endif + +sighandler_t crm_signal(int sig, sighandler_t dispatch); gboolean mainloop_add_signal(int sig, void (*dispatch) (int sig)); gboolean mainloop_destroy_signal(int sig); bool mainloop_timer_running(mainloop_timer_t *t); void mainloop_timer_start(mainloop_timer_t *t); void mainloop_timer_stop(mainloop_timer_t *t); guint mainloop_timer_set_period(mainloop_timer_t *t, guint period_ms); mainloop_timer_t *mainloop_timer_add(const char *name, guint period_ms, bool repeat, GSourceFunc cb, void *userdata); void mainloop_timer_del(mainloop_timer_t *t); # include # include struct ipc_client_callbacks { int (*dispatch) (const char *buffer, ssize_t length, gpointer userdata); void (*destroy) (gpointer); }; qb_ipcs_service_t *mainloop_add_ipc_server(const char *name, enum qb_ipc_type type, struct qb_ipcs_service_handlers *callbacks); void mainloop_del_ipc_server(qb_ipcs_service_t * server); mainloop_io_t *mainloop_add_ipc_client(const char *name, int priority, size_t max_size, void *userdata, struct ipc_client_callbacks *callbacks); void mainloop_del_ipc_client(mainloop_io_t * client); crm_ipc_t *mainloop_get_ipc_client(mainloop_io_t * client); struct mainloop_fd_callbacks { int (*dispatch) (gpointer userdata); void (*destroy) (gpointer userdata); }; mainloop_io_t *mainloop_add_fd(const char *name, int priority, int fd, void *userdata, struct mainloop_fd_callbacks *callbacks); void mainloop_del_fd(mainloop_io_t * client); /* * Create a new tracked process * To track a process group, use -pid */ void mainloop_child_add(pid_t pid, int timeout, const char *desc, void *userdata, void (*callback) (mainloop_child_t * p, pid_t pid, int core, int signo, int exitcode)); void mainloop_child_add_with_flags(pid_t pid, int timeout, const char *desc, void *userdata, enum mainloop_child_flags, void (*callback) (mainloop_child_t * p, pid_t pid, int core, int signo, int exitcode)); void *mainloop_child_userdata(mainloop_child_t * child); int mainloop_child_timeout(mainloop_child_t * child); const char *mainloop_child_name(mainloop_child_t * child); pid_t mainloop_child_pid(mainloop_child_t * child); void mainloop_clear_child_userdata(mainloop_child_t * child); gboolean mainloop_child_kill(pid_t pid); void pcmk_drain_main_loop(GMainLoop *mloop, guint timer_ms, bool (*check)(guint)); # define G_PRIORITY_MEDIUM (G_PRIORITY_HIGH/2) #ifdef __cplusplus } #endif #endif diff --git a/include/pcmki/pcmki_sched_allocate.h b/include/pcmki/pcmki_sched_allocate.h index a81e3008c8..eb5f3ec9b0 100644 --- a/include/pcmki/pcmki_sched_allocate.h +++ b/include/pcmki/pcmki_sched_allocate.h @@ -1,189 +1,189 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef SCHED_ALLOCATE__H # define SCHED_ALLOCATE__H # include # include # include # include # include # include struct resource_alloc_functions_s { GHashTable *(*merge_weights) (resource_t *, const char *, GHashTable *, const char *, float, enum pe_weights); node_t *(*allocate) (resource_t *, node_t *, pe_working_set_t *); void (*create_actions) (resource_t *, pe_working_set_t *); gboolean(*create_probe) (resource_t *, node_t *, action_t *, gboolean, pe_working_set_t *); void (*internal_constraints) (resource_t *, pe_working_set_t *); void (*rsc_colocation_lh) (pe_resource_t *, pe_resource_t *, rsc_colocation_t *, pe_working_set_t *); void (*rsc_colocation_rh) (pe_resource_t *, pe_resource_t *, rsc_colocation_t *, pe_working_set_t *); void (*rsc_location) (pe_resource_t *, pe__location_t *); enum pe_action_flags (*action_flags) (action_t *, node_t *); enum pe_graph_flags (*update_actions) (pe_action_t *, pe_action_t *, pe_node_t *, enum pe_action_flags, enum pe_action_flags, enum pe_ordering, pe_working_set_t *data_set); void (*expand) (resource_t *, pe_working_set_t *); void (*append_meta) (resource_t * rsc, xmlNode * xml); }; extern GHashTable *rsc_merge_weights(resource_t * rsc, const char *rhs, GHashTable * nodes, const char *attr, float factor, enum pe_weights flags); extern GHashTable *clone_merge_weights(resource_t * rsc, const char *rhs, GHashTable * nodes, const char *attr, float factor, enum pe_weights flags); GHashTable *pcmk__bundle_merge_weights(pe_resource_t *rsc, const char *rhs, GHashTable *nodes, const char *attr, float factor, enum pe_weights flags); extern GHashTable *native_merge_weights(resource_t * rsc, const char *rhs, GHashTable * nodes, const char *attr, float factor, enum pe_weights flags); extern GHashTable *group_merge_weights(resource_t * rsc, const char *rhs, GHashTable * nodes, const char *attr, float factor, enum pe_weights flags); extern node_t *native_color(resource_t * rsc, node_t * preferred, pe_working_set_t * data_set); extern void native_create_actions(resource_t * rsc, pe_working_set_t * data_set); extern void native_internal_constraints(resource_t * rsc, pe_working_set_t * data_set); void native_rsc_colocation_lh(pe_resource_t *lh_rsc, pe_resource_t *rh_rsc, rsc_colocation_t *constraint, pe_working_set_t *data_set); void native_rsc_colocation_rh(pe_resource_t *lh_rsc, pe_resource_t *rh_rsc, rsc_colocation_t *constraint, pe_working_set_t *data_set); extern void rsc_ticket_constraint(resource_t * lh_rsc, rsc_ticket_t * rsc_ticket, pe_working_set_t * data_set); extern enum pe_action_flags native_action_flags(action_t * action, node_t * node); void native_rsc_location(pe_resource_t *rsc, pe__location_t *constraint); extern void native_expand(resource_t * rsc, pe_working_set_t * data_set); extern gboolean native_create_probe(resource_t * rsc, node_t * node, action_t * complete, gboolean force, pe_working_set_t * data_set); extern void native_append_meta(resource_t * rsc, xmlNode * xml); extern node_t *group_color(resource_t * rsc, node_t * preferred, pe_working_set_t * data_set); extern void group_create_actions(resource_t * rsc, pe_working_set_t * data_set); extern void group_internal_constraints(resource_t * rsc, pe_working_set_t * data_set); void group_rsc_colocation_lh(pe_resource_t *lh_rsc, pe_resource_t *rh_rsc, rsc_colocation_t *constraint, pe_working_set_t *data_set); void group_rsc_colocation_rh(pe_resource_t *lh_rsc, pe_resource_t *rh_rsc, rsc_colocation_t *constraint, pe_working_set_t *data_set); extern enum pe_action_flags group_action_flags(action_t * action, node_t * node); void group_rsc_location(pe_resource_t *rsc, pe__location_t *constraint); extern void group_expand(resource_t * rsc, pe_working_set_t * data_set); extern void group_append_meta(resource_t * rsc, xmlNode * xml); pe_node_t *pcmk__bundle_color(pe_resource_t *rsc, pe_node_t *preferred, pe_working_set_t *data_set); void pcmk__bundle_create_actions(pe_resource_t *rsc, pe_working_set_t *data_set); gboolean pcmk__bundle_create_probe(pe_resource_t *rsc, pe_node_t *node, pe_action_t *complete, gboolean force, pe_working_set_t *data_set); void pcmk__bundle_internal_constraints(pe_resource_t *rsc, pe_working_set_t *data_set); void pcmk__bundle_rsc_colocation_lh(pe_resource_t *lh_rsc, pe_resource_t *rh_rsc, rsc_colocation_t *constraint, pe_working_set_t *data_set); void pcmk__bundle_rsc_colocation_rh(pe_resource_t *lh_rsc, pe_resource_t *rh_rsc, rsc_colocation_t *constraint, pe_working_set_t *data_set); void pcmk__bundle_rsc_location(pe_resource_t *rsc, pe__location_t *constraint); enum pe_action_flags pcmk__bundle_action_flags(pe_action_t *action, pe_node_t *node); void pcmk__bundle_expand(pe_resource_t *rsc, pe_working_set_t *data_set); void pcmk__bundle_append_meta(pe_resource_t *rsc, xmlNode *xml); extern node_t *clone_color(resource_t * rsc, node_t * preferred, pe_working_set_t * data_set); extern void clone_create_actions(resource_t * rsc, pe_working_set_t * data_set); extern void clone_internal_constraints(resource_t * rsc, pe_working_set_t * data_set); void clone_rsc_colocation_lh(pe_resource_t *lh_rsc, pe_resource_t *rh_rsc, rsc_colocation_t *constraint, pe_working_set_t *data_set); void clone_rsc_colocation_rh(pe_resource_t *lh_rsc, pe_resource_t *rh_rsc, rsc_colocation_t *constraint, pe_working_set_t *data_set); void clone_rsc_location(pe_resource_t *rsc, pe__location_t *constraint); extern enum pe_action_flags clone_action_flags(action_t * action, node_t * node); extern void clone_expand(resource_t * rsc, pe_working_set_t * data_set); extern gboolean clone_create_probe(resource_t * rsc, node_t * node, action_t * complete, gboolean force, pe_working_set_t * data_set); extern void clone_append_meta(resource_t * rsc, xmlNode * xml); void apply_master_prefs(resource_t *rsc); node_t *color_promotable(resource_t *rsc, pe_working_set_t *data_set); void create_promotable_actions(resource_t *rsc, pe_working_set_t *data_set); void promote_demote_constraints(resource_t *rsc, pe_working_set_t *data_set); void promotable_constraints(resource_t *rsc, pe_working_set_t *data_set); void promotable_colocation_rh(resource_t *lh_rsc, resource_t *rh_rsc, rsc_colocation_t *constraint, pe_working_set_t *data_set); /* extern resource_object_functions_t resource_variants[]; */ extern resource_alloc_functions_t resource_class_alloc_functions[]; gboolean is_active(pe__location_t *cons); extern gboolean unpack_rsc_order(xmlNode * xml_obj, pe_working_set_t * data_set); extern gboolean unpack_rsc_colocation(xmlNode * xml_obj, pe_working_set_t * data_set); extern gboolean unpack_location(xmlNode * xml_obj, pe_working_set_t * data_set); extern gboolean unpack_rsc_ticket(xmlNode * xml_obj, pe_working_set_t * data_set); void LogNodeActions(pe_working_set_t * data_set, gboolean terminal); void LogActions(resource_t * rsc, pe_working_set_t * data_set, gboolean terminal); void pcmk__bundle_log_actions(pe_resource_t *rsc, pe_working_set_t *data_set, gboolean terminal); extern void rsc_stonith_ordering(resource_t * rsc, action_t * stonith_op, pe_working_set_t * data_set); enum pe_graph_flags native_update_actions(pe_action_t *first, pe_action_t *then, pe_node_t *node, enum pe_action_flags flags, enum pe_action_flags filter, enum pe_ordering type, pe_working_set_t *data_set); enum pe_graph_flags group_update_actions(pe_action_t *first, pe_action_t *then, pe_node_t *node, enum pe_action_flags flags, enum pe_action_flags filter, enum pe_ordering type, pe_working_set_t *data_set); enum pe_graph_flags pcmk__multi_update_actions(pe_action_t *first, pe_action_t *then, pe_node_t *node, enum pe_action_flags flags, enum pe_action_flags filter, enum pe_ordering type, pe_working_set_t *data_set); gboolean update_action_flags(action_t * action, enum pe_action_flags flags, const char *source, int line); gboolean update_action(pe_action_t *action, pe_working_set_t *data_set); void complex_set_cmds(resource_t * rsc); - +void pcmk__log_transition_summary(const char *filename); void clone_create_pseudo_actions( resource_t * rsc, GListPtr children, notify_data_t **start_notify, notify_data_t **stop_notify, pe_working_set_t * data_set); #endif diff --git a/include/pcmki/pcmki_scheduler.h b/include/pcmki/pcmki_scheduler.h index 0d49f21c04..0c94458139 100644 --- a/include/pcmki/pcmki_scheduler.h +++ b/include/pcmki/pcmki_scheduler.h @@ -1,112 +1,114 @@ /* * Copyright 2014-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef PENGINE__H # define PENGINE__H typedef struct rsc_colocation_s rsc_colocation_t; typedef struct rsc_ticket_s rsc_ticket_t; typedef struct lrm_agent_s lrm_agent_t; # include # include # include # include # include # include # include enum pe_stop_fail { pesf_block, pesf_stonith, pesf_ignore }; enum pe_weights { pe_weights_none = 0x0, pe_weights_init = 0x1, pe_weights_forward = 0x4, pe_weights_positive = 0x8, pe_weights_rollback = 0x10, }; struct rsc_colocation_s { const char *id; const char *node_attribute; resource_t *rsc_lh; resource_t *rsc_rh; int role_lh; int role_rh; int score; }; enum loss_ticket_policy_e { loss_ticket_stop, loss_ticket_demote, loss_ticket_fence, loss_ticket_freeze }; struct rsc_ticket_s { const char *id; resource_t *rsc_lh; ticket_t *ticket; enum loss_ticket_policy_e loss_policy; int role_lh; }; extern gboolean stage0(pe_working_set_t * data_set); extern gboolean probe_resources(pe_working_set_t * data_set); extern gboolean stage2(pe_working_set_t * data_set); extern gboolean stage3(pe_working_set_t * data_set); extern gboolean stage4(pe_working_set_t * data_set); extern gboolean stage5(pe_working_set_t * data_set); extern gboolean stage6(pe_working_set_t * data_set); extern gboolean stage7(pe_working_set_t * data_set); extern gboolean stage8(pe_working_set_t * data_set); extern gboolean summary(GListPtr resources); extern gboolean unpack_constraints(xmlNode * xml_constraints, pe_working_set_t * data_set); extern gboolean update_action_states(GListPtr actions); extern gboolean shutdown_constraints(node_t * node, action_t * shutdown_op, pe_working_set_t * data_set); extern gboolean stonith_constraints(node_t * node, action_t * stonith_op, pe_working_set_t * data_set); extern int custom_action_order(resource_t * lh_rsc, char *lh_task, action_t * lh_action, resource_t * rh_rsc, char *rh_task, action_t * rh_action, enum pe_ordering type, pe_working_set_t * data_set); extern int new_rsc_order(resource_t * lh_rsc, const char *lh_task, resource_t * rh_rsc, const char *rh_task, enum pe_ordering type, pe_working_set_t * data_set); # define order_start_start(rsc1,rsc2, type) \ new_rsc_order(rsc1, CRMD_ACTION_START, rsc2, CRMD_ACTION_START, type, data_set) # define order_stop_stop(rsc1, rsc2, type) \ new_rsc_order(rsc1, CRMD_ACTION_STOP, rsc2, CRMD_ACTION_STOP, type, data_set) extern void graph_element_from_action(action_t * action, pe_working_set_t * data_set); extern void add_maintenance_update(pe_working_set_t *data_set); +xmlNode *pcmk__schedule_actions(pe_working_set_t *data_set, xmlNode *xml_input, + crm_time_t *now); extern gboolean show_scores; extern int scores_log_level; extern gboolean show_utilization; extern int utilization_log_level; extern const char *transition_idle_timeout; #endif diff --git a/lib/common/mainloop.c b/lib/common/mainloop.c index 4d650791e4..8b66f18ece 100644 --- a/lib/common/mainloop.c +++ b/lib/common/mainloop.c @@ -1,1339 +1,1341 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #ifndef _GNU_SOURCE # define _GNU_SOURCE #endif #include #include #include #include #include #include #include #include #include struct mainloop_child_s { pid_t pid; char *desc; unsigned timerid; gboolean timeout; void *privatedata; enum mainloop_child_flags flags; /* Called when a process dies */ void (*callback) (mainloop_child_t * p, pid_t pid, int core, int signo, int exitcode); }; struct trigger_s { GSource source; gboolean running; gboolean trigger; void *user_data; guint id; }; static gboolean crm_trigger_prepare(GSource * source, gint * timeout) { crm_trigger_t *trig = (crm_trigger_t *) source; /* cluster-glue's FD and IPC related sources make use of * g_source_add_poll() but do not set a timeout in their prepare * functions * * This means mainloop's poll() will block until an event for one * of these sources occurs - any /other/ type of source, such as * this one or g_idle_*, that doesn't use g_source_add_poll() is * S-O-L and won't be processed until there is something fd-based * happens. * * Luckily the timeout we can set here affects all sources and * puts an upper limit on how long poll() can take. * * So unconditionally set a small-ish timeout, not too small that * we're in constant motion, which will act as an upper bound on * how long the signal handling might be delayed for. */ *timeout = 500; /* Timeout in ms */ return trig->trigger; } static gboolean crm_trigger_check(GSource * source) { crm_trigger_t *trig = (crm_trigger_t *) source; return trig->trigger; } static gboolean crm_trigger_dispatch(GSource * source, GSourceFunc callback, gpointer userdata) { int rc = TRUE; crm_trigger_t *trig = (crm_trigger_t *) source; if (trig->running) { /* Wait until the existing job is complete before starting the next one */ return TRUE; } trig->trigger = FALSE; if (callback) { rc = callback(trig->user_data); if (rc < 0) { crm_trace("Trigger handler %p not yet complete", trig); trig->running = TRUE; rc = TRUE; } } return rc; } static void crm_trigger_finalize(GSource * source) { crm_trace("Trigger %p destroyed", source); } #if 0 struct _GSourceCopy { gpointer callback_data; GSourceCallbackFuncs *callback_funcs; const GSourceFuncs *source_funcs; guint ref_count; GMainContext *context; gint priority; guint flags; guint source_id; GSList *poll_fds; GSource *prev; GSource *next; char *name; void *priv; }; static int g_source_refcount(GSource * source) { /* Duplicating the contents of private header files is a necessary evil */ if (source) { struct _GSourceCopy *evil = (struct _GSourceCopy*)source; return evil->ref_count; } return 0; } #else static int g_source_refcount(GSource * source) { return 0; } #endif static GSourceFuncs crm_trigger_funcs = { crm_trigger_prepare, crm_trigger_check, crm_trigger_dispatch, crm_trigger_finalize, }; static crm_trigger_t * mainloop_setup_trigger(GSource * source, int priority, int (*dispatch) (gpointer user_data), gpointer userdata) { crm_trigger_t *trigger = NULL; trigger = (crm_trigger_t *) source; trigger->id = 0; trigger->trigger = FALSE; trigger->user_data = userdata; if (dispatch) { g_source_set_callback(source, dispatch, trigger, NULL); } g_source_set_priority(source, priority); g_source_set_can_recurse(source, FALSE); crm_trace("Setup %p with ref-count=%u", source, g_source_refcount(source)); trigger->id = g_source_attach(source, NULL); crm_trace("Attached %p with ref-count=%u", source, g_source_refcount(source)); return trigger; } void mainloop_trigger_complete(crm_trigger_t * trig) { crm_trace("Trigger handler %p complete", trig); trig->running = FALSE; } /* If dispatch returns: * -1: Job running but not complete * 0: Remove the trigger from mainloop * 1: Leave the trigger in mainloop */ crm_trigger_t * mainloop_add_trigger(int priority, int (*dispatch) (gpointer user_data), gpointer userdata) { GSource *source = NULL; CRM_ASSERT(sizeof(crm_trigger_t) > sizeof(GSource)); source = g_source_new(&crm_trigger_funcs, sizeof(crm_trigger_t)); CRM_ASSERT(source != NULL); return mainloop_setup_trigger(source, priority, dispatch, userdata); } void mainloop_set_trigger(crm_trigger_t * source) { if(source) { source->trigger = TRUE; } } gboolean mainloop_destroy_trigger(crm_trigger_t * source) { GSource *gs = NULL; if(source == NULL) { return TRUE; } gs = (GSource *)source; if(g_source_refcount(gs) > 2) { crm_info("Trigger %p is still referenced %u times", gs, g_source_refcount(gs)); } g_source_destroy(gs); /* Remove from mainloop, ref_count-- */ g_source_unref(gs); /* The caller no longer carries a reference to source * * At this point the source should be free'd, * unless we're currently processing said * source, in which case mainloop holds an * additional reference and it will be free'd * once our processing completes */ return TRUE; } // Define a custom glib source for signal handling // Data structure for custom glib source typedef struct signal_s { crm_trigger_t trigger; // trigger that invoked source (must be first) void (*handler) (int sig); // signal handler int signal; // signal that was received } crm_signal_t; // Table to associate signal handlers with signal numbers static crm_signal_t *crm_signals[NSIG]; /*! * \internal * \brief Dispatch an event from custom glib source for signals * * Given an signal event, clear the event trigger and call any registered * signal handler. * * \param[in] source glib source that triggered this dispatch * \param[in] callback (ignored) * \param[in] userdata (ignored) */ static gboolean crm_signal_dispatch(GSource * source, GSourceFunc callback, gpointer userdata) { crm_signal_t *sig = (crm_signal_t *) source; if(sig->signal != SIGCHLD) { crm_notice("Caught '%s' signal "CRM_XS" %d (%s handler)", strsignal(sig->signal), sig->signal, (sig->handler? "invoking" : "no")); } sig->trigger.trigger = FALSE; if (sig->handler) { sig->handler(sig->signal); } return TRUE; } /*! * \internal * \brief Handle a signal by setting a trigger for signal source * * \param[in] sig Signal number that was received * * \note This is the true signal handler for the mainloop signal source, and * must be async-safe. */ static void mainloop_signal_handler(int sig) { if (sig > 0 && sig < NSIG && crm_signals[sig] != NULL) { mainloop_set_trigger((crm_trigger_t *) crm_signals[sig]); } } // Functions implementing our custom glib source for signal handling static GSourceFuncs crm_signal_funcs = { crm_trigger_prepare, crm_trigger_check, crm_signal_dispatch, crm_trigger_finalize, }; /*! * \internal * \brief Set a true signal handler * * signal()-like interface to sigaction() * * \param[in] sig Signal number to register handler for * \param[in] dispatch Signal handler * + * \return The previous value of the signal handler, or SIG_ERR on error * \note The dispatch function must be async-safe. */ -gboolean -crm_signal(int sig, void (*dispatch) (int sig)) +sighandler_t +crm_signal(int sig, sighandler_t dispatch) { sigset_t mask; struct sigaction sa; struct sigaction old; if (sigemptyset(&mask) < 0) { - crm_perror(LOG_ERR, "Call to sigemptyset failed"); - return FALSE; + crm_err("Could not set handler for signal %d: %s", + sig, pcmk_strerror(errno)); + return SIG_ERR; } memset(&sa, 0, sizeof(struct sigaction)); sa.sa_handler = dispatch; sa.sa_flags = SA_RESTART; sa.sa_mask = mask; if (sigaction(sig, &sa, &old) < 0) { - crm_perror(LOG_ERR, "Could not install signal handler for signal %d", sig); - return FALSE; + crm_err("Could not set handler for signal %d: %s", + sig, pcmk_strerror(errno)); + return SIG_ERR; } - - return TRUE; + return old.sa_handler; } static void mainloop_destroy_signal_entry(int sig) { crm_signal_t *tmp = crm_signals[sig]; crm_signals[sig] = NULL; crm_trace("Destroying signal %d", sig); mainloop_destroy_trigger((crm_trigger_t *) tmp); } /*! * \internal * \brief Add a signal handler to a mainloop * * \param[in] sig Signal number to handle * \param[in] dispatch Signal handler function * * \note The true signal handler merely sets a mainloop trigger to call this * dispatch function via the mainloop. Therefore, the dispatch function * does not need to be async-safe. */ gboolean mainloop_add_signal(int sig, void (*dispatch) (int sig)) { GSource *source = NULL; int priority = G_PRIORITY_HIGH - 1; if (sig == SIGTERM) { /* TERM is higher priority than other signals, * signals are higher priority than other ipc. * Yes, minus: smaller is "higher" */ priority--; } if (sig >= NSIG || sig < 0) { crm_err("Signal %d is out of range", sig); return FALSE; } else if (crm_signals[sig] != NULL && crm_signals[sig]->handler == dispatch) { crm_trace("Signal handler for %d is already installed", sig); return TRUE; } else if (crm_signals[sig] != NULL) { crm_err("Different signal handler for %d is already installed", sig); return FALSE; } CRM_ASSERT(sizeof(crm_signal_t) > sizeof(GSource)); source = g_source_new(&crm_signal_funcs, sizeof(crm_signal_t)); crm_signals[sig] = (crm_signal_t *) mainloop_setup_trigger(source, priority, NULL, NULL); CRM_ASSERT(crm_signals[sig] != NULL); crm_signals[sig]->handler = dispatch; crm_signals[sig]->signal = sig; - if (crm_signal(sig, mainloop_signal_handler) == FALSE) { + if (crm_signal(sig, mainloop_signal_handler) == SIG_ERR) { mainloop_destroy_signal_entry(sig); return FALSE; } #if 0 /* If we want signals to interrupt mainloop's poll(), instead of waiting for * the timeout, then we should call siginterrupt() below * * For now, just enforce a low timeout */ if (siginterrupt(sig, 1) < 0) { crm_perror(LOG_INFO, "Could not enable system call interruptions for signal %d", sig); } #endif return TRUE; } gboolean mainloop_destroy_signal(int sig) { if (sig >= NSIG || sig < 0) { crm_err("Signal %d is out of range", sig); return FALSE; - } else if (crm_signal(sig, NULL) == FALSE) { + } else if (crm_signal(sig, NULL) == SIG_ERR) { crm_perror(LOG_ERR, "Could not uninstall signal handler for signal %d", sig); return FALSE; } else if (crm_signals[sig] == NULL) { return TRUE; } mainloop_destroy_signal_entry(sig); return TRUE; } static qb_array_t *gio_map = NULL; void mainloop_cleanup(void) { if (gio_map) { qb_array_free(gio_map); } for (int sig = 0; sig < NSIG; ++sig) { mainloop_destroy_signal_entry(sig); } } /* * libqb... */ struct gio_to_qb_poll { int32_t is_used; guint source; int32_t events; void *data; qb_ipcs_dispatch_fn_t fn; enum qb_loop_priority p; }; static gboolean gio_read_socket(GIOChannel * gio, GIOCondition condition, gpointer data) { struct gio_to_qb_poll *adaptor = (struct gio_to_qb_poll *)data; gint fd = g_io_channel_unix_get_fd(gio); crm_trace("%p.%d %d", data, fd, condition); /* if this assert get's hit, then there is a race condition between * when we destroy a fd and when mainloop actually gives it up */ CRM_ASSERT(adaptor->is_used > 0); return (adaptor->fn(fd, condition, adaptor->data) == 0); } static void gio_poll_destroy(gpointer data) { struct gio_to_qb_poll *adaptor = (struct gio_to_qb_poll *)data; adaptor->is_used--; CRM_ASSERT(adaptor->is_used >= 0); if (adaptor->is_used == 0) { crm_trace("Marking adaptor %p unused", adaptor); adaptor->source = 0; } } static int32_t gio_poll_dispatch_update(enum qb_loop_priority p, int32_t fd, int32_t evts, void *data, qb_ipcs_dispatch_fn_t fn, int32_t add) { struct gio_to_qb_poll *adaptor; GIOChannel *channel; int32_t res = 0; res = qb_array_index(gio_map, fd, (void **)&adaptor); if (res < 0) { crm_err("Array lookup failed for fd=%d: %d", fd, res); return res; } crm_trace("Adding fd=%d to mainloop as adaptor %p", fd, adaptor); if (add && adaptor->source) { crm_err("Adaptor for descriptor %d is still in-use", fd); return -EEXIST; } if (!add && !adaptor->is_used) { crm_err("Adaptor for descriptor %d is not in-use", fd); return -ENOENT; } /* channel is created with ref_count = 1 */ channel = g_io_channel_unix_new(fd); if (!channel) { crm_err("No memory left to add fd=%d", fd); return -ENOMEM; } if (adaptor->source) { g_source_remove(adaptor->source); adaptor->source = 0; } /* Because unlike the poll() API, glib doesn't tell us about HUPs by default */ evts |= (G_IO_HUP | G_IO_NVAL | G_IO_ERR); adaptor->fn = fn; adaptor->events = evts; adaptor->data = data; adaptor->p = p; adaptor->is_used++; adaptor->source = g_io_add_watch_full(channel, G_PRIORITY_DEFAULT, evts, gio_read_socket, adaptor, gio_poll_destroy); /* Now that mainloop now holds a reference to channel, * thanks to g_io_add_watch_full(), drop ours from g_io_channel_unix_new(). * * This means that channel will be free'd by: * g_main_context_dispatch() * -> g_source_destroy_internal() * -> g_source_callback_unref() * shortly after gio_poll_destroy() completes */ g_io_channel_unref(channel); crm_trace("Added to mainloop with gsource id=%d", adaptor->source); if (adaptor->source > 0) { return 0; } return -EINVAL; } static int32_t gio_poll_dispatch_add(enum qb_loop_priority p, int32_t fd, int32_t evts, void *data, qb_ipcs_dispatch_fn_t fn) { return gio_poll_dispatch_update(p, fd, evts, data, fn, QB_TRUE); } static int32_t gio_poll_dispatch_mod(enum qb_loop_priority p, int32_t fd, int32_t evts, void *data, qb_ipcs_dispatch_fn_t fn) { return gio_poll_dispatch_update(p, fd, evts, data, fn, QB_FALSE); } static int32_t gio_poll_dispatch_del(int32_t fd) { struct gio_to_qb_poll *adaptor; crm_trace("Looking for fd=%d", fd); if (qb_array_index(gio_map, fd, (void **)&adaptor) == 0) { if (adaptor->source) { g_source_remove(adaptor->source); adaptor->source = 0; } } return 0; } struct qb_ipcs_poll_handlers gio_poll_funcs = { .job_add = NULL, .dispatch_add = gio_poll_dispatch_add, .dispatch_mod = gio_poll_dispatch_mod, .dispatch_del = gio_poll_dispatch_del, }; static enum qb_ipc_type pick_ipc_type(enum qb_ipc_type requested) { const char *env = getenv("PCMK_ipc_type"); if (env && strcmp("shared-mem", env) == 0) { return QB_IPC_SHM; } else if (env && strcmp("socket", env) == 0) { return QB_IPC_SOCKET; } else if (env && strcmp("posix", env) == 0) { return QB_IPC_POSIX_MQ; } else if (env && strcmp("sysv", env) == 0) { return QB_IPC_SYSV_MQ; } else if (requested == QB_IPC_NATIVE) { /* We prefer shared memory because the server never blocks on * send. If part of a message fits into the socket, libqb * needs to block until the remainder can be sent also. * Otherwise the client will wait forever for the remaining * bytes. */ return QB_IPC_SHM; } return requested; } qb_ipcs_service_t * mainloop_add_ipc_server(const char *name, enum qb_ipc_type type, struct qb_ipcs_service_handlers * callbacks) { int rc = 0; qb_ipcs_service_t *server = NULL; if (gio_map == NULL) { gio_map = qb_array_create_2(64, sizeof(struct gio_to_qb_poll), 1); } crm_client_init(); server = qb_ipcs_create(name, 0, pick_ipc_type(type), callbacks); #ifdef HAVE_IPCS_GET_BUFFER_SIZE /* All clients should use at least ipc_buffer_max as their buffer size */ qb_ipcs_enforce_buffer_size(server, crm_ipc_default_buffer_size()); #endif qb_ipcs_poll_handlers_set(server, &gio_poll_funcs); rc = qb_ipcs_run(server); if (rc < 0) { crm_err("Could not start %s IPC server: %s (%d)", name, pcmk_strerror(rc), rc); return NULL; } return server; } void mainloop_del_ipc_server(qb_ipcs_service_t * server) { if (server) { qb_ipcs_destroy(server); } } struct mainloop_io_s { char *name; void *userdata; int fd; guint source; crm_ipc_t *ipc; GIOChannel *channel; int (*dispatch_fn_ipc) (const char *buffer, ssize_t length, gpointer userdata); int (*dispatch_fn_io) (gpointer userdata); void (*destroy_fn) (gpointer userdata); }; static gboolean mainloop_gio_callback(GIOChannel * gio, GIOCondition condition, gpointer data) { gboolean keep = TRUE; mainloop_io_t *client = data; CRM_ASSERT(client->fd == g_io_channel_unix_get_fd(gio)); if (condition & G_IO_IN) { if (client->ipc) { long rc = 0; int max = 10; do { rc = crm_ipc_read(client->ipc); if (rc <= 0) { crm_trace("Message acquisition from %s[%p] failed: %s (%ld)", client->name, client, pcmk_strerror(rc), rc); } else if (client->dispatch_fn_ipc) { const char *buffer = crm_ipc_buffer(client->ipc); crm_trace("New message from %s[%p] = %ld (I/O condition=%d)", client->name, client, rc, condition); if (client->dispatch_fn_ipc(buffer, rc, client->userdata) < 0) { crm_trace("Connection to %s no longer required", client->name); keep = FALSE; } } } while (keep && rc > 0 && --max > 0); } else { crm_trace("New message from %s[%p] %u", client->name, client, condition); if (client->dispatch_fn_io) { if (client->dispatch_fn_io(client->userdata) < 0) { crm_trace("Connection to %s no longer required", client->name); keep = FALSE; } } } } if (client->ipc && crm_ipc_connected(client->ipc) == FALSE) { crm_err("Connection to %s closed " CRM_XS "client=%p condition=%d", client->name, client, condition); keep = FALSE; } else if (condition & (G_IO_HUP | G_IO_NVAL | G_IO_ERR)) { crm_trace("The connection %s[%p] has been closed (I/O condition=%d)", client->name, client, condition); keep = FALSE; } else if ((condition & G_IO_IN) == 0) { /* #define GLIB_SYSDEF_POLLIN =1 #define GLIB_SYSDEF_POLLPRI =2 #define GLIB_SYSDEF_POLLOUT =4 #define GLIB_SYSDEF_POLLERR =8 #define GLIB_SYSDEF_POLLHUP =16 #define GLIB_SYSDEF_POLLNVAL =32 typedef enum { G_IO_IN GLIB_SYSDEF_POLLIN, G_IO_OUT GLIB_SYSDEF_POLLOUT, G_IO_PRI GLIB_SYSDEF_POLLPRI, G_IO_ERR GLIB_SYSDEF_POLLERR, G_IO_HUP GLIB_SYSDEF_POLLHUP, G_IO_NVAL GLIB_SYSDEF_POLLNVAL } GIOCondition; A bitwise combination representing a condition to watch for on an event source. G_IO_IN There is data to read. G_IO_OUT Data can be written (without blocking). G_IO_PRI There is urgent data to read. G_IO_ERR Error condition. G_IO_HUP Hung up (the connection has been broken, usually for pipes and sockets). G_IO_NVAL Invalid request. The file descriptor is not open. */ crm_err("Strange condition: %d", condition); } /* keep == FALSE results in mainloop_gio_destroy() being called * just before the source is removed from mainloop */ return keep; } static void mainloop_gio_destroy(gpointer c) { mainloop_io_t *client = c; char *c_name = strdup(client->name); /* client->source is valid but about to be destroyed (ref_count == 0) in gmain.c * client->channel will still have ref_count > 0... should be == 1 */ crm_trace("Destroying client %s[%p]", c_name, c); if (client->ipc) { crm_ipc_close(client->ipc); } if (client->destroy_fn) { void (*destroy_fn) (gpointer userdata) = client->destroy_fn; client->destroy_fn = NULL; destroy_fn(client->userdata); } if (client->ipc) { crm_ipc_t *ipc = client->ipc; client->ipc = NULL; crm_ipc_destroy(ipc); } crm_trace("Destroyed client %s[%p]", c_name, c); free(client->name); client->name = NULL; free(client); free(c_name); } mainloop_io_t * mainloop_add_ipc_client(const char *name, int priority, size_t max_size, void *userdata, struct ipc_client_callbacks *callbacks) { mainloop_io_t *client = NULL; crm_ipc_t *conn = crm_ipc_new(name, max_size); if (conn && crm_ipc_connect(conn)) { int32_t fd = crm_ipc_get_fd(conn); client = mainloop_add_fd(name, priority, fd, userdata, NULL); } if (client == NULL) { crm_perror(LOG_TRACE, "Connection to %s failed", name); if (conn) { crm_ipc_close(conn); crm_ipc_destroy(conn); } return NULL; } client->ipc = conn; client->destroy_fn = callbacks->destroy; client->dispatch_fn_ipc = callbacks->dispatch; return client; } void mainloop_del_ipc_client(mainloop_io_t * client) { mainloop_del_fd(client); } crm_ipc_t * mainloop_get_ipc_client(mainloop_io_t * client) { if (client) { return client->ipc; } return NULL; } mainloop_io_t * mainloop_add_fd(const char *name, int priority, int fd, void *userdata, struct mainloop_fd_callbacks * callbacks) { mainloop_io_t *client = NULL; if (fd >= 0) { client = calloc(1, sizeof(mainloop_io_t)); if (client == NULL) { return NULL; } client->name = strdup(name); client->userdata = userdata; if (callbacks) { client->destroy_fn = callbacks->destroy; client->dispatch_fn_io = callbacks->dispatch; } client->fd = fd; client->channel = g_io_channel_unix_new(fd); client->source = g_io_add_watch_full(client->channel, priority, (G_IO_IN | G_IO_HUP | G_IO_NVAL | G_IO_ERR), mainloop_gio_callback, client, mainloop_gio_destroy); /* Now that mainloop now holds a reference to channel, * thanks to g_io_add_watch_full(), drop ours from g_io_channel_unix_new(). * * This means that channel will be free'd by: * g_main_context_dispatch() or g_source_remove() * -> g_source_destroy_internal() * -> g_source_callback_unref() * shortly after mainloop_gio_destroy() completes */ g_io_channel_unref(client->channel); crm_trace("Added connection %d for %s[%p].%d", client->source, client->name, client, fd); } else { errno = EINVAL; } return client; } void mainloop_del_fd(mainloop_io_t * client) { if (client != NULL) { crm_trace("Removing client %s[%p]", client->name, client); if (client->source) { /* Results in mainloop_gio_destroy() being called just * before the source is removed from mainloop */ g_source_remove(client->source); } } } static GListPtr child_list = NULL; pid_t mainloop_child_pid(mainloop_child_t * child) { return child->pid; } const char * mainloop_child_name(mainloop_child_t * child) { return child->desc; } int mainloop_child_timeout(mainloop_child_t * child) { return child->timeout; } void * mainloop_child_userdata(mainloop_child_t * child) { return child->privatedata; } void mainloop_clear_child_userdata(mainloop_child_t * child) { child->privatedata = NULL; } /* good function name */ static void child_free(mainloop_child_t *child) { if (child->timerid != 0) { crm_trace("Removing timer %d", child->timerid); g_source_remove(child->timerid); child->timerid = 0; } free(child->desc); free(child); } /* terrible function name */ static int child_kill_helper(mainloop_child_t *child) { int rc; if (child->flags & mainloop_leave_pid_group) { crm_debug("Kill pid %d only. leave group intact.", child->pid); rc = kill(child->pid, SIGKILL); } else { crm_debug("Kill pid %d's group", child->pid); rc = kill(-child->pid, SIGKILL); } if (rc < 0) { if (errno != ESRCH) { crm_perror(LOG_ERR, "kill(%d, KILL) failed", child->pid); } return -errno; } return 0; } static gboolean child_timeout_callback(gpointer p) { mainloop_child_t *child = p; int rc = 0; child->timerid = 0; if (child->timeout) { crm_crit("%s process (PID %d) will not die!", child->desc, (int)child->pid); return FALSE; } rc = child_kill_helper(child); if (rc == ESRCH) { /* Nothing left to do. pid doesn't exist */ return FALSE; } child->timeout = TRUE; crm_warn("%s process (PID %d) timed out", child->desc, (int)child->pid); child->timerid = g_timeout_add(5000, child_timeout_callback, child); return FALSE; } static gboolean child_waitpid(mainloop_child_t *child, int flags) { int rc = 0; int core = 0; int signo = 0; int status = 0; int exitcode = 0; rc = waitpid(child->pid, &status, flags); if(rc == 0) { crm_perror(LOG_DEBUG, "wait(%d) = %d", child->pid, rc); return FALSE; } else if(rc != child->pid) { signo = SIGCHLD; exitcode = 1; status = 1; crm_perror(LOG_ERR, "Call to waitpid(%d) failed", child->pid); } else { crm_trace("Managed process %d exited: %p", child->pid, child); if (WIFEXITED(status)) { exitcode = WEXITSTATUS(status); crm_trace("Managed process %d (%s) exited with rc=%d", child->pid, child->desc, exitcode); } else if (WIFSIGNALED(status)) { signo = WTERMSIG(status); crm_trace("Managed process %d (%s) exited with signal=%d", child->pid, child->desc, signo); } #ifdef WCOREDUMP if (WCOREDUMP(status)) { core = 1; crm_err("Managed process %d (%s) dumped core", child->pid, child->desc); } #endif } if (child->callback) { child->callback(child, child->pid, core, signo, exitcode); } return TRUE; } static void child_death_dispatch(int signal) { GListPtr iter = child_list; gboolean exited; while(iter) { GListPtr saved = NULL; mainloop_child_t *child = iter->data; exited = child_waitpid(child, WNOHANG); saved = iter; iter = iter->next; if (exited == FALSE) { continue; } crm_trace("Removing process entry %p for %d", child, child->pid); child_list = g_list_remove_link(child_list, saved); g_list_free(saved); child_free(child); } } static gboolean child_signal_init(gpointer p) { crm_trace("Installed SIGCHLD handler"); /* Do NOT use g_child_watch_add() and friends, they rely on pthreads */ mainloop_add_signal(SIGCHLD, child_death_dispatch); /* In case they terminated before the signal handler was installed */ child_death_dispatch(SIGCHLD); return FALSE; } int mainloop_child_kill(pid_t pid) { GListPtr iter; mainloop_child_t *child = NULL; mainloop_child_t *match = NULL; /* It is impossible to block SIGKILL, this allows us to * call waitpid without WNOHANG flag.*/ int waitflags = 0, rc = 0; for (iter = child_list; iter != NULL && match == NULL; iter = iter->next) { child = iter->data; if (pid == child->pid) { match = child; } } if (match == NULL) { return FALSE; } rc = child_kill_helper(match); if(rc == -ESRCH) { /* It's gone, but hasn't shown up in waitpid() yet * * Wait until we get SIGCHLD and let child_death_dispatch() * clean it up as normal (so we get the correct return * code/status) * * The blocking alternative would be to call: * child_waitpid(match, 0); */ crm_trace("Waiting for child %d to be reaped by child_death_dispatch()", match->pid); return TRUE; } else if(rc != 0) { /* If KILL for some other reason set the WNOHANG flag since we * can't be certain what happened. */ waitflags = WNOHANG; } if (child_waitpid(match, waitflags) == FALSE) { /* not much we can do if this occurs */ return FALSE; } child_list = g_list_remove(child_list, match); child_free(match); return TRUE; } /* Create/Log a new tracked process * To track a process group, use -pid */ void mainloop_child_add_with_flags(pid_t pid, int timeout, const char *desc, void *privatedata, enum mainloop_child_flags flags, void (*callback) (mainloop_child_t * p, pid_t pid, int core, int signo, int exitcode)) { static bool need_init = TRUE; mainloop_child_t *child = g_new(mainloop_child_t, 1); child->pid = pid; child->timerid = 0; child->timeout = FALSE; child->privatedata = privatedata; child->callback = callback; child->flags = flags; if(desc) { child->desc = strdup(desc); } if (timeout) { child->timerid = g_timeout_add(timeout, child_timeout_callback, child); } child_list = g_list_append(child_list, child); if(need_init) { need_init = FALSE; /* SIGCHLD processing has to be invoked from mainloop. * We do not want it to be possible to both add a child pid * to mainloop, and have the pid's exit callback invoked within * the same callstack. */ g_timeout_add(1, child_signal_init, NULL); } } void mainloop_child_add(pid_t pid, int timeout, const char *desc, void *privatedata, void (*callback) (mainloop_child_t * p, pid_t pid, int core, int signo, int exitcode)) { mainloop_child_add_with_flags(pid, timeout, desc, privatedata, 0, callback); } struct mainloop_timer_s { guint id; guint period_ms; bool repeat; char *name; GSourceFunc cb; void *userdata; }; static gboolean mainloop_timer_cb(gpointer user_data) { int id = 0; bool repeat = FALSE; struct mainloop_timer_s *t = user_data; CRM_ASSERT(t != NULL); id = t->id; t->id = 0; /* Ensure it's unset during callbacks so that * mainloop_timer_running() works as expected */ if(t->cb) { crm_trace("Invoking callbacks for timer %s", t->name); repeat = t->repeat; if(t->cb(t->userdata) == FALSE) { crm_trace("Timer %s complete", t->name); repeat = FALSE; } } if(repeat) { /* Restore if repeating */ t->id = id; } return repeat; } bool mainloop_timer_running(mainloop_timer_t *t) { if(t && t->id != 0) { return TRUE; } return FALSE; } void mainloop_timer_start(mainloop_timer_t *t) { mainloop_timer_stop(t); if(t && t->period_ms > 0) { crm_trace("Starting timer %s", t->name); t->id = g_timeout_add(t->period_ms, mainloop_timer_cb, t); } } void mainloop_timer_stop(mainloop_timer_t *t) { if(t && t->id != 0) { crm_trace("Stopping timer %s", t->name); g_source_remove(t->id); t->id = 0; } } guint mainloop_timer_set_period(mainloop_timer_t *t, guint period_ms) { guint last = 0; if(t) { last = t->period_ms; t->period_ms = period_ms; } if(t && t->id != 0 && last != t->period_ms) { mainloop_timer_start(t); } return last; } mainloop_timer_t * mainloop_timer_add(const char *name, guint period_ms, bool repeat, GSourceFunc cb, void *userdata) { mainloop_timer_t *t = calloc(1, sizeof(mainloop_timer_t)); if(t) { if(name) { t->name = crm_strdup_printf("%s-%u-%d", name, period_ms, repeat); } else { t->name = crm_strdup_printf("%p-%u-%d", t, period_ms, repeat); } t->id = 0; t->period_ms = period_ms; t->repeat = repeat; t->cb = cb; t->userdata = userdata; crm_trace("Created timer %s with %p %p", t->name, userdata, t->userdata); } return t; } void mainloop_timer_del(mainloop_timer_t *t) { if(t) { crm_trace("Destroying timer %s", t->name); mainloop_timer_stop(t); free(t->name); free(t); } } /* * Helpers to make sure certain events aren't lost at shutdown */ static gboolean drain_timeout_cb(gpointer user_data) { bool *timeout_popped = (bool*) user_data; *timeout_popped = TRUE; return FALSE; } /*! * \brief Process main loop events while a certain condition is met * * \param[in] mloop Main loop to process * \param[in] timer_ms Don't process longer than this amount of time * \param[in] check Function that returns TRUE if events should be processed * * \note This function is intended to be called at shutdown if certain important * events should not be missed. The caller would likely quit the main loop * or exit after calling this function. The check() function will be * passed the remaining timeout in milliseconds. */ void pcmk_drain_main_loop(GMainLoop *mloop, guint timer_ms, bool (*check)(guint)) { bool timeout_popped = FALSE; guint timer = 0; GMainContext *ctx = NULL; CRM_CHECK(mloop && check, return); ctx = g_main_loop_get_context(mloop); if (ctx) { time_t start_time = time(NULL); timer = g_timeout_add(timer_ms, drain_timeout_cb, &timeout_popped); while (!timeout_popped && check(timer_ms - (time(NULL) - start_time) * 1000)) { g_main_context_iteration(ctx, TRUE); } } if (!timeout_popped && (timer > 0)) { g_source_remove(timer); } } diff --git a/lib/common/operations.c b/lib/common/operations.c index e6b89103f1..2144cc681a 100644 --- a/lib/common/operations.c +++ b/lib/common/operations.c @@ -1,583 +1,636 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #ifndef _GNU_SOURCE # define _GNU_SOURCE #endif #include #include #include #include #include #include #include #include #include /*! * \brief Generate an operation key * * \param[in] rsc_id ID of resource being operated on * \param[in] op_type Operation name * \param[in] interval_ms Operation interval * * \return Newly allocated memory containing operation key as string * * \note It is the caller's responsibility to free() the result. */ char * generate_op_key(const char *rsc_id, const char *op_type, guint interval_ms) { CRM_ASSERT(rsc_id != NULL); CRM_ASSERT(op_type != NULL); return crm_strdup_printf(CRM_OP_FMT, rsc_id, op_type, interval_ms); } gboolean parse_op_key(const char *key, char **rsc_id, char **op_type, guint *interval_ms) { char *notify = NULL; char *mutable_key = NULL; char *mutable_key_ptr = NULL; size_t len = 0, offset = 0; unsigned long long ch = 0; guint local_interval_ms = 0; // Initialize output variables in case of early return if (rsc_id) { *rsc_id = NULL; } if (op_type) { *op_type = NULL; } if (interval_ms) { *interval_ms = 0; } CRM_CHECK(key && *key, return FALSE); // Parse interval at end of string len = strlen(key); offset = len - 1; while ((offset > 0) && isdigit(key[offset])) { ch = key[offset] - '0'; for (int digits = len - offset; digits > 1; --digits) { ch = ch * 10; } local_interval_ms += ch; offset--; } crm_trace("Operation key '%s' has interval %ums", key, local_interval_ms); if (interval_ms) { *interval_ms = local_interval_ms; } CRM_CHECK((offset != (len - 1)) && (key[offset] == '_'), return FALSE); mutable_key = strndup(key, offset); offset--; while (offset > 0 && key[offset] != '_') { offset--; } CRM_CHECK(key[offset] == '_', free(mutable_key); return FALSE); mutable_key_ptr = mutable_key + offset + 1; crm_trace(" Action: %s", mutable_key_ptr); if (op_type) { *op_type = strdup(mutable_key_ptr); } mutable_key[offset] = 0; offset--; notify = strstr(mutable_key, "_post_notify"); if (notify && safe_str_eq(notify, "_post_notify")) { notify[0] = 0; } notify = strstr(mutable_key, "_pre_notify"); if (notify && safe_str_eq(notify, "_pre_notify")) { notify[0] = 0; } crm_trace(" Resource: %s", mutable_key); if (rsc_id) { *rsc_id = mutable_key; } else { free(mutable_key); } return TRUE; } char * generate_notify_key(const char *rsc_id, const char *notify_type, const char *op_type) { CRM_CHECK(rsc_id != NULL, return NULL); CRM_CHECK(op_type != NULL, return NULL); CRM_CHECK(notify_type != NULL, return NULL); return crm_strdup_printf("%s_%s_notify_%s_0", rsc_id, notify_type, op_type); } static char * generate_transition_magic(const char *transition_key, int op_status, int op_rc) { CRM_CHECK(transition_key != NULL, return NULL); return crm_strdup_printf("%d:%d;%s", op_status, op_rc, transition_key); } +/*! + * \brief Parse a transition magic string into its constituent parts + * + * \param[in] magic Magic string to parse (must be non-NULL) + * \param[out] uuid If non-NULL, where to store copy of parsed UUID + * \param[out] transition_id If non-NULL, where to store parsed transition ID + * \param[out] action_id If non-NULL, where to store parsed action ID + * \param[out] op_status If non-NULL, where to store parsed result status + * \param[out] op_rc If non-NULL, where to store parsed actual rc + * \param[out] target_rc If non-NULL, where to stored parsed target rc + * + * \return TRUE if key was valid, FALSE otherwise + * \note If uuid is supplied and this returns TRUE, the caller is responsible + * for freeing the memory for *uuid using free(). + */ gboolean decode_transition_magic(const char *magic, char **uuid, int *transition_id, int *action_id, int *op_status, int *op_rc, int *target_rc) { int res = 0; char *key = NULL; gboolean result = TRUE; + int local_op_status = -1; + int local_op_rc = -1; CRM_CHECK(magic != NULL, return FALSE); - CRM_CHECK(op_rc != NULL, return FALSE); - CRM_CHECK(op_status != NULL, return FALSE); #ifdef SSCANF_HAS_M - res = sscanf(magic, "%d:%d;%ms", op_status, op_rc, &key); + res = sscanf(magic, "%d:%d;%ms", &local_op_status, &local_op_rc, &key); #else key = calloc(1, strlen(magic) - 3); // magic must have >=4 other characters - res = sscanf(magic, "%d:%d;%s", op_status, op_rc, key); + CRM_ASSERT(key); + res = sscanf(magic, "%d:%d;%s", &local_op_status, &local_op_rc, key); #endif if (res == EOF) { crm_err("Could not decode transition information '%s': %s", magic, pcmk_strerror(errno)); result = FALSE; } else if (res < 3) { crm_warn("Transition information '%s' incomplete (%d of 3 expected items)", magic, res); result = FALSE; } else { - CRM_CHECK(decode_transition_key(key, uuid, transition_id, action_id, - target_rc), result = FALSE); + if (op_status) { + *op_status = local_op_status; + } + if (op_rc) { + *op_rc = local_op_rc; + } + result = decode_transition_key(key, uuid, transition_id, action_id, + target_rc); } free(key); return result; } char * generate_transition_key(int transition_id, int action_id, int target_rc, const char *node) { CRM_CHECK(node != NULL, return NULL); return crm_strdup_printf("%d:%d:%d:%-*s", action_id, transition_id, target_rc, 36, node); } +/*! + * \brief Parse a transition key into its constituent parts + * + * \param[in] key Transition key to parse (must be non-NULL) + * \param[out] uuid If non-NULL, where to store copy of parsed UUID + * \param[out] transition_id If non-NULL, where to store parsed transition ID + * \param[out] action_id If non-NULL, where to store parsed action ID + * \param[out] target_rc If non-NULL, where to stored parsed target rc + * + * \return TRUE if key was valid, FALSE otherwise + * \note If uuid is supplied and this returns TRUE, the caller is responsible + * for freeing the memory for *uuid using free(). + */ gboolean decode_transition_key(const char *key, char **uuid, int *transition_id, int *action_id, int *target_rc) { - CRM_CHECK(uuid != NULL, return FALSE); - CRM_CHECK(target_rc != NULL, return FALSE); - CRM_CHECK(action_id != NULL, return FALSE); - CRM_CHECK(transition_id != NULL, return FALSE); - - *uuid = calloc(37, sizeof(char)); - if (sscanf(key, "%d:%d:%d:%36s", - action_id, transition_id, target_rc, *uuid) != 4) { - crm_err("Invalid transition key '%s'", key); - free(*uuid); + int local_transition_id = -1; + int local_action_id = -1; + int local_target_rc = -1; + char local_uuid[37] = { '\0' }; + + // Initialize any supplied output arguments + if (uuid) { *uuid = NULL; - *target_rc = -1; - *action_id = -1; + } + if (transition_id) { *transition_id = -1; + } + if (action_id) { + *action_id = -1; + } + if (target_rc) { + *target_rc = -1; + } + + CRM_CHECK(key != NULL, return FALSE); + if (sscanf(key, "%d:%d:%d:%36s", &local_action_id, &local_transition_id, + &local_target_rc, local_uuid) != 4) { + crm_err("Invalid transition key '%s'", key); return FALSE; } - if (strlen(*uuid) != 36) { - crm_warn("Invalid UUID '%s' in transition key '%s'", *uuid, key); + if (strlen(local_uuid) != 36) { + crm_warn("Invalid UUID '%s' in transition key '%s'", local_uuid, key); + } + if (uuid) { + *uuid = strdup(local_uuid); + CRM_ASSERT(*uuid); + } + if (transition_id) { + *transition_id = local_transition_id; + } + if (action_id) { + *action_id = local_action_id; + } + if (target_rc) { + *target_rc = local_target_rc; } return TRUE; } void filter_action_parameters(xmlNode * param_set, const char *version) { char *key = NULL; char *timeout = NULL; char *interval_ms_s = NULL; const char *attr_filter[] = { XML_ATTR_ID, XML_ATTR_CRM_VERSION, XML_LRM_ATTR_OP_DIGEST, XML_LRM_ATTR_TARGET, XML_LRM_ATTR_TARGET_UUID, "pcmk_external_ip" }; gboolean do_delete = FALSE; int lpc = 0; static int meta_len = 0; if (meta_len == 0) { meta_len = strlen(CRM_META); } if (param_set == NULL) { return; } for (lpc = 0; lpc < DIMOF(attr_filter); lpc++) { xml_remove_prop(param_set, attr_filter[lpc]); } key = crm_meta_name(XML_LRM_ATTR_INTERVAL_MS); interval_ms_s = crm_element_value_copy(param_set, key); free(key); key = crm_meta_name(XML_ATTR_TIMEOUT); timeout = crm_element_value_copy(param_set, key); if (param_set) { xmlAttrPtr xIter = param_set->properties; while (xIter) { const char *prop_name = (const char *)xIter->name; xIter = xIter->next; do_delete = FALSE; if (strncasecmp(prop_name, CRM_META, meta_len) == 0) { do_delete = TRUE; } if (do_delete) { xml_remove_prop(param_set, prop_name); } } } if (interval_ms_s && strcmp(interval_ms_s, "0")) { /* Re-instate the operation's timeout value */ if (timeout != NULL) { crm_xml_add(param_set, key, timeout); } } free(interval_ms_s); free(timeout); free(key); } #define FAKE_TE_ID "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx" static void append_digest(lrmd_event_data_t * op, xmlNode * update, const char *version, const char *magic, int level) { /* this will enable us to later determine that the * resource's parameters have changed and we should force * a restart */ char *digest = NULL; xmlNode *args_xml = NULL; if (op->params == NULL) { return; } args_xml = create_xml_node(NULL, XML_TAG_PARAMS); g_hash_table_foreach(op->params, hash2field, args_xml); filter_action_parameters(args_xml, version); digest = calculate_operation_digest(args_xml, version); #if 0 if (level < get_crm_log_level() && op->interval_ms == 0 && crm_str_eq(op->op_type, CRMD_ACTION_START, TRUE)) { char *digest_source = dump_xml_unformatted(args_xml); do_crm_log(level, "Calculated digest %s for %s (%s). Source: %s\n", digest, ID(update), magic, digest_source); free(digest_source); } #endif crm_xml_add(update, XML_LRM_ATTR_OP_DIGEST, digest); free_xml(args_xml); free(digest); } int rsc_op_expected_rc(lrmd_event_data_t * op) { int rc = 0; if (op && op->user_data) { - int dummy = 0; - char *uuid = NULL; - - decode_transition_key(op->user_data, &uuid, &dummy, &dummy, &rc); - free(uuid); + decode_transition_key(op->user_data, NULL, NULL, NULL, &rc); } return rc; } gboolean did_rsc_op_fail(lrmd_event_data_t * op, int target_rc) { switch (op->op_status) { case PCMK_LRM_OP_CANCELLED: case PCMK_LRM_OP_PENDING: return FALSE; break; case PCMK_LRM_OP_NOTSUPPORTED: case PCMK_LRM_OP_TIMEOUT: case PCMK_LRM_OP_ERROR: return TRUE; break; default: if (target_rc != op->rc) { return TRUE; } } return FALSE; } /*! * \brief Create a CIB XML element for an operation * * \param[in] parent If not NULL, make new XML node a child of this one * \param[in] prefix Generate an ID using this prefix * \param[in] task Operation task to set * \param[in] interval_spec Operation interval to set * \param[in] timeout If not NULL, operation timeout to set * * \return New XML object on success, NULL otherwise */ xmlNode * crm_create_op_xml(xmlNode *parent, const char *prefix, const char *task, const char *interval_spec, const char *timeout) { xmlNode *xml_op; CRM_CHECK(prefix && task && interval_spec, return NULL); xml_op = create_xml_node(parent, XML_ATTR_OP); crm_xml_set_id(xml_op, "%s-%s-%s", prefix, task, interval_spec); crm_xml_add(xml_op, XML_LRM_ATTR_INTERVAL, interval_spec); crm_xml_add(xml_op, "name", task); if (timeout) { crm_xml_add(xml_op, XML_ATTR_TIMEOUT, timeout); } return xml_op; } xmlNode * create_operation_update(xmlNode * parent, lrmd_event_data_t * op, const char * caller_version, int target_rc, const char * node, const char * origin, int level) { char *key = NULL; char *magic = NULL; char *op_id = NULL; char *op_id_additional = NULL; char *local_user_data = NULL; const char *exit_reason = NULL; xmlNode *xml_op = NULL; const char *task = NULL; CRM_CHECK(op != NULL, return NULL); do_crm_log(level, "%s: Updating resource %s after %s op %s (interval=%u)", origin, op->rsc_id, op->op_type, services_lrm_status_str(op->op_status), op->interval_ms); crm_trace("DC version: %s", caller_version); task = op->op_type; /* Record a successful reload as a start, and a failed reload as a monitor, * to make life easier for the scheduler when determining the current state. */ if (crm_str_eq(task, "reload", TRUE)) { if (op->op_status == PCMK_LRM_OP_DONE) { task = CRMD_ACTION_START; } else { task = CRMD_ACTION_STATUS; } } key = generate_op_key(op->rsc_id, task, op->interval_ms); if (crm_str_eq(task, CRMD_ACTION_NOTIFY, TRUE)) { const char *n_type = crm_meta_value(op->params, "notify_type"); const char *n_task = crm_meta_value(op->params, "notify_operation"); CRM_LOG_ASSERT(n_type != NULL); CRM_LOG_ASSERT(n_task != NULL); op_id = generate_notify_key(op->rsc_id, n_type, n_task); if (op->op_status != PCMK_LRM_OP_PENDING) { /* Ignore notify errors. * * @TODO It might be better to keep the correct result here, and * ignore it in process_graph_event(). */ op->op_status = PCMK_LRM_OP_DONE; op->rc = 0; } } else if (did_rsc_op_fail(op, target_rc)) { op_id = generate_op_key(op->rsc_id, "last_failure", 0); if (op->interval_ms == 0) { // Ensure 'last' gets updated, in case record-pending is true op_id_additional = generate_op_key(op->rsc_id, "last", 0); } exit_reason = op->exit_reason; } else if (op->interval_ms > 0) { op_id = strdup(key); } else { op_id = generate_op_key(op->rsc_id, "last", 0); } again: xml_op = find_entity(parent, XML_LRM_TAG_RSC_OP, op_id); if (xml_op == NULL) { xml_op = create_xml_node(parent, XML_LRM_TAG_RSC_OP); } if (op->user_data == NULL) { crm_debug("Generating fake transition key for: " CRM_OP_FMT " %d from %s", op->rsc_id, op->op_type, op->interval_ms, op->call_id, origin); local_user_data = generate_transition_key(-1, op->call_id, target_rc, FAKE_TE_ID); op->user_data = local_user_data; } if(magic == NULL) { magic = generate_transition_magic(op->user_data, op->op_status, op->rc); } crm_xml_add(xml_op, XML_ATTR_ID, op_id); crm_xml_add(xml_op, XML_LRM_ATTR_TASK_KEY, key); crm_xml_add(xml_op, XML_LRM_ATTR_TASK, task); crm_xml_add(xml_op, XML_ATTR_ORIGIN, origin); crm_xml_add(xml_op, XML_ATTR_CRM_VERSION, caller_version); crm_xml_add(xml_op, XML_ATTR_TRANSITION_KEY, op->user_data); crm_xml_add(xml_op, XML_ATTR_TRANSITION_MAGIC, magic); crm_xml_add(xml_op, XML_LRM_ATTR_EXIT_REASON, exit_reason == NULL ? "" : exit_reason); crm_xml_add(xml_op, XML_LRM_ATTR_TARGET, node); /* For context during triage */ crm_xml_add_int(xml_op, XML_LRM_ATTR_CALLID, op->call_id); crm_xml_add_int(xml_op, XML_LRM_ATTR_RC, op->rc); crm_xml_add_int(xml_op, XML_LRM_ATTR_OPSTATUS, op->op_status); crm_xml_add_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, op->interval_ms); if (compare_version("2.1", caller_version) <= 0) { if (op->t_run || op->t_rcchange || op->exec_time || op->queue_time) { crm_trace("Timing data (" CRM_OP_FMT "): last=%u change=%u exec=%u queue=%u", op->rsc_id, op->op_type, op->interval_ms, op->t_run, op->t_rcchange, op->exec_time, op->queue_time); if (op->interval_ms == 0) { /* The values are the same for non-recurring ops */ crm_xml_add_int(xml_op, XML_RSC_OP_LAST_RUN, op->t_run); crm_xml_add_int(xml_op, XML_RSC_OP_LAST_CHANGE, op->t_run); } else if(op->t_rcchange) { /* last-run is not accurate for recurring ops */ crm_xml_add_int(xml_op, XML_RSC_OP_LAST_CHANGE, op->t_rcchange); } else { /* ...but is better than nothing otherwise */ crm_xml_add_int(xml_op, XML_RSC_OP_LAST_CHANGE, op->t_run); } crm_xml_add_int(xml_op, XML_RSC_OP_T_EXEC, op->exec_time); crm_xml_add_int(xml_op, XML_RSC_OP_T_QUEUE, op->queue_time); } } if (crm_str_eq(op->op_type, CRMD_ACTION_MIGRATE, TRUE) || crm_str_eq(op->op_type, CRMD_ACTION_MIGRATED, TRUE)) { /* * Record migrate_source and migrate_target always for migrate ops. */ const char *name = XML_LRM_ATTR_MIGRATE_SOURCE; crm_xml_add(xml_op, name, crm_meta_value(op->params, name)); name = XML_LRM_ATTR_MIGRATE_TARGET; crm_xml_add(xml_op, name, crm_meta_value(op->params, name)); } append_digest(op, xml_op, caller_version, magic, LOG_DEBUG); if (op_id_additional) { free(op_id); op_id = op_id_additional; op_id_additional = NULL; goto again; } if (local_user_data) { free(local_user_data); op->user_data = NULL; } free(magic); free(op_id); free(key); return xml_op; } /*! * \brief Check whether an operation requires resource agent meta-data * * \param[in] rsc_class Resource agent class (or NULL to skip class check) * \param[in] op Operation action (or NULL to skip op check) * * \return TRUE if operation needs meta-data, FALSE otherwise * \note At least one of rsc_class and op must be specified. */ bool crm_op_needs_metadata(const char *rsc_class, const char *op) { /* Agent meta-data is used to determine whether a reload is possible, and to * evaluate versioned parameters -- so if this op is not relevant to those * features, we don't need the meta-data. */ CRM_CHECK(rsc_class || op, return FALSE); if (rsc_class && is_not_set(pcmk_get_ra_caps(rsc_class), pcmk_ra_cap_params)) { /* Meta-data is only needed for resource classes that use parameters */ return FALSE; } /* Meta-data is only needed for these actions */ if (op && strcmp(op, CRMD_ACTION_START) && strcmp(op, CRMD_ACTION_STATUS) && strcmp(op, CRMD_ACTION_PROMOTE) && strcmp(op, CRMD_ACTION_DEMOTE) && strcmp(op, CRMD_ACTION_RELOAD) && strcmp(op, CRMD_ACTION_MIGRATE) && strcmp(op, CRMD_ACTION_MIGRATED) && strcmp(op, CRMD_ACTION_NOTIFY)) { return FALSE; } return TRUE; } diff --git a/lib/pacemaker/pcmk_sched_allocate.c b/lib/pacemaker/pcmk_sched_allocate.c index 42b29ccfb4..3363a72221 100644 --- a/lib/pacemaker/pcmk_sched_allocate.c +++ b/lib/pacemaker/pcmk_sched_allocate.c @@ -1,2728 +1,2755 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include CRM_TRACE_INIT_DATA(pe_allocate); void set_alloc_actions(pe_working_set_t * data_set); extern void ReloadRsc(resource_t * rsc, node_t *node, pe_working_set_t * data_set); extern gboolean DeleteRsc(resource_t * rsc, node_t * node, gboolean optional, pe_working_set_t * data_set); static void apply_remote_node_ordering(pe_working_set_t *data_set); static enum remote_connection_state get_remote_node_state(pe_node_t *node); enum remote_connection_state { remote_state_unknown = 0, remote_state_alive = 1, remote_state_resting = 2, remote_state_failed = 3, remote_state_stopped = 4 }; static const char * state2text(enum remote_connection_state state) { switch (state) { case remote_state_unknown: return "unknown"; case remote_state_alive: return "alive"; case remote_state_resting: return "resting"; case remote_state_failed: return "failed"; case remote_state_stopped: return "stopped"; } return "impossible"; } resource_alloc_functions_t resource_class_alloc_functions[] = { { native_merge_weights, native_color, native_create_actions, native_create_probe, native_internal_constraints, native_rsc_colocation_lh, native_rsc_colocation_rh, native_rsc_location, native_action_flags, native_update_actions, native_expand, native_append_meta, }, { group_merge_weights, group_color, group_create_actions, native_create_probe, group_internal_constraints, group_rsc_colocation_lh, group_rsc_colocation_rh, group_rsc_location, group_action_flags, group_update_actions, group_expand, group_append_meta, }, { clone_merge_weights, clone_color, clone_create_actions, clone_create_probe, clone_internal_constraints, clone_rsc_colocation_lh, clone_rsc_colocation_rh, clone_rsc_location, clone_action_flags, pcmk__multi_update_actions, clone_expand, clone_append_meta, }, { pcmk__bundle_merge_weights, pcmk__bundle_color, pcmk__bundle_create_actions, pcmk__bundle_create_probe, pcmk__bundle_internal_constraints, pcmk__bundle_rsc_colocation_lh, pcmk__bundle_rsc_colocation_rh, pcmk__bundle_rsc_location, pcmk__bundle_action_flags, pcmk__multi_update_actions, pcmk__bundle_expand, pcmk__bundle_append_meta, } }; gboolean update_action_flags(action_t * action, enum pe_action_flags flags, const char *source, int line) { static unsigned long calls = 0; gboolean changed = FALSE; gboolean clear = is_set(flags, pe_action_clear); enum pe_action_flags last = action->flags; if (clear) { action->flags = crm_clear_bit(source, line, action->uuid, action->flags, flags); } else { action->flags = crm_set_bit(source, line, action->uuid, action->flags, flags); } if (last != action->flags) { calls++; changed = TRUE; /* Useful for tracking down _who_ changed a specific flag */ /* CRM_ASSERT(calls != 534); */ clear_bit(flags, pe_action_clear); crm_trace("%s on %s: %sset flags 0x%.6x (was 0x%.6x, now 0x%.6x, %lu, %s)", action->uuid, action->node ? action->node->details->uname : "[none]", clear ? "un-" : "", flags, last, action->flags, calls, source); } return changed; } static gboolean check_rsc_parameters(resource_t * rsc, node_t * node, xmlNode * rsc_entry, gboolean active_here, pe_working_set_t * data_set) { int attr_lpc = 0; gboolean force_restart = FALSE; gboolean delete_resource = FALSE; gboolean changed = FALSE; const char *value = NULL; const char *old_value = NULL; const char *attr_list[] = { XML_ATTR_TYPE, XML_AGENT_ATTR_CLASS, XML_AGENT_ATTR_PROVIDER }; for (; attr_lpc < DIMOF(attr_list); attr_lpc++) { value = crm_element_value(rsc->xml, attr_list[attr_lpc]); old_value = crm_element_value(rsc_entry, attr_list[attr_lpc]); if (value == old_value /* i.e. NULL */ || crm_str_eq(value, old_value, TRUE)) { continue; } changed = TRUE; trigger_unfencing(rsc, node, "Device definition changed", NULL, data_set); if (active_here) { force_restart = TRUE; crm_notice("Forcing restart of %s on %s, %s changed: %s -> %s", rsc->id, node->details->uname, attr_list[attr_lpc], crm_str(old_value), crm_str(value)); } } if (force_restart) { /* make sure the restart happens */ stop_action(rsc, node, FALSE); set_bit(rsc->flags, pe_rsc_start_pending); delete_resource = TRUE; } else if (changed) { delete_resource = TRUE; } return delete_resource; } static void CancelXmlOp(resource_t * rsc, xmlNode * xml_op, node_t * active_node, const char *reason, pe_working_set_t * data_set) { guint interval_ms = 0; action_t *cancel = NULL; const char *task = NULL; const char *call_id = NULL; const char *interval_ms_s = NULL; CRM_CHECK(xml_op != NULL, return); CRM_CHECK(active_node != NULL, return); task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); call_id = crm_element_value(xml_op, XML_LRM_ATTR_CALLID); interval_ms_s = crm_element_value(xml_op, XML_LRM_ATTR_INTERVAL_MS); interval_ms = crm_parse_ms(interval_ms_s); crm_info("Action " CRM_OP_FMT " on %s will be stopped: %s", rsc->id, task, interval_ms, active_node->details->uname, (reason? reason : "unknown")); cancel = pe_cancel_op(rsc, task, interval_ms, active_node, data_set); add_hash_param(cancel->meta, XML_LRM_ATTR_CALLID, call_id); custom_action_order(rsc, stop_key(rsc), NULL, rsc, NULL, cancel, pe_order_optional, data_set); } static gboolean check_action_definition(resource_t * rsc, node_t * active_node, xmlNode * xml_op, pe_working_set_t * data_set) { char *key = NULL; guint interval_ms = 0; const char *interval_ms_s = NULL; const op_digest_cache_t *digest_data = NULL; gboolean did_change = FALSE; const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); const char *digest_secure = NULL; CRM_CHECK(active_node != NULL, return FALSE); interval_ms_s = crm_element_value(xml_op, XML_LRM_ATTR_INTERVAL_MS); interval_ms = crm_parse_ms(interval_ms_s); if (interval_ms > 0) { xmlNode *op_match = NULL; /* we need to reconstruct the key because of the way we used to construct resource IDs */ key = generate_op_key(rsc->id, task, interval_ms); pe_rsc_trace(rsc, "Checking parameters for %s", key); op_match = find_rsc_op_entry(rsc, key); if (op_match == NULL && is_set(data_set->flags, pe_flag_stop_action_orphans)) { CancelXmlOp(rsc, xml_op, active_node, "orphan", data_set); free(key); return TRUE; } else if (op_match == NULL) { pe_rsc_debug(rsc, "Orphan action detected: %s on %s", key, active_node->details->uname); free(key); return TRUE; } free(key); key = NULL; } crm_trace("Testing " CRM_OP_FMT " on %s", rsc->id, task, interval_ms, active_node->details->uname); if ((interval_ms == 0) && safe_str_eq(task, RSC_STATUS)) { /* Reload based on the start action not a probe */ task = RSC_START; } else if ((interval_ms == 0) && safe_str_eq(task, RSC_MIGRATED)) { /* Reload based on the start action not a migrate */ task = RSC_START; } else if ((interval_ms == 0) && safe_str_eq(task, RSC_PROMOTE)) { /* Reload based on the start action not a promote */ task = RSC_START; } digest_data = rsc_action_digest_cmp(rsc, xml_op, active_node, data_set); if(is_set(data_set->flags, pe_flag_sanitized)) { digest_secure = crm_element_value(xml_op, XML_LRM_ATTR_SECURE_DIGEST); } if(digest_data->rc != RSC_DIGEST_MATCH && digest_secure && digest_data->digest_secure_calc && strcmp(digest_data->digest_secure_calc, digest_secure) == 0) { if (is_set(data_set->flags, pe_flag_stdout)) { printf("Only 'private' parameters to " CRM_OP_FMT " on %s changed: %s\n", rsc->id, task, interval_ms, active_node->details->uname, crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC)); } } else if (digest_data->rc == RSC_DIGEST_RESTART) { /* Changes that force a restart */ pe_action_t *required = NULL; did_change = TRUE; key = generate_op_key(rsc->id, task, interval_ms); crm_log_xml_info(digest_data->params_restart, "params:restart"); required = custom_action(rsc, key, task, NULL, TRUE, TRUE, data_set); pe_action_set_flag_reason(__FUNCTION__, __LINE__, required, NULL, "resource definition change", pe_action_optional, TRUE); trigger_unfencing(rsc, active_node, "Device parameters changed", NULL, data_set); } else if ((digest_data->rc == RSC_DIGEST_ALL) || (digest_data->rc == RSC_DIGEST_UNKNOWN)) { /* Changes that can potentially be handled by a reload */ const char *digest_restart = crm_element_value(xml_op, XML_LRM_ATTR_RESTART_DIGEST); did_change = TRUE; trigger_unfencing(rsc, active_node, "Device parameters changed (reload)", NULL, data_set); crm_log_xml_info(digest_data->params_all, "params:reload"); key = generate_op_key(rsc->id, task, interval_ms); if (interval_ms > 0) { action_t *op = NULL; #if 0 /* Always reload/restart the entire resource */ ReloadRsc(rsc, active_node, data_set); #else /* Re-sending the recurring op is sufficient - the old one will be cancelled automatically */ op = custom_action(rsc, key, task, active_node, TRUE, TRUE, data_set); set_bit(op->flags, pe_action_reschedule); #endif } else if (digest_restart) { pe_rsc_trace(rsc, "Reloading '%s' action for resource %s", task, rsc->id); /* Reload this resource */ ReloadRsc(rsc, active_node, data_set); free(key); } else { pe_action_t *required = NULL; pe_rsc_trace(rsc, "Resource %s doesn't know how to reload", rsc->id); /* Re-send the start/demote/promote op * Recurring ops will be detected independently */ required = custom_action(rsc, key, task, NULL, TRUE, TRUE, data_set); pe_action_set_flag_reason(__FUNCTION__, __LINE__, required, NULL, "resource definition change", pe_action_optional, TRUE); } } return did_change; } /*! * \internal * \brief Do deferred action checks after allocation * * \param[in] data_set Working set for cluster */ static void check_params(pe_resource_t *rsc, pe_node_t *node, xmlNode *rsc_op, enum pe_check_parameters check, pe_working_set_t *data_set) { const char *reason = NULL; op_digest_cache_t *digest_data = NULL; switch (check) { case pe_check_active: if (check_action_definition(rsc, node, rsc_op, data_set) && pe_get_failcount(node, rsc, NULL, pe_fc_effective, NULL, data_set)) { reason = "action definition changed"; } break; case pe_check_last_failure: digest_data = rsc_action_digest_cmp(rsc, rsc_op, node, data_set); switch (digest_data->rc) { case RSC_DIGEST_UNKNOWN: crm_trace("Resource %s history entry %s on %s has no digest to compare", rsc->id, ID(rsc_op), node->details->id); break; case RSC_DIGEST_MATCH: break; default: reason = "resource parameters have changed"; break; } break; } if (reason) { pe__clear_failcount(rsc, node, reason, data_set); } } static void check_actions_for(xmlNode * rsc_entry, resource_t * rsc, node_t * node, pe_working_set_t * data_set) { GListPtr gIter = NULL; int offset = -1; guint interval_ms = 0; int stop_index = 0; int start_index = 0; const char *task = NULL; const char *interval_ms_s = NULL; xmlNode *rsc_op = NULL; GListPtr op_list = NULL; GListPtr sorted_op_list = NULL; CRM_CHECK(node != NULL, return); if (is_set(rsc->flags, pe_rsc_orphan)) { resource_t *parent = uber_parent(rsc); if(parent == NULL || pe_rsc_is_clone(parent) == FALSE || is_set(parent->flags, pe_rsc_unique)) { pe_rsc_trace(rsc, "Skipping param check for %s and deleting: orphan", rsc->id); DeleteRsc(rsc, node, FALSE, data_set); } else { pe_rsc_trace(rsc, "Skipping param check for %s (orphan clone)", rsc->id); } return; } else if (pe_find_node_id(rsc->running_on, node->details->id) == NULL) { if (check_rsc_parameters(rsc, node, rsc_entry, FALSE, data_set)) { DeleteRsc(rsc, node, FALSE, data_set); } pe_rsc_trace(rsc, "Skipping param check for %s: no longer active on %s", rsc->id, node->details->uname); return; } pe_rsc_trace(rsc, "Processing %s on %s", rsc->id, node->details->uname); if (check_rsc_parameters(rsc, node, rsc_entry, TRUE, data_set)) { DeleteRsc(rsc, node, FALSE, data_set); } for (rsc_op = __xml_first_child(rsc_entry); rsc_op != NULL; rsc_op = __xml_next_element(rsc_op)) { if (crm_str_eq((const char *)rsc_op->name, XML_LRM_TAG_RSC_OP, TRUE)) { op_list = g_list_prepend(op_list, rsc_op); } } sorted_op_list = g_list_sort(op_list, sort_op_by_callid); calculate_active_ops(sorted_op_list, &start_index, &stop_index); for (gIter = sorted_op_list; gIter != NULL; gIter = gIter->next) { xmlNode *rsc_op = (xmlNode *) gIter->data; offset++; if (start_index < stop_index) { /* stopped */ continue; } else if (offset < start_index) { /* action occurred prior to a start */ continue; } task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); interval_ms_s = crm_element_value(rsc_op, XML_LRM_ATTR_INTERVAL_MS); interval_ms = crm_parse_ms(interval_ms_s); if ((interval_ms > 0) && (is_set(rsc->flags, pe_rsc_maintenance) || node->details->maintenance)) { // Maintenance mode cancels recurring operations CancelXmlOp(rsc, rsc_op, node, "maintenance mode", data_set); } else if ((interval_ms > 0) || safe_str_eq(task, RSC_STATUS) || safe_str_eq(task, RSC_START) || safe_str_eq(task, RSC_PROMOTE) || safe_str_eq(task, RSC_MIGRATED)) { /* If a resource operation failed, and the operation's definition * has changed, clear any fail count so they can be retried fresh. */ if (pe__bundle_needs_remote_name(rsc)) { /* We haven't allocated resources to nodes yet, so if the * REMOTE_CONTAINER_HACK is used, we may calculate the digest * based on the literal "#uname" value rather than the properly * substituted value. That would mistakenly make the action * definition appear to have been changed. Defer the check until * later in this case. */ pe__add_param_check(rsc_op, rsc, node, pe_check_active, data_set); } else if (check_action_definition(rsc, node, rsc_op, data_set) && pe_get_failcount(node, rsc, NULL, pe_fc_effective, NULL, data_set)) { pe__clear_failcount(rsc, node, "action definition changed", data_set); } } } g_list_free(sorted_op_list); } static GListPtr find_rsc_list(GListPtr result, resource_t * rsc, const char *id, gboolean renamed_clones, gboolean partial, pe_working_set_t * data_set) { GListPtr gIter = NULL; gboolean match = FALSE; if (id == NULL) { return NULL; } else if (rsc == NULL && data_set) { for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) { resource_t *child = (resource_t *) gIter->data; result = find_rsc_list(result, child, id, renamed_clones, partial, NULL); } return result; } else if (rsc == NULL) { return NULL; } if (partial) { if (strstr(rsc->id, id)) { match = TRUE; } else if (renamed_clones && rsc->clone_name && strstr(rsc->clone_name, id)) { match = TRUE; } } else { if (strcmp(rsc->id, id) == 0) { match = TRUE; } else if (renamed_clones && rsc->clone_name && strcmp(rsc->clone_name, id) == 0) { match = TRUE; } } if (match) { result = g_list_prepend(result, rsc); } if (rsc->children) { gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { resource_t *child = (resource_t *) gIter->data; result = find_rsc_list(result, child, id, renamed_clones, partial, NULL); } } return result; } static void check_actions(pe_working_set_t * data_set) { const char *id = NULL; node_t *node = NULL; xmlNode *lrm_rscs = NULL; xmlNode *status = get_object_root(XML_CIB_TAG_STATUS, data_set->input); xmlNode *node_state = NULL; for (node_state = __xml_first_child(status); node_state != NULL; node_state = __xml_next_element(node_state)) { if (crm_str_eq((const char *)node_state->name, XML_CIB_TAG_STATE, TRUE)) { id = crm_element_value(node_state, XML_ATTR_ID); lrm_rscs = find_xml_node(node_state, XML_CIB_TAG_LRM, FALSE); lrm_rscs = find_xml_node(lrm_rscs, XML_LRM_TAG_RESOURCES, FALSE); node = pe_find_node_id(data_set->nodes, id); if (node == NULL) { continue; /* Still need to check actions for a maintenance node to cancel existing monitor operations */ } else if (can_run_resources(node) == FALSE && node->details->maintenance == FALSE) { crm_trace("Skipping param check for %s: can't run resources", node->details->uname); continue; } crm_trace("Processing node %s", node->details->uname); if (node->details->online || is_set(data_set->flags, pe_flag_stonith_enabled)) { xmlNode *rsc_entry = NULL; for (rsc_entry = __xml_first_child(lrm_rscs); rsc_entry != NULL; rsc_entry = __xml_next_element(rsc_entry)) { if (crm_str_eq((const char *)rsc_entry->name, XML_LRM_TAG_RESOURCE, TRUE)) { if (xml_has_children(rsc_entry)) { GListPtr gIter = NULL; GListPtr result = NULL; const char *rsc_id = ID(rsc_entry); CRM_CHECK(rsc_id != NULL, return); result = find_rsc_list(NULL, NULL, rsc_id, TRUE, FALSE, data_set); for (gIter = result; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; if (rsc->variant != pe_native) { continue; } check_actions_for(rsc_entry, rsc, node, data_set); } g_list_free(result); } } } } } } } static gboolean apply_placement_constraints(pe_working_set_t * data_set) { GListPtr gIter = NULL; crm_trace("Applying constraints..."); for (gIter = data_set->placement_constraints; gIter != NULL; gIter = gIter->next) { pe__location_t *cons = gIter->data; cons->rsc_lh->cmds->rsc_location(cons->rsc_lh, cons); } return TRUE; } static gboolean failcount_clear_action_exists(node_t * node, resource_t * rsc) { gboolean rc = FALSE; GList *list = pe__resource_actions(rsc, node, CRM_OP_CLEAR_FAILCOUNT, TRUE); if (list) { rc = TRUE; } g_list_free(list); return rc; } /*! * \internal * \brief Force resource away if failures hit migration threshold * * \param[in,out] rsc Resource to check for failures * \param[in,out] node Node to check for failures * \param[in,out] data_set Cluster working set to update */ static void check_migration_threshold(resource_t *rsc, node_t *node, pe_working_set_t *data_set) { int fail_count, countdown; resource_t *failed; /* Migration threshold of 0 means never force away */ if (rsc->migration_threshold == 0) { return; } // If we're ignoring failures, also ignore the migration threshold if (is_set(rsc->flags, pe_rsc_failure_ignored)) { return; } /* If there are no failures, there's no need to force away */ fail_count = pe_get_failcount(node, rsc, NULL, pe_fc_effective|pe_fc_fillers, NULL, data_set); if (fail_count <= 0) { return; } /* How many more times recovery will be tried on this node */ countdown = QB_MAX(rsc->migration_threshold - fail_count, 0); /* If failed resource has a parent, we'll force the parent away */ failed = rsc; if (is_not_set(rsc->flags, pe_rsc_unique)) { failed = uber_parent(rsc); } if (countdown == 0) { resource_location(failed, node, -INFINITY, "__fail_limit__", data_set); crm_warn("Forcing %s away from %s after %d failures (max=%d)", failed->id, node->details->uname, fail_count, rsc->migration_threshold); } else { crm_info("%s can fail %d more times on %s before being forced off", failed->id, countdown, node->details->uname); } } static void common_apply_stickiness(resource_t * rsc, node_t * node, pe_working_set_t * data_set) { if (rsc->children) { GListPtr gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { resource_t *child_rsc = (resource_t *) gIter->data; common_apply_stickiness(child_rsc, node, data_set); } return; } if (is_set(rsc->flags, pe_rsc_managed) && rsc->stickiness != 0 && g_list_length(rsc->running_on) == 1) { node_t *current = pe_find_node_id(rsc->running_on, node->details->id); node_t *match = pe_hash_table_lookup(rsc->allowed_nodes, node->details->id); if (current == NULL) { } else if (match != NULL || is_set(data_set->flags, pe_flag_symmetric_cluster)) { resource_t *sticky_rsc = rsc; resource_location(sticky_rsc, node, rsc->stickiness, "stickiness", data_set); pe_rsc_debug(sticky_rsc, "Resource %s: preferring current location" " (node=%s, weight=%d)", sticky_rsc->id, node->details->uname, rsc->stickiness); } else { GHashTableIter iter; node_t *nIter = NULL; pe_rsc_debug(rsc, "Ignoring stickiness for %s: the cluster is asymmetric" " and node %s is not explicitly allowed", rsc->id, node->details->uname); g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&nIter)) { crm_err("%s[%s] = %d", rsc->id, nIter->details->uname, nIter->weight); } } } /* Check the migration threshold only if a failcount clear action * has not already been placed for this resource on the node. * There is no sense in potentially forcing the resource from this * node if the failcount is being reset anyway. * * @TODO A clear_failcount operation can be scheduled in stage4() via * check_actions_for(), or in stage5() via check_params(). This runs in * stage2(), so it cannot detect those, meaning we might check the migration * threshold when we shouldn't -- worst case, we stop or move the resource, * then move it back next transition. */ if (failcount_clear_action_exists(node, rsc) == FALSE) { check_migration_threshold(rsc, node, data_set); } } void complex_set_cmds(resource_t * rsc) { GListPtr gIter = rsc->children; rsc->cmds = &resource_class_alloc_functions[rsc->variant]; for (; gIter != NULL; gIter = gIter->next) { resource_t *child_rsc = (resource_t *) gIter->data; complex_set_cmds(child_rsc); } } void set_alloc_actions(pe_working_set_t * data_set) { GListPtr gIter = data_set->resources; for (; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; complex_set_cmds(rsc); } } static void calculate_system_health(gpointer gKey, gpointer gValue, gpointer user_data) { const char *key = (const char *)gKey; const char *value = (const char *)gValue; int *system_health = (int *)user_data; if (!gKey || !gValue || !user_data) { return; } if (crm_starts_with(key, "#health")) { int score; /* Convert the value into an integer */ score = char2score(value); /* Add it to the running total */ *system_health = merge_weights(score, *system_health); } } static gboolean apply_system_health(pe_working_set_t * data_set) { GListPtr gIter = NULL; const char *health_strategy = pe_pref(data_set->config_hash, "node-health-strategy"); int base_health = 0; if (health_strategy == NULL || safe_str_eq(health_strategy, "none")) { /* Prevent any accidental health -> score translation */ node_score_red = 0; node_score_yellow = 0; node_score_green = 0; return TRUE; } else if (safe_str_eq(health_strategy, "migrate-on-red")) { /* Resources on nodes which have health values of red are * weighted away from that node. */ node_score_red = -INFINITY; node_score_yellow = 0; node_score_green = 0; } else if (safe_str_eq(health_strategy, "only-green")) { /* Resources on nodes which have health values of red or yellow * are forced away from that node. */ node_score_red = -INFINITY; node_score_yellow = -INFINITY; node_score_green = 0; } else if (safe_str_eq(health_strategy, "progressive")) { /* Same as the above, but use the r/y/g scores provided by the user * Defaults are provided by the pe_prefs table * Also, custom health "base score" can be used */ base_health = crm_parse_int(pe_pref(data_set->config_hash, "node-health-base"), "0"); } else if (safe_str_eq(health_strategy, "custom")) { /* Requires the admin to configure the rsc_location constaints for * processing the stored health scores */ /* TODO: Check for the existence of appropriate node health constraints */ return TRUE; } else { crm_err("Unknown node health strategy: %s", health_strategy); return FALSE; } crm_info("Applying automated node health strategy: %s", health_strategy); for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { int system_health = base_health; node_t *node = (node_t *) gIter->data; /* Search through the node hash table for system health entries. */ g_hash_table_foreach(node->details->attrs, calculate_system_health, &system_health); crm_info(" Node %s has an combined system health of %d", node->details->uname, system_health); /* If the health is non-zero, then create a new rsc2node so that the * weight will be added later on. */ if (system_health != 0) { GListPtr gIter2 = data_set->resources; for (; gIter2 != NULL; gIter2 = gIter2->next) { resource_t *rsc = (resource_t *) gIter2->data; rsc2node_new(health_strategy, rsc, system_health, NULL, node, data_set); } } } return TRUE; } gboolean stage0(pe_working_set_t * data_set) { xmlNode *cib_constraints = get_object_root(XML_CIB_TAG_CONSTRAINTS, data_set->input); if (data_set->input == NULL) { return FALSE; } if (is_set(data_set->flags, pe_flag_have_status) == FALSE) { crm_trace("Calculating status"); cluster_status(data_set); } set_alloc_actions(data_set); apply_system_health(data_set); unpack_constraints(cib_constraints, data_set); return TRUE; } /* * Check nodes for resources started outside of the LRM */ gboolean probe_resources(pe_working_set_t * data_set) { action_t *probe_node_complete = NULL; for (GListPtr gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; const char *probed = pe_node_attribute_raw(node, CRM_OP_PROBED); if (node->details->online == FALSE) { if (pe__is_remote_node(node) && node->details->remote_rsc && (get_remote_node_state(node) == remote_state_failed)) { pe_fence_node(data_set, node, "the connection is unrecoverable"); } continue; } else if (node->details->unclean) { continue; } else if (node->details->rsc_discovery_enabled == FALSE) { /* resource discovery is disabled for this node */ continue; } if (probed != NULL && crm_is_true(probed) == FALSE) { action_t *probe_op = custom_action(NULL, crm_strdup_printf("%s-%s", CRM_OP_REPROBE, node->details->uname), CRM_OP_REPROBE, node, FALSE, TRUE, data_set); add_hash_param(probe_op->meta, XML_ATTR_TE_NOWAIT, XML_BOOLEAN_TRUE); continue; } for (GListPtr gIter2 = data_set->resources; gIter2 != NULL; gIter2 = gIter2->next) { resource_t *rsc = (resource_t *) gIter2->data; rsc->cmds->create_probe(rsc, node, probe_node_complete, FALSE, data_set); } } return TRUE; } static void rsc_discover_filter(resource_t *rsc, node_t *node) { GListPtr gIter = rsc->children; resource_t *top = uber_parent(rsc); node_t *match; if (rsc->exclusive_discover == FALSE && top->exclusive_discover == FALSE) { return; } for (; gIter != NULL; gIter = gIter->next) { resource_t *child_rsc = (resource_t *) gIter->data; rsc_discover_filter(child_rsc, node); } match = g_hash_table_lookup(rsc->allowed_nodes, node->details->id); if (match && match->rsc_discover_mode != pe_discover_exclusive) { match->weight = -INFINITY; } } /* * Count how many valid nodes we have (so we know the maximum number of * colors we can resolve). * * Apply node constraints (i.e. filter the "allowed_nodes" part of resources) */ gboolean stage2(pe_working_set_t * data_set) { GListPtr gIter = NULL; crm_trace("Applying placement constraints"); gIter = data_set->nodes; for (; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; if (node == NULL) { /* error */ } else if (node->weight >= 0.0 /* global weight */ && node->details->online && node->details->type != node_ping) { data_set->max_valid_nodes++; } } apply_placement_constraints(data_set); gIter = data_set->nodes; for (; gIter != NULL; gIter = gIter->next) { GListPtr gIter2 = NULL; node_t *node = (node_t *) gIter->data; gIter2 = data_set->resources; for (; gIter2 != NULL; gIter2 = gIter2->next) { resource_t *rsc = (resource_t *) gIter2->data; common_apply_stickiness(rsc, node, data_set); rsc_discover_filter(rsc, node); } } return TRUE; } /* * Create internal resource constraints before allocation */ gboolean stage3(pe_working_set_t * data_set) { GListPtr gIter = data_set->resources; for (; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; rsc->cmds->internal_constraints(rsc, data_set); } return TRUE; } /* * Check for orphaned or redefined actions */ gboolean stage4(pe_working_set_t * data_set) { check_actions(data_set); return TRUE; } static void * convert_const_pointer(const void *ptr) { /* Worst function ever */ return (void *)ptr; } static gint sort_rsc_process_order(gconstpointer a, gconstpointer b, gpointer data) { int rc = 0; int r1_weight = -INFINITY; int r2_weight = -INFINITY; const char *reason = "existence"; const GListPtr nodes = (GListPtr) data; const resource_t *resource1 = a; const resource_t *resource2 = b; node_t *r1_node = NULL; node_t *r2_node = NULL; GListPtr gIter = NULL; GHashTable *r1_nodes = NULL; GHashTable *r2_nodes = NULL; if (a == NULL && b == NULL) { goto done; } if (a == NULL) { return 1; } if (b == NULL) { return -1; } reason = "priority"; r1_weight = resource1->priority; r2_weight = resource2->priority; if (r1_weight > r2_weight) { rc = -1; goto done; } if (r1_weight < r2_weight) { rc = 1; goto done; } reason = "no node list"; if (nodes == NULL) { goto done; } r1_nodes = rsc_merge_weights(convert_const_pointer(resource1), resource1->id, NULL, NULL, 1, pe_weights_forward | pe_weights_init); dump_node_scores(LOG_TRACE, NULL, resource1->id, r1_nodes); r2_nodes = rsc_merge_weights(convert_const_pointer(resource2), resource2->id, NULL, NULL, 1, pe_weights_forward | pe_weights_init); dump_node_scores(LOG_TRACE, NULL, resource2->id, r2_nodes); /* Current location score */ reason = "current location"; r1_weight = -INFINITY; r2_weight = -INFINITY; if (resource1->running_on) { r1_node = pe__current_node(resource1); r1_node = g_hash_table_lookup(r1_nodes, r1_node->details->id); if (r1_node != NULL) { r1_weight = r1_node->weight; } } if (resource2->running_on) { r2_node = pe__current_node(resource2); r2_node = g_hash_table_lookup(r2_nodes, r2_node->details->id); if (r2_node != NULL) { r2_weight = r2_node->weight; } } if (r1_weight > r2_weight) { rc = -1; goto done; } if (r1_weight < r2_weight) { rc = 1; goto done; } reason = "score"; for (gIter = nodes; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; r1_node = NULL; r2_node = NULL; r1_weight = -INFINITY; if (r1_nodes) { r1_node = g_hash_table_lookup(r1_nodes, node->details->id); } if (r1_node) { r1_weight = r1_node->weight; } r2_weight = -INFINITY; if (r2_nodes) { r2_node = g_hash_table_lookup(r2_nodes, node->details->id); } if (r2_node) { r2_weight = r2_node->weight; } if (r1_weight > r2_weight) { rc = -1; goto done; } if (r1_weight < r2_weight) { rc = 1; goto done; } } done: crm_trace("%s (%d) on %s %c %s (%d) on %s: %s", resource1->id, r1_weight, r1_node ? r1_node->details->id : "n/a", rc < 0 ? '>' : rc > 0 ? '<' : '=', resource2->id, r2_weight, r2_node ? r2_node->details->id : "n/a", reason); if (r1_nodes) { g_hash_table_destroy(r1_nodes); } if (r2_nodes) { g_hash_table_destroy(r2_nodes); } return rc; } static void allocate_resources(pe_working_set_t * data_set) { GListPtr gIter = NULL; if (is_set(data_set->flags, pe_flag_have_remote_nodes)) { /* Force remote connection resources to be allocated first. This * also forces any colocation dependencies to be allocated as well */ for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; if (rsc->is_remote_node == FALSE) { continue; } pe_rsc_trace(rsc, "Allocating: %s", rsc->id); /* For remote node connection resources, always prefer the partial * migration target during resource allocation, if the rsc is in the * middle of a migration. */ rsc->cmds->allocate(rsc, rsc->partial_migration_target, data_set); } } /* now do the rest of the resources */ for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; if (rsc->is_remote_node == TRUE) { continue; } pe_rsc_trace(rsc, "Allocating: %s", rsc->id); rsc->cmds->allocate(rsc, NULL, data_set); } } /* We always use pe_order_preserve with these convenience functions to exempt * internally generated constraints from the prohibition of user constraints * involving remote connection resources. * * The start ordering additionally uses pe_order_runnable_left so that the * specified action is not runnable if the start is not runnable. */ static inline void order_start_then_action(resource_t *lh_rsc, action_t *rh_action, enum pe_ordering extra, pe_working_set_t *data_set) { if (lh_rsc && rh_action && data_set) { custom_action_order(lh_rsc, start_key(lh_rsc), NULL, rh_action->rsc, NULL, rh_action, pe_order_preserve | pe_order_runnable_left | extra, data_set); } } static inline void order_action_then_stop(action_t *lh_action, resource_t *rh_rsc, enum pe_ordering extra, pe_working_set_t *data_set) { if (lh_action && rh_rsc && data_set) { custom_action_order(lh_action->rsc, NULL, lh_action, rh_rsc, stop_key(rh_rsc), NULL, pe_order_preserve | extra, data_set); } } static void cleanup_orphans(resource_t * rsc, pe_working_set_t * data_set) { GListPtr gIter = NULL; if (is_set(data_set->flags, pe_flag_stop_rsc_orphans) == FALSE) { return; } /* Don't recurse into ->children, those are just unallocated clone instances */ if(is_not_set(rsc->flags, pe_rsc_orphan)) { return; } for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; if (node->details->online && pe_get_failcount(node, rsc, NULL, pe_fc_effective, NULL, data_set)) { pe_action_t *clear_op = NULL; clear_op = pe__clear_failcount(rsc, node, "it is orphaned", data_set); /* We can't use order_action_then_stop() here because its * pe_order_preserve breaks things */ custom_action_order(clear_op->rsc, NULL, clear_op, rsc, stop_key(rsc), NULL, pe_order_optional, data_set); } } } gboolean stage5(pe_working_set_t * data_set) { GListPtr gIter = NULL; if (safe_str_neq(data_set->placement_strategy, "default")) { GListPtr nodes = g_list_copy(data_set->nodes); nodes = sort_nodes_by_weight(nodes, NULL, data_set); data_set->resources = g_list_sort_with_data(data_set->resources, sort_rsc_process_order, nodes); g_list_free(nodes); } gIter = data_set->nodes; for (; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; dump_node_capacity(show_utilization ? 0 : utilization_log_level, "Original", node); } crm_trace("Allocating services"); /* Take (next) highest resource, assign it and create its actions */ allocate_resources(data_set); gIter = data_set->nodes; for (; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; dump_node_capacity(show_utilization ? 0 : utilization_log_level, "Remaining", node); } // Process deferred action checks pe__foreach_param_check(data_set, check_params); pe__free_param_checks(data_set); if (is_set(data_set->flags, pe_flag_startup_probes)) { crm_trace("Calculating needed probes"); /* This code probably needs optimization * ptest -x with 100 nodes, 100 clones and clone-max=100: With probes: ptest[14781]: 2010/09/27_17:56:46 notice: TRACE: do_calculations: pengine.c:258 Calculate cluster status ptest[14781]: 2010/09/27_17:56:46 notice: TRACE: do_calculations: pengine.c:278 Applying placement constraints ptest[14781]: 2010/09/27_17:56:47 notice: TRACE: do_calculations: pengine.c:285 Create internal constraints ptest[14781]: 2010/09/27_17:56:47 notice: TRACE: do_calculations: pengine.c:292 Check actions ptest[14781]: 2010/09/27_17:56:48 notice: TRACE: do_calculations: pengine.c:299 Allocate resources ptest[14781]: 2010/09/27_17:56:48 notice: TRACE: stage5: allocate.c:881 Allocating services ptest[14781]: 2010/09/27_17:56:49 notice: TRACE: stage5: allocate.c:894 Calculating needed probes ptest[14781]: 2010/09/27_17:56:51 notice: TRACE: stage5: allocate.c:899 Creating actions ptest[14781]: 2010/09/27_17:56:52 notice: TRACE: stage5: allocate.c:905 Creating done ptest[14781]: 2010/09/27_17:56:52 notice: TRACE: do_calculations: pengine.c:306 Processing fencing and shutdown cases ptest[14781]: 2010/09/27_17:56:52 notice: TRACE: do_calculations: pengine.c:313 Applying ordering constraints 36s ptest[14781]: 2010/09/27_17:57:28 notice: TRACE: do_calculations: pengine.c:320 Create transition graph Without probes: ptest[14637]: 2010/09/27_17:56:21 notice: TRACE: do_calculations: pengine.c:258 Calculate cluster status ptest[14637]: 2010/09/27_17:56:22 notice: TRACE: do_calculations: pengine.c:278 Applying placement constraints ptest[14637]: 2010/09/27_17:56:22 notice: TRACE: do_calculations: pengine.c:285 Create internal constraints ptest[14637]: 2010/09/27_17:56:22 notice: TRACE: do_calculations: pengine.c:292 Check actions ptest[14637]: 2010/09/27_17:56:23 notice: TRACE: do_calculations: pengine.c:299 Allocate resources ptest[14637]: 2010/09/27_17:56:23 notice: TRACE: stage5: allocate.c:881 Allocating services ptest[14637]: 2010/09/27_17:56:24 notice: TRACE: stage5: allocate.c:899 Creating actions ptest[14637]: 2010/09/27_17:56:25 notice: TRACE: stage5: allocate.c:905 Creating done ptest[14637]: 2010/09/27_17:56:25 notice: TRACE: do_calculations: pengine.c:306 Processing fencing and shutdown cases ptest[14637]: 2010/09/27_17:56:25 notice: TRACE: do_calculations: pengine.c:313 Applying ordering constraints ptest[14637]: 2010/09/27_17:56:25 notice: TRACE: do_calculations: pengine.c:320 Create transition graph */ probe_resources(data_set); } crm_trace("Handle orphans"); for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; cleanup_orphans(rsc, data_set); } crm_trace("Creating actions"); for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; rsc->cmds->create_actions(rsc, data_set); } crm_trace("Creating done"); return TRUE; } static gboolean is_managed(const resource_t * rsc) { GListPtr gIter = rsc->children; if (is_set(rsc->flags, pe_rsc_managed)) { return TRUE; } for (; gIter != NULL; gIter = gIter->next) { resource_t *child_rsc = (resource_t *) gIter->data; if (is_managed(child_rsc)) { return TRUE; } } return FALSE; } static gboolean any_managed_resources(pe_working_set_t * data_set) { GListPtr gIter = data_set->resources; for (; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; if (is_managed(rsc)) { return TRUE; } } return FALSE; } /*! * \internal * \brief Create pseudo-op for guest node fence, and order relative to it * * \param[in] node Guest node to fence * \param[in] data_set Working set of CIB state */ static void fence_guest(pe_node_t *node, pe_working_set_t *data_set) { resource_t *container = node->details->remote_rsc->container; pe_action_t *stop = NULL; pe_action_t *stonith_op = NULL; /* The fence action is just a label; we don't do anything differently for * off vs. reboot. We specify it explicitly, rather than let it default to * cluster's default action, because we are not _initiating_ fencing -- we * are creating a pseudo-event to describe fencing that is already occurring * by other means (container recovery). */ const char *fence_action = "off"; /* Check whether guest's container resource has any explicit stop or * start (the stop may be implied by fencing of the guest's host). */ if (container) { stop = find_first_action(container->actions, NULL, CRMD_ACTION_STOP, NULL); if (find_first_action(container->actions, NULL, CRMD_ACTION_START, NULL)) { fence_action = "reboot"; } } /* Create a fence pseudo-event, so we have an event to order actions * against, and the controller can always detect it. */ stonith_op = pe_fence_op(node, fence_action, FALSE, "guest is unclean", data_set); update_action_flags(stonith_op, pe_action_pseudo | pe_action_runnable, __FUNCTION__, __LINE__); /* We want to imply stops/demotes after the guest is stopped, not wait until * it is restarted, so we always order pseudo-fencing after stop, not start * (even though start might be closer to what is done for a real reboot). */ if(stop && is_set(stop->flags, pe_action_pseudo)) { pe_action_t *parent_stonith_op = pe_fence_op(stop->node, NULL, FALSE, NULL, data_set); crm_info("Implying guest node %s is down (action %d) after %s fencing", node->details->uname, stonith_op->id, stop->node->details->uname); order_actions(parent_stonith_op, stonith_op, pe_order_runnable_left|pe_order_implies_then); } else if (stop) { order_actions(stop, stonith_op, pe_order_runnable_left|pe_order_implies_then); crm_info("Implying guest node %s is down (action %d) " "after container %s is stopped (action %d)", node->details->uname, stonith_op->id, container->id, stop->id); } else { /* If we're fencing the guest node but there's no stop for the guest * resource, we must think the guest is already stopped. However, we may * think so because its resource history was just cleaned. To avoid * unnecessarily considering the guest node down if it's really up, * order the pseudo-fencing after any stop of the connection resource, * which will be ordered after any container (re-)probe. */ stop = find_first_action(node->details->remote_rsc->actions, NULL, RSC_STOP, NULL); if (stop) { order_actions(stop, stonith_op, pe_order_optional); crm_info("Implying guest node %s is down (action %d) " "after connection is stopped (action %d)", node->details->uname, stonith_op->id, stop->id); } else { /* Not sure why we're fencing, but everything must already be * cleanly stopped. */ crm_info("Implying guest node %s is down (action %d) ", node->details->uname, stonith_op->id); } } /* Order/imply other actions relative to pseudo-fence as with real fence */ stonith_constraints(node, stonith_op, data_set); } /* * Create dependencies for stonith and shutdown operations */ gboolean stage6(pe_working_set_t * data_set) { action_t *dc_down = NULL; action_t *stonith_op = NULL; gboolean integrity_lost = FALSE; gboolean need_stonith = TRUE; GListPtr gIter; GListPtr stonith_ops = NULL; GList *shutdown_ops = NULL; /* Remote ordering constraints need to happen prior to calculate * fencing because it is one more place we will mark the node as * dirty. * * A nice side-effect of doing it first is that we can remove a * bunch of special logic from apply_*_ordering() because its * already part of pe_fence_node() */ crm_trace("Creating remote ordering constraints"); apply_remote_node_ordering(data_set); crm_trace("Processing fencing and shutdown cases"); if (any_managed_resources(data_set) == FALSE) { crm_notice("Delaying fencing operations until there are resources to manage"); need_stonith = FALSE; } /* Check each node for stonith/shutdown */ for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; /* Guest nodes are "fenced" by recovering their container resource, * so handle them separately. */ if (pe__is_guest_node(node)) { if (node->details->remote_requires_reset && need_stonith) { fence_guest(node, data_set); } continue; } stonith_op = NULL; if (node->details->unclean && need_stonith && pe_can_fence(data_set, node)) { stonith_op = pe_fence_op(node, NULL, FALSE, "node is unclean", data_set); pe_warn("Scheduling Node %s for STONITH", node->details->uname); stonith_constraints(node, stonith_op, data_set); if (node->details->is_dc) { // Remember if the DC is being fenced dc_down = stonith_op; } else { if (is_not_set(data_set->flags, pe_flag_concurrent_fencing) && (stonith_ops != NULL)) { /* Concurrent fencing is disabled, so order each non-DC * fencing in a chain. If there is any DC fencing or * shutdown, it will be ordered after the last action in the * chain later. */ order_actions((pe_action_t *) stonith_ops->data, stonith_op, pe_order_optional); } // Remember all non-DC fencing actions in a separate list stonith_ops = g_list_prepend(stonith_ops, stonith_op); } } else if (node->details->online && node->details->shutdown && /* TODO define what a shutdown op means for a remote node. * For now we do not send shutdown operations for remote nodes, but * if we can come up with a good use for this in the future, we will. */ pe__is_guest_or_remote_node(node) == FALSE) { action_t *down_op = sched_shutdown_op(node, data_set); if (node->details->is_dc) { // Remember if the DC is being shut down dc_down = down_op; } else { // Remember non-DC shutdowns for later ordering shutdown_ops = g_list_prepend(shutdown_ops, down_op); } } if (node->details->unclean && stonith_op == NULL) { integrity_lost = TRUE; pe_warn("Node %s is unclean!", node->details->uname); } } if (integrity_lost) { if (is_set(data_set->flags, pe_flag_stonith_enabled) == FALSE) { pe_warn("YOUR RESOURCES ARE NOW LIKELY COMPROMISED"); pe_err("ENABLE STONITH TO KEEP YOUR RESOURCES SAFE"); } else if (is_set(data_set->flags, pe_flag_have_quorum) == FALSE) { crm_notice("Cannot fence unclean nodes until quorum is" " attained (or no-quorum-policy is set to ignore)"); } } if (dc_down != NULL) { /* Order any non-DC shutdowns before any DC shutdown, to avoid repeated * DC elections. However, we don't want to order non-DC shutdowns before * a DC *fencing*, because even though we don't want a node that's * shutting down to become DC, the DC fencing could be ordered before a * clone stop that's also ordered before the shutdowns, thus leading to * a graph loop. */ if (safe_str_eq(dc_down->task, CRM_OP_SHUTDOWN)) { for (gIter = shutdown_ops; gIter != NULL; gIter = gIter->next) { action_t *node_stop = (action_t *) gIter->data; crm_debug("Ordering shutdown on %s before %s on DC %s", node_stop->node->details->uname, dc_down->task, dc_down->node->details->uname); order_actions(node_stop, dc_down, pe_order_optional); } } // Order any non-DC fencing before any DC fencing or shutdown if (is_set(data_set->flags, pe_flag_concurrent_fencing)) { /* With concurrent fencing, order each non-DC fencing action * separately before any DC fencing or shutdown. */ for (gIter = stonith_ops; gIter != NULL; gIter = gIter->next) { order_actions((pe_action_t *) gIter->data, dc_down, pe_order_optional); } } else if (stonith_ops) { /* Without concurrent fencing, the non-DC fencing actions are * already ordered relative to each other, so we just need to order * the DC fencing after the last action in the chain (which is the * first item in the list). */ order_actions((pe_action_t *) stonith_ops->data, dc_down, pe_order_optional); } } g_list_free(stonith_ops); g_list_free(shutdown_ops); return TRUE; } /* * Determine the sets of independent actions and the correct order for the * actions in each set. * * Mark dependencies of un-runnable actions un-runnable * */ static GListPtr find_actions_by_task(GListPtr actions, resource_t * rsc, const char *original_key) { GListPtr list = NULL; list = find_actions(actions, original_key, NULL); if (list == NULL) { /* we're potentially searching a child of the original resource */ char *key = NULL; char *task = NULL; guint interval_ms = 0; if (parse_op_key(original_key, NULL, &task, &interval_ms)) { key = generate_op_key(rsc->id, task, interval_ms); list = find_actions(actions, key, NULL); } else { crm_err("search key: %s", original_key); } free(key); free(task); } return list; } static void rsc_order_then(pe_action_t *lh_action, pe_resource_t *rsc, pe__ordering_t *order) { GListPtr gIter = NULL; GListPtr rh_actions = NULL; action_t *rh_action = NULL; enum pe_ordering type; CRM_CHECK(rsc != NULL, return); CRM_CHECK(order != NULL, return); type = order->type; rh_action = order->rh_action; crm_trace("Processing RH of ordering constraint %d", order->id); if (rh_action != NULL) { rh_actions = g_list_prepend(NULL, rh_action); } else if (rsc != NULL) { rh_actions = find_actions_by_task(rsc->actions, rsc, order->rh_action_task); } if (rh_actions == NULL) { pe_rsc_trace(rsc, "No RH-Side (%s/%s) found for constraint..." " ignoring", rsc->id, order->rh_action_task); if (lh_action) { pe_rsc_trace(rsc, "LH-Side was: %s", lh_action->uuid); } return; } if (lh_action && lh_action->rsc == rsc && is_set(lh_action->flags, pe_action_dangle)) { pe_rsc_trace(rsc, "Detected dangling operation %s -> %s", lh_action->uuid, order->rh_action_task); clear_bit(type, pe_order_implies_then); } gIter = rh_actions; for (; gIter != NULL; gIter = gIter->next) { action_t *rh_action_iter = (action_t *) gIter->data; if (lh_action) { order_actions(lh_action, rh_action_iter, type); } else if (type & pe_order_implies_then) { update_action_flags(rh_action_iter, pe_action_runnable | pe_action_clear, __FUNCTION__, __LINE__); crm_warn("Unrunnable %s 0x%.6x", rh_action_iter->uuid, type); } else { crm_warn("neither %s 0x%.6x", rh_action_iter->uuid, type); } } g_list_free(rh_actions); } static void rsc_order_first(pe_resource_t *lh_rsc, pe__ordering_t *order, pe_working_set_t *data_set) { GListPtr gIter = NULL; GListPtr lh_actions = NULL; action_t *lh_action = order->lh_action; resource_t *rh_rsc = order->rh_rsc; crm_trace("Processing LH of ordering constraint %d", order->id); CRM_ASSERT(lh_rsc != NULL); if (lh_action != NULL) { lh_actions = g_list_prepend(NULL, lh_action); } else { lh_actions = find_actions_by_task(lh_rsc->actions, lh_rsc, order->lh_action_task); } if (lh_actions == NULL && lh_rsc != rh_rsc) { char *key = NULL; char *op_type = NULL; guint interval_ms = 0; parse_op_key(order->lh_action_task, NULL, &op_type, &interval_ms); key = generate_op_key(lh_rsc->id, op_type, interval_ms); if (lh_rsc->fns->state(lh_rsc, TRUE) == RSC_ROLE_STOPPED && safe_str_eq(op_type, RSC_STOP)) { free(key); pe_rsc_trace(lh_rsc, "No LH-Side (%s/%s) found for constraint %d with %s - ignoring", lh_rsc->id, order->lh_action_task, order->id, order->rh_action_task); } else if (lh_rsc->fns->state(lh_rsc, TRUE) == RSC_ROLE_SLAVE && safe_str_eq(op_type, RSC_DEMOTE)) { free(key); pe_rsc_trace(lh_rsc, "No LH-Side (%s/%s) found for constraint %d with %s - ignoring", lh_rsc->id, order->lh_action_task, order->id, order->rh_action_task); } else { pe_rsc_trace(lh_rsc, "No LH-Side (%s/%s) found for constraint %d with %s - creating", lh_rsc->id, order->lh_action_task, order->id, order->rh_action_task); lh_action = custom_action(lh_rsc, key, op_type, NULL, TRUE, TRUE, data_set); lh_actions = g_list_prepend(NULL, lh_action); } free(op_type); } gIter = lh_actions; for (; gIter != NULL; gIter = gIter->next) { action_t *lh_action_iter = (action_t *) gIter->data; if (rh_rsc == NULL && order->rh_action) { rh_rsc = order->rh_action->rsc; } if (rh_rsc) { rsc_order_then(lh_action_iter, rh_rsc, order); } else if (order->rh_action) { order_actions(lh_action_iter, order->rh_action, order->type); } } g_list_free(lh_actions); } extern void update_colo_start_chain(pe_action_t *action, pe_working_set_t *data_set); static int is_recurring_action(action_t *action) { const char *interval_ms_s = g_hash_table_lookup(action->meta, XML_LRM_ATTR_INTERVAL_MS); guint interval_ms = crm_parse_ms(interval_ms_s); return (interval_ms > 0); } static void apply_container_ordering(action_t *action, pe_working_set_t *data_set) { /* VMs are also classified as containers for these purposes... in * that they both involve a 'thing' running on a real or remote * cluster node. * * This allows us to be smarter about the type and extent of * recovery actions required in various scenarios */ resource_t *remote_rsc = NULL; resource_t *container = NULL; enum action_tasks task = text2task(action->task); CRM_ASSERT(action->rsc); CRM_ASSERT(action->node); CRM_ASSERT(pe__is_guest_or_remote_node(action->node)); remote_rsc = action->node->details->remote_rsc; CRM_ASSERT(remote_rsc); container = remote_rsc->container; CRM_ASSERT(container); if(is_set(container->flags, pe_rsc_failed)) { pe_fence_node(data_set, action->node, "container failed"); } crm_trace("Order %s action %s relative to %s%s for %s%s", action->task, action->uuid, is_set(remote_rsc->flags, pe_rsc_failed)? "failed " : "", remote_rsc->id, is_set(container->flags, pe_rsc_failed)? "failed " : "", container->id); if (safe_str_eq(action->task, CRMD_ACTION_MIGRATE) || safe_str_eq(action->task, CRMD_ACTION_MIGRATED)) { /* Migration ops map to "no_action", but we need to apply the same * ordering as for stop or demote (see get_router_node()). */ task = stop_rsc; } switch (task) { case start_rsc: case action_promote: /* Force resource recovery if the container is recovered */ order_start_then_action(container, action, pe_order_implies_then, data_set); /* Wait for the connection resource to be up too */ order_start_then_action(remote_rsc, action, pe_order_none, data_set); break; case stop_rsc: case action_demote: if (is_set(container->flags, pe_rsc_failed)) { /* When the container representing a guest node fails, any stop * or demote actions for resources running on the guest node * are implied by the container stopping. This is similar to * how fencing operations work for cluster nodes and remote * nodes. */ } else { /* Ensure the operation happens before the connection is brought * down. * * If we really wanted to, we could order these after the * connection start, IFF the container's current role was * stopped (otherwise we re-introduce an ordering loop when the * connection is restarting). */ order_action_then_stop(action, remote_rsc, pe_order_none, data_set); } break; default: /* Wait for the connection resource to be up */ if (is_recurring_action(action)) { /* In case we ever get the recovery logic wrong, force * recurring monitors to be restarted, even if just * the connection was re-established */ if(task != no_action) { order_start_then_action(remote_rsc, action, pe_order_implies_then, data_set); } } else { order_start_then_action(remote_rsc, action, pe_order_none, data_set); } break; } } static enum remote_connection_state get_remote_node_state(pe_node_t *node) { resource_t *remote_rsc = NULL; node_t *cluster_node = NULL; CRM_ASSERT(node); remote_rsc = node->details->remote_rsc; CRM_ASSERT(remote_rsc); cluster_node = pe__current_node(remote_rsc); /* If the cluster node the remote connection resource resides on * is unclean or went offline, we can't process any operations * on that remote node until after it starts elsewhere. */ if(remote_rsc->next_role == RSC_ROLE_STOPPED || remote_rsc->allocated_to == NULL) { /* The connection resource is not going to run anywhere */ if (cluster_node && cluster_node->details->unclean) { /* The remote connection is failed because its resource is on a * failed node and can't be recovered elsewhere, so we must fence. */ return remote_state_failed; } if (is_not_set(remote_rsc->flags, pe_rsc_failed)) { /* Connection resource is cleanly stopped */ return remote_state_stopped; } /* Connection resource is failed */ if ((remote_rsc->next_role == RSC_ROLE_STOPPED) && remote_rsc->remote_reconnect_ms && node->details->remote_was_fenced) { /* We won't know whether the connection is recoverable until the * reconnect interval expires and we reattempt connection. */ return remote_state_unknown; } /* The remote connection is in a failed state. If there are any * resources known to be active on it (stop) or in an unknown state * (probe), we must assume the worst and fence it. */ return remote_state_failed; } else if (cluster_node == NULL) { /* Connection is recoverable but not currently running anywhere, see if we can recover it first */ return remote_state_unknown; } else if(cluster_node->details->unclean == TRUE || cluster_node->details->online == FALSE) { /* Connection is running on a dead node, see if we can recover it first */ return remote_state_resting; } else if (g_list_length(remote_rsc->running_on) > 1 && remote_rsc->partial_migration_source && remote_rsc->partial_migration_target) { /* We're in the middle of migrating a connection resource, * wait until after the resource migrates before performing * any actions. */ return remote_state_resting; } return remote_state_alive; } /*! * \internal * \brief Order actions on remote node relative to actions for the connection */ static void apply_remote_ordering(action_t *action, pe_working_set_t *data_set) { resource_t *remote_rsc = NULL; enum action_tasks task = text2task(action->task); enum remote_connection_state state = get_remote_node_state(action->node); enum pe_ordering order_opts = pe_order_none; if (action->rsc == NULL) { return; } CRM_ASSERT(action->node); CRM_ASSERT(pe__is_guest_or_remote_node(action->node)); remote_rsc = action->node->details->remote_rsc; CRM_ASSERT(remote_rsc); crm_trace("Order %s action %s relative to %s%s (state: %s)", action->task, action->uuid, is_set(remote_rsc->flags, pe_rsc_failed)? "failed " : "", remote_rsc->id, state2text(state)); if (safe_str_eq(action->task, CRMD_ACTION_MIGRATE) || safe_str_eq(action->task, CRMD_ACTION_MIGRATED)) { /* Migration ops map to "no_action", but we need to apply the same * ordering as for stop or demote (see get_router_node()). */ task = stop_rsc; } switch (task) { case start_rsc: case action_promote: order_opts = pe_order_none; if (state == remote_state_failed) { /* Force recovery, by making this action required */ order_opts |= pe_order_implies_then; } /* Ensure connection is up before running this action */ order_start_then_action(remote_rsc, action, order_opts, data_set); break; case stop_rsc: if(state == remote_state_alive) { order_action_then_stop(action, remote_rsc, pe_order_implies_first, data_set); } else if(state == remote_state_failed) { /* We would only be here if the resource is * running on the remote node. Since we have no * way to stop it, it is necessary to fence the * node. */ pe_fence_node(data_set, action->node, "resources are active and the connection is unrecoverable"); order_action_then_stop(action, remote_rsc, pe_order_implies_first, data_set); } else if(remote_rsc->next_role == RSC_ROLE_STOPPED) { /* State must be remote_state_unknown or remote_state_stopped. * Since the connection is not coming back up in this * transition, stop this resource first. */ order_action_then_stop(action, remote_rsc, pe_order_implies_first, data_set); } else { /* The connection is going to be started somewhere else, so * stop this resource after that completes. */ order_start_then_action(remote_rsc, action, pe_order_none, data_set); } break; case action_demote: /* Only order this demote relative to the connection start if the * connection isn't being torn down. Otherwise, the demote would be * blocked because the connection start would not be allowed. */ if(state == remote_state_resting || state == remote_state_unknown) { order_start_then_action(remote_rsc, action, pe_order_none, data_set); } /* Otherwise we can rely on the stop ordering */ break; default: /* Wait for the connection resource to be up */ if (is_recurring_action(action)) { /* In case we ever get the recovery logic wrong, force * recurring monitors to be restarted, even if just * the connection was re-established */ order_start_then_action(remote_rsc, action, pe_order_implies_then, data_set); } else { node_t *cluster_node = pe__current_node(remote_rsc); if(task == monitor_rsc && state == remote_state_failed) { /* We would only be here if we do not know the * state of the resource on the remote node. * Since we have no way to find out, it is * necessary to fence the node. */ pe_fence_node(data_set, action->node, "resources are in an unknown state and the connection is unrecoverable"); } if(cluster_node && state == remote_state_stopped) { /* The connection is currently up, but is going * down permanently. * * Make sure we check services are actually * stopped _before_ we let the connection get * closed */ order_action_then_stop(action, remote_rsc, pe_order_runnable_left, data_set); } else { order_start_then_action(remote_rsc, action, pe_order_none, data_set); } } break; } } static void apply_remote_node_ordering(pe_working_set_t *data_set) { if (is_set(data_set->flags, pe_flag_have_remote_nodes) == FALSE) { return; } for (GListPtr gIter = data_set->actions; gIter != NULL; gIter = gIter->next) { action_t *action = (action_t *) gIter->data; resource_t *remote = NULL; // We are only interested in resource actions if (action->rsc == NULL) { continue; } /* Special case: If we are clearing the failcount of an actual * remote connection resource, then make sure this happens before * any start of the resource in this transition. */ if (action->rsc->is_remote_node && safe_str_eq(action->task, CRM_OP_CLEAR_FAILCOUNT)) { custom_action_order(action->rsc, NULL, action, action->rsc, generate_op_key(action->rsc->id, RSC_START, 0), NULL, pe_order_optional, data_set); continue; } // We are only interested in actions allocated to a node if (action->node == NULL) { continue; } if (!pe__is_guest_or_remote_node(action->node)) { continue; } /* We are only interested in real actions. * * @TODO This is probably wrong; pseudo-actions might be converted to * real actions and vice versa later in update_actions() at the end of * stage7(). */ if (is_set(action->flags, pe_action_pseudo)) { continue; } remote = action->node->details->remote_rsc; if (remote == NULL) { // Orphaned continue; } /* Another special case: if a resource is moving to a Pacemaker Remote * node, order the stop on the original node after any start of the * remote connection. This ensures that if the connection fails to * start, we leave the resource running on the original node. */ if (safe_str_eq(action->task, RSC_START)) { for (GList *item = action->rsc->actions; item != NULL; item = item->next) { pe_action_t *rsc_action = item->data; if ((rsc_action->node->details != action->node->details) && safe_str_eq(rsc_action->task, RSC_STOP)) { custom_action_order(remote, start_key(remote), NULL, action->rsc, NULL, rsc_action, pe_order_optional, data_set); } } } /* The action occurs across a remote connection, so create * ordering constraints that guarantee the action occurs while the node * is active (after start, before stop ... things like that). * * This is somewhat brittle in that we need to make sure the results of * this ordering are compatible with the result of get_router_node(). * It would probably be better to add XML_LRM_ATTR_ROUTER_NODE as part * of this logic rather than action2xml(). */ if (remote->container) { crm_trace("Container ordering for %s", action->uuid); apply_container_ordering(action, data_set); } else { crm_trace("Remote ordering for %s", action->uuid); apply_remote_ordering(action, data_set); } } } static gboolean order_first_probe_unneeded(pe_action_t * probe, pe_action_t * rh_action) { /* No need to probe the resource on the node that is being * unfenced. Otherwise it might introduce transition loop * since probe will be performed after the node is * unfenced. */ if (safe_str_eq(rh_action->task, CRM_OP_FENCE) && probe->node && rh_action->node && probe->node->details == rh_action->node->details) { const char *op = g_hash_table_lookup(rh_action->meta, "stonith_action"); if (safe_str_eq(op, "on")) { return TRUE; } } // Shutdown waits for probe to complete only if it's on the same node if ((safe_str_eq(rh_action->task, CRM_OP_SHUTDOWN)) && probe->node && rh_action->node && probe->node->details != rh_action->node->details) { return TRUE; } return FALSE; } static void order_first_probes(pe_working_set_t * data_set) { GListPtr gIter = NULL; for (gIter = data_set->ordering_constraints; gIter != NULL; gIter = gIter->next) { pe__ordering_t *order = gIter->data; enum pe_ordering order_type = pe_order_optional; pe_resource_t *lh_rsc = order->lh_rsc; pe_resource_t *rh_rsc = order->rh_rsc; pe_action_t *lh_action = order->lh_action; pe_action_t *rh_action = order->rh_action; const char *lh_action_task = order->lh_action_task; const char *rh_action_task = order->rh_action_task; GListPtr probes = NULL; GListPtr rh_actions = NULL; GListPtr pIter = NULL; if (lh_rsc == NULL) { continue; } else if (rh_rsc && lh_rsc == rh_rsc) { continue; } if (lh_action == NULL && lh_action_task == NULL) { continue; } if (rh_action == NULL && rh_action_task == NULL) { continue; } /* Technically probe is expected to return "not running", which could be * the alternative of stop action if the status of the resource is * unknown yet. */ if (lh_action && safe_str_neq(lh_action->task, RSC_STOP)) { continue; } else if (lh_action == NULL && lh_action_task && crm_ends_with(lh_action_task, "_" RSC_STOP "_0") == FALSE) { continue; } /* Do not probe the resource inside of a stopping container. Otherwise * it might introduce transition loop since probe will be performed * after the container starts again. */ if (rh_rsc && lh_rsc->container == rh_rsc) { if (rh_action && safe_str_eq(rh_action->task, RSC_STOP)) { continue; } else if (rh_action == NULL && rh_action_task && crm_ends_with(rh_action_task,"_" RSC_STOP "_0")) { continue; } } if (order->type == pe_order_none) { continue; } // Preserve the order options for future filtering if (is_set(order->type, pe_order_apply_first_non_migratable)) { set_bit(order_type, pe_order_apply_first_non_migratable); } if (is_set(order->type, pe_order_same_node)) { set_bit(order_type, pe_order_same_node); } // Keep the order types for future filtering if (order->type == pe_order_anti_colocation || order->type == pe_order_load) { order_type = order->type; } probes = pe__resource_actions(lh_rsc, NULL, RSC_STATUS, FALSE); if (probes == NULL) { continue; } if (rh_action) { rh_actions = g_list_prepend(rh_actions, rh_action); } else if (rh_rsc && rh_action_task) { rh_actions = find_actions(rh_rsc->actions, rh_action_task, NULL); } if (rh_actions == NULL) { g_list_free(probes); continue; } crm_trace("Processing for LH probe based on ordering constraint %s -> %s" " (id=%d, type=%.6x)", lh_action ? lh_action->uuid : lh_action_task, rh_action ? rh_action->uuid : rh_action_task, order->id, order->type); for (pIter = probes; pIter != NULL; pIter = pIter->next) { pe_action_t *probe = (pe_action_t *) pIter->data; GListPtr rIter = NULL; for (rIter = rh_actions; rIter != NULL; rIter = rIter->next) { pe_action_t *rh_action_iter = (pe_action_t *) rIter->data; if (order_first_probe_unneeded(probe, rh_action_iter)) { continue; } order_actions(probe, rh_action_iter, order_type); } } g_list_free(rh_actions); g_list_free(probes); } } static void order_then_probes(pe_working_set_t * data_set) { #if 0 GListPtr gIter = NULL; for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; /* Given "A then B", we would prefer to wait for A to be * started before probing B. * * If A was a filesystem on which the binaries and data for B * lived, it would have been useful if the author of B's agent * could assume that A is running before B.monitor will be * called. * * However we can't _only_ probe once A is running, otherwise * we'd not detect the state of B if A could not be started * for some reason. * * In practice however, we cannot even do an opportunistic * version of this because B may be moving: * * B.probe -> B.start * B.probe -> B.stop * B.stop -> B.start * A.stop -> A.start * A.start -> B.probe * * So far so good, but if we add the result of this code: * * B.stop -> A.stop * * Then we get a loop: * * B.probe -> B.stop -> A.stop -> A.start -> B.probe * * We could kill the 'B.probe -> B.stop' dependency, but that * could mean stopping B "too" soon, because B.start must wait * for the probes to complete. * * Another option is to allow it only if A is a non-unique * clone with clone-max == node-max (since we'll never be * moving it). However, we could still be stopping one * instance at the same time as starting another. * The complexity of checking for allowed conditions combined * with the ever narrowing usecase suggests that this code * should remain disabled until someone gets smarter. */ action_t *start = NULL; GListPtr actions = NULL; GListPtr probes = NULL; actions = pe__resource_actions(rsc, NULL, RSC_START, FALSE); if (actions) { start = actions->data; g_list_free(actions); } if(start == NULL) { crm_err("No start action for %s", rsc->id); continue; } probes = pe__resource_actions(rsc, NULL, RSC_STATUS, FALSE); for (actions = start->actions_before; actions != NULL; actions = actions->next) { action_wrapper_t *before = (action_wrapper_t *) actions->data; GListPtr pIter = NULL; action_t *first = before->action; resource_t *first_rsc = first->rsc; if(first->required_runnable_before) { GListPtr clone_actions = NULL; for (clone_actions = first->actions_before; clone_actions != NULL; clone_actions = clone_actions->next) { before = (action_wrapper_t *) clone_actions->data; crm_trace("Testing %s -> %s (%p) for %s", first->uuid, before->action->uuid, before->action->rsc, start->uuid); CRM_ASSERT(before->action->rsc); first_rsc = before->action->rsc; break; } } else if(safe_str_neq(first->task, RSC_START)) { crm_trace("Not a start op %s for %s", first->uuid, start->uuid); } if(first_rsc == NULL) { continue; } else if(uber_parent(first_rsc) == uber_parent(start->rsc)) { crm_trace("Same parent %s for %s", first_rsc->id, start->uuid); continue; } else if(FALSE && pe_rsc_is_clone(uber_parent(first_rsc)) == FALSE) { crm_trace("Not a clone %s for %s", first_rsc->id, start->uuid); continue; } crm_err("Applying %s before %s %d", first->uuid, start->uuid, uber_parent(first_rsc)->variant); for (pIter = probes; pIter != NULL; pIter = pIter->next) { action_t *probe = (action_t *) pIter->data; crm_err("Ordering %s before %s", first->uuid, probe->uuid); order_actions(first, probe, pe_order_optional); } } } #endif } static void order_probes(pe_working_set_t * data_set) { order_first_probes(data_set); order_then_probes(data_set); } gboolean stage7(pe_working_set_t * data_set) { GListPtr gIter = NULL; crm_trace("Applying ordering constraints"); /* Don't ask me why, but apparently they need to be processed in * the order they were created in... go figure * * Also g_list_append() has horrendous performance characteristics * So we need to use g_list_prepend() and then reverse the list here */ data_set->ordering_constraints = g_list_reverse(data_set->ordering_constraints); for (gIter = data_set->ordering_constraints; gIter != NULL; gIter = gIter->next) { pe__ordering_t *order = gIter->data; resource_t *rsc = order->lh_rsc; crm_trace("Applying ordering constraint: %d", order->id); if (rsc != NULL) { crm_trace("rsc_action-to-*"); rsc_order_first(rsc, order, data_set); continue; } rsc = order->rh_rsc; if (rsc != NULL) { crm_trace("action-to-rsc_action"); rsc_order_then(order->lh_action, rsc, order); } else { crm_trace("action-to-action"); order_actions(order->lh_action, order->rh_action, order->type); } } for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) { action_t *action = (action_t *) gIter->data; update_colo_start_chain(action, data_set); } crm_trace("Ordering probes"); order_probes(data_set); crm_trace("Updating %d actions", g_list_length(data_set->actions)); for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) { action_t *action = (action_t *) gIter->data; update_action(action, data_set); } LogNodeActions(data_set, FALSE); for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; LogActions(rsc, data_set, FALSE); } return TRUE; } -int transition_id = -1; +static int transition_id = -1; + +/*! + * \internal + * \brief Log a message after calculating a transition + * + * \param[in] filename Where transition input is stored + */ +void +pcmk__log_transition_summary(const char *filename) +{ + if (was_processing_error) { + crm_err("Calculated transition %d (with errors), saving inputs in %s", + transition_id, filename); + + } else if (was_processing_warning) { + crm_warn("Calculated transition %d (with warnings), saving inputs in %s", + transition_id, filename); + + } else { + crm_notice("Calculated transition %d, saving inputs in %s", + transition_id, filename); + } + if (crm_config_error) { + crm_notice("Configuration errors found during scheduler processing," + " please run \"crm_verify -L\" to identify issues"); + } +} /* * Create a dependency graph to send to the transitioner (via the controller) */ gboolean stage8(pe_working_set_t * data_set) { GListPtr gIter = NULL; const char *value = NULL; transition_id++; crm_trace("Creating transition graph %d.", transition_id); data_set->graph = create_xml_node(NULL, XML_TAG_GRAPH); value = pe_pref(data_set->config_hash, "cluster-delay"); crm_xml_add(data_set->graph, "cluster-delay", value); value = pe_pref(data_set->config_hash, "stonith-timeout"); crm_xml_add(data_set->graph, "stonith-timeout", value); crm_xml_add(data_set->graph, "failed-stop-offset", "INFINITY"); if (is_set(data_set->flags, pe_flag_start_failure_fatal)) { crm_xml_add(data_set->graph, "failed-start-offset", "INFINITY"); } else { crm_xml_add(data_set->graph, "failed-start-offset", "1"); } value = pe_pref(data_set->config_hash, "batch-limit"); crm_xml_add(data_set->graph, "batch-limit", value); crm_xml_add_int(data_set->graph, "transition_id", transition_id); value = pe_pref(data_set->config_hash, "migration-limit"); if (crm_int_helper(value, NULL) > 0) { crm_xml_add(data_set->graph, "migration-limit", value); } /* errors... slist_iter(action, action_t, action_list, lpc, if(action->optional == FALSE && action->runnable == FALSE) { print_action("Ignoring", action, TRUE); } ); */ gIter = data_set->resources; for (; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; pe_rsc_trace(rsc, "processing actions for rsc=%s", rsc->id); rsc->cmds->expand(rsc, data_set); } crm_log_xml_trace(data_set->graph, "created resource-driven action list"); /* pseudo action to distribute list of nodes with maintenance state update */ add_maintenance_update(data_set); /* catch any non-resource specific actions */ crm_trace("processing non-resource actions"); gIter = data_set->actions; for (; gIter != NULL; gIter = gIter->next) { action_t *action = (action_t *) gIter->data; if (action->rsc && action->node && action->node->details->shutdown && is_not_set(action->rsc->flags, pe_rsc_maintenance) && is_not_set(action->flags, pe_action_optional) && is_not_set(action->flags, pe_action_runnable) && crm_str_eq(action->task, RSC_STOP, TRUE) ) { /* Eventually we should just ignore the 'fence' case * But for now it's the best way to detect (in CTS) when * CIB resource updates are being lost */ if (is_set(data_set->flags, pe_flag_have_quorum) || data_set->no_quorum_policy == no_quorum_ignore) { crm_crit("Cannot %s node '%s' because of %s:%s%s (%s)", action->node->details->unclean ? "fence" : "shut down", action->node->details->uname, action->rsc->id, is_not_set(action->rsc->flags, pe_rsc_managed) ? " unmanaged" : " blocked", is_set(action->rsc->flags, pe_rsc_failed) ? " failed" : "", action->uuid); } } graph_element_from_action(action, data_set); } crm_log_xml_trace(data_set->graph, "created generic action list"); crm_trace("Created transition graph %d.", transition_id); return TRUE; } void LogNodeActions(pe_working_set_t * data_set, gboolean terminal) { GListPtr gIter = NULL; for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) { char *node_name = NULL; char *task = NULL; action_t *action = (action_t *) gIter->data; if (action->rsc != NULL) { continue; } else if (is_set(action->flags, pe_action_optional)) { continue; } if (pe__is_guest_node(action->node)) { node_name = crm_strdup_printf("%s (resource: %s)", action->node->details->uname, action->node->details->remote_rsc->container->id); } else if(action->node) { node_name = crm_strdup_printf("%s", action->node->details->uname); } if (safe_str_eq(action->task, CRM_OP_SHUTDOWN)) { task = strdup("Shutdown"); } else if (safe_str_eq(action->task, CRM_OP_FENCE)) { const char *op = g_hash_table_lookup(action->meta, "stonith_action"); task = crm_strdup_printf("Fence (%s)", op); } if(task == NULL) { /* Nothing to report */ } else if(terminal && action->reason) { printf(" * %s %s '%s'\n", task, node_name, action->reason); } else if(terminal) { printf(" * %s %s\n", task, node_name); } else if(action->reason) { crm_notice(" * %s %s '%s'\n", task, node_name, action->reason); } else { crm_notice(" * %s %s\n", task, node_name); } free(node_name); free(task); } } diff --git a/lib/pacemaker/pcmk_sched_messages.c b/lib/pacemaker/pcmk_sched_messages.c index 780fe826b7..524600d93a 100644 --- a/lib/pacemaker/pcmk_sched_messages.c +++ b/lib/pacemaker/pcmk_sched_messages.c @@ -1,109 +1,118 @@ /* - * Copyright 2004-2018 Andrew Beekhof + * Copyright 2004-2019 the Pacemaker project contributors + * + * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include -xmlNode *do_calculations(pe_working_set_t * data_set, xmlNode * xml_input, crm_time_t * now); - gboolean show_scores = FALSE; int scores_log_level = LOG_TRACE; gboolean show_utilization = FALSE; int utilization_log_level = LOG_TRACE; +/*! + * \internal + * \brief Run the scheduler for a given CIB + * + * \param[in,out] data_set Cluster working set + * \param[in] xml_input CIB XML to use as scheduler input + * \param[in] now Time to use for rule evaluation (or NULL for now) + */ xmlNode * -do_calculations(pe_working_set_t * data_set, xmlNode * xml_input, crm_time_t * now) +pcmk__schedule_actions(pe_working_set_t *data_set, xmlNode *xml_input, + crm_time_t *now) { GListPtr gIter = NULL; int rsc_log_level = LOG_INFO; /* pe_debug_on(); */ CRM_ASSERT(xml_input || is_set(data_set->flags, pe_flag_have_status)); if (is_set(data_set->flags, pe_flag_have_status) == FALSE) { set_working_set_defaults(data_set); data_set->input = xml_input; data_set->now = now; } else { crm_trace("Already have status - reusing"); } if (data_set->now == NULL) { data_set->now = crm_time_new(NULL); } crm_trace("Calculate cluster status"); stage0(data_set); if(is_not_set(data_set->flags, pe_flag_quick_location)) { gIter = data_set->resources; for (; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; if (is_set(rsc->flags, pe_rsc_orphan) && rsc->role == RSC_ROLE_STOPPED) { continue; } rsc->fns->print(rsc, NULL, pe_print_log, &rsc_log_level); } } crm_trace("Applying placement constraints"); stage2(data_set); if(is_set(data_set->flags, pe_flag_quick_location)){ return NULL; } crm_trace("Create internal constraints"); stage3(data_set); crm_trace("Check actions"); stage4(data_set); crm_trace("Allocate resources"); stage5(data_set); crm_trace("Processing fencing and shutdown cases"); stage6(data_set); crm_trace("Applying ordering constraints"); stage7(data_set); crm_trace("Create transition graph"); stage8(data_set); crm_trace("=#=#=#=#= Summary =#=#=#=#="); crm_trace("\t========= Set %d (Un-runnable) =========", -1); if (get_crm_log_level() >= LOG_TRACE) { gIter = data_set->actions; for (; gIter != NULL; gIter = gIter->next) { action_t *action = (action_t *) gIter->data; if (is_set(action->flags, pe_action_optional) == FALSE && is_set(action->flags, pe_action_runnable) == FALSE && is_set(action->flags, pe_action_pseudo) == FALSE) { log_action(LOG_TRACE, "\t", action, TRUE); } } } return data_set->graph; } diff --git a/lib/pengine/unpack.c b/lib/pengine/unpack.c index 7a4e04c960..02cef2c3fd 100644 --- a/lib/pengine/unpack.c +++ b/lib/pengine/unpack.c @@ -1,3529 +1,3525 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include CRM_TRACE_INIT_DATA(pe_status); #define set_config_flag(data_set, option, flag) do { \ const char *tmp = pe_pref(data_set->config_hash, option); \ if(tmp) { \ if(crm_is_true(tmp)) { \ set_bit(data_set->flags, flag); \ } else { \ clear_bit(data_set->flags, flag); \ } \ } \ } while(0) gboolean unpack_rsc_op(resource_t * rsc, node_t * node, xmlNode * xml_op, xmlNode ** last_failure, enum action_fail_response *failed, pe_working_set_t * data_set); static gboolean determine_remote_online_status(pe_working_set_t * data_set, node_t * this_node); // Bitmask for warnings we only want to print once uint32_t pe_wo = 0; static gboolean is_dangling_guest_node(node_t *node) { /* we are looking for a remote-node that was supposed to be mapped to a * container resource, but all traces of that container have disappeared * from both the config and the status section. */ if (pe__is_guest_or_remote_node(node) && node->details->remote_rsc && node->details->remote_rsc->container == NULL && is_set(node->details->remote_rsc->flags, pe_rsc_orphan_container_filler)) { return TRUE; } return FALSE; } /*! * \brief Schedule a fence action for a node * * \param[in,out] data_set Current working set of cluster * \param[in,out] node Node to fence * \param[in] reason Text description of why fencing is needed */ void pe_fence_node(pe_working_set_t * data_set, node_t * node, const char *reason) { CRM_CHECK(node, return); /* A guest node is fenced by marking its container as failed */ if (pe__is_guest_node(node)) { resource_t *rsc = node->details->remote_rsc->container; if (is_set(rsc->flags, pe_rsc_failed) == FALSE) { if (!is_set(rsc->flags, pe_rsc_managed)) { crm_notice("Not fencing guest node %s " "(otherwise would because %s): " "its guest resource %s is unmanaged", node->details->uname, reason, rsc->id); } else { crm_warn("Guest node %s will be fenced " "(by recovering its guest resource %s): %s", node->details->uname, rsc->id, reason); /* We don't mark the node as unclean because that would prevent the * node from running resources. We want to allow it to run resources * in this transition if the recovery succeeds. */ node->details->remote_requires_reset = TRUE; set_bit(rsc->flags, pe_rsc_failed); } } } else if (is_dangling_guest_node(node)) { crm_info("Cleaning up dangling connection for guest node %s: " "fencing was already done because %s, " "and guest resource no longer exists", node->details->uname, reason); set_bit(node->details->remote_rsc->flags, pe_rsc_failed); } else if (pe__is_remote_node(node)) { resource_t *rsc = node->details->remote_rsc; if (rsc && (!is_set(rsc->flags, pe_rsc_managed))) { crm_notice("Not fencing remote node %s " "(otherwise would because %s): connection is unmanaged", node->details->uname, reason); } else if(node->details->remote_requires_reset == FALSE) { node->details->remote_requires_reset = TRUE; crm_warn("Remote node %s %s: %s", node->details->uname, pe_can_fence(data_set, node)? "will be fenced" : "is unclean", reason); } node->details->unclean = TRUE; pe_fence_op(node, NULL, TRUE, reason, data_set); } else if (node->details->unclean) { crm_trace("Cluster node %s %s because %s", node->details->uname, pe_can_fence(data_set, node)? "would also be fenced" : "also is unclean", reason); } else { crm_warn("Cluster node %s %s: %s", node->details->uname, pe_can_fence(data_set, node)? "will be fenced" : "is unclean", reason); node->details->unclean = TRUE; pe_fence_op(node, NULL, TRUE, reason, data_set); } } // @TODO xpaths can't handle templates, rules, or id-refs // nvpair with provides or requires set to unfencing #define XPATH_UNFENCING_NVPAIR XML_CIB_TAG_NVPAIR \ "[(@" XML_NVPAIR_ATTR_NAME "='" XML_RSC_ATTR_PROVIDES "'" \ "or @" XML_NVPAIR_ATTR_NAME "='" XML_RSC_ATTR_REQUIRES "') " \ "and @" XML_NVPAIR_ATTR_VALUE "='unfencing']" // unfencing in rsc_defaults or any resource #define XPATH_ENABLE_UNFENCING \ "/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_RESOURCES \ "//" XML_TAG_META_SETS "/" XPATH_UNFENCING_NVPAIR \ "|/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_RSCCONFIG \ "/" XML_TAG_META_SETS "/" XPATH_UNFENCING_NVPAIR static void set_if_xpath(unsigned long long flag, const char *xpath, pe_working_set_t *data_set) { xmlXPathObjectPtr result = NULL; if (is_not_set(data_set->flags, flag)) { result = xpath_search(data_set->input, xpath); if (result && (numXpathResults(result) > 0)) { set_bit(data_set->flags, flag); } freeXpathObject(result); } } gboolean unpack_config(xmlNode * config, pe_working_set_t * data_set) { const char *value = NULL; GHashTable *config_hash = crm_str_table_new(); data_set->config_hash = config_hash; unpack_instance_attributes(data_set->input, config, XML_CIB_TAG_PROPSET, NULL, config_hash, CIB_OPTIONS_FIRST, FALSE, data_set->now); verify_pe_options(data_set->config_hash); set_config_flag(data_set, "enable-startup-probes", pe_flag_startup_probes); if(is_not_set(data_set->flags, pe_flag_startup_probes)) { crm_info("Startup probes: disabled (dangerous)"); } value = pe_pref(data_set->config_hash, XML_ATTR_HAVE_WATCHDOG); if (value && crm_is_true(value)) { crm_notice("Watchdog will be used via SBD if fencing is required"); set_bit(data_set->flags, pe_flag_have_stonith_resource); } /* Set certain flags via xpath here, so they can be used before the relevant * configuration sections are unpacked. */ set_if_xpath(pe_flag_enable_unfencing, XPATH_ENABLE_UNFENCING, data_set); value = pe_pref(data_set->config_hash, "stonith-timeout"); data_set->stonith_timeout = crm_get_msec(value); crm_debug("STONITH timeout: %d", data_set->stonith_timeout); set_config_flag(data_set, "stonith-enabled", pe_flag_stonith_enabled); crm_debug("STONITH of failed nodes is %s", is_set(data_set->flags, pe_flag_stonith_enabled) ? "enabled" : "disabled"); data_set->stonith_action = pe_pref(data_set->config_hash, "stonith-action"); if (!strcmp(data_set->stonith_action, "poweroff")) { pe_warn_once(pe_wo_poweroff, "Support for stonith-action of 'poweroff' is deprecated " "and will be removed in a future release (use 'off' instead)"); data_set->stonith_action = "off"; } crm_trace("STONITH will %s nodes", data_set->stonith_action); set_config_flag(data_set, "concurrent-fencing", pe_flag_concurrent_fencing); crm_debug("Concurrent fencing is %s", is_set(data_set->flags, pe_flag_concurrent_fencing) ? "enabled" : "disabled"); set_config_flag(data_set, "stop-all-resources", pe_flag_stop_everything); crm_debug("Stop all active resources: %s", is_set(data_set->flags, pe_flag_stop_everything) ? "true" : "false"); set_config_flag(data_set, "symmetric-cluster", pe_flag_symmetric_cluster); if (is_set(data_set->flags, pe_flag_symmetric_cluster)) { crm_debug("Cluster is symmetric" " - resources can run anywhere by default"); } value = pe_pref(data_set->config_hash, "no-quorum-policy"); if (safe_str_eq(value, "ignore")) { data_set->no_quorum_policy = no_quorum_ignore; } else if (safe_str_eq(value, "freeze")) { data_set->no_quorum_policy = no_quorum_freeze; } else if (safe_str_eq(value, "suicide")) { if (is_set(data_set->flags, pe_flag_stonith_enabled)) { int do_panic = 0; crm_element_value_int(data_set->input, XML_ATTR_QUORUM_PANIC, &do_panic); if (do_panic || is_set(data_set->flags, pe_flag_have_quorum)) { data_set->no_quorum_policy = no_quorum_suicide; } else { crm_notice("Resetting no-quorum-policy to 'stop': cluster has never had quorum"); data_set->no_quorum_policy = no_quorum_stop; } } else { crm_config_err("Resetting no-quorum-policy to 'stop': stonith is not configured"); data_set->no_quorum_policy = no_quorum_stop; } } else { data_set->no_quorum_policy = no_quorum_stop; } switch (data_set->no_quorum_policy) { case no_quorum_freeze: crm_debug("On loss of quorum: Freeze resources"); break; case no_quorum_stop: crm_debug("On loss of quorum: Stop ALL resources"); break; case no_quorum_suicide: crm_notice("On loss of quorum: Fence all remaining nodes"); break; case no_quorum_ignore: crm_notice("On loss of quorum: Ignore"); break; } set_config_flag(data_set, "stop-orphan-resources", pe_flag_stop_rsc_orphans); crm_trace("Orphan resources are %s", is_set(data_set->flags, pe_flag_stop_rsc_orphans) ? "stopped" : "ignored"); set_config_flag(data_set, "stop-orphan-actions", pe_flag_stop_action_orphans); crm_trace("Orphan resource actions are %s", is_set(data_set->flags, pe_flag_stop_action_orphans) ? "stopped" : "ignored"); set_config_flag(data_set, "remove-after-stop", pe_flag_remove_after_stop); crm_trace("Stopped resources are removed from the status section: %s", is_set(data_set->flags, pe_flag_remove_after_stop) ? "true" : "false"); set_config_flag(data_set, "maintenance-mode", pe_flag_maintenance_mode); crm_trace("Maintenance mode: %s", is_set(data_set->flags, pe_flag_maintenance_mode) ? "true" : "false"); set_config_flag(data_set, "start-failure-is-fatal", pe_flag_start_failure_fatal); crm_trace("Start failures are %s", is_set(data_set->flags, pe_flag_start_failure_fatal) ? "always fatal" : "handled by failcount"); if (is_set(data_set->flags, pe_flag_stonith_enabled)) { set_config_flag(data_set, "startup-fencing", pe_flag_startup_fencing); } if (is_set(data_set->flags, pe_flag_startup_fencing)) { crm_trace("Unseen nodes will be fenced"); } else { pe_warn_once(pe_wo_blind, "Blind faith: not fencing unseen nodes"); } node_score_red = char2score(pe_pref(data_set->config_hash, "node-health-red")); node_score_green = char2score(pe_pref(data_set->config_hash, "node-health-green")); node_score_yellow = char2score(pe_pref(data_set->config_hash, "node-health-yellow")); crm_debug("Node scores: 'red' = %s, 'yellow' = %s, 'green' = %s", pe_pref(data_set->config_hash, "node-health-red"), pe_pref(data_set->config_hash, "node-health-yellow"), pe_pref(data_set->config_hash, "node-health-green")); data_set->placement_strategy = pe_pref(data_set->config_hash, "placement-strategy"); crm_trace("Placement strategy: %s", data_set->placement_strategy); return TRUE; } static void destroy_digest_cache(gpointer ptr) { op_digest_cache_t *data = ptr; free_xml(data->params_all); free_xml(data->params_secure); free_xml(data->params_restart); free(data->digest_all_calc); free(data->digest_restart_calc); free(data->digest_secure_calc); free(data); } node_t * pe_create_node(const char *id, const char *uname, const char *type, const char *score, pe_working_set_t * data_set) { node_t *new_node = NULL; if (pe_find_node(data_set->nodes, uname) != NULL) { crm_config_warn("Detected multiple node entries with uname=%s" " - this is rarely intended", uname); } new_node = calloc(1, sizeof(node_t)); if (new_node == NULL) { return NULL; } new_node->weight = char2score(score); new_node->fixed = FALSE; new_node->details = calloc(1, sizeof(struct pe_node_shared_s)); if (new_node->details == NULL) { free(new_node); return NULL; } crm_trace("Creating node for entry %s/%s", uname, id); new_node->details->id = id; new_node->details->uname = uname; new_node->details->online = FALSE; new_node->details->shutdown = FALSE; new_node->details->rsc_discovery_enabled = TRUE; new_node->details->running_rsc = NULL; new_node->details->type = node_ping; if (safe_str_eq(type, "remote")) { new_node->details->type = node_remote; set_bit(data_set->flags, pe_flag_have_remote_nodes); } else if ((type == NULL) || safe_str_eq(type, "member")) { new_node->details->type = node_member; } new_node->details->attrs = crm_str_table_new(); if (pe__is_guest_or_remote_node(new_node)) { g_hash_table_insert(new_node->details->attrs, strdup(CRM_ATTR_KIND), strdup("remote")); } else { g_hash_table_insert(new_node->details->attrs, strdup(CRM_ATTR_KIND), strdup("cluster")); } new_node->details->utilization = crm_str_table_new(); new_node->details->digest_cache = g_hash_table_new_full(crm_str_hash, g_str_equal, free, destroy_digest_cache); data_set->nodes = g_list_insert_sorted(data_set->nodes, new_node, sort_node_uname); return new_node; } bool remote_id_conflict(const char *remote_name, pe_working_set_t *data) { bool match = FALSE; #if 1 pe_find_resource(data->resources, remote_name); #else if (data->name_check == NULL) { data->name_check = g_hash_table_new(crm_str_hash, g_str_equal); for (xml_rsc = __xml_first_child(parent); xml_rsc != NULL; xml_rsc = __xml_next_element(xml_rsc)) { const char *id = ID(xml_rsc); /* avoiding heap allocation here because we know the duration of this hashtable allows us to */ g_hash_table_insert(data->name_check, (char *) id, (char *) id); } } if (g_hash_table_lookup(data->name_check, remote_name)) { match = TRUE; } #endif if (match) { crm_err("Invalid remote-node name, a resource called '%s' already exists.", remote_name); return NULL; } return match; } static const char * expand_remote_rsc_meta(xmlNode *xml_obj, xmlNode *parent, pe_working_set_t *data) { xmlNode *attr_set = NULL; xmlNode *attr = NULL; const char *container_id = ID(xml_obj); const char *remote_name = NULL; const char *remote_server = NULL; const char *remote_port = NULL; const char *connect_timeout = "60s"; const char *remote_allow_migrate=NULL; const char *is_managed = NULL; for (attr_set = __xml_first_child(xml_obj); attr_set != NULL; attr_set = __xml_next_element(attr_set)) { if (safe_str_neq((const char *)attr_set->name, XML_TAG_META_SETS)) { continue; } for (attr = __xml_first_child(attr_set); attr != NULL; attr = __xml_next_element(attr)) { const char *value = crm_element_value(attr, XML_NVPAIR_ATTR_VALUE); const char *name = crm_element_value(attr, XML_NVPAIR_ATTR_NAME); if (safe_str_eq(name, XML_RSC_ATTR_REMOTE_NODE)) { remote_name = value; } else if (safe_str_eq(name, "remote-addr")) { remote_server = value; } else if (safe_str_eq(name, "remote-port")) { remote_port = value; } else if (safe_str_eq(name, "remote-connect-timeout")) { connect_timeout = value; } else if (safe_str_eq(name, "remote-allow-migrate")) { remote_allow_migrate=value; } else if (safe_str_eq(name, XML_RSC_ATTR_MANAGED)) { is_managed = value; } } } if (remote_name == NULL) { return NULL; } if (remote_id_conflict(remote_name, data)) { return NULL; } pe_create_remote_xml(parent, remote_name, container_id, remote_allow_migrate, is_managed, connect_timeout, remote_server, remote_port); return remote_name; } static void handle_startup_fencing(pe_working_set_t *data_set, node_t *new_node) { if ((new_node->details->type == node_remote) && (new_node->details->remote_rsc == NULL)) { /* Ignore fencing for remote nodes that don't have a connection resource * associated with them. This happens when remote node entries get left * in the nodes section after the connection resource is removed. */ return; } if (is_set(data_set->flags, pe_flag_startup_fencing)) { // All nodes are unclean until we've seen their status entry new_node->details->unclean = TRUE; } else { // Blind faith ... new_node->details->unclean = FALSE; } /* We need to be able to determine if a node's status section * exists or not separate from whether the node is unclean. */ new_node->details->unseen = TRUE; } gboolean unpack_nodes(xmlNode * xml_nodes, pe_working_set_t * data_set) { xmlNode *xml_obj = NULL; node_t *new_node = NULL; const char *id = NULL; const char *uname = NULL; const char *type = NULL; const char *score = NULL; for (xml_obj = __xml_first_child(xml_nodes); xml_obj != NULL; xml_obj = __xml_next_element(xml_obj)) { if (crm_str_eq((const char *)xml_obj->name, XML_CIB_TAG_NODE, TRUE)) { new_node = NULL; id = crm_element_value(xml_obj, XML_ATTR_ID); uname = crm_element_value(xml_obj, XML_ATTR_UNAME); type = crm_element_value(xml_obj, XML_ATTR_TYPE); score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE); crm_trace("Processing node %s/%s", uname, id); if (id == NULL) { crm_config_err("Must specify id tag in "); continue; } new_node = pe_create_node(id, uname, type, score, data_set); if (new_node == NULL) { return FALSE; } /* if(data_set->have_quorum == FALSE */ /* && data_set->no_quorum_policy == no_quorum_stop) { */ /* /\* start shutting resources down *\/ */ /* new_node->weight = -INFINITY; */ /* } */ handle_startup_fencing(data_set, new_node); add_node_attrs(xml_obj, new_node, FALSE, data_set); unpack_instance_attributes(data_set->input, xml_obj, XML_TAG_UTILIZATION, NULL, new_node->details->utilization, NULL, FALSE, data_set->now); crm_trace("Done with node %s", crm_element_value(xml_obj, XML_ATTR_UNAME)); } } if (data_set->localhost && pe_find_node(data_set->nodes, data_set->localhost) == NULL) { crm_info("Creating a fake local node"); pe_create_node(data_set->localhost, data_set->localhost, NULL, 0, data_set); } return TRUE; } static void setup_container(resource_t * rsc, pe_working_set_t * data_set) { const char *container_id = NULL; if (rsc->children) { GListPtr gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { resource_t *child_rsc = (resource_t *) gIter->data; setup_container(child_rsc, data_set); } return; } container_id = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_CONTAINER); if (container_id && safe_str_neq(container_id, rsc->id)) { resource_t *container = pe_find_resource(data_set->resources, container_id); if (container) { rsc->container = container; set_bit(container->flags, pe_rsc_is_container); container->fillers = g_list_append(container->fillers, rsc); pe_rsc_trace(rsc, "Resource %s's container is %s", rsc->id, container_id); } else { pe_err("Resource %s: Unknown resource container (%s)", rsc->id, container_id); } } } gboolean unpack_remote_nodes(xmlNode * xml_resources, pe_working_set_t * data_set) { xmlNode *xml_obj = NULL; /* Create remote nodes and guest nodes from the resource configuration * before unpacking resources. */ for (xml_obj = __xml_first_child(xml_resources); xml_obj != NULL; xml_obj = __xml_next_element(xml_obj)) { const char *new_node_id = NULL; /* Check for remote nodes, which are defined by ocf:pacemaker:remote * primitives. */ if (xml_contains_remote_node(xml_obj)) { new_node_id = ID(xml_obj); /* The "pe_find_node" check is here to make sure we don't iterate over * an expanded node that has already been added to the node list. */ if (new_node_id && pe_find_node(data_set->nodes, new_node_id) == NULL) { crm_trace("Found remote node %s defined by resource %s", new_node_id, ID(xml_obj)); pe_create_node(new_node_id, new_node_id, "remote", NULL, data_set); } continue; } /* Check for guest nodes, which are defined by special meta-attributes * of a primitive of any type (for example, VirtualDomain or Xen). */ if (crm_str_eq((const char *)xml_obj->name, XML_CIB_TAG_RESOURCE, TRUE)) { /* This will add an ocf:pacemaker:remote primitive to the * configuration for the guest node's connection, to be unpacked * later. */ new_node_id = expand_remote_rsc_meta(xml_obj, xml_resources, data_set); if (new_node_id && pe_find_node(data_set->nodes, new_node_id) == NULL) { crm_trace("Found guest node %s in resource %s", new_node_id, ID(xml_obj)); pe_create_node(new_node_id, new_node_id, "remote", NULL, data_set); } continue; } /* Check for guest nodes inside a group. Clones are currently not * supported as guest nodes. */ if (crm_str_eq((const char *)xml_obj->name, XML_CIB_TAG_GROUP, TRUE)) { xmlNode *xml_obj2 = NULL; for (xml_obj2 = __xml_first_child(xml_obj); xml_obj2 != NULL; xml_obj2 = __xml_next_element(xml_obj2)) { new_node_id = expand_remote_rsc_meta(xml_obj2, xml_resources, data_set); if (new_node_id && pe_find_node(data_set->nodes, new_node_id) == NULL) { crm_trace("Found guest node %s in resource %s inside group %s", new_node_id, ID(xml_obj2), ID(xml_obj)); pe_create_node(new_node_id, new_node_id, "remote", NULL, data_set); } } } } return TRUE; } /* Call this after all the nodes and resources have been * unpacked, but before the status section is read. * * A remote node's online status is reflected by the state * of the remote node's connection resource. We need to link * the remote node to this connection resource so we can have * easy access to the connection resource during the PE calculations. */ static void link_rsc2remotenode(pe_working_set_t *data_set, resource_t *new_rsc) { node_t *remote_node = NULL; if (new_rsc->is_remote_node == FALSE) { return; } if (is_set(data_set->flags, pe_flag_quick_location)) { /* remote_nodes and remote_resources are not linked in quick location calculations */ return; } print_resource(LOG_TRACE, "Linking remote-node connection resource, ", new_rsc, FALSE); remote_node = pe_find_node(data_set->nodes, new_rsc->id); CRM_CHECK(remote_node != NULL, return;); remote_node->details->remote_rsc = new_rsc; if (new_rsc->container == NULL) { /* Handle start-up fencing for remote nodes (as opposed to guest nodes) * the same as is done for cluster nodes. */ handle_startup_fencing(data_set, remote_node); } else { /* pe_create_node() marks the new node as "remote" or "cluster"; now * that we know the node is a guest node, update it correctly. */ g_hash_table_replace(remote_node->details->attrs, strdup(CRM_ATTR_KIND), strdup("container")); } } static void destroy_tag(gpointer data) { tag_t *tag = data; if (tag) { free(tag->id); g_list_free_full(tag->refs, free); free(tag); } } /*! * \internal * \brief Parse configuration XML for resource information * * \param[in] xml_resources Top of resource configuration XML * \param[in,out] data_set Where to put resource information * * \return TRUE * * \note unpack_remote_nodes() MUST be called before this, so that the nodes can * be used when common_unpack() calls resource_location() */ gboolean unpack_resources(xmlNode * xml_resources, pe_working_set_t * data_set) { xmlNode *xml_obj = NULL; GListPtr gIter = NULL; data_set->template_rsc_sets = g_hash_table_new_full(crm_str_hash, g_str_equal, free, destroy_tag); for (xml_obj = __xml_first_child(xml_resources); xml_obj != NULL; xml_obj = __xml_next_element(xml_obj)) { resource_t *new_rsc = NULL; if (crm_str_eq((const char *)xml_obj->name, XML_CIB_TAG_RSC_TEMPLATE, TRUE)) { const char *template_id = ID(xml_obj); if (template_id && g_hash_table_lookup_extended(data_set->template_rsc_sets, template_id, NULL, NULL) == FALSE) { /* Record the template's ID for the knowledge of its existence anyway. */ g_hash_table_insert(data_set->template_rsc_sets, strdup(template_id), NULL); } continue; } crm_trace("Beginning unpack... <%s id=%s... >", crm_element_name(xml_obj), ID(xml_obj)); if (common_unpack(xml_obj, &new_rsc, NULL, data_set)) { data_set->resources = g_list_append(data_set->resources, new_rsc); print_resource(LOG_TRACE, "Added ", new_rsc, FALSE); } else { crm_config_err("Failed unpacking %s %s", crm_element_name(xml_obj), crm_element_value(xml_obj, XML_ATTR_ID)); if (new_rsc != NULL && new_rsc->fns != NULL) { new_rsc->fns->free(new_rsc); } } } for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; setup_container(rsc, data_set); link_rsc2remotenode(data_set, rsc); } data_set->resources = g_list_sort(data_set->resources, sort_rsc_priority); if (is_set(data_set->flags, pe_flag_quick_location)) { /* Ignore */ } else if (is_set(data_set->flags, pe_flag_stonith_enabled) && is_set(data_set->flags, pe_flag_have_stonith_resource) == FALSE) { crm_config_err("Resource start-up disabled since no STONITH resources have been defined"); crm_config_err("Either configure some or disable STONITH with the stonith-enabled option"); crm_config_err("NOTE: Clusters with shared data need STONITH to ensure data integrity"); } return TRUE; } gboolean unpack_tags(xmlNode * xml_tags, pe_working_set_t * data_set) { xmlNode *xml_tag = NULL; data_set->tags = g_hash_table_new_full(crm_str_hash, g_str_equal, free, destroy_tag); for (xml_tag = __xml_first_child(xml_tags); xml_tag != NULL; xml_tag = __xml_next_element(xml_tag)) { xmlNode *xml_obj_ref = NULL; const char *tag_id = ID(xml_tag); if (crm_str_eq((const char *)xml_tag->name, XML_CIB_TAG_TAG, TRUE) == FALSE) { continue; } if (tag_id == NULL) { crm_config_err("Failed unpacking %s: %s should be specified", crm_element_name(xml_tag), XML_ATTR_ID); continue; } for (xml_obj_ref = __xml_first_child(xml_tag); xml_obj_ref != NULL; xml_obj_ref = __xml_next_element(xml_obj_ref)) { const char *obj_ref = ID(xml_obj_ref); if (crm_str_eq((const char *)xml_obj_ref->name, XML_CIB_TAG_OBJ_REF, TRUE) == FALSE) { continue; } if (obj_ref == NULL) { crm_config_err("Failed unpacking %s for tag %s: %s should be specified", crm_element_name(xml_obj_ref), tag_id, XML_ATTR_ID); continue; } if (add_tag_ref(data_set->tags, tag_id, obj_ref) == FALSE) { return FALSE; } } } return TRUE; } /* The ticket state section: * "/cib/status/tickets/ticket_state" */ static gboolean unpack_ticket_state(xmlNode * xml_ticket, pe_working_set_t * data_set) { const char *ticket_id = NULL; const char *granted = NULL; const char *last_granted = NULL; const char *standby = NULL; xmlAttrPtr xIter = NULL; ticket_t *ticket = NULL; ticket_id = ID(xml_ticket); if (ticket_id == NULL || strlen(ticket_id) == 0) { return FALSE; } crm_trace("Processing ticket state for %s", ticket_id); ticket = g_hash_table_lookup(data_set->tickets, ticket_id); if (ticket == NULL) { ticket = ticket_new(ticket_id, data_set); if (ticket == NULL) { return FALSE; } } for (xIter = xml_ticket->properties; xIter; xIter = xIter->next) { const char *prop_name = (const char *)xIter->name; const char *prop_value = crm_element_value(xml_ticket, prop_name); if (crm_str_eq(prop_name, XML_ATTR_ID, TRUE)) { continue; } g_hash_table_replace(ticket->state, strdup(prop_name), strdup(prop_value)); } granted = g_hash_table_lookup(ticket->state, "granted"); if (granted && crm_is_true(granted)) { ticket->granted = TRUE; crm_info("We have ticket '%s'", ticket->id); } else { ticket->granted = FALSE; crm_info("We do not have ticket '%s'", ticket->id); } last_granted = g_hash_table_lookup(ticket->state, "last-granted"); if (last_granted) { ticket->last_granted = crm_parse_int(last_granted, 0); } standby = g_hash_table_lookup(ticket->state, "standby"); if (standby && crm_is_true(standby)) { ticket->standby = TRUE; if (ticket->granted) { crm_info("Granted ticket '%s' is in standby-mode", ticket->id); } } else { ticket->standby = FALSE; } crm_trace("Done with ticket state for %s", ticket_id); return TRUE; } static gboolean unpack_tickets_state(xmlNode * xml_tickets, pe_working_set_t * data_set) { xmlNode *xml_obj = NULL; for (xml_obj = __xml_first_child(xml_tickets); xml_obj != NULL; xml_obj = __xml_next_element(xml_obj)) { if (crm_str_eq((const char *)xml_obj->name, XML_CIB_TAG_TICKET_STATE, TRUE) == FALSE) { continue; } unpack_ticket_state(xml_obj, data_set); } return TRUE; } static void unpack_handle_remote_attrs(node_t *this_node, xmlNode *state, pe_working_set_t * data_set) { const char *resource_discovery_enabled = NULL; xmlNode *attrs = NULL; resource_t *rsc = NULL; const char *shutdown = NULL; if (crm_str_eq((const char *)state->name, XML_CIB_TAG_STATE, TRUE) == FALSE) { return; } if ((this_node == NULL) || !pe__is_guest_or_remote_node(this_node)) { return; } crm_trace("Processing remote node id=%s, uname=%s", this_node->details->id, this_node->details->uname); this_node->details->remote_maintenance = crm_atoi(crm_element_value(state, XML_NODE_IS_MAINTENANCE), "0"); rsc = this_node->details->remote_rsc; if (this_node->details->remote_requires_reset == FALSE) { this_node->details->unclean = FALSE; this_node->details->unseen = FALSE; } attrs = find_xml_node(state, XML_TAG_TRANSIENT_NODEATTRS, FALSE); add_node_attrs(attrs, this_node, TRUE, data_set); shutdown = pe_node_attribute_raw(this_node, XML_CIB_ATTR_SHUTDOWN); if (shutdown != NULL && safe_str_neq("0", shutdown)) { crm_info("Node %s is shutting down", this_node->details->uname); this_node->details->shutdown = TRUE; if (rsc) { rsc->next_role = RSC_ROLE_STOPPED; } } if (crm_is_true(pe_node_attribute_raw(this_node, "standby"))) { crm_info("Node %s is in standby-mode", this_node->details->uname); this_node->details->standby = TRUE; } if (crm_is_true(pe_node_attribute_raw(this_node, "maintenance")) || (rsc && !is_set(rsc->flags, pe_rsc_managed))) { crm_info("Node %s is in maintenance-mode", this_node->details->uname); this_node->details->maintenance = TRUE; } resource_discovery_enabled = pe_node_attribute_raw(this_node, XML_NODE_ATTR_RSC_DISCOVERY); if (resource_discovery_enabled && !crm_is_true(resource_discovery_enabled)) { if (pe__is_remote_node(this_node) && is_not_set(data_set->flags, pe_flag_stonith_enabled)) { crm_warn("Ignoring %s attribute on remote node %s because stonith is disabled", XML_NODE_ATTR_RSC_DISCOVERY, this_node->details->uname); } else { /* This is either a remote node with fencing enabled, or a guest * node. We don't care whether fencing is enabled when fencing guest * nodes, because they are "fenced" by recovering their containing * resource. */ crm_info("Node %s has resource discovery disabled", this_node->details->uname); this_node->details->rsc_discovery_enabled = FALSE; } } } static bool unpack_node_loop(xmlNode * status, bool fence, pe_working_set_t * data_set) { bool changed = false; xmlNode *lrm_rsc = NULL; for (xmlNode *state = __xml_first_child(status); state != NULL; state = __xml_next_element(state)) { const char *id = NULL; const char *uname = NULL; node_t *this_node = NULL; bool process = FALSE; if (crm_str_eq((const char *)state->name, XML_CIB_TAG_STATE, TRUE) == FALSE) { continue; } id = crm_element_value(state, XML_ATTR_ID); uname = crm_element_value(state, XML_ATTR_UNAME); this_node = pe_find_node_any(data_set->nodes, id, uname); if (this_node == NULL) { crm_info("Node %s is unknown", id); continue; } else if (this_node->details->unpacked) { crm_info("Node %s is already processed", id); continue; } else if (!pe__is_guest_or_remote_node(this_node) && is_set(data_set->flags, pe_flag_stonith_enabled)) { // A redundant test, but preserves the order for regression tests process = TRUE; } else if (pe__is_guest_or_remote_node(this_node)) { bool check = FALSE; resource_t *rsc = this_node->details->remote_rsc; if(fence) { check = TRUE; } else if(rsc == NULL) { /* Not ready yet */ } else if (pe__is_guest_node(this_node) && rsc->role == RSC_ROLE_STARTED && rsc->container->role == RSC_ROLE_STARTED) { /* Both the connection and its containing resource need to be * known to be up before we process resources running in it. */ check = TRUE; crm_trace("Checking node %s/%s/%s status %d/%d/%d", id, rsc->id, rsc->container->id, fence, rsc->role, RSC_ROLE_STARTED); } else if (!pe__is_guest_node(this_node) && rsc->role == RSC_ROLE_STARTED) { check = TRUE; crm_trace("Checking node %s/%s status %d/%d/%d", id, rsc->id, fence, rsc->role, RSC_ROLE_STARTED); } if (check) { determine_remote_online_status(data_set, this_node); unpack_handle_remote_attrs(this_node, state, data_set); process = TRUE; } } else if (this_node->details->online) { process = TRUE; } else if (fence) { process = TRUE; } if(process) { crm_trace("Processing lrm resource entries on %shealthy%s node: %s", fence?"un":"", (pe__is_guest_or_remote_node(this_node)? " remote" : ""), this_node->details->uname); changed = TRUE; this_node->details->unpacked = TRUE; lrm_rsc = find_xml_node(state, XML_CIB_TAG_LRM, FALSE); lrm_rsc = find_xml_node(lrm_rsc, XML_LRM_TAG_RESOURCES, FALSE); unpack_lrm_resources(this_node, lrm_rsc, data_set); } } return changed; } /* remove nodes that are down, stopping */ /* create positive rsc_to_node constraints between resources and the nodes they are running on */ /* anything else? */ gboolean unpack_status(xmlNode * status, pe_working_set_t * data_set) { const char *id = NULL; const char *uname = NULL; xmlNode *state = NULL; node_t *this_node = NULL; crm_trace("Beginning unpack"); if (data_set->tickets == NULL) { data_set->tickets = g_hash_table_new_full(crm_str_hash, g_str_equal, free, destroy_ticket); } for (state = __xml_first_child(status); state != NULL; state = __xml_next_element(state)) { if (crm_str_eq((const char *)state->name, XML_CIB_TAG_TICKETS, TRUE)) { unpack_tickets_state((xmlNode *) state, data_set); } else if (crm_str_eq((const char *)state->name, XML_CIB_TAG_STATE, TRUE)) { xmlNode *attrs = NULL; const char *resource_discovery_enabled = NULL; id = crm_element_value(state, XML_ATTR_ID); uname = crm_element_value(state, XML_ATTR_UNAME); this_node = pe_find_node_any(data_set->nodes, id, uname); if (uname == NULL) { /* error */ continue; } else if (this_node == NULL) { crm_config_warn("Node %s in status section no longer exists", uname); continue; } else if (pe__is_guest_or_remote_node(this_node)) { /* online state for remote nodes is determined by the * rsc state after all the unpacking is done. we do however * need to mark whether or not the node has been fenced as this plays * a role during unpacking cluster node resource state */ this_node->details->remote_was_fenced = crm_atoi(crm_element_value(state, XML_NODE_IS_FENCED), "0"); continue; } crm_trace("Processing node id=%s, uname=%s", id, uname); /* Mark the node as provisionally clean * - at least we have seen it in the current cluster's lifetime */ this_node->details->unclean = FALSE; this_node->details->unseen = FALSE; attrs = find_xml_node(state, XML_TAG_TRANSIENT_NODEATTRS, FALSE); add_node_attrs(attrs, this_node, TRUE, data_set); if (crm_is_true(pe_node_attribute_raw(this_node, "standby"))) { crm_info("Node %s is in standby-mode", this_node->details->uname); this_node->details->standby = TRUE; } if (crm_is_true(pe_node_attribute_raw(this_node, "maintenance"))) { crm_info("Node %s is in maintenance-mode", this_node->details->uname); this_node->details->maintenance = TRUE; } resource_discovery_enabled = pe_node_attribute_raw(this_node, XML_NODE_ATTR_RSC_DISCOVERY); if (resource_discovery_enabled && !crm_is_true(resource_discovery_enabled)) { crm_warn("ignoring %s attribute on node %s, disabling resource discovery is not allowed on cluster nodes", XML_NODE_ATTR_RSC_DISCOVERY, this_node->details->uname); } crm_trace("determining node state"); determine_online_status(state, this_node, data_set); if (is_not_set(data_set->flags, pe_flag_have_quorum) && this_node->details->online && (data_set->no_quorum_policy == no_quorum_suicide)) { /* Everything else should flow from this automatically * At least until the PE becomes able to migrate off healthy resources */ pe_fence_node(data_set, this_node, "cluster does not have quorum"); } } } while(unpack_node_loop(status, FALSE, data_set)) { crm_trace("Start another loop"); } // Now catch any nodes we didn't see unpack_node_loop(status, is_set(data_set->flags, pe_flag_stonith_enabled), data_set); /* Now that we know where resources are, we can schedule stops of containers * with failed bundle connections */ if (data_set->stop_needed != NULL) { for (GList *item = data_set->stop_needed; item; item = item->next) { pe_resource_t *container = item->data; pe_node_t *node = pe__current_node(container); if (node) { stop_action(container, node, FALSE); } } g_list_free(data_set->stop_needed); data_set->stop_needed = NULL; } for (GListPtr gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { node_t *this_node = gIter->data; if (this_node == NULL) { continue; } else if (!pe__is_guest_or_remote_node(this_node)) { continue; } else if(this_node->details->unpacked) { continue; } determine_remote_online_status(data_set, this_node); } return TRUE; } static gboolean determine_online_status_no_fencing(pe_working_set_t * data_set, xmlNode * node_state, node_t * this_node) { gboolean online = FALSE; const char *join = crm_element_value(node_state, XML_NODE_JOIN_STATE); const char *is_peer = crm_element_value(node_state, XML_NODE_IS_PEER); const char *in_cluster = crm_element_value(node_state, XML_NODE_IN_CLUSTER); const char *exp_state = crm_element_value(node_state, XML_NODE_EXPECTED); if (!crm_is_true(in_cluster)) { crm_trace("Node is down: in_cluster=%s", crm_str(in_cluster)); } else if (safe_str_eq(is_peer, ONLINESTATUS)) { if (safe_str_eq(join, CRMD_JOINSTATE_MEMBER)) { online = TRUE; } else { crm_debug("Node is not ready to run resources: %s", join); } } else if (this_node->details->expected_up == FALSE) { crm_trace("Controller is down: in_cluster=%s", crm_str(in_cluster)); crm_trace("\tis_peer=%s, join=%s, expected=%s", crm_str(is_peer), crm_str(join), crm_str(exp_state)); } else { /* mark it unclean */ pe_fence_node(data_set, this_node, "peer is unexpectedly down"); crm_info("\tin_cluster=%s, is_peer=%s, join=%s, expected=%s", crm_str(in_cluster), crm_str(is_peer), crm_str(join), crm_str(exp_state)); } return online; } static gboolean determine_online_status_fencing(pe_working_set_t * data_set, xmlNode * node_state, node_t * this_node) { gboolean online = FALSE; gboolean do_terminate = FALSE; bool crmd_online = FALSE; const char *join = crm_element_value(node_state, XML_NODE_JOIN_STATE); const char *is_peer = crm_element_value(node_state, XML_NODE_IS_PEER); const char *in_cluster = crm_element_value(node_state, XML_NODE_IN_CLUSTER); const char *exp_state = crm_element_value(node_state, XML_NODE_EXPECTED); const char *terminate = pe_node_attribute_raw(this_node, "terminate"); /* - XML_NODE_IN_CLUSTER ::= true|false - XML_NODE_IS_PEER ::= online|offline - XML_NODE_JOIN_STATE ::= member|down|pending|banned - XML_NODE_EXPECTED ::= member|down */ if (crm_is_true(terminate)) { do_terminate = TRUE; } else if (terminate != NULL && strlen(terminate) > 0) { /* could be a time() value */ char t = terminate[0]; if (t != '0' && isdigit(t)) { do_terminate = TRUE; } } crm_trace("%s: in_cluster=%s, is_peer=%s, join=%s, expected=%s, term=%d", this_node->details->uname, crm_str(in_cluster), crm_str(is_peer), crm_str(join), crm_str(exp_state), do_terminate); online = crm_is_true(in_cluster); crmd_online = safe_str_eq(is_peer, ONLINESTATUS); if (exp_state == NULL) { exp_state = CRMD_JOINSTATE_DOWN; } if (this_node->details->shutdown) { crm_debug("%s is shutting down", this_node->details->uname); /* Slightly different criteria since we can't shut down a dead peer */ online = crmd_online; } else if (in_cluster == NULL) { pe_fence_node(data_set, this_node, "peer has not been seen by the cluster"); } else if (safe_str_eq(join, CRMD_JOINSTATE_NACK)) { pe_fence_node(data_set, this_node, "peer failed the pacemaker membership criteria"); } else if (do_terminate == FALSE && safe_str_eq(exp_state, CRMD_JOINSTATE_DOWN)) { if (crm_is_true(in_cluster) || crmd_online) { crm_info("- Node %s is not ready to run resources", this_node->details->uname); this_node->details->standby = TRUE; this_node->details->pending = TRUE; } else { crm_trace("%s is down or still coming up", this_node->details->uname); } } else if (do_terminate && safe_str_eq(join, CRMD_JOINSTATE_DOWN) && crm_is_true(in_cluster) == FALSE && !crmd_online) { crm_info("Node %s was just shot", this_node->details->uname); online = FALSE; } else if (crm_is_true(in_cluster) == FALSE) { pe_fence_node(data_set, this_node, "peer is no longer part of the cluster"); } else if (!crmd_online) { pe_fence_node(data_set, this_node, "peer process is no longer available"); /* Everything is running at this point, now check join state */ } else if (do_terminate) { pe_fence_node(data_set, this_node, "termination was requested"); } else if (safe_str_eq(join, CRMD_JOINSTATE_MEMBER)) { crm_info("Node %s is active", this_node->details->uname); } else if (safe_str_eq(join, CRMD_JOINSTATE_PENDING) || safe_str_eq(join, CRMD_JOINSTATE_DOWN)) { crm_info("Node %s is not ready to run resources", this_node->details->uname); this_node->details->standby = TRUE; this_node->details->pending = TRUE; } else { pe_fence_node(data_set, this_node, "peer was in an unknown state"); crm_warn("%s: in-cluster=%s, is-peer=%s, join=%s, expected=%s, term=%d, shutdown=%d", this_node->details->uname, crm_str(in_cluster), crm_str(is_peer), crm_str(join), crm_str(exp_state), do_terminate, this_node->details->shutdown); } return online; } static gboolean determine_remote_online_status(pe_working_set_t * data_set, node_t * this_node) { resource_t *rsc = this_node->details->remote_rsc; resource_t *container = NULL; pe_node_t *host = NULL; /* If there is a node state entry for a (former) Pacemaker Remote node * but no resource creating that node, the node's connection resource will * be NULL. Consider it an offline remote node in that case. */ if (rsc == NULL) { this_node->details->online = FALSE; goto remote_online_done; } container = rsc->container; if (container && (g_list_length(rsc->running_on) == 1)) { host = rsc->running_on->data; } /* If the resource is currently started, mark it online. */ if (rsc->role == RSC_ROLE_STARTED) { crm_trace("%s node %s presumed ONLINE because connection resource is started", (container? "Guest" : "Remote"), this_node->details->id); this_node->details->online = TRUE; } /* consider this node shutting down if transitioning start->stop */ if (rsc->role == RSC_ROLE_STARTED && rsc->next_role == RSC_ROLE_STOPPED) { crm_trace("%s node %s shutting down because connection resource is stopping", (container? "Guest" : "Remote"), this_node->details->id); this_node->details->shutdown = TRUE; } /* Now check all the failure conditions. */ if(container && is_set(container->flags, pe_rsc_failed)) { crm_trace("Guest node %s UNCLEAN because guest resource failed", this_node->details->id); this_node->details->online = FALSE; this_node->details->remote_requires_reset = TRUE; } else if(is_set(rsc->flags, pe_rsc_failed)) { crm_trace("%s node %s OFFLINE because connection resource failed", (container? "Guest" : "Remote"), this_node->details->id); this_node->details->online = FALSE; } else if (rsc->role == RSC_ROLE_STOPPED || (container && container->role == RSC_ROLE_STOPPED)) { crm_trace("%s node %s OFFLINE because its resource is stopped", (container? "Guest" : "Remote"), this_node->details->id); this_node->details->online = FALSE; this_node->details->remote_requires_reset = FALSE; } else if (host && (host->details->online == FALSE) && host->details->unclean) { crm_trace("Guest node %s UNCLEAN because host is unclean", this_node->details->id); this_node->details->online = FALSE; this_node->details->remote_requires_reset = TRUE; } remote_online_done: crm_trace("Remote node %s online=%s", this_node->details->id, this_node->details->online ? "TRUE" : "FALSE"); return this_node->details->online; } gboolean determine_online_status(xmlNode * node_state, node_t * this_node, pe_working_set_t * data_set) { gboolean online = FALSE; const char *shutdown = NULL; const char *exp_state = crm_element_value(node_state, XML_NODE_EXPECTED); if (this_node == NULL) { crm_config_err("No node to check"); return online; } this_node->details->shutdown = FALSE; this_node->details->expected_up = FALSE; shutdown = pe_node_attribute_raw(this_node, XML_CIB_ATTR_SHUTDOWN); if (shutdown != NULL && safe_str_neq("0", shutdown)) { this_node->details->shutdown = TRUE; } else if (safe_str_eq(exp_state, CRMD_JOINSTATE_MEMBER)) { this_node->details->expected_up = TRUE; } if (this_node->details->type == node_ping) { this_node->details->unclean = FALSE; online = FALSE; /* As far as resource management is concerned, * the node is safely offline. * Anyone caught abusing this logic will be shot */ } else if (is_set(data_set->flags, pe_flag_stonith_enabled) == FALSE) { online = determine_online_status_no_fencing(data_set, node_state, this_node); } else { online = determine_online_status_fencing(data_set, node_state, this_node); } if (online) { this_node->details->online = TRUE; } else { /* remove node from contention */ this_node->fixed = TRUE; this_node->weight = -INFINITY; } if (online && this_node->details->shutdown) { /* don't run resources here */ this_node->fixed = TRUE; this_node->weight = -INFINITY; } if (this_node->details->type == node_ping) { crm_info("Node %s is not a pacemaker node", this_node->details->uname); } else if (this_node->details->unclean) { pe_proc_warn("Node %s is unclean", this_node->details->uname); } else if (this_node->details->online) { crm_info("Node %s is %s", this_node->details->uname, this_node->details->shutdown ? "shutting down" : this_node->details->pending ? "pending" : this_node->details->standby ? "standby" : this_node->details->maintenance ? "maintenance" : "online"); } else { crm_trace("Node %s is offline", this_node->details->uname); } return online; } /*! * \internal * \brief Find the end of a resource's name, excluding any clone suffix * * \param[in] id Resource ID to check * * \return Pointer to last character of resource's base name */ const char * pe_base_name_end(const char *id) { if (!crm_strlen_zero(id)) { const char *end = id + strlen(id) - 1; for (const char *s = end; s > id; --s) { switch (*s) { case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': break; case ':': return (s == end)? s : (s - 1); default: return end; } } return end; } return NULL; } /*! * \internal * \brief Get a resource name excluding any clone suffix * * \param[in] last_rsc_id Resource ID to check * * \return Pointer to newly allocated string with resource's base name * \note It is the caller's responsibility to free() the result. * This asserts on error, so callers can assume result is not NULL. */ char * clone_strip(const char *last_rsc_id) { const char *end = pe_base_name_end(last_rsc_id); char *basename = NULL; CRM_ASSERT(end); basename = strndup(last_rsc_id, end - last_rsc_id + 1); CRM_ASSERT(basename); return basename; } /*! * \internal * \brief Get the name of the first instance of a cloned resource * * \param[in] last_rsc_id Resource ID to check * * \return Pointer to newly allocated string with resource's base name plus :0 * \note It is the caller's responsibility to free() the result. * This asserts on error, so callers can assume result is not NULL. */ char * clone_zero(const char *last_rsc_id) { const char *end = pe_base_name_end(last_rsc_id); size_t base_name_len = end - last_rsc_id + 1; char *zero = NULL; CRM_ASSERT(end); zero = calloc(base_name_len + 3, sizeof(char)); CRM_ASSERT(zero); memcpy(zero, last_rsc_id, base_name_len); zero[base_name_len] = ':'; zero[base_name_len + 1] = '0'; return zero; } static resource_t * create_fake_resource(const char *rsc_id, xmlNode * rsc_entry, pe_working_set_t * data_set) { resource_t *rsc = NULL; xmlNode *xml_rsc = create_xml_node(NULL, XML_CIB_TAG_RESOURCE); copy_in_properties(xml_rsc, rsc_entry); crm_xml_add(xml_rsc, XML_ATTR_ID, rsc_id); crm_log_xml_debug(xml_rsc, "Orphan resource"); if (!common_unpack(xml_rsc, &rsc, NULL, data_set)) { return NULL; } if (xml_contains_remote_node(xml_rsc)) { node_t *node; crm_debug("Detected orphaned remote node %s", rsc_id); node = pe_find_node(data_set->nodes, rsc_id); if (node == NULL) { node = pe_create_node(rsc_id, rsc_id, "remote", NULL, data_set); } link_rsc2remotenode(data_set, rsc); if (node) { crm_trace("Setting node %s as shutting down due to orphaned connection resource", rsc_id); node->details->shutdown = TRUE; } } if (crm_element_value(rsc_entry, XML_RSC_ATTR_CONTAINER)) { /* This orphaned rsc needs to be mapped to a container. */ crm_trace("Detected orphaned container filler %s", rsc_id); set_bit(rsc->flags, pe_rsc_orphan_container_filler); } set_bit(rsc->flags, pe_rsc_orphan); data_set->resources = g_list_append(data_set->resources, rsc); return rsc; } /*! * \internal * \brief Create orphan instance for anonymous clone resource history */ static pe_resource_t * create_anonymous_orphan(pe_resource_t *parent, const char *rsc_id, pe_node_t *node, pe_working_set_t *data_set) { pe_resource_t *top = pe__create_clone_child(parent, data_set); // find_rsc() because we might be a cloned group pe_resource_t *orphan = top->fns->find_rsc(top, rsc_id, NULL, pe_find_clone); pe_rsc_debug(parent, "Created orphan %s for %s: %s on %s", top->id, parent->id, rsc_id, node->details->uname); return orphan; } /*! * \internal * \brief Check a node for an instance of an anonymous clone * * Return a child instance of the specified anonymous clone, in order of * preference: (1) the instance running on the specified node, if any; * (2) an inactive instance (i.e. within the total of clone-max instances); * (3) a newly created orphan (i.e. clone-max instances are already active). * * \param[in] data_set Cluster information * \param[in] node Node on which to check for instance * \param[in] parent Clone to check * \param[in] rsc_id Name of cloned resource in history (without instance) */ static resource_t * find_anonymous_clone(pe_working_set_t * data_set, node_t * node, resource_t * parent, const char *rsc_id) { GListPtr rIter = NULL; pe_resource_t *rsc = NULL; pe_resource_t *inactive_instance = NULL; gboolean skip_inactive = FALSE; CRM_ASSERT(parent != NULL); CRM_ASSERT(pe_rsc_is_clone(parent)); CRM_ASSERT(is_not_set(parent->flags, pe_rsc_unique)); // Check for active (or partially active, for cloned groups) instance pe_rsc_trace(parent, "Looking for %s on %s in %s", rsc_id, node->details->uname, parent->id); for (rIter = parent->children; rsc == NULL && rIter; rIter = rIter->next) { GListPtr locations = NULL; resource_t *child = rIter->data; /* Check whether this instance is already known to be active or pending * anywhere, at this stage of unpacking. Because this function is called * for a resource before the resource's individual operation history * entries are unpacked, locations will generally not contain the * desired node. * * However, there are three exceptions: * (1) when child is a cloned group and we have already unpacked the * history of another member of the group on the same node; * (2) when we've already unpacked the history of another numbered * instance on the same node (which can happen if globally-unique * was flipped from true to false); and * (3) when we re-run calculations on the same data set as part of a * simulation. */ child->fns->location(child, &locations, 2); if (locations) { /* We should never associate the same numbered anonymous clone * instance with multiple nodes, and clone instances can't migrate, * so there must be only one location, regardless of history. */ CRM_LOG_ASSERT(locations->next == NULL); if (((pe_node_t *)locations->data)->details == node->details) { /* This child instance is active on the requested node, so check * for a corresponding configured resource. We use find_rsc() * instead of child because child may be a cloned group, and we * need the particular member corresponding to rsc_id. * * If the history entry is orphaned, rsc will be NULL. */ rsc = parent->fns->find_rsc(child, rsc_id, NULL, pe_find_clone); if (rsc) { /* If there are multiple instance history entries for an * anonymous clone in a single node's history (which can * happen if globally-unique is switched from true to * false), we want to consider the instances beyond the * first as orphans, even if there are inactive instance * numbers available. */ if (rsc->running_on) { crm_notice("Active (now-)anonymous clone %s has " "multiple (orphan) instance histories on %s", parent->id, node->details->uname); skip_inactive = TRUE; rsc = NULL; } else { pe_rsc_trace(parent, "Resource %s, active", rsc->id); } } } g_list_free(locations); } else { pe_rsc_trace(parent, "Resource %s, skip inactive", child->id); if (!skip_inactive && !inactive_instance && is_not_set(child->flags, pe_rsc_block)) { // Remember one inactive instance in case we don't find active inactive_instance = parent->fns->find_rsc(child, rsc_id, NULL, pe_find_clone); /* ... but don't use it if it was already associated with a * pending action on another node */ if (inactive_instance && inactive_instance->pending_node && (inactive_instance->pending_node->details != node->details)) { inactive_instance = NULL; } } } } if ((rsc == NULL) && !skip_inactive && (inactive_instance != NULL)) { pe_rsc_trace(parent, "Resource %s, empty slot", inactive_instance->id); rsc = inactive_instance; } /* If the resource has "requires" set to "quorum" or "nothing", and we don't * have a clone instance for every node, we don't want to consume a valid * instance number for unclean nodes. Such instances may appear to be active * according to the history, but should be considered inactive, so we can * start an instance elsewhere. Treat such instances as orphans. * * An exception is instances running on guest nodes -- since guest node * "fencing" is actually just a resource stop, requires shouldn't apply. * * @TODO Ideally, we'd use an inactive instance number if it is not needed * for any clean instances. However, we don't know that at this point. */ if ((rsc != NULL) && is_not_set(rsc->flags, pe_rsc_needs_fencing) && (!node->details->online || node->details->unclean) && !pe__is_guest_node(node) && !pe__is_universal_clone(parent, data_set)) { rsc = NULL; } if (rsc == NULL) { rsc = create_anonymous_orphan(parent, rsc_id, node, data_set); pe_rsc_trace(parent, "Resource %s, orphan", rsc->id); } return rsc; } static resource_t * unpack_find_resource(pe_working_set_t * data_set, node_t * node, const char *rsc_id, xmlNode * rsc_entry) { resource_t *rsc = NULL; resource_t *parent = NULL; crm_trace("looking for %s", rsc_id); rsc = pe_find_resource(data_set->resources, rsc_id); if (rsc == NULL) { /* If we didn't find the resource by its name in the operation history, * check it again as a clone instance. Even when clone-max=0, we create * a single :0 orphan to match against here. */ char *clone0_id = clone_zero(rsc_id); resource_t *clone0 = pe_find_resource(data_set->resources, clone0_id); if (clone0 && is_not_set(clone0->flags, pe_rsc_unique)) { rsc = clone0; parent = uber_parent(clone0); crm_trace("%s found as %s (%s)", rsc_id, clone0_id, parent->id); } else { crm_trace("%s is not known as %s either (orphan)", rsc_id, clone0_id); } free(clone0_id); } else if (rsc->variant > pe_native) { crm_trace("Resource history for %s is orphaned because it is no longer primitive", rsc_id); return NULL; } else { parent = uber_parent(rsc); } if (pe_rsc_is_anon_clone(parent)) { if (pe_rsc_is_bundled(parent)) { rsc = pe__find_bundle_replica(parent->parent, node); } else { char *base = clone_strip(rsc_id); rsc = find_anonymous_clone(data_set, node, parent, base); free(base); CRM_ASSERT(rsc != NULL); } } if (rsc && safe_str_neq(rsc_id, rsc->id) && safe_str_neq(rsc_id, rsc->clone_name)) { free(rsc->clone_name); rsc->clone_name = strdup(rsc_id); pe_rsc_debug(rsc, "Internally renamed %s on %s to %s%s", rsc_id, node->details->uname, rsc->id, (is_set(rsc->flags, pe_rsc_orphan)? " (ORPHAN)" : "")); } return rsc; } static resource_t * process_orphan_resource(xmlNode * rsc_entry, node_t * node, pe_working_set_t * data_set) { resource_t *rsc = NULL; const char *rsc_id = crm_element_value(rsc_entry, XML_ATTR_ID); crm_debug("Detected orphan resource %s on %s", rsc_id, node->details->uname); rsc = create_fake_resource(rsc_id, rsc_entry, data_set); if (is_set(data_set->flags, pe_flag_stop_rsc_orphans) == FALSE) { clear_bit(rsc->flags, pe_rsc_managed); } else { print_resource(LOG_TRACE, "Added orphan", rsc, FALSE); CRM_CHECK(rsc != NULL, return NULL); resource_location(rsc, NULL, -INFINITY, "__orphan_dont_run__", data_set); } return rsc; } static void process_rsc_state(resource_t * rsc, node_t * node, enum action_fail_response on_fail, xmlNode * migrate_op, pe_working_set_t * data_set) { node_t *tmpnode = NULL; char *reason = NULL; CRM_ASSERT(rsc); pe_rsc_trace(rsc, "Resource %s is %s on %s: on_fail=%s", rsc->id, role2text(rsc->role), node->details->uname, fail2text(on_fail)); /* process current state */ if (rsc->role != RSC_ROLE_UNKNOWN) { resource_t *iter = rsc; while (iter) { if (g_hash_table_lookup(iter->known_on, node->details->id) == NULL) { node_t *n = node_copy(node); pe_rsc_trace(rsc, "%s (aka. %s) known on %s", rsc->id, rsc->clone_name, n->details->uname); g_hash_table_insert(iter->known_on, (gpointer) n->details->id, n); } if (is_set(iter->flags, pe_rsc_unique)) { break; } iter = iter->parent; } } /* If a managed resource is believed to be running, but node is down ... */ if (rsc->role > RSC_ROLE_STOPPED && node->details->online == FALSE && node->details->maintenance == FALSE && is_set(rsc->flags, pe_rsc_managed)) { gboolean should_fence = FALSE; /* If this is a guest node, fence it (regardless of whether fencing is * enabled, because guest node fencing is done by recovery of the * container resource rather than by the fencer). Mark the resource * we're processing as failed. When the guest comes back up, its * operation history in the CIB will be cleared, freeing the affected * resource to run again once we are sure we know its state. */ if (pe__is_guest_node(node)) { set_bit(rsc->flags, pe_rsc_failed); should_fence = TRUE; } else if (is_set(data_set->flags, pe_flag_stonith_enabled)) { if (pe__is_remote_node(node) && node->details->remote_rsc && is_not_set(node->details->remote_rsc->flags, pe_rsc_failed)) { /* Setting unseen means that fencing of the remote node will * occur only if the connection resource is not going to start * somewhere. This allows connection resources on a failed * cluster node to move to another node without requiring the * remote nodes to be fenced as well. */ node->details->unseen = TRUE; reason = crm_strdup_printf("%s is active there (fencing will be" " revoked if remote connection can " "be re-established elsewhere)", rsc->id); } should_fence = TRUE; } if (should_fence) { if (reason == NULL) { reason = crm_strdup_printf("%s is thought to be active there", rsc->id); } pe_fence_node(data_set, node, reason); } free(reason); } if (node->details->unclean) { /* No extra processing needed * Also allows resources to be started again after a node is shot */ on_fail = action_fail_ignore; } switch (on_fail) { case action_fail_ignore: /* nothing to do */ break; case action_fail_fence: /* treat it as if it is still running * but also mark the node as unclean */ reason = crm_strdup_printf("%s failed there", rsc->id); pe_fence_node(data_set, node, reason); free(reason); break; case action_fail_standby: node->details->standby = TRUE; node->details->standby_onfail = TRUE; break; case action_fail_block: /* is_managed == FALSE will prevent any * actions being sent for the resource */ clear_bit(rsc->flags, pe_rsc_managed); set_bit(rsc->flags, pe_rsc_block); break; case action_fail_migrate: /* make sure it comes up somewhere else * or not at all */ resource_location(rsc, node, -INFINITY, "__action_migration_auto__", data_set); break; case action_fail_stop: rsc->next_role = RSC_ROLE_STOPPED; break; case action_fail_recover: if (rsc->role != RSC_ROLE_STOPPED && rsc->role != RSC_ROLE_UNKNOWN) { set_bit(rsc->flags, pe_rsc_failed); stop_action(rsc, node, FALSE); } break; case action_fail_restart_container: set_bit(rsc->flags, pe_rsc_failed); if (rsc->container && pe_rsc_is_bundled(rsc)) { /* A bundle's remote connection can run on a different node than * the bundle's container. We don't necessarily know where the * container is running yet, so remember it and add a stop * action for it later. */ data_set->stop_needed = g_list_prepend(data_set->stop_needed, rsc->container); } else if (rsc->container) { stop_action(rsc->container, node, FALSE); } else if (rsc->role != RSC_ROLE_STOPPED && rsc->role != RSC_ROLE_UNKNOWN) { stop_action(rsc, node, FALSE); } break; case action_fail_reset_remote: set_bit(rsc->flags, pe_rsc_failed); if (is_set(data_set->flags, pe_flag_stonith_enabled)) { tmpnode = NULL; if (rsc->is_remote_node) { tmpnode = pe_find_node(data_set->nodes, rsc->id); } if (tmpnode && pe__is_remote_node(tmpnode) && tmpnode->details->remote_was_fenced == 0) { /* The remote connection resource failed in a way that * should result in fencing the remote node. */ pe_fence_node(data_set, tmpnode, "remote connection is unrecoverable"); } } /* require the stop action regardless if fencing is occurring or not. */ if (rsc->role > RSC_ROLE_STOPPED) { stop_action(rsc, node, FALSE); } /* if reconnect delay is in use, prevent the connection from exiting the * "STOPPED" role until the failure is cleared by the delay timeout. */ if (rsc->remote_reconnect_ms) { rsc->next_role = RSC_ROLE_STOPPED; } break; } /* ensure a remote-node connection failure forces an unclean remote-node * to be fenced. By setting unseen = FALSE, the remote-node failure will * result in a fencing operation regardless if we're going to attempt to * reconnect to the remote-node in this transition or not. */ if (is_set(rsc->flags, pe_rsc_failed) && rsc->is_remote_node) { tmpnode = pe_find_node(data_set->nodes, rsc->id); if (tmpnode && tmpnode->details->unclean) { tmpnode->details->unseen = FALSE; } } if (rsc->role != RSC_ROLE_STOPPED && rsc->role != RSC_ROLE_UNKNOWN) { if (is_set(rsc->flags, pe_rsc_orphan)) { if (is_set(rsc->flags, pe_rsc_managed)) { crm_config_warn("Detected active orphan %s running on %s", rsc->id, node->details->uname); } else { crm_config_warn("Cluster configured not to stop active orphans." " %s must be stopped manually on %s", rsc->id, node->details->uname); } } native_add_running(rsc, node, data_set); if (on_fail != action_fail_ignore) { set_bit(rsc->flags, pe_rsc_failed); } } else if (rsc->clone_name && strchr(rsc->clone_name, ':') != NULL) { /* Only do this for older status sections that included instance numbers * Otherwise stopped instances will appear as orphans */ pe_rsc_trace(rsc, "Resetting clone_name %s for %s (stopped)", rsc->clone_name, rsc->id); free(rsc->clone_name); rsc->clone_name = NULL; } else { GList *possible_matches = pe__resource_actions(rsc, node, RSC_STOP, FALSE); GListPtr gIter = possible_matches; for (; gIter != NULL; gIter = gIter->next) { action_t *stop = (action_t *) gIter->data; stop->flags |= pe_action_optional; } g_list_free(possible_matches); } } /* create active recurring operations as optional */ static void process_recurring(node_t * node, resource_t * rsc, int start_index, int stop_index, GListPtr sorted_op_list, pe_working_set_t * data_set) { int counter = -1; const char *task = NULL; const char *status = NULL; GListPtr gIter = sorted_op_list; CRM_ASSERT(rsc); pe_rsc_trace(rsc, "%s: Start index %d, stop index = %d", rsc->id, start_index, stop_index); for (; gIter != NULL; gIter = gIter->next) { xmlNode *rsc_op = (xmlNode *) gIter->data; guint interval_ms = 0; char *key = NULL; const char *id = ID(rsc_op); const char *interval_ms_s = NULL; counter++; if (node->details->online == FALSE) { pe_rsc_trace(rsc, "Skipping %s/%s: node is offline", rsc->id, node->details->uname); break; /* Need to check if there's a monitor for role="Stopped" */ } else if (start_index < stop_index && counter <= stop_index) { pe_rsc_trace(rsc, "Skipping %s/%s: resource is not active", id, node->details->uname); continue; } else if (counter < start_index) { pe_rsc_trace(rsc, "Skipping %s/%s: old %d", id, node->details->uname, counter); continue; } interval_ms_s = crm_element_value(rsc_op, XML_LRM_ATTR_INTERVAL_MS); interval_ms = crm_parse_ms(interval_ms_s); if (interval_ms == 0) { pe_rsc_trace(rsc, "Skipping %s/%s: non-recurring", id, node->details->uname); continue; } status = crm_element_value(rsc_op, XML_LRM_ATTR_OPSTATUS); if (safe_str_eq(status, "-1")) { pe_rsc_trace(rsc, "Skipping %s/%s: status", id, node->details->uname); continue; } task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); /* create the action */ key = generate_op_key(rsc->id, task, interval_ms); pe_rsc_trace(rsc, "Creating %s/%s", key, node->details->uname); custom_action(rsc, key, task, node, TRUE, TRUE, data_set); } } void calculate_active_ops(GListPtr sorted_op_list, int *start_index, int *stop_index) { int counter = -1; int implied_monitor_start = -1; int implied_clone_start = -1; const char *task = NULL; const char *status = NULL; GListPtr gIter = sorted_op_list; *stop_index = -1; *start_index = -1; for (; gIter != NULL; gIter = gIter->next) { xmlNode *rsc_op = (xmlNode *) gIter->data; counter++; task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); status = crm_element_value(rsc_op, XML_LRM_ATTR_OPSTATUS); if (safe_str_eq(task, CRMD_ACTION_STOP) && safe_str_eq(status, "0")) { *stop_index = counter; } else if (safe_str_eq(task, CRMD_ACTION_START) || safe_str_eq(task, CRMD_ACTION_MIGRATED)) { *start_index = counter; } else if ((implied_monitor_start <= *stop_index) && safe_str_eq(task, CRMD_ACTION_STATUS)) { const char *rc = crm_element_value(rsc_op, XML_LRM_ATTR_RC); if (safe_str_eq(rc, "0") || safe_str_eq(rc, "8")) { implied_monitor_start = counter; } } else if (safe_str_eq(task, CRMD_ACTION_PROMOTE) || safe_str_eq(task, CRMD_ACTION_DEMOTE)) { implied_clone_start = counter; } } if (*start_index == -1) { if (implied_clone_start != -1) { *start_index = implied_clone_start; } else if (implied_monitor_start != -1) { *start_index = implied_monitor_start; } } } static resource_t * unpack_lrm_rsc_state(node_t * node, xmlNode * rsc_entry, pe_working_set_t * data_set) { GListPtr gIter = NULL; int stop_index = -1; int start_index = -1; enum rsc_role_e req_role = RSC_ROLE_UNKNOWN; const char *task = NULL; const char *rsc_id = crm_element_value(rsc_entry, XML_ATTR_ID); resource_t *rsc = NULL; GListPtr op_list = NULL; GListPtr sorted_op_list = NULL; xmlNode *migrate_op = NULL; xmlNode *rsc_op = NULL; xmlNode *last_failure = NULL; enum action_fail_response on_fail = FALSE; enum rsc_role_e saved_role = RSC_ROLE_UNKNOWN; crm_trace("[%s] Processing %s on %s", crm_element_name(rsc_entry), rsc_id, node->details->uname); /* extract operations */ op_list = NULL; sorted_op_list = NULL; for (rsc_op = __xml_first_child(rsc_entry); rsc_op != NULL; rsc_op = __xml_next_element(rsc_op)) { if (crm_str_eq((const char *)rsc_op->name, XML_LRM_TAG_RSC_OP, TRUE)) { op_list = g_list_prepend(op_list, rsc_op); } } if (op_list == NULL) { /* if there are no operations, there is nothing to do */ return NULL; } /* find the resource */ rsc = unpack_find_resource(data_set, node, rsc_id, rsc_entry); if (rsc == NULL) { rsc = process_orphan_resource(rsc_entry, node, data_set); } CRM_ASSERT(rsc != NULL); /* process operations */ saved_role = rsc->role; on_fail = action_fail_ignore; rsc->role = RSC_ROLE_UNKNOWN; sorted_op_list = g_list_sort(op_list, sort_op_by_callid); for (gIter = sorted_op_list; gIter != NULL; gIter = gIter->next) { xmlNode *rsc_op = (xmlNode *) gIter->data; task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); if (safe_str_eq(task, CRMD_ACTION_MIGRATED)) { migrate_op = rsc_op; } unpack_rsc_op(rsc, node, rsc_op, &last_failure, &on_fail, data_set); } /* create active recurring operations as optional */ calculate_active_ops(sorted_op_list, &start_index, &stop_index); process_recurring(node, rsc, start_index, stop_index, sorted_op_list, data_set); /* no need to free the contents */ g_list_free(sorted_op_list); process_rsc_state(rsc, node, on_fail, migrate_op, data_set); if (get_target_role(rsc, &req_role)) { if (rsc->next_role == RSC_ROLE_UNKNOWN || req_role < rsc->next_role) { pe_rsc_debug(rsc, "%s: Overwriting calculated next role %s" " with requested next role %s", rsc->id, role2text(rsc->next_role), role2text(req_role)); rsc->next_role = req_role; } else if (req_role > rsc->next_role) { pe_rsc_info(rsc, "%s: Not overwriting calculated next role %s" " with requested next role %s", rsc->id, role2text(rsc->next_role), role2text(req_role)); } } if (saved_role > rsc->role) { rsc->role = saved_role; } return rsc; } static void handle_orphaned_container_fillers(xmlNode * lrm_rsc_list, pe_working_set_t * data_set) { xmlNode *rsc_entry = NULL; for (rsc_entry = __xml_first_child(lrm_rsc_list); rsc_entry != NULL; rsc_entry = __xml_next_element(rsc_entry)) { resource_t *rsc; resource_t *container; const char *rsc_id; const char *container_id; if (safe_str_neq((const char *)rsc_entry->name, XML_LRM_TAG_RESOURCE)) { continue; } container_id = crm_element_value(rsc_entry, XML_RSC_ATTR_CONTAINER); rsc_id = crm_element_value(rsc_entry, XML_ATTR_ID); if (container_id == NULL || rsc_id == NULL) { continue; } container = pe_find_resource(data_set->resources, container_id); if (container == NULL) { continue; } rsc = pe_find_resource(data_set->resources, rsc_id); if (rsc == NULL || is_set(rsc->flags, pe_rsc_orphan_container_filler) == FALSE || rsc->container != NULL) { continue; } pe_rsc_trace(rsc, "Mapped container of orphaned resource %s to %s", rsc->id, container_id); rsc->container = container; container->fillers = g_list_append(container->fillers, rsc); } } gboolean unpack_lrm_resources(node_t * node, xmlNode * lrm_rsc_list, pe_working_set_t * data_set) { xmlNode *rsc_entry = NULL; gboolean found_orphaned_container_filler = FALSE; CRM_CHECK(node != NULL, return FALSE); crm_trace("Unpacking resources on %s", node->details->uname); for (rsc_entry = __xml_first_child(lrm_rsc_list); rsc_entry != NULL; rsc_entry = __xml_next_element(rsc_entry)) { if (crm_str_eq((const char *)rsc_entry->name, XML_LRM_TAG_RESOURCE, TRUE)) { resource_t *rsc = unpack_lrm_rsc_state(node, rsc_entry, data_set); if (!rsc) { continue; } if (is_set(rsc->flags, pe_rsc_orphan_container_filler)) { found_orphaned_container_filler = TRUE; } } } /* now that all the resource state has been unpacked for this node * we have to go back and map any orphaned container fillers to their * container resource */ if (found_orphaned_container_filler) { handle_orphaned_container_fillers(lrm_rsc_list, data_set); } return TRUE; } static void set_active(resource_t * rsc) { resource_t *top = uber_parent(rsc); if (top && is_set(top->flags, pe_rsc_promotable)) { rsc->role = RSC_ROLE_SLAVE; } else { rsc->role = RSC_ROLE_STARTED; } } static void set_node_score(gpointer key, gpointer value, gpointer user_data) { node_t *node = value; int *score = user_data; node->weight = *score; } #define STATUS_PATH_MAX 1024 static xmlNode * find_lrm_op(const char *resource, const char *op, const char *node, const char *source, pe_working_set_t * data_set) { int offset = 0; char xpath[STATUS_PATH_MAX]; offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "//node_state[@uname='%s']", node); offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "//" XML_LRM_TAG_RESOURCE "[@id='%s']", resource); /* Need to check against transition_magic too? */ if (source && safe_str_eq(op, CRMD_ACTION_MIGRATE)) { offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "/" XML_LRM_TAG_RSC_OP "[@operation='%s' and @migrate_target='%s']", op, source); } else if (source && safe_str_eq(op, CRMD_ACTION_MIGRATED)) { offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "/" XML_LRM_TAG_RSC_OP "[@operation='%s' and @migrate_source='%s']", op, source); } else { offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "/" XML_LRM_TAG_RSC_OP "[@operation='%s']", op); } CRM_LOG_ASSERT(offset > 0); return get_xpath_object(xpath, data_set->input, LOG_DEBUG); } static bool stop_happened_after(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op, pe_working_set_t *data_set) { xmlNode *stop_op = find_lrm_op(rsc->id, CRMD_ACTION_STOP, node->details->id, NULL, data_set); if (stop_op) { int stop_id = 0; int task_id = 0; crm_element_value_int(stop_op, XML_LRM_ATTR_CALLID, &stop_id); crm_element_value_int(xml_op, XML_LRM_ATTR_CALLID, &task_id); if (stop_id > task_id) { return TRUE; } } return FALSE; } static void unpack_rsc_migration(resource_t *rsc, node_t *node, xmlNode *xml_op, pe_working_set_t * data_set) { /* A successful migration sequence is: * migrate_to on source node * migrate_from on target node * stop on source node * * If a migrate_to is followed by a stop, the entire migration (successful * or failed) is complete, and we don't care what happened on the target. * * If no migrate_from has happened, the migration is considered to be * "partial". If the migrate_from failed, make sure the resource gets * stopped on both source and target (if up). * * If the migrate_to and migrate_from both succeeded (which also implies the * resource is no longer running on the source), but there is no stop, the * migration is considered to be "dangling". */ int from_rc = 0; int from_status = 0; const char *migrate_source = NULL; const char *migrate_target = NULL; pe_node_t *target = NULL; pe_node_t *source = NULL; xmlNode *migrate_from = NULL; if (stop_happened_after(rsc, node, xml_op, data_set)) { return; } // Clones are not allowed to migrate, so role can't be master rsc->role = RSC_ROLE_STARTED; migrate_source = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_SOURCE); migrate_target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET); target = pe_find_node(data_set->nodes, migrate_target); source = pe_find_node(data_set->nodes, migrate_source); // Check whether there was a migrate_from action migrate_from = find_lrm_op(rsc->id, CRMD_ACTION_MIGRATED, migrate_target, migrate_source, data_set); if (migrate_from) { crm_element_value_int(migrate_from, XML_LRM_ATTR_RC, &from_rc); crm_element_value_int(migrate_from, XML_LRM_ATTR_OPSTATUS, &from_status); pe_rsc_trace(rsc, "%s op on %s exited with status=%d, rc=%d", ID(migrate_from), migrate_target, from_status, from_rc); } if (migrate_from && from_rc == PCMK_OCF_OK && from_status == PCMK_LRM_OP_DONE) { /* The migrate_to and migrate_from both succeeded, so mark the migration * as "dangling". This will be used to schedule a stop action on the * source without affecting the target. */ pe_rsc_trace(rsc, "Detected dangling migration op: %s on %s", ID(xml_op), migrate_source); rsc->role = RSC_ROLE_STOPPED; rsc->dangling_migrations = g_list_prepend(rsc->dangling_migrations, node); } else if (migrate_from && (from_status != PCMK_LRM_OP_PENDING)) { // Failed if (target && target->details->online) { pe_rsc_trace(rsc, "Marking active on %s %p %d", migrate_target, target, target->details->online); native_add_running(rsc, target, data_set); } } else { // Pending, or complete but erased if (target && target->details->online) { pe_rsc_trace(rsc, "Marking active on %s %p %d", migrate_target, target, target->details->online); native_add_running(rsc, target, data_set); if (source && source->details->online) { /* This is a partial migration: the migrate_to completed * successfully on the source, but the migrate_from has not * completed. Remember the source and target; if the newly * chosen target remains the same when we schedule actions * later, we may continue with the migration. */ rsc->partial_migration_target = target; rsc->partial_migration_source = source; } } else { /* Consider it failed here - forces a restart, prevents migration */ set_bit(rsc->flags, pe_rsc_failed); clear_bit(rsc->flags, pe_rsc_allow_migrate); } } } static void unpack_rsc_migration_failure(resource_t *rsc, node_t *node, xmlNode *xml_op, pe_working_set_t * data_set) { const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); CRM_ASSERT(rsc); if (safe_str_eq(task, CRMD_ACTION_MIGRATED)) { int stop_id = 0; int migrate_id = 0; const char *migrate_source = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_SOURCE); const char *migrate_target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET); xmlNode *stop_op = find_lrm_op(rsc->id, CRMD_ACTION_STOP, migrate_source, NULL, data_set); xmlNode *migrate_op = find_lrm_op(rsc->id, CRMD_ACTION_MIGRATE, migrate_source, migrate_target, data_set); if (stop_op) { crm_element_value_int(stop_op, XML_LRM_ATTR_CALLID, &stop_id); } if (migrate_op) { crm_element_value_int(migrate_op, XML_LRM_ATTR_CALLID, &migrate_id); } /* Get our state right */ rsc->role = RSC_ROLE_STARTED; /* can be master? */ if (stop_op == NULL || stop_id < migrate_id) { node_t *source = pe_find_node(data_set->nodes, migrate_source); if (source && source->details->online) { native_add_running(rsc, source, data_set); } } } else if (safe_str_eq(task, CRMD_ACTION_MIGRATE)) { int stop_id = 0; int migrate_id = 0; const char *migrate_source = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_SOURCE); const char *migrate_target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET); xmlNode *stop_op = find_lrm_op(rsc->id, CRMD_ACTION_STOP, migrate_target, NULL, data_set); xmlNode *migrate_op = find_lrm_op(rsc->id, CRMD_ACTION_MIGRATED, migrate_target, migrate_source, data_set); if (stop_op) { crm_element_value_int(stop_op, XML_LRM_ATTR_CALLID, &stop_id); } if (migrate_op) { crm_element_value_int(migrate_op, XML_LRM_ATTR_CALLID, &migrate_id); } /* Get our state right */ rsc->role = RSC_ROLE_STARTED; /* can be master? */ if (stop_op == NULL || stop_id < migrate_id) { node_t *target = pe_find_node(data_set->nodes, migrate_target); pe_rsc_trace(rsc, "Stop: %p %d, Migrated: %p %d", stop_op, stop_id, migrate_op, migrate_id); if (target && target->details->online) { native_add_running(rsc, target, data_set); } } else if (migrate_op == NULL) { /* Make sure it gets cleaned up, the stop may pre-date the migrate_from */ rsc->dangling_migrations = g_list_prepend(rsc->dangling_migrations, node); } } } static void record_failed_op(xmlNode *op, node_t* node, resource_t *rsc, pe_working_set_t * data_set) { xmlNode *xIter = NULL; const char *op_key = crm_element_value(op, XML_LRM_ATTR_TASK_KEY); if (node->details->online == FALSE) { return; } for (xIter = data_set->failed->children; xIter; xIter = xIter->next) { const char *key = crm_element_value(xIter, XML_LRM_ATTR_TASK_KEY); const char *uname = crm_element_value(xIter, XML_ATTR_UNAME); if(safe_str_eq(op_key, key) && safe_str_eq(uname, node->details->uname)) { crm_trace("Skipping duplicate entry %s on %s", op_key, node->details->uname); return; } } crm_trace("Adding entry %s on %s", op_key, node->details->uname); crm_xml_add(op, XML_ATTR_UNAME, node->details->uname); crm_xml_add(op, XML_LRM_ATTR_RSCID, rsc->id); add_node_copy(data_set->failed, op); } static const char *get_op_key(xmlNode *xml_op) { const char *key = crm_element_value(xml_op, XML_LRM_ATTR_TASK_KEY); if(key == NULL) { key = ID(xml_op); } return key; } static void unpack_rsc_op_failure(resource_t * rsc, node_t * node, int rc, xmlNode * xml_op, xmlNode ** last_failure, enum action_fail_response * on_fail, pe_working_set_t * data_set) { guint interval_ms = 0; bool is_probe = FALSE; action_t *action = NULL; const char *key = get_op_key(xml_op); const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); CRM_ASSERT(rsc); *last_failure = xml_op; crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); if ((interval_ms == 0) && safe_str_eq(task, CRMD_ACTION_STATUS)) { is_probe = TRUE; pe_rsc_trace(rsc, "is a probe: %s", key); } if (rc != PCMK_OCF_NOT_INSTALLED || is_set(data_set->flags, pe_flag_symmetric_cluster)) { crm_warn("Processing failed %s of %s on %s: %s " CRM_XS " rc=%d", (is_probe? "probe" : task), rsc->id, node->details->uname, services_ocf_exitcode_str(rc), rc); if (is_probe && (rc != PCMK_OCF_OK) && (rc != PCMK_OCF_NOT_RUNNING) && (rc != PCMK_OCF_RUNNING_MASTER)) { /* A failed (not just unexpected) probe result could mean the user * didn't know resources will be probed even where they can't run. */ crm_notice("If it is not possible for %s to run on %s, see " "the resource-discovery option for location constraints", rsc->id, node->details->uname); } record_failed_op(xml_op, node, rsc, data_set); } else { crm_trace("Processing failed op %s for %s on %s: %s (%d)", task, rsc->id, node->details->uname, services_ocf_exitcode_str(rc), rc); } action = custom_action(rsc, strdup(key), task, NULL, TRUE, FALSE, data_set); if ((action->on_fail <= action_fail_fence && *on_fail < action->on_fail) || (action->on_fail == action_fail_reset_remote && *on_fail <= action_fail_recover) || (action->on_fail == action_fail_restart_container && *on_fail <= action_fail_recover) || (*on_fail == action_fail_restart_container && action->on_fail >= action_fail_migrate)) { pe_rsc_trace(rsc, "on-fail %s -> %s for %s (%s)", fail2text(*on_fail), fail2text(action->on_fail), action->uuid, key); *on_fail = action->on_fail; } if (safe_str_eq(task, CRMD_ACTION_STOP)) { resource_location(rsc, node, -INFINITY, "__stop_fail__", data_set); } else if (safe_str_eq(task, CRMD_ACTION_MIGRATE) || safe_str_eq(task, CRMD_ACTION_MIGRATED)) { unpack_rsc_migration_failure(rsc, node, xml_op, data_set); } else if (safe_str_eq(task, CRMD_ACTION_PROMOTE)) { rsc->role = RSC_ROLE_MASTER; } else if (safe_str_eq(task, CRMD_ACTION_DEMOTE)) { if (action->on_fail == action_fail_block) { rsc->role = RSC_ROLE_MASTER; rsc->next_role = RSC_ROLE_STOPPED; } else if(rc == PCMK_OCF_NOT_RUNNING) { rsc->role = RSC_ROLE_STOPPED; } else { /* * Staying in master role would put the PE/TE into a loop. Setting * slave role is not dangerous because the resource will be stopped * as part of recovery, and any master promotion will be ordered * after that stop. */ rsc->role = RSC_ROLE_SLAVE; } } if(is_probe && rc == PCMK_OCF_NOT_INSTALLED) { /* leave stopped */ pe_rsc_trace(rsc, "Leaving %s stopped", rsc->id); rsc->role = RSC_ROLE_STOPPED; } else if (rsc->role < RSC_ROLE_STARTED) { pe_rsc_trace(rsc, "Setting %s active", rsc->id); set_active(rsc); } pe_rsc_trace(rsc, "Resource %s: role=%s, unclean=%s, on_fail=%s, fail_role=%s", rsc->id, role2text(rsc->role), node->details->unclean ? "true" : "false", fail2text(action->on_fail), role2text(action->fail_role)); if (action->fail_role != RSC_ROLE_STARTED && rsc->next_role < action->fail_role) { rsc->next_role = action->fail_role; } if (action->fail_role == RSC_ROLE_STOPPED) { int score = -INFINITY; resource_t *fail_rsc = rsc; if (fail_rsc->parent) { resource_t *parent = uber_parent(fail_rsc); if (pe_rsc_is_clone(parent) && is_not_set(parent->flags, pe_rsc_unique)) { /* For clone resources, if a child fails on an operation * with on-fail = stop, all the resources fail. Do this by preventing * the parent from coming up again. */ fail_rsc = parent; } } crm_warn("Making sure %s doesn't come up again", fail_rsc->id); /* make sure it doesn't come up again */ if (fail_rsc->allowed_nodes != NULL) { g_hash_table_destroy(fail_rsc->allowed_nodes); } fail_rsc->allowed_nodes = node_hash_from_list(data_set->nodes); g_hash_table_foreach(fail_rsc->allowed_nodes, set_node_score, &score); } pe_free_action(action); } /*! * \internal * \brief Remap operation status based on action result * * Given an action result, determine an appropriate operation status for the * purposes of responding to the action (the status provided by the executor is * not directly usable since the executor does not know what was expected). * * \param[in,out] rsc Resource that operation history entry is for * \param[in] rc Actual return code of operation * \param[in] target_rc Expected return code of operation * \param[in] node Node where operation was executed * \param[in] xml_op Operation history entry XML from CIB status * \param[in,out] on_fail What should be done about the result * \param[in] data_set Current cluster working set * * \return Operation status based on return code and action info * \note This may update the resource's current and next role. */ static int determine_op_status( resource_t *rsc, int rc, int target_rc, node_t * node, xmlNode * xml_op, enum action_fail_response * on_fail, pe_working_set_t * data_set) { guint interval_ms = 0; int result = PCMK_LRM_OP_DONE; const char *key = get_op_key(xml_op); const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); bool is_probe = FALSE; CRM_ASSERT(rsc); crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); if ((interval_ms == 0) && safe_str_eq(task, CRMD_ACTION_STATUS)) { is_probe = TRUE; } if (target_rc < 0) { /* Pre-1.0 Pacemaker versions, and Pacemaker 1.1.6 or earlier with * Heartbeat 2.0.7 or earlier as the cluster layer, did not include the * target_rc in the transition key, which (along with the similar case * of a corrupted transition key in the CIB) will be reported to this * function as -1. Pacemaker 2.0+ does not support rolling upgrades from * those versions or processing of saved CIB files from those versions, * so we do not need to care much about this case. */ result = PCMK_LRM_OP_ERROR; crm_warn("Expected result not found for %s on %s (corrupt or obsolete CIB?)", key, node->details->uname); } else if (target_rc != rc) { result = PCMK_LRM_OP_ERROR; pe_rsc_debug(rsc, "%s on %s returned '%s' (%d) instead of the expected value: '%s' (%d)", key, node->details->uname, services_ocf_exitcode_str(rc), rc, services_ocf_exitcode_str(target_rc), target_rc); } switch (rc) { case PCMK_OCF_OK: // @TODO Should this be (rc != target_rc)? if (is_probe && (target_rc == PCMK_OCF_NOT_RUNNING)) { result = PCMK_LRM_OP_DONE; pe_rsc_info(rsc, "Operation %s found resource %s active on %s", task, rsc->id, node->details->uname); } break; case PCMK_OCF_NOT_RUNNING: if (is_probe || target_rc == rc || is_not_set(rsc->flags, pe_rsc_managed)) { result = PCMK_LRM_OP_DONE; rsc->role = RSC_ROLE_STOPPED; /* clear any previous failure actions */ *on_fail = action_fail_ignore; rsc->next_role = RSC_ROLE_UNKNOWN; } break; case PCMK_OCF_RUNNING_MASTER: if (is_probe && (rc != target_rc)) { result = PCMK_LRM_OP_DONE; pe_rsc_info(rsc, "Operation %s found resource %s active in master mode on %s", task, rsc->id, node->details->uname); } rsc->role = RSC_ROLE_MASTER; break; case PCMK_OCF_DEGRADED_MASTER: case PCMK_OCF_FAILED_MASTER: rsc->role = RSC_ROLE_MASTER; result = PCMK_LRM_OP_ERROR; break; case PCMK_OCF_NOT_CONFIGURED: result = PCMK_LRM_OP_ERROR_FATAL; break; case PCMK_OCF_UNIMPLEMENT_FEATURE: if (interval_ms > 0) { result = PCMK_LRM_OP_NOTSUPPORTED; break; } // fall through case PCMK_OCF_NOT_INSTALLED: case PCMK_OCF_INVALID_PARAM: case PCMK_OCF_INSUFFICIENT_PRIV: if (!pe_can_fence(data_set, node) && safe_str_eq(task, CRMD_ACTION_STOP)) { /* If a stop fails and we can't fence, there's nothing else we can do */ pe_proc_err("No further recovery can be attempted for %s: %s action failed with '%s' (%d)", rsc->id, task, services_ocf_exitcode_str(rc), rc); clear_bit(rsc->flags, pe_rsc_managed); set_bit(rsc->flags, pe_rsc_block); } result = PCMK_LRM_OP_ERROR_HARD; break; default: if (result == PCMK_LRM_OP_DONE) { crm_info("Treating unknown return code %d for %s on %s as failure", rc, key, node->details->uname); result = PCMK_LRM_OP_ERROR; } break; } return result; } static bool check_operation_expiry(resource_t *rsc, node_t *node, int rc, xmlNode *xml_op, pe_working_set_t * data_set) { bool expired = FALSE; time_t last_failure = 0; guint interval_ms = 0; int failure_timeout = rsc->failure_timeout; const char *key = get_op_key(xml_op); const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); const char *clear_reason = NULL; crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); /* clearing recurring monitor operation failures automatically * needs to be carefully considered */ if ((interval_ms != 0) && safe_str_eq(task, "monitor")) { /* TODO, in the future we should consider not clearing recurring monitor * op failures unless the last action for a resource was a "stop" action. * otherwise it is possible that clearing the monitor failure will result * in the resource being in an undeterministic state. * * For now we handle this potential undeterministic condition for remote * node connection resources by not clearing a recurring monitor op failure * until after the node has been fenced. */ if (is_set(data_set->flags, pe_flag_stonith_enabled) && rsc->remote_reconnect_ms) { node_t *remote_node = pe_find_node(data_set->nodes, rsc->id); if (remote_node && remote_node->details->remote_was_fenced == 0) { if (strstr(ID(xml_op), "last_failure")) { crm_info("Waiting to clear monitor failure for remote node %s until fencing has occurred", rsc->id); } /* disabling failure timeout for this operation because we believe * fencing of the remote node should occur first. */ failure_timeout = 0; } } } if (failure_timeout > 0) { int last_run = 0; if (crm_element_value_int(xml_op, XML_RSC_OP_LAST_CHANGE, &last_run) == 0) { time_t now = get_effective_time(data_set); if (now > (last_run + failure_timeout)) { expired = TRUE; } } } if (expired) { if (pe_get_failcount(node, rsc, &last_failure, pe_fc_default, xml_op, data_set)) { // There is a fail count ignoring timeout if (pe_get_failcount(node, rsc, &last_failure, pe_fc_effective, xml_op, data_set) == 0) { // There is no fail count considering timeout clear_reason = "it expired"; } else { expired = FALSE; } } else if (rsc->remote_reconnect_ms && strstr(ID(xml_op), "last_failure")) { // Always clear last failure when reconnect interval is set clear_reason = "reconnect interval is set"; } } else if (strstr(ID(xml_op), "last_failure") && ((strcmp(task, "start") == 0) || (strcmp(task, "monitor") == 0))) { if (pe__bundle_needs_remote_name(rsc)) { /* We haven't allocated resources yet, so we can't reliably * substitute addr parameters for the REMOTE_CONTAINER_HACK. * When that's needed, defer the check until later. */ pe__add_param_check(xml_op, rsc, node, pe_check_last_failure, data_set); } else { op_digest_cache_t *digest_data = NULL; digest_data = rsc_action_digest_cmp(rsc, xml_op, node, data_set); switch (digest_data->rc) { case RSC_DIGEST_UNKNOWN: crm_trace("Resource %s history entry %s on %s has no digest to compare", rsc->id, key, node->details->id); break; case RSC_DIGEST_MATCH: break; default: clear_reason = "resource parameters have changed"; break; } } } if (clear_reason != NULL) { // Schedule clearing of the fail count pe_action_t *clear_op = pe__clear_failcount(rsc, node, clear_reason, data_set); if (is_set(data_set->flags, pe_flag_stonith_enabled) && rsc->remote_reconnect_ms) { pe_node_t *remote_node = pe_find_node(data_set->nodes, rsc->id); if (remote_node) { /* If we're clearing a remote connection due to a reconnect * interval, we want to wait until any scheduled fencing * completes. * * We could limit this to remote_node->details->unclean, but at * this point, that's always true (it won't be reliable until * after unpack_node_loop() is done). */ pe_action_t *fence = pe_fence_op(remote_node, NULL, TRUE, NULL, data_set); crm_info("Clearing %s failure will wait until any scheduled " "fencing of %s completes", task, rsc->id); order_actions(fence, clear_op, pe_order_implies_then); } } } if (expired && (interval_ms == 0) && safe_str_eq(task, CRMD_ACTION_STATUS)) { switch(rc) { case PCMK_OCF_OK: case PCMK_OCF_NOT_RUNNING: case PCMK_OCF_RUNNING_MASTER: case PCMK_OCF_DEGRADED: case PCMK_OCF_DEGRADED_MASTER: /* Don't expire probes that return these values */ expired = FALSE; break; } } return expired; } int get_target_rc(xmlNode *xml_op) { - int dummy = 0; int target_rc = 0; - char *dummy_string = NULL; const char *key = crm_element_value(xml_op, XML_ATTR_TRANSITION_KEY); + if (key == NULL) { return -1; } - - decode_transition_key(key, &dummy_string, &dummy, &dummy, &target_rc); - free(dummy_string); - + decode_transition_key(key, NULL, NULL, NULL, &target_rc); return target_rc; } static enum action_fail_response get_action_on_fail(resource_t *rsc, const char *key, const char *task, pe_working_set_t * data_set) { int result = action_fail_recover; action_t *action = custom_action(rsc, strdup(key), task, NULL, TRUE, FALSE, data_set); result = action->on_fail; pe_free_action(action); return result; } static void update_resource_state(resource_t * rsc, node_t * node, xmlNode * xml_op, const char * task, int rc, xmlNode * last_failure, enum action_fail_response * on_fail, pe_working_set_t * data_set) { gboolean clear_past_failure = FALSE; CRM_ASSERT(rsc); CRM_ASSERT(xml_op); if (rc == PCMK_OCF_NOT_RUNNING) { clear_past_failure = TRUE; } else if (rc == PCMK_OCF_NOT_INSTALLED) { rsc->role = RSC_ROLE_STOPPED; } else if (safe_str_eq(task, CRMD_ACTION_STATUS)) { if (last_failure) { const char *op_key = get_op_key(xml_op); const char *last_failure_key = get_op_key(last_failure); if (safe_str_eq(op_key, last_failure_key)) { clear_past_failure = TRUE; } } if (rsc->role < RSC_ROLE_STARTED) { set_active(rsc); } } else if (safe_str_eq(task, CRMD_ACTION_START)) { rsc->role = RSC_ROLE_STARTED; clear_past_failure = TRUE; } else if (safe_str_eq(task, CRMD_ACTION_STOP)) { rsc->role = RSC_ROLE_STOPPED; clear_past_failure = TRUE; } else if (safe_str_eq(task, CRMD_ACTION_PROMOTE)) { rsc->role = RSC_ROLE_MASTER; clear_past_failure = TRUE; } else if (safe_str_eq(task, CRMD_ACTION_DEMOTE)) { /* Demote from Master does not clear an error */ rsc->role = RSC_ROLE_SLAVE; } else if (safe_str_eq(task, CRMD_ACTION_MIGRATED)) { rsc->role = RSC_ROLE_STARTED; clear_past_failure = TRUE; } else if (safe_str_eq(task, CRMD_ACTION_MIGRATE)) { unpack_rsc_migration(rsc, node, xml_op, data_set); } else if (rsc->role < RSC_ROLE_STARTED) { pe_rsc_trace(rsc, "%s active on %s", rsc->id, node->details->uname); set_active(rsc); } /* clear any previous failure actions */ if (clear_past_failure) { switch (*on_fail) { case action_fail_stop: case action_fail_fence: case action_fail_migrate: case action_fail_standby: pe_rsc_trace(rsc, "%s.%s is not cleared by a completed stop", rsc->id, fail2text(*on_fail)); break; case action_fail_block: case action_fail_ignore: case action_fail_recover: case action_fail_restart_container: *on_fail = action_fail_ignore; rsc->next_role = RSC_ROLE_UNKNOWN; break; case action_fail_reset_remote: if (rsc->remote_reconnect_ms == 0) { /* With no reconnect interval, the connection is allowed to * start again after the remote node is fenced and * completely stopped. (With a reconnect interval, we wait * for the failure to be cleared entirely before attempting * to reconnect.) */ *on_fail = action_fail_ignore; rsc->next_role = RSC_ROLE_UNKNOWN; } break; } } } gboolean unpack_rsc_op(resource_t * rsc, node_t * node, xmlNode * xml_op, xmlNode ** last_failure, enum action_fail_response * on_fail, pe_working_set_t * data_set) { int task_id = 0; const char *key = NULL; const char *task = NULL; const char *task_key = NULL; int rc = 0; int status = PCMK_LRM_OP_UNKNOWN; int target_rc = get_target_rc(xml_op); guint interval_ms = 0; gboolean expired = FALSE; resource_t *parent = rsc; enum action_fail_response failure_strategy = action_fail_recover; CRM_CHECK(rsc != NULL, return FALSE); CRM_CHECK(node != NULL, return FALSE); CRM_CHECK(xml_op != NULL, return FALSE); task_key = get_op_key(xml_op); task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); key = crm_element_value(xml_op, XML_ATTR_TRANSITION_KEY); crm_element_value_int(xml_op, XML_LRM_ATTR_RC, &rc); crm_element_value_int(xml_op, XML_LRM_ATTR_CALLID, &task_id); crm_element_value_int(xml_op, XML_LRM_ATTR_OPSTATUS, &status); crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); CRM_CHECK(task != NULL, return FALSE); CRM_CHECK(status <= PCMK_LRM_OP_NOT_INSTALLED, return FALSE); CRM_CHECK(status >= PCMK_LRM_OP_PENDING, return FALSE); if (safe_str_eq(task, CRMD_ACTION_NOTIFY) || safe_str_eq(task, CRMD_ACTION_METADATA)) { /* safe to ignore these */ return TRUE; } if (is_not_set(rsc->flags, pe_rsc_unique)) { parent = uber_parent(rsc); } pe_rsc_trace(rsc, "Unpacking task %s/%s (call_id=%d, status=%d, rc=%d) on %s (role=%s)", task_key, task, task_id, status, rc, node->details->uname, role2text(rsc->role)); if (node->details->unclean) { pe_rsc_trace(rsc, "Node %s (where %s is running) is unclean." " Further action depends on the value of the stop's on-fail attribute", node->details->uname, rsc->id); } if(status != PCMK_LRM_OP_NOT_INSTALLED) { expired = check_operation_expiry(rsc, node, rc, xml_op, data_set); } /* Degraded results are informational only, re-map them to their error-free equivalents */ if (rc == PCMK_OCF_DEGRADED && safe_str_eq(task, CRMD_ACTION_STATUS)) { rc = PCMK_OCF_OK; /* Add them to the failed list to highlight them for the user */ if ((node->details->shutdown == FALSE) || (node->details->online == TRUE)) { crm_trace("Remapping %d to %d", PCMK_OCF_DEGRADED, PCMK_OCF_OK); record_failed_op(xml_op, node, rsc, data_set); } } else if (rc == PCMK_OCF_DEGRADED_MASTER && safe_str_eq(task, CRMD_ACTION_STATUS)) { rc = PCMK_OCF_RUNNING_MASTER; /* Add them to the failed list to highlight them for the user */ if ((node->details->shutdown == FALSE) || (node->details->online == TRUE)) { crm_trace("Remapping %d to %d", PCMK_OCF_DEGRADED_MASTER, PCMK_OCF_RUNNING_MASTER); record_failed_op(xml_op, node, rsc, data_set); } } if (expired && target_rc != rc) { const char *magic = crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC); pe_rsc_debug(rsc, "Expired operation '%s' on %s returned '%s' (%d) instead of the expected value: '%s' (%d)", key, node->details->uname, services_ocf_exitcode_str(rc), rc, services_ocf_exitcode_str(target_rc), target_rc); if (interval_ms == 0) { crm_notice("Ignoring expired calculated failure %s (rc=%d, magic=%s) on %s", task_key, rc, magic, node->details->uname); goto done; } else if(node->details->online && node->details->unclean == FALSE) { crm_notice("Re-initiated expired calculated failure %s (rc=%d, magic=%s) on %s", task_key, rc, magic, node->details->uname); /* This is SO horrible, but we don't have access to CancelXmlOp() yet */ crm_xml_add(xml_op, XML_LRM_ATTR_RESTART_DIGEST, "calculated-failure-timeout"); goto done; } } /* If the executor reported an operation status of anything but done or * error, consider that final. But for done or error, we know better whether * it should be treated as a failure or not, because we know the expected * result. */ if(status == PCMK_LRM_OP_DONE || status == PCMK_LRM_OP_ERROR) { status = determine_op_status(rsc, rc, target_rc, node, xml_op, on_fail, data_set); } pe_rsc_trace(rsc, "Handling status: %d", status); switch (status) { case PCMK_LRM_OP_CANCELLED: /* do nothing?? */ pe_err("Don't know what to do for cancelled ops yet"); break; case PCMK_LRM_OP_PENDING: if (safe_str_eq(task, CRMD_ACTION_START)) { set_bit(rsc->flags, pe_rsc_start_pending); set_active(rsc); } else if (safe_str_eq(task, CRMD_ACTION_PROMOTE)) { rsc->role = RSC_ROLE_MASTER; } else if (safe_str_eq(task, CRMD_ACTION_MIGRATE) && node->details->unclean) { /* If a pending migrate_to action is out on a unclean node, * we have to force the stop action on the target. */ const char *migrate_target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET); node_t *target = pe_find_node(data_set->nodes, migrate_target); if (target) { stop_action(rsc, target, FALSE); } } if (rsc->pending_task == NULL) { if (safe_str_eq(task, CRMD_ACTION_STATUS) && (interval_ms == 0)) { /* Pending probes are not printed, even if pending * operations are requested. If someone ever requests that * behavior, uncomment this and the corresponding part of * native.c:native_pending_task(). */ /*rsc->pending_task = strdup("probe");*/ /*rsc->pending_node = node;*/ } else { rsc->pending_task = strdup(task); rsc->pending_node = node; } } break; case PCMK_LRM_OP_DONE: pe_rsc_trace(rsc, "%s/%s completed on %s", rsc->id, task, node->details->uname); update_resource_state(rsc, node, xml_op, task, rc, *last_failure, on_fail, data_set); break; case PCMK_LRM_OP_NOT_INSTALLED: failure_strategy = get_action_on_fail(rsc, task_key, task, data_set); if (failure_strategy == action_fail_ignore) { crm_warn("Cannot ignore failed %s (status=%d, rc=%d) on %s: " "Resource agent doesn't exist", task_key, status, rc, node->details->uname); /* Also for printing it as "FAILED" by marking it as pe_rsc_failed later */ *on_fail = action_fail_migrate; } resource_location(parent, node, -INFINITY, "hard-error", data_set); unpack_rsc_op_failure(rsc, node, rc, xml_op, last_failure, on_fail, data_set); break; case PCMK_LRM_OP_ERROR: case PCMK_LRM_OP_ERROR_HARD: case PCMK_LRM_OP_ERROR_FATAL: case PCMK_LRM_OP_TIMEOUT: case PCMK_LRM_OP_NOTSUPPORTED: failure_strategy = get_action_on_fail(rsc, task_key, task, data_set); if ((failure_strategy == action_fail_ignore) || (failure_strategy == action_fail_restart_container && safe_str_eq(task, CRMD_ACTION_STOP))) { crm_warn("Pretending the failure of %s (rc=%d) on %s succeeded", task_key, rc, node->details->uname); update_resource_state(rsc, node, xml_op, task, target_rc, *last_failure, on_fail, data_set); crm_xml_add(xml_op, XML_ATTR_UNAME, node->details->uname); set_bit(rsc->flags, pe_rsc_failure_ignored); record_failed_op(xml_op, node, rsc, data_set); if (failure_strategy == action_fail_restart_container && *on_fail <= action_fail_recover) { *on_fail = failure_strategy; } } else { unpack_rsc_op_failure(rsc, node, rc, xml_op, last_failure, on_fail, data_set); if(status == PCMK_LRM_OP_ERROR_HARD) { do_crm_log(rc != PCMK_OCF_NOT_INSTALLED?LOG_ERR:LOG_NOTICE, "Preventing %s from re-starting on %s: operation %s failed '%s' (%d)", parent->id, node->details->uname, task, services_ocf_exitcode_str(rc), rc); resource_location(parent, node, -INFINITY, "hard-error", data_set); } else if(status == PCMK_LRM_OP_ERROR_FATAL) { crm_err("Preventing %s from re-starting anywhere: operation %s failed '%s' (%d)", parent->id, task, services_ocf_exitcode_str(rc), rc); resource_location(parent, NULL, -INFINITY, "fatal-error", data_set); } } break; } done: pe_rsc_trace(rsc, "Resource %s after %s: role=%s, next=%s", rsc->id, task, role2text(rsc->role), role2text(rsc->next_role)); return TRUE; } gboolean add_node_attrs(xmlNode * xml_obj, node_t * node, gboolean overwrite, pe_working_set_t * data_set) { const char *cluster_name = NULL; g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_UNAME), strdup(node->details->uname)); g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_ID), strdup(node->details->id)); if (safe_str_eq(node->details->id, data_set->dc_uuid)) { data_set->dc_node = node; node->details->is_dc = TRUE; g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_IS_DC), strdup(XML_BOOLEAN_TRUE)); } else { g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_IS_DC), strdup(XML_BOOLEAN_FALSE)); } cluster_name = g_hash_table_lookup(data_set->config_hash, "cluster-name"); if (cluster_name) { g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_CLUSTER_NAME), strdup(cluster_name)); } unpack_instance_attributes(data_set->input, xml_obj, XML_TAG_ATTR_SETS, NULL, node->details->attrs, NULL, overwrite, data_set->now); if (pe_node_attribute_raw(node, CRM_ATTR_SITE_NAME) == NULL) { const char *site_name = pe_node_attribute_raw(node, "site-name"); if (site_name) { g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_SITE_NAME), strdup(site_name)); } else if (cluster_name) { /* Default to cluster-name if unset */ g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_SITE_NAME), strdup(cluster_name)); } } return TRUE; } static GListPtr extract_operations(const char *node, const char *rsc, xmlNode * rsc_entry, gboolean active_filter) { int counter = -1; int stop_index = -1; int start_index = -1; xmlNode *rsc_op = NULL; GListPtr gIter = NULL; GListPtr op_list = NULL; GListPtr sorted_op_list = NULL; /* extract operations */ op_list = NULL; sorted_op_list = NULL; for (rsc_op = __xml_first_child(rsc_entry); rsc_op != NULL; rsc_op = __xml_next_element(rsc_op)) { if (crm_str_eq((const char *)rsc_op->name, XML_LRM_TAG_RSC_OP, TRUE)) { crm_xml_add(rsc_op, "resource", rsc); crm_xml_add(rsc_op, XML_ATTR_UNAME, node); op_list = g_list_prepend(op_list, rsc_op); } } if (op_list == NULL) { /* if there are no operations, there is nothing to do */ return NULL; } sorted_op_list = g_list_sort(op_list, sort_op_by_callid); /* create active recurring operations as optional */ if (active_filter == FALSE) { return sorted_op_list; } op_list = NULL; calculate_active_ops(sorted_op_list, &start_index, &stop_index); for (gIter = sorted_op_list; gIter != NULL; gIter = gIter->next) { xmlNode *rsc_op = (xmlNode *) gIter->data; counter++; if (start_index < stop_index) { crm_trace("Skipping %s: not active", ID(rsc_entry)); break; } else if (counter < start_index) { crm_trace("Skipping %s: old", ID(rsc_op)); continue; } op_list = g_list_append(op_list, rsc_op); } g_list_free(sorted_op_list); return op_list; } GListPtr find_operations(const char *rsc, const char *node, gboolean active_filter, pe_working_set_t * data_set) { GListPtr output = NULL; GListPtr intermediate = NULL; xmlNode *tmp = NULL; xmlNode *status = find_xml_node(data_set->input, XML_CIB_TAG_STATUS, TRUE); node_t *this_node = NULL; xmlNode *node_state = NULL; for (node_state = __xml_first_child(status); node_state != NULL; node_state = __xml_next_element(node_state)) { if (crm_str_eq((const char *)node_state->name, XML_CIB_TAG_STATE, TRUE)) { const char *uname = crm_element_value(node_state, XML_ATTR_UNAME); if (node != NULL && safe_str_neq(uname, node)) { continue; } this_node = pe_find_node(data_set->nodes, uname); if(this_node == NULL) { CRM_LOG_ASSERT(this_node != NULL); continue; } else if (pe__is_guest_or_remote_node(this_node)) { determine_remote_online_status(data_set, this_node); } else { determine_online_status(node_state, this_node, data_set); } if (this_node->details->online || is_set(data_set->flags, pe_flag_stonith_enabled)) { /* offline nodes run no resources... * unless stonith is enabled in which case we need to * make sure rsc start events happen after the stonith */ xmlNode *lrm_rsc = NULL; tmp = find_xml_node(node_state, XML_CIB_TAG_LRM, FALSE); tmp = find_xml_node(tmp, XML_LRM_TAG_RESOURCES, FALSE); for (lrm_rsc = __xml_first_child(tmp); lrm_rsc != NULL; lrm_rsc = __xml_next_element(lrm_rsc)) { if (crm_str_eq((const char *)lrm_rsc->name, XML_LRM_TAG_RESOURCE, TRUE)) { const char *rsc_id = crm_element_value(lrm_rsc, XML_ATTR_ID); if (rsc != NULL && safe_str_neq(rsc_id, rsc)) { continue; } intermediate = extract_operations(uname, rsc_id, lrm_rsc, active_filter); output = g_list_concat(output, intermediate); } } } } } return output; } diff --git a/lib/pengine/utils.c b/lib/pengine/utils.c index a788309f6a..2f4dc1e84e 100644 --- a/lib/pengine/utils.c +++ b/lib/pengine/utils.c @@ -1,2417 +1,2418 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include extern xmlNode *get_object_root(const char *object_type, xmlNode * the_root); void print_str_str(gpointer key, gpointer value, gpointer user_data); gboolean ghash_free_str_str(gpointer key, gpointer value, gpointer user_data); void unpack_operation(action_t * action, xmlNode * xml_obj, resource_t * container, pe_working_set_t * data_set); static xmlNode *find_rsc_op_entry_helper(resource_t * rsc, const char *key, gboolean include_disabled); #if ENABLE_VERSIONED_ATTRS pe_rsc_action_details_t * pe_rsc_action_details(pe_action_t *action) { pe_rsc_action_details_t *details; CRM_CHECK(action != NULL, return NULL); if (action->action_details == NULL) { action->action_details = calloc(1, sizeof(pe_rsc_action_details_t)); CRM_CHECK(action->action_details != NULL, return NULL); } details = (pe_rsc_action_details_t *) action->action_details; if (details->versioned_parameters == NULL) { details->versioned_parameters = create_xml_node(NULL, XML_TAG_OP_VER_ATTRS); } if (details->versioned_meta == NULL) { details->versioned_meta = create_xml_node(NULL, XML_TAG_OP_VER_META); } return details; } static void pe_free_rsc_action_details(pe_action_t *action) { pe_rsc_action_details_t *details; if ((action == NULL) || (action->action_details == NULL)) { return; } details = (pe_rsc_action_details_t *) action->action_details; if (details->versioned_parameters) { free_xml(details->versioned_parameters); } if (details->versioned_meta) { free_xml(details->versioned_meta); } action->action_details = NULL; } #endif /*! * \internal * \brief Check whether we can fence a particular node * * \param[in] data_set Working set for cluster * \param[in] node Name of node to check * * \return TRUE if node can be fenced, FALSE otherwise * * \note This function should only be called for cluster nodes and * remote nodes; guest nodes are fenced by stopping their container * resource, so fence execution requirements do not apply to them. */ bool pe_can_fence(pe_working_set_t * data_set, node_t *node) { if(is_not_set(data_set->flags, pe_flag_stonith_enabled)) { return FALSE; /* Turned off */ } else if (is_not_set(data_set->flags, pe_flag_have_stonith_resource)) { return FALSE; /* No devices */ } else if (is_set(data_set->flags, pe_flag_have_quorum)) { return TRUE; } else if (data_set->no_quorum_policy == no_quorum_ignore) { return TRUE; } else if(node == NULL) { return FALSE; } else if(node->details->online) { crm_notice("We can fence %s without quorum because they're in our membership", node->details->uname); return TRUE; } crm_trace("Cannot fence %s", node->details->uname); return FALSE; } node_t * node_copy(const node_t *this_node) { node_t *new_node = NULL; CRM_CHECK(this_node != NULL, return NULL); new_node = calloc(1, sizeof(node_t)); CRM_ASSERT(new_node != NULL); crm_trace("Copying %p (%s) to %p", this_node, this_node->details->uname, new_node); new_node->rsc_discover_mode = this_node->rsc_discover_mode; new_node->weight = this_node->weight; new_node->fixed = this_node->fixed; new_node->details = this_node->details; return new_node; } /* any node in list1 or list2 and not in the other gets a score of -INFINITY */ void node_list_exclude(GHashTable * hash, GListPtr list, gboolean merge_scores) { GHashTable *result = hash; node_t *other_node = NULL; GListPtr gIter = list; GHashTableIter iter; node_t *node = NULL; g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { other_node = pe_find_node_id(list, node->details->id); if (other_node == NULL) { node->weight = -INFINITY; } else if (merge_scores) { node->weight = merge_weights(node->weight, other_node->weight); } } for (; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; other_node = pe_hash_table_lookup(result, node->details->id); if (other_node == NULL) { node_t *new_node = node_copy(node); new_node->weight = -INFINITY; g_hash_table_insert(result, (gpointer) new_node->details->id, new_node); } } } GHashTable * node_hash_from_list(GListPtr list) { GListPtr gIter = list; GHashTable *result = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, free); for (; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; node_t *n = node_copy(node); g_hash_table_insert(result, (gpointer) n->details->id, n); } return result; } GListPtr node_list_dup(GListPtr list1, gboolean reset, gboolean filter) { GListPtr result = NULL; GListPtr gIter = list1; for (; gIter != NULL; gIter = gIter->next) { node_t *new_node = NULL; node_t *this_node = (node_t *) gIter->data; if (filter && this_node->weight < 0) { continue; } new_node = node_copy(this_node); if (reset) { new_node->weight = 0; } if (new_node != NULL) { result = g_list_prepend(result, new_node); } } return result; } gint sort_node_uname(gconstpointer a, gconstpointer b) { const char *name_a = ((const node_t *) a)->details->uname; const char *name_b = ((const node_t *) b)->details->uname; while (*name_a && *name_b) { if (isdigit(*name_a) && isdigit(*name_b)) { // If node names contain a number, sort numerically char *end_a = NULL; char *end_b = NULL; long num_a = strtol(name_a, &end_a, 10); long num_b = strtol(name_b, &end_b, 10); // allow ordering e.g. 007 > 7 size_t len_a = end_a - name_a; size_t len_b = end_b - name_b; if (num_a < num_b) { return -1; } else if (num_a > num_b) { return 1; } else if (len_a < len_b) { return -1; } else if (len_a > len_b) { return 1; } name_a = end_a; name_b = end_b; } else { // Compare non-digits case-insensitively int lower_a = tolower(*name_a); int lower_b = tolower(*name_b); if (lower_a < lower_b) { return -1; } else if (lower_a > lower_b) { return 1; } ++name_a; ++name_b; } } if (!*name_a && *name_b) { return -1; } else if (*name_a && !*name_b) { return 1; } return 0; } void dump_node_scores_worker(int level, const char *file, const char *function, int line, resource_t * rsc, const char *comment, GHashTable * nodes) { GHashTable *hash = nodes; GHashTableIter iter; node_t *node = NULL; if (rsc) { hash = rsc->allowed_nodes; } if (rsc && is_set(rsc->flags, pe_rsc_orphan)) { /* Don't show the allocation scores for orphans */ return; } if (level == 0) { char score[128]; int len = sizeof(score); /* For now we want this in sorted order to keep the regression tests happy */ GListPtr gIter = NULL; GListPtr list = g_hash_table_get_values(hash); list = g_list_sort(list, sort_node_uname); gIter = list; for (; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; /* This function is called a whole lot, use stack allocated score */ score2char_stack(node->weight, score, len); if (rsc) { printf("%s: %s allocation score on %s: %s\n", comment, rsc->id, node->details->uname, score); } else { printf("%s: %s = %s\n", comment, node->details->uname, score); } } g_list_free(list); } else if (hash) { char score[128]; int len = sizeof(score); g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { /* This function is called a whole lot, use stack allocated score */ score2char_stack(node->weight, score, len); if (rsc) { do_crm_log_alias(LOG_TRACE, file, function, line, "%s: %s allocation score on %s: %s", comment, rsc->id, node->details->uname, score); } else { do_crm_log_alias(LOG_TRACE, file, function, line + 1, "%s: %s = %s", comment, node->details->uname, score); } } } if (rsc && rsc->children) { GListPtr gIter = NULL; gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { resource_t *child = (resource_t *) gIter->data; dump_node_scores_worker(level, file, function, line, child, comment, nodes); } } } static void append_dump_text(gpointer key, gpointer value, gpointer user_data) { char **dump_text = user_data; char *new_text = crm_strdup_printf("%s %s=%s", *dump_text, (char *)key, (char *)value); free(*dump_text); *dump_text = new_text; } void dump_node_capacity(int level, const char *comment, node_t * node) { char *dump_text = crm_strdup_printf("%s: %s capacity:", comment, node->details->uname); g_hash_table_foreach(node->details->utilization, append_dump_text, &dump_text); if (level == 0) { fprintf(stdout, "%s\n", dump_text); } else { crm_trace("%s", dump_text); } free(dump_text); } void dump_rsc_utilization(int level, const char *comment, resource_t * rsc, node_t * node) { char *dump_text = crm_strdup_printf("%s: %s utilization on %s:", comment, rsc->id, node->details->uname); g_hash_table_foreach(rsc->utilization, append_dump_text, &dump_text); if (level == 0) { fprintf(stdout, "%s\n", dump_text); } else { crm_trace("%s", dump_text); } free(dump_text); } gint sort_rsc_index(gconstpointer a, gconstpointer b) { const resource_t *resource1 = (const resource_t *)a; const resource_t *resource2 = (const resource_t *)b; if (a == NULL && b == NULL) { return 0; } if (a == NULL) { return 1; } if (b == NULL) { return -1; } if (resource1->sort_index > resource2->sort_index) { return -1; } if (resource1->sort_index < resource2->sort_index) { return 1; } return 0; } gint sort_rsc_priority(gconstpointer a, gconstpointer b) { const resource_t *resource1 = (const resource_t *)a; const resource_t *resource2 = (const resource_t *)b; if (a == NULL && b == NULL) { return 0; } if (a == NULL) { return 1; } if (b == NULL) { return -1; } if (resource1->priority > resource2->priority) { return -1; } if (resource1->priority < resource2->priority) { return 1; } return 0; } action_t * custom_action(resource_t * rsc, char *key, const char *task, node_t * on_node, gboolean optional, gboolean save_action, pe_working_set_t * data_set) { action_t *action = NULL; GListPtr possible_matches = NULL; CRM_CHECK(key != NULL, return NULL); CRM_CHECK(task != NULL, free(key); return NULL); if (save_action && rsc != NULL) { possible_matches = find_actions(rsc->actions, key, on_node); } else if(save_action) { #if 0 action = g_hash_table_lookup(data_set->singletons, key); #else /* More expensive but takes 'node' into account */ possible_matches = find_actions(data_set->actions, key, on_node); #endif } if(data_set->singletons == NULL) { data_set->singletons = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, NULL); } if (possible_matches != NULL) { if (g_list_length(possible_matches) > 1) { pe_warn("Action %s for %s on %s exists %d times", task, rsc ? rsc->id : "", on_node ? on_node->details->uname : "", g_list_length(possible_matches)); } action = g_list_nth_data(possible_matches, 0); pe_rsc_trace(rsc, "Found existing action %d (%s) for %s (%s) on %s", action->id, action->uuid, (rsc? rsc->id : "no resource"), task, (on_node? on_node->details->uname : "no node")); g_list_free(possible_matches); } if (action == NULL) { if (save_action) { pe_rsc_trace(rsc, "Creating %s action %d: %s for %s (%s) on %s", (optional? "optional" : " mandatory"), data_set->action_id, key, (rsc? rsc->id : "no resource"), task, (on_node? on_node->details->uname : "no node")); } action = calloc(1, sizeof(action_t)); if (save_action) { action->id = data_set->action_id++; } else { action->id = 0; } action->rsc = rsc; CRM_ASSERT(task != NULL); action->task = strdup(task); if (on_node) { action->node = node_copy(on_node); } action->uuid = strdup(key); pe_set_action_bit(action, pe_action_runnable); if (optional) { pe_set_action_bit(action, pe_action_optional); } else { pe_clear_action_bit(action, pe_action_optional); } action->extra = crm_str_table_new(); action->meta = crm_str_table_new(); if (save_action) { data_set->actions = g_list_prepend(data_set->actions, action); if(rsc == NULL) { g_hash_table_insert(data_set->singletons, action->uuid, action); } } if (rsc != NULL) { action->op_entry = find_rsc_op_entry_helper(rsc, key, TRUE); unpack_operation(action, action->op_entry, rsc->container, data_set); if (save_action) { rsc->actions = g_list_prepend(rsc->actions, action); } } if (save_action) { pe_rsc_trace(rsc, "Action %d created", action->id); } } if (!optional && is_set(action->flags, pe_action_optional)) { pe_rsc_trace(rsc, "Unset optional on action %d", action->id); pe_clear_action_bit(action, pe_action_optional); } if (rsc != NULL) { enum action_tasks a_task = text2task(action->task); int warn_level = LOG_TRACE; if (save_action) { warn_level = LOG_WARNING; } if (is_set(action->flags, pe_action_have_node_attrs) == FALSE && action->node != NULL && action->op_entry != NULL) { pe_set_action_bit(action, pe_action_have_node_attrs); unpack_instance_attributes(data_set->input, action->op_entry, XML_TAG_ATTR_SETS, action->node->details->attrs, action->extra, NULL, FALSE, data_set->now); } if (is_set(action->flags, pe_action_pseudo)) { /* leave untouched */ } else if (action->node == NULL) { pe_rsc_trace(rsc, "Unset runnable on %s", action->uuid); pe_clear_action_bit(action, pe_action_runnable); } else if (is_not_set(rsc->flags, pe_rsc_managed) && g_hash_table_lookup(action->meta, XML_LRM_ATTR_INTERVAL_MS) == NULL) { crm_debug("Action %s (unmanaged)", action->uuid); pe_rsc_trace(rsc, "Set optional on %s", action->uuid); pe_set_action_bit(action, pe_action_optional); /* action->runnable = FALSE; */ } else if (action->node->details->online == FALSE && (!pe__is_guest_node(action->node) || action->node->details->remote_requires_reset)) { pe_clear_action_bit(action, pe_action_runnable); do_crm_log(warn_level, "Action %s on %s is unrunnable (offline)", action->uuid, action->node->details->uname); if (is_set(action->rsc->flags, pe_rsc_managed) && save_action && a_task == stop_rsc && action->node->details->unclean == FALSE) { pe_fence_node(data_set, action->node, "resource actions are unrunnable"); } } else if (action->node->details->pending) { pe_clear_action_bit(action, pe_action_runnable); do_crm_log(warn_level, "Action %s on %s is unrunnable (pending)", action->uuid, action->node->details->uname); } else if (action->needs == rsc_req_nothing) { pe_rsc_trace(rsc, "Action %s does not require anything", action->uuid); pe_action_set_reason(action, NULL, TRUE); pe_set_action_bit(action, pe_action_runnable); #if 0 /* * No point checking this * - if we don't have quorum we can't stonith anyway */ } else if (action->needs == rsc_req_stonith) { crm_trace("Action %s requires only stonith", action->uuid); action->runnable = TRUE; #endif } else if (is_set(data_set->flags, pe_flag_have_quorum) == FALSE && data_set->no_quorum_policy == no_quorum_stop) { pe_action_set_flag_reason(__FUNCTION__, __LINE__, action, NULL, "no quorum", pe_action_runnable, TRUE); crm_debug("%s\t%s (cancelled : quorum)", action->node->details->uname, action->uuid); } else if (is_set(data_set->flags, pe_flag_have_quorum) == FALSE && data_set->no_quorum_policy == no_quorum_freeze) { pe_rsc_trace(rsc, "Check resource is already active: %s %s %s %s", rsc->id, action->uuid, role2text(rsc->next_role), role2text(rsc->role)); if (rsc->fns->active(rsc, TRUE) == FALSE || rsc->next_role > rsc->role) { pe_action_set_flag_reason(__FUNCTION__, __LINE__, action, NULL, "quorum freeze", pe_action_runnable, TRUE); pe_rsc_debug(rsc, "%s\t%s (cancelled : quorum freeze)", action->node->details->uname, action->uuid); } } else if(is_not_set(action->flags, pe_action_runnable)) { pe_rsc_trace(rsc, "Action %s is runnable", action->uuid); //pe_action_set_reason(action, NULL, TRUE); pe_set_action_bit(action, pe_action_runnable); } if (save_action) { switch (a_task) { case stop_rsc: set_bit(rsc->flags, pe_rsc_stopping); break; case start_rsc: clear_bit(rsc->flags, pe_rsc_starting); if (is_set(action->flags, pe_action_runnable)) { set_bit(rsc->flags, pe_rsc_starting); } break; default: break; } } } free(key); return action; } static const char * unpack_operation_on_fail(action_t * action) { const char *value = g_hash_table_lookup(action->meta, XML_OP_ATTR_ON_FAIL); if (safe_str_eq(action->task, CRMD_ACTION_STOP) && safe_str_eq(value, "standby")) { crm_config_err("on-fail=standby is not allowed for stop actions: %s", action->rsc->id); return NULL; } else if (safe_str_eq(action->task, CRMD_ACTION_DEMOTE) && !value) { /* demote on_fail defaults to master monitor value if present */ xmlNode *operation = NULL; const char *name = NULL; const char *role = NULL; const char *on_fail = NULL; const char *interval_spec = NULL; const char *enabled = NULL; CRM_CHECK(action->rsc != NULL, return NULL); for (operation = __xml_first_child(action->rsc->ops_xml); operation && !value; operation = __xml_next_element(operation)) { if (!crm_str_eq((const char *)operation->name, "op", TRUE)) { continue; } name = crm_element_value(operation, "name"); role = crm_element_value(operation, "role"); on_fail = crm_element_value(operation, XML_OP_ATTR_ON_FAIL); enabled = crm_element_value(operation, "enabled"); interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL); if (!on_fail) { continue; } else if (enabled && !crm_is_true(enabled)) { continue; } else if (safe_str_neq(name, "monitor") || safe_str_neq(role, "Master")) { continue; } else if (crm_parse_interval_spec(interval_spec) == 0) { continue; } value = on_fail; } } return value; } static xmlNode * find_min_interval_mon(resource_t * rsc, gboolean include_disabled) { guint interval_ms = 0; guint min_interval_ms = G_MAXUINT; const char *name = NULL; const char *value = NULL; const char *interval_spec = NULL; xmlNode *op = NULL; xmlNode *operation = NULL; for (operation = __xml_first_child(rsc->ops_xml); operation != NULL; operation = __xml_next_element(operation)) { if (crm_str_eq((const char *)operation->name, "op", TRUE)) { name = crm_element_value(operation, "name"); interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL); value = crm_element_value(operation, "enabled"); if (!include_disabled && value && crm_is_true(value) == FALSE) { continue; } if (safe_str_neq(name, RSC_STATUS)) { continue; } interval_ms = crm_parse_interval_spec(interval_spec); if (interval_ms && (interval_ms < min_interval_ms)) { min_interval_ms = interval_ms; op = operation; } } } return op; } static int unpack_start_delay(const char *value, GHashTable *meta) { int start_delay = 0; if (value != NULL) { start_delay = crm_get_msec(value); if (start_delay < 0) { start_delay = 0; } if (meta) { g_hash_table_replace(meta, strdup(XML_OP_ATTR_START_DELAY), crm_itoa(start_delay)); } } return start_delay; } static int unpack_interval_origin(const char *value, GHashTable *meta, xmlNode *xml_obj, guint interval_ms, crm_time_t *now) { int start_delay = 0; if ((interval_ms > 0) && (value != NULL)) { crm_time_t *origin = crm_time_new(value); if (origin && now) { crm_time_t *delay = NULL; int rc = crm_time_compare(origin, now); long long delay_s = 0; int interval_sec = interval_ms / 1000; crm_trace("Origin: %s, interval: %d", value, interval_sec); /* If 'origin' is in the future, find the most recent "multiple" that occurred in the past */ while(rc > 0) { crm_time_add_seconds(origin, -interval_sec); rc = crm_time_compare(origin, now); } /* Now find the first "multiple" that occurs after 'now' */ while (rc < 0) { crm_time_add_seconds(origin, interval_sec); rc = crm_time_compare(origin, now); } delay = crm_time_calculate_duration(origin, now); crm_time_log(LOG_TRACE, "origin", origin, crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone); crm_time_log(LOG_TRACE, "now", now, crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone); crm_time_log(LOG_TRACE, "delay", delay, crm_time_log_duration); delay_s = crm_time_get_seconds(delay); if (delay_s < 0) { delay_s = 0; } start_delay = delay_s * 1000; if (xml_obj) { crm_info("Calculated a start delay of %llds for %s", delay_s, ID(xml_obj)); } if (meta) { g_hash_table_replace(meta, strdup(XML_OP_ATTR_START_DELAY), crm_itoa(start_delay)); } crm_time_free(origin); crm_time_free(delay); } else if (!origin && xml_obj) { crm_config_err("Operation %s contained an invalid " XML_OP_ATTR_ORIGIN ": %s", ID(xml_obj), value); } } return start_delay; } static int unpack_timeout(const char *value) { int timeout = crm_get_msec(value); if (timeout < 0) { timeout = crm_get_msec(CRM_DEFAULT_OP_TIMEOUT_S); } return timeout; } int pe_get_configured_timeout(resource_t *rsc, const char *action, pe_working_set_t *data_set) { xmlNode *child = NULL; const char *timeout = NULL; int timeout_ms = 0; for (child = first_named_child(rsc->ops_xml, XML_ATTR_OP); child != NULL; child = crm_next_same_xml(child)) { if (safe_str_eq(action, crm_element_value(child, XML_NVPAIR_ATTR_NAME))) { timeout = crm_element_value(child, XML_ATTR_TIMEOUT); break; } } if (timeout == NULL && data_set->op_defaults) { GHashTable *action_meta = crm_str_table_new(); unpack_instance_attributes(data_set->input, data_set->op_defaults, XML_TAG_META_SETS, NULL, action_meta, NULL, FALSE, data_set->now); timeout = g_hash_table_lookup(action_meta, XML_ATTR_TIMEOUT); } // @TODO check meta-attributes (including versioned meta-attributes) // @TODO maybe use min-interval monitor timeout as default for monitors timeout_ms = crm_get_msec(timeout); if (timeout_ms < 0) { timeout_ms = crm_get_msec(CRM_DEFAULT_OP_TIMEOUT_S); } return timeout_ms; } #if ENABLE_VERSIONED_ATTRS static void unpack_versioned_meta(xmlNode *versioned_meta, xmlNode *xml_obj, guint interval_ms, crm_time_t *now) { xmlNode *attrs = NULL; xmlNode *attr = NULL; for (attrs = __xml_first_child(versioned_meta); attrs != NULL; attrs = __xml_next_element(attrs)) { for (attr = __xml_first_child(attrs); attr != NULL; attr = __xml_next_element(attr)) { const char *name = crm_element_value(attr, XML_NVPAIR_ATTR_NAME); const char *value = crm_element_value(attr, XML_NVPAIR_ATTR_VALUE); if (safe_str_eq(name, XML_OP_ATTR_START_DELAY)) { int start_delay = unpack_start_delay(value, NULL); crm_xml_add_int(attr, XML_NVPAIR_ATTR_VALUE, start_delay); } else if (safe_str_eq(name, XML_OP_ATTR_ORIGIN)) { int start_delay = unpack_interval_origin(value, NULL, xml_obj, interval_ms, now); crm_xml_add(attr, XML_NVPAIR_ATTR_NAME, XML_OP_ATTR_START_DELAY); crm_xml_add_int(attr, XML_NVPAIR_ATTR_VALUE, start_delay); } else if (safe_str_eq(name, XML_ATTR_TIMEOUT)) { int timeout = unpack_timeout(value); crm_xml_add_int(attr, XML_NVPAIR_ATTR_VALUE, timeout); } } } } #endif /*! * \brief Unpack operation XML into an action structure * * Unpack an operation's meta-attributes (normalizing the interval, timeout, * and start delay values as integer milliseconds), requirements, and * failure policy. * * \param[in,out] action Action to unpack into * \param[in] xml_obj Operation XML (or NULL if all defaults) * \param[in] container Resource that contains affected resource, if any * \param[in] data_set Cluster state */ void unpack_operation(action_t * action, xmlNode * xml_obj, resource_t * container, pe_working_set_t * data_set) { guint interval_ms = 0; int timeout = 0; char *value_ms = NULL; const char *value = NULL; const char *field = NULL; char *default_timeout = NULL; #if ENABLE_VERSIONED_ATTRS pe_rsc_action_details_t *rsc_details = NULL; #endif CRM_CHECK(action && action->rsc, return); // Cluster-wide unpack_instance_attributes(data_set->input, data_set->op_defaults, XML_TAG_META_SETS, NULL, action->meta, NULL, FALSE, data_set->now); // Probe timeouts default differently, so handle timeout default later default_timeout = g_hash_table_lookup(action->meta, XML_ATTR_TIMEOUT); if (default_timeout) { default_timeout = strdup(default_timeout); g_hash_table_remove(action->meta, XML_ATTR_TIMEOUT); } if (xml_obj) { xmlAttrPtr xIter = NULL; // take precedence over defaults unpack_instance_attributes(data_set->input, xml_obj, XML_TAG_META_SETS, NULL, action->meta, NULL, TRUE, data_set->now); #if ENABLE_VERSIONED_ATTRS rsc_details = pe_rsc_action_details(action); pe_unpack_versioned_attributes(data_set->input, xml_obj, XML_TAG_ATTR_SETS, NULL, rsc_details->versioned_parameters, data_set->now); pe_unpack_versioned_attributes(data_set->input, xml_obj, XML_TAG_META_SETS, NULL, rsc_details->versioned_meta, data_set->now); #endif /* Anything set as an XML property has highest precedence. * This ensures we use the name and interval from the tag. */ for (xIter = xml_obj->properties; xIter; xIter = xIter->next) { const char *prop_name = (const char *)xIter->name; const char *prop_value = crm_element_value(xml_obj, prop_name); g_hash_table_replace(action->meta, strdup(prop_name), strdup(prop_value)); } } g_hash_table_remove(action->meta, "id"); // Normalize interval to milliseconds field = XML_LRM_ATTR_INTERVAL; value = g_hash_table_lookup(action->meta, field); if (value != NULL) { interval_ms = crm_parse_interval_spec(value); } else if ((xml_obj == NULL) && !strcmp(action->task, RSC_STATUS)) { /* An orphaned recurring monitor will not have any XML. However, we * want the interval to be set, so the action can be properly detected * as a recurring monitor. Parse it from the key in this case. */ parse_op_key(action->uuid, NULL, NULL, &interval_ms); } if (interval_ms > 0) { value_ms = crm_strdup_printf("%u", interval_ms); g_hash_table_replace(action->meta, strdup(field), value_ms); } else if (value) { g_hash_table_remove(action->meta, field); } // Handle timeout default, now that we know the interval if (g_hash_table_lookup(action->meta, XML_ATTR_TIMEOUT)) { free(default_timeout); } else { // Probe timeouts default to minimum-interval monitor's if (safe_str_eq(action->task, RSC_STATUS) && (interval_ms == 0)) { xmlNode *min_interval_mon = find_min_interval_mon(action->rsc, FALSE); if (min_interval_mon) { value = crm_element_value(min_interval_mon, XML_ATTR_TIMEOUT); if (value) { crm_trace("\t%s defaults to minimum-interval monitor's timeout '%s'", action->uuid, value); free(default_timeout); default_timeout = strdup(value); } } } if (default_timeout) { g_hash_table_insert(action->meta, strdup(XML_ATTR_TIMEOUT), default_timeout); } } if (safe_str_neq(action->task, RSC_START) && safe_str_neq(action->task, RSC_PROMOTE)) { action->needs = rsc_req_nothing; value = "nothing (not start/promote)"; } else if (is_set(action->rsc->flags, pe_rsc_needs_fencing)) { action->needs = rsc_req_stonith; value = "fencing (resource)"; } else if (is_set(action->rsc->flags, pe_rsc_needs_quorum)) { action->needs = rsc_req_quorum; value = "quorum (resource)"; } else { action->needs = rsc_req_nothing; value = "nothing (resource)"; } pe_rsc_trace(action->rsc, "\tAction %s requires: %s", action->uuid, value); value = unpack_operation_on_fail(action); if (value == NULL) { } else if (safe_str_eq(value, "block")) { action->on_fail = action_fail_block; g_hash_table_insert(action->meta, strdup(XML_OP_ATTR_ON_FAIL), strdup("block")); value = "block"; // The above could destroy the original string } else if (safe_str_eq(value, "fence")) { action->on_fail = action_fail_fence; value = "node fencing"; if (is_set(data_set->flags, pe_flag_stonith_enabled) == FALSE) { crm_config_err("Specifying on_fail=fence and" " stonith-enabled=false makes no sense"); action->on_fail = action_fail_stop; action->fail_role = RSC_ROLE_STOPPED; value = "stop resource"; } } else if (safe_str_eq(value, "standby")) { action->on_fail = action_fail_standby; value = "node standby"; } else if (safe_str_eq(value, "ignore") || safe_str_eq(value, "nothing")) { action->on_fail = action_fail_ignore; value = "ignore"; } else if (safe_str_eq(value, "migrate")) { action->on_fail = action_fail_migrate; value = "force migration"; } else if (safe_str_eq(value, "stop")) { action->on_fail = action_fail_stop; action->fail_role = RSC_ROLE_STOPPED; value = "stop resource"; } else if (safe_str_eq(value, "restart")) { action->on_fail = action_fail_recover; value = "restart (and possibly migrate)"; } else if (safe_str_eq(value, "restart-container")) { if (container) { action->on_fail = action_fail_restart_container; value = "restart container (and possibly migrate)"; } else { value = NULL; } } else { pe_err("Resource %s: Unknown failure type (%s)", action->rsc->id, value); value = NULL; } /* defaults */ if (value == NULL && container) { action->on_fail = action_fail_restart_container; value = "restart container (and possibly migrate) (default)"; /* For remote nodes, ensure that any failure that results in dropping an * active connection to the node results in fencing of the node. * * There are only two action failures that don't result in fencing. * 1. probes - probe failures are expected. * 2. start - a start failure indicates that an active connection does not already * exist. The user can set op on-fail=fence if they really want to fence start * failures. */ } else if (((value == NULL) || !is_set(action->rsc->flags, pe_rsc_managed)) && (pe__resource_is_remote_conn(action->rsc, data_set) && !(safe_str_eq(action->task, CRMD_ACTION_STATUS) && (interval_ms == 0)) && (safe_str_neq(action->task, CRMD_ACTION_START)))) { if (!is_set(action->rsc->flags, pe_rsc_managed)) { action->on_fail = action_fail_stop; action->fail_role = RSC_ROLE_STOPPED; value = "stop unmanaged remote node (enforcing default)"; } else { if (is_set(data_set->flags, pe_flag_stonith_enabled)) { value = "fence remote node (default)"; } else { value = "recover remote node connection (default)"; } if (action->rsc->remote_reconnect_ms) { action->fail_role = RSC_ROLE_STOPPED; } action->on_fail = action_fail_reset_remote; } } else if (value == NULL && safe_str_eq(action->task, CRMD_ACTION_STOP)) { if (is_set(data_set->flags, pe_flag_stonith_enabled)) { action->on_fail = action_fail_fence; value = "resource fence (default)"; } else { action->on_fail = action_fail_block; value = "resource block (default)"; } } else if (value == NULL) { action->on_fail = action_fail_recover; value = "restart (and possibly migrate) (default)"; } pe_rsc_trace(action->rsc, "\t%s failure handling: %s", action->task, value); value = NULL; if (xml_obj != NULL) { value = g_hash_table_lookup(action->meta, "role_after_failure"); if (value) { pe_warn_once(pe_wo_role_after, "Support for role_after_failure is deprecated and will be removed in a future release"); } } if (value != NULL && action->fail_role == RSC_ROLE_UNKNOWN) { action->fail_role = text2role(value); } /* defaults */ if (action->fail_role == RSC_ROLE_UNKNOWN) { if (safe_str_eq(action->task, CRMD_ACTION_PROMOTE)) { action->fail_role = RSC_ROLE_SLAVE; } else { action->fail_role = RSC_ROLE_STARTED; } } pe_rsc_trace(action->rsc, "\t%s failure results in: %s", action->task, role2text(action->fail_role)); value = g_hash_table_lookup(action->meta, XML_OP_ATTR_START_DELAY); if (value) { unpack_start_delay(value, action->meta); } else { value = g_hash_table_lookup(action->meta, XML_OP_ATTR_ORIGIN); unpack_interval_origin(value, action->meta, xml_obj, interval_ms, data_set->now); } value = g_hash_table_lookup(action->meta, XML_ATTR_TIMEOUT); timeout = unpack_timeout(value); g_hash_table_replace(action->meta, strdup(XML_ATTR_TIMEOUT), crm_itoa(timeout)); #if ENABLE_VERSIONED_ATTRS unpack_versioned_meta(rsc_details->versioned_meta, xml_obj, interval_ms, data_set->now); #endif } static xmlNode * find_rsc_op_entry_helper(resource_t * rsc, const char *key, gboolean include_disabled) { guint interval_ms = 0; gboolean do_retry = TRUE; char *local_key = NULL; const char *name = NULL; const char *value = NULL; const char *interval_spec = NULL; char *match_key = NULL; xmlNode *op = NULL; xmlNode *operation = NULL; retry: for (operation = __xml_first_child(rsc->ops_xml); operation != NULL; operation = __xml_next_element(operation)) { if (crm_str_eq((const char *)operation->name, "op", TRUE)) { name = crm_element_value(operation, "name"); interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL); value = crm_element_value(operation, "enabled"); if (!include_disabled && value && crm_is_true(value) == FALSE) { continue; } interval_ms = crm_parse_interval_spec(interval_spec); match_key = generate_op_key(rsc->id, name, interval_ms); if (safe_str_eq(key, match_key)) { op = operation; } free(match_key); if (rsc->clone_name) { match_key = generate_op_key(rsc->clone_name, name, interval_ms); if (safe_str_eq(key, match_key)) { op = operation; } free(match_key); } if (op != NULL) { free(local_key); return op; } } } free(local_key); if (do_retry == FALSE) { return NULL; } do_retry = FALSE; if (strstr(key, CRMD_ACTION_MIGRATE) || strstr(key, CRMD_ACTION_MIGRATED)) { local_key = generate_op_key(rsc->id, "migrate", 0); key = local_key; goto retry; } else if (strstr(key, "_notify_")) { local_key = generate_op_key(rsc->id, "notify", 0); key = local_key; goto retry; } return NULL; } xmlNode * find_rsc_op_entry(resource_t * rsc, const char *key) { return find_rsc_op_entry_helper(rsc, key, FALSE); } void print_node(const char *pre_text, node_t * node, gboolean details) { if (node == NULL) { crm_trace("%s%s: ", pre_text == NULL ? "" : pre_text, pre_text == NULL ? "" : ": "); return; } CRM_ASSERT(node->details); crm_trace("%s%s%sNode %s: (weight=%d, fixed=%s)", pre_text == NULL ? "" : pre_text, pre_text == NULL ? "" : ": ", node->details->online ? "" : "Unavailable/Unclean ", node->details->uname, node->weight, node->fixed ? "True" : "False"); if (details) { char *pe_mutable = strdup("\t\t"); GListPtr gIter = node->details->running_rsc; crm_trace("\t\t===Node Attributes"); g_hash_table_foreach(node->details->attrs, print_str_str, pe_mutable); free(pe_mutable); crm_trace("\t\t=== Resources"); for (; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; print_resource(LOG_TRACE, "\t\t", rsc, FALSE); } } } /* * Used by the HashTable for-loop */ void print_str_str(gpointer key, gpointer value, gpointer user_data) { crm_trace("%s%s %s ==> %s", user_data == NULL ? "" : (char *)user_data, user_data == NULL ? "" : ": ", (char *)key, (char *)value); } void print_resource(int log_level, const char *pre_text, resource_t * rsc, gboolean details) { long options = pe_print_log | pe_print_pending; if (rsc == NULL) { do_crm_log(log_level - 1, "%s%s: ", pre_text == NULL ? "" : pre_text, pre_text == NULL ? "" : ": "); return; } if (details) { options |= pe_print_details; } rsc->fns->print(rsc, pre_text, options, &log_level); } void pe_free_action(action_t * action) { if (action == NULL) { return; } g_list_free_full(action->actions_before, free); /* action_wrapper_t* */ g_list_free_full(action->actions_after, free); /* action_wrapper_t* */ if (action->extra) { g_hash_table_destroy(action->extra); } if (action->meta) { g_hash_table_destroy(action->meta); } #if ENABLE_VERSIONED_ATTRS if (action->rsc) { pe_free_rsc_action_details(action); } #endif free(action->cancel_task); free(action->reason); free(action->task); free(action->uuid); free(action->node); free(action); } GListPtr find_recurring_actions(GListPtr input, node_t * not_on_node) { const char *value = NULL; GListPtr result = NULL; GListPtr gIter = input; CRM_CHECK(input != NULL, return NULL); for (; gIter != NULL; gIter = gIter->next) { action_t *action = (action_t *) gIter->data; value = g_hash_table_lookup(action->meta, XML_LRM_ATTR_INTERVAL_MS); if (value == NULL) { /* skip */ } else if (safe_str_eq(value, "0")) { /* skip */ } else if (safe_str_eq(CRMD_ACTION_CANCEL, action->task)) { /* skip */ } else if (not_on_node == NULL) { crm_trace("(null) Found: %s", action->uuid); result = g_list_prepend(result, action); } else if (action->node == NULL) { /* skip */ } else if (action->node->details != not_on_node->details) { crm_trace("Found: %s", action->uuid); result = g_list_prepend(result, action); } } return result; } enum action_tasks get_complex_task(resource_t * rsc, const char *name, gboolean allow_non_atomic) { enum action_tasks task = text2task(name); if (rsc == NULL) { return task; } else if (allow_non_atomic == FALSE || rsc->variant == pe_native) { switch (task) { case stopped_rsc: case started_rsc: case action_demoted: case action_promoted: crm_trace("Folding %s back into its atomic counterpart for %s", name, rsc->id); return task - 1; break; default: break; } } return task; } action_t * find_first_action(GListPtr input, const char *uuid, const char *task, node_t * on_node) { GListPtr gIter = NULL; CRM_CHECK(uuid || task, return NULL); for (gIter = input; gIter != NULL; gIter = gIter->next) { action_t *action = (action_t *) gIter->data; if (uuid != NULL && safe_str_neq(uuid, action->uuid)) { continue; } else if (task != NULL && safe_str_neq(task, action->task)) { continue; } else if (on_node == NULL) { return action; } else if (action->node == NULL) { continue; } else if (on_node->details == action->node->details) { return action; } } return NULL; } GListPtr find_actions(GListPtr input, const char *key, const node_t *on_node) { GListPtr gIter = input; GListPtr result = NULL; CRM_CHECK(key != NULL, return NULL); for (; gIter != NULL; gIter = gIter->next) { action_t *action = (action_t *) gIter->data; if (safe_str_neq(key, action->uuid)) { crm_trace("%s does not match action %s", key, action->uuid); continue; } else if (on_node == NULL) { crm_trace("Action %s matches (ignoring node)", key); result = g_list_prepend(result, action); } else if (action->node == NULL) { crm_trace("Action %s matches (unallocated, assigning to %s)", key, on_node->details->uname); action->node = node_copy(on_node); result = g_list_prepend(result, action); } else if (on_node->details == action->node->details) { crm_trace("Action %s on %s matches", key, on_node->details->uname); result = g_list_prepend(result, action); } else { crm_trace("Action %s on node %s does not match requested node %s", key, action->node->details->uname, on_node->details->uname); } } return result; } GList * find_actions_exact(GList *input, const char *key, const pe_node_t *on_node) { GList *result = NULL; CRM_CHECK(key != NULL, return NULL); if (on_node == NULL) { crm_trace("Not searching for action %s because node not specified", key); return NULL; } for (GList *gIter = input; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; if (action->node == NULL) { crm_trace("Skipping comparison of %s vs action %s without node", key, action->uuid); } else if (safe_str_neq(key, action->uuid)) { crm_trace("Desired action %s doesn't match %s", key, action->uuid); } else if (safe_str_neq(on_node->details->id, action->node->details->id)) { crm_trace("Action %s desired node ID %s doesn't match %s", key, on_node->details->id, action->node->details->id); } else { crm_trace("Action %s matches", key); result = g_list_prepend(result, action); } } return result; } /*! * \brief Find all actions of given type for a resource * * \param[in] rsc Resource to search * \param[in] node Find only actions scheduled on this node * \param[in] task Action name to search for * \param[in] require_node If TRUE, NULL node or action node will not match * * \return List of actions found (or NULL if none) * \note If node is not NULL and require_node is FALSE, matching actions * without a node will be assigned to node. */ GList * pe__resource_actions(const pe_resource_t *rsc, const pe_node_t *node, const char *task, bool require_node) { GList *result = NULL; char *key = generate_op_key(rsc->id, task, 0); if (require_node) { result = find_actions_exact(rsc->actions, key, node); } else { result = find_actions(rsc->actions, key, node); } free(key); return result; } static void resource_node_score(resource_t * rsc, node_t * node, int score, const char *tag) { node_t *match = NULL; if ((rsc->exclusive_discover || (node->rsc_discover_mode == pe_discover_never)) && safe_str_eq(tag, "symmetric_default")) { /* This string comparision may be fragile, but exclusive resources and * exclusive nodes should not have the symmetric_default constraint * applied to them. */ return; } else if (rsc->children) { GListPtr gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { resource_t *child_rsc = (resource_t *) gIter->data; resource_node_score(child_rsc, node, score, tag); } } pe_rsc_trace(rsc, "Setting %s for %s on %s: %d", tag, rsc->id, node->details->uname, score); match = pe_hash_table_lookup(rsc->allowed_nodes, node->details->id); if (match == NULL) { match = node_copy(node); g_hash_table_insert(rsc->allowed_nodes, (gpointer) match->details->id, match); } match->weight = merge_weights(match->weight, score); } void resource_location(resource_t * rsc, node_t * node, int score, const char *tag, pe_working_set_t * data_set) { if (node != NULL) { resource_node_score(rsc, node, score, tag); } else if (data_set != NULL) { GListPtr gIter = data_set->nodes; for (; gIter != NULL; gIter = gIter->next) { node_t *node_iter = (node_t *) gIter->data; resource_node_score(rsc, node_iter, score, tag); } } else { GHashTableIter iter; node_t *node_iter = NULL; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node_iter)) { resource_node_score(rsc, node_iter, score, tag); } } if (node == NULL && score == -INFINITY) { if (rsc->allocated_to) { crm_info("Deallocating %s from %s", rsc->id, rsc->allocated_to->details->uname); free(rsc->allocated_to); rsc->allocated_to = NULL; } } } #define sort_return(an_int, why) do { \ free(a_uuid); \ free(b_uuid); \ crm_trace("%s (%d) %c %s (%d) : %s", \ a_xml_id, a_call_id, an_int>0?'>':an_int<0?'<':'=', \ b_xml_id, b_call_id, why); \ return an_int; \ } while(0) gint sort_op_by_callid(gconstpointer a, gconstpointer b) { int a_call_id = -1; int b_call_id = -1; char *a_uuid = NULL; char *b_uuid = NULL; const xmlNode *xml_a = a; const xmlNode *xml_b = b; const char *a_xml_id = crm_element_value(xml_a, XML_ATTR_ID); const char *b_xml_id = crm_element_value(xml_b, XML_ATTR_ID); if (safe_str_eq(a_xml_id, b_xml_id)) { /* We have duplicate lrm_rsc_op entries in the status * section which is unliklely to be a good thing * - we can handle it easily enough, but we need to get * to the bottom of why it's happening. */ pe_err("Duplicate lrm_rsc_op entries named %s", a_xml_id); sort_return(0, "duplicate"); } crm_element_value_int(xml_a, XML_LRM_ATTR_CALLID, &a_call_id); crm_element_value_int(xml_b, XML_LRM_ATTR_CALLID, &b_call_id); if (a_call_id == -1 && b_call_id == -1) { /* both are pending ops so it doesn't matter since * stops are never pending */ sort_return(0, "pending"); } else if (a_call_id >= 0 && a_call_id < b_call_id) { sort_return(-1, "call id"); } else if (b_call_id >= 0 && a_call_id > b_call_id) { sort_return(1, "call id"); } else if (b_call_id >= 0 && a_call_id == b_call_id) { /* * The op and last_failed_op are the same * Order on last-rc-change */ int last_a = -1; int last_b = -1; crm_element_value_int(xml_a, XML_RSC_OP_LAST_CHANGE, &last_a); crm_element_value_int(xml_b, XML_RSC_OP_LAST_CHANGE, &last_b); crm_trace("rc-change: %d vs %d", last_a, last_b); if (last_a >= 0 && last_a < last_b) { sort_return(-1, "rc-change"); } else if (last_b >= 0 && last_a > last_b) { sort_return(1, "rc-change"); } sort_return(0, "rc-change"); } else { /* One of the inputs is a pending operation * Attempt to use XML_ATTR_TRANSITION_MAGIC to determine its age relative to the other */ int a_id = -1; int b_id = -1; - int dummy = -1; const char *a_magic = crm_element_value(xml_a, XML_ATTR_TRANSITION_MAGIC); const char *b_magic = crm_element_value(xml_b, XML_ATTR_TRANSITION_MAGIC); CRM_CHECK(a_magic != NULL && b_magic != NULL, sort_return(0, "No magic")); - if(!decode_transition_magic(a_magic, &a_uuid, &a_id, &dummy, &dummy, &dummy, &dummy)) { + if (!decode_transition_magic(a_magic, &a_uuid, &a_id, NULL, NULL, NULL, + NULL)) { sort_return(0, "bad magic a"); } - if(!decode_transition_magic(b_magic, &b_uuid, &b_id, &dummy, &dummy, &dummy, &dummy)) { + if (!decode_transition_magic(b_magic, &b_uuid, &b_id, NULL, NULL, NULL, + NULL)) { sort_return(0, "bad magic b"); } /* try to determine the relative age of the operation... * some pending operations (e.g. a start) may have been superseded * by a subsequent stop * * [a|b]_id == -1 means it's a shutdown operation and _always_ comes last */ if (safe_str_neq(a_uuid, b_uuid) || a_id == b_id) { /* * some of the logic in here may be redundant... * * if the UUID from the TE doesn't match then one better * be a pending operation. * pending operations don't survive between elections and joins * because we query the LRM directly */ if (b_call_id == -1) { sort_return(-1, "transition + call"); } else if (a_call_id == -1) { sort_return(1, "transition + call"); } } else if ((a_id >= 0 && a_id < b_id) || b_id == -1) { sort_return(-1, "transition"); } else if ((b_id >= 0 && a_id > b_id) || a_id == -1) { sort_return(1, "transition"); } } /* we should never end up here */ CRM_CHECK(FALSE, sort_return(0, "default")); } time_t get_effective_time(pe_working_set_t * data_set) { if(data_set) { if (data_set->now == NULL) { crm_trace("Recording a new 'now'"); data_set->now = crm_time_new(NULL); } return crm_time_get_seconds_since_epoch(data_set->now); } crm_trace("Defaulting to 'now'"); return time(NULL); } gboolean get_target_role(resource_t * rsc, enum rsc_role_e * role) { enum rsc_role_e local_role = RSC_ROLE_UNKNOWN; const char *value = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE); CRM_CHECK(role != NULL, return FALSE); if (value == NULL || safe_str_eq("started", value) || safe_str_eq("default", value)) { return FALSE; } local_role = text2role(value); if (local_role == RSC_ROLE_UNKNOWN) { crm_config_err("%s: Unknown value for %s: %s", rsc->id, XML_RSC_ATTR_TARGET_ROLE, value); return FALSE; } else if (local_role > RSC_ROLE_STARTED) { if (is_set(uber_parent(rsc)->flags, pe_rsc_promotable)) { if (local_role > RSC_ROLE_SLAVE) { /* This is what we'd do anyway, just leave the default to avoid messing up the placement algorithm */ return FALSE; } } else { crm_config_err("%s is not part of a promotable clone resource, a %s of '%s' makes no sense", rsc->id, XML_RSC_ATTR_TARGET_ROLE, value); return FALSE; } } *role = local_role; return TRUE; } gboolean order_actions(action_t * lh_action, action_t * rh_action, enum pe_ordering order) { GListPtr gIter = NULL; action_wrapper_t *wrapper = NULL; GListPtr list = NULL; if (order == pe_order_none) { return FALSE; } if (lh_action == NULL || rh_action == NULL) { return FALSE; } crm_trace("Ordering Action %s before %s", lh_action->uuid, rh_action->uuid); /* Ensure we never create a dependency on ourselves... it's happened */ CRM_ASSERT(lh_action != rh_action); /* Filter dups, otherwise update_action_states() has too much work to do */ gIter = lh_action->actions_after; for (; gIter != NULL; gIter = gIter->next) { action_wrapper_t *after = (action_wrapper_t *) gIter->data; if (after->action == rh_action && (after->type & order)) { return FALSE; } } wrapper = calloc(1, sizeof(action_wrapper_t)); wrapper->action = rh_action; wrapper->type = order; list = lh_action->actions_after; list = g_list_prepend(list, wrapper); lh_action->actions_after = list; wrapper = NULL; /* order |= pe_order_implies_then; */ /* order ^= pe_order_implies_then; */ wrapper = calloc(1, sizeof(action_wrapper_t)); wrapper->action = lh_action; wrapper->type = order; list = rh_action->actions_before; list = g_list_prepend(list, wrapper); rh_action->actions_before = list; return TRUE; } action_t * get_pseudo_op(const char *name, pe_working_set_t * data_set) { action_t *op = NULL; if(data_set->singletons) { op = g_hash_table_lookup(data_set->singletons, name); } if (op == NULL) { op = custom_action(NULL, strdup(name), name, NULL, TRUE, TRUE, data_set); set_bit(op->flags, pe_action_pseudo); set_bit(op->flags, pe_action_runnable); } return op; } void destroy_ticket(gpointer data) { ticket_t *ticket = data; if (ticket->state) { g_hash_table_destroy(ticket->state); } free(ticket->id); free(ticket); } ticket_t * ticket_new(const char *ticket_id, pe_working_set_t * data_set) { ticket_t *ticket = NULL; if (ticket_id == NULL || strlen(ticket_id) == 0) { return NULL; } if (data_set->tickets == NULL) { data_set->tickets = g_hash_table_new_full(crm_str_hash, g_str_equal, free, destroy_ticket); } ticket = g_hash_table_lookup(data_set->tickets, ticket_id); if (ticket == NULL) { ticket = calloc(1, sizeof(ticket_t)); if (ticket == NULL) { crm_err("Cannot allocate ticket '%s'", ticket_id); return NULL; } crm_trace("Creaing ticket entry for %s", ticket_id); ticket->id = strdup(ticket_id); ticket->granted = FALSE; ticket->last_granted = -1; ticket->standby = FALSE; ticket->state = crm_str_table_new(); g_hash_table_insert(data_set->tickets, strdup(ticket->id), ticket); } return ticket; } static void filter_parameters(xmlNode * param_set, const char *param_string, bool need_present) { if (param_set && param_string) { xmlAttrPtr xIter = param_set->properties; while (xIter) { const char *prop_name = (const char *)xIter->name; char *name = crm_strdup_printf(" %s ", prop_name); char *match = strstr(param_string, name); free(name); // Do now, because current entry might get removed below xIter = xIter->next; if (need_present && match == NULL) { crm_trace("%s not found in %s", prop_name, param_string); xml_remove_prop(param_set, prop_name); } else if (need_present == FALSE && match) { crm_trace("%s found in %s", prop_name, param_string); xml_remove_prop(param_set, prop_name); } } } } #if ENABLE_VERSIONED_ATTRS static void append_versioned_params(xmlNode *versioned_params, const char *ra_version, xmlNode *params) { GHashTable *hash = pe_unpack_versioned_parameters(versioned_params, ra_version); char *key = NULL; char *value = NULL; GHashTableIter iter; g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) { crm_xml_add(params, key, value); } g_hash_table_destroy(hash); } #endif static op_digest_cache_t * rsc_action_digest(resource_t * rsc, const char *task, const char *key, node_t * node, xmlNode * xml_op, pe_working_set_t * data_set) { op_digest_cache_t *data = NULL; data = g_hash_table_lookup(node->details->digest_cache, key); if (data == NULL) { GHashTable *local_rsc_params = crm_str_table_new(); action_t *action = custom_action(rsc, strdup(key), task, node, TRUE, FALSE, data_set); #if ENABLE_VERSIONED_ATTRS xmlNode *local_versioned_params = create_xml_node(NULL, XML_TAG_RSC_VER_ATTRS); const char *ra_version = NULL; #endif const char *op_version; const char *restart_list = NULL; const char *secure_list = " passwd password "; data = calloc(1, sizeof(op_digest_cache_t)); CRM_ASSERT(data != NULL); get_rsc_attributes(local_rsc_params, rsc, node, data_set); #if ENABLE_VERSIONED_ATTRS pe_get_versioned_attributes(local_versioned_params, rsc, node, data_set); #endif data->params_all = create_xml_node(NULL, XML_TAG_PARAMS); // REMOTE_CONTAINER_HACK: Allow remote nodes that start containers with pacemaker remote inside if (pe__add_bundle_remote_name(rsc, data->params_all, XML_RSC_ATTR_REMOTE_RA_ADDR)) { crm_trace("Set address for bundle connection %s (on %s)", rsc->id, node->details->uname); } g_hash_table_foreach(local_rsc_params, hash2field, data->params_all); g_hash_table_foreach(action->extra, hash2field, data->params_all); g_hash_table_foreach(rsc->parameters, hash2field, data->params_all); g_hash_table_foreach(action->meta, hash2metafield, data->params_all); if(xml_op) { secure_list = crm_element_value(xml_op, XML_LRM_ATTR_OP_SECURE); restart_list = crm_element_value(xml_op, XML_LRM_ATTR_OP_RESTART); op_version = crm_element_value(xml_op, XML_ATTR_CRM_VERSION); #if ENABLE_VERSIONED_ATTRS ra_version = crm_element_value(xml_op, XML_ATTR_RA_VERSION); #endif } else { op_version = CRM_FEATURE_SET; } #if ENABLE_VERSIONED_ATTRS append_versioned_params(local_versioned_params, ra_version, data->params_all); append_versioned_params(rsc->versioned_parameters, ra_version, data->params_all); { pe_rsc_action_details_t *details = pe_rsc_action_details(action); append_versioned_params(details->versioned_parameters, ra_version, data->params_all); } #endif filter_action_parameters(data->params_all, op_version); g_hash_table_destroy(local_rsc_params); pe_free_action(action); data->digest_all_calc = calculate_operation_digest(data->params_all, op_version); if (is_set(data_set->flags, pe_flag_sanitized)) { data->params_secure = copy_xml(data->params_all); if(secure_list) { filter_parameters(data->params_secure, secure_list, FALSE); } data->digest_secure_calc = calculate_operation_digest(data->params_secure, op_version); } if(xml_op && crm_element_value(xml_op, XML_LRM_ATTR_RESTART_DIGEST) != NULL) { data->params_restart = copy_xml(data->params_all); if (restart_list) { filter_parameters(data->params_restart, restart_list, TRUE); } data->digest_restart_calc = calculate_operation_digest(data->params_restart, op_version); } g_hash_table_insert(node->details->digest_cache, strdup(key), data); } return data; } op_digest_cache_t * rsc_action_digest_cmp(resource_t * rsc, xmlNode * xml_op, node_t * node, pe_working_set_t * data_set) { op_digest_cache_t *data = NULL; char *key = NULL; guint interval_ms = 0; const char *op_version; const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); const char *interval_ms_s = crm_element_value(xml_op, XML_LRM_ATTR_INTERVAL_MS); const char *digest_all; const char *digest_restart; CRM_ASSERT(node != NULL); op_version = crm_element_value(xml_op, XML_ATTR_CRM_VERSION); digest_all = crm_element_value(xml_op, XML_LRM_ATTR_OP_DIGEST); digest_restart = crm_element_value(xml_op, XML_LRM_ATTR_RESTART_DIGEST); interval_ms = crm_parse_ms(interval_ms_s); key = generate_op_key(rsc->id, task, interval_ms); data = rsc_action_digest(rsc, task, key, node, xml_op, data_set); data->rc = RSC_DIGEST_MATCH; if (digest_restart && data->digest_restart_calc && strcmp(data->digest_restart_calc, digest_restart) != 0) { pe_rsc_info(rsc, "Parameters to %s on %s changed: was %s vs. now %s (restart:%s) %s", key, node->details->uname, crm_str(digest_restart), data->digest_restart_calc, op_version, crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC)); data->rc = RSC_DIGEST_RESTART; } else if (digest_all == NULL) { /* it is unknown what the previous op digest was */ data->rc = RSC_DIGEST_UNKNOWN; } else if (strcmp(digest_all, data->digest_all_calc) != 0) { pe_rsc_info(rsc, "Parameters to %s on %s changed: was %s vs. now %s (%s:%s) %s", key, node->details->uname, crm_str(digest_all), data->digest_all_calc, (interval_ms > 0)? "reschedule" : "reload", op_version, crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC)); data->rc = RSC_DIGEST_ALL; } free(key); return data; } #define STONITH_DIGEST_TASK "stonith-on" static op_digest_cache_t * fencing_action_digest_cmp(resource_t * rsc, node_t * node, pe_working_set_t * data_set) { char *key = generate_op_key(rsc->id, STONITH_DIGEST_TASK, 0); op_digest_cache_t *data = rsc_action_digest(rsc, STONITH_DIGEST_TASK, key, node, NULL, data_set); const char *digest_all = pe_node_attribute_raw(node, CRM_ATTR_DIGESTS_ALL); const char *digest_secure = pe_node_attribute_raw(node, CRM_ATTR_DIGESTS_SECURE); /* No 'reloads' for fencing device changes * * We use the resource id + agent + digest so that we can detect * changes to the agent and/or the parameters used */ char *search_all = crm_strdup_printf("%s:%s:%s", rsc->id, (const char*)g_hash_table_lookup(rsc->meta, XML_ATTR_TYPE), data->digest_all_calc); char *search_secure = crm_strdup_printf("%s:%s:%s", rsc->id, (const char*)g_hash_table_lookup(rsc->meta, XML_ATTR_TYPE), data->digest_secure_calc); data->rc = RSC_DIGEST_ALL; if (digest_all == NULL) { /* it is unknown what the previous op digest was */ data->rc = RSC_DIGEST_UNKNOWN; } else if (strstr(digest_all, search_all)) { data->rc = RSC_DIGEST_MATCH; } else if(digest_secure && data->digest_secure_calc) { if(strstr(digest_secure, search_secure)) { if (is_set(data_set->flags, pe_flag_stdout)) { printf("Only 'private' parameters to %s for unfencing %s changed\n", rsc->id, node->details->uname); } data->rc = RSC_DIGEST_MATCH; } } if (is_set(data_set->flags, pe_flag_sanitized) && is_set(data_set->flags, pe_flag_stdout) && (data->rc == RSC_DIGEST_ALL) && data->digest_secure_calc) { printf("Parameters to %s for unfencing %s changed, try '%s:%s:%s'\n", rsc->id, node->details->uname, rsc->id, (const char *) g_hash_table_lookup(rsc->meta, XML_ATTR_TYPE), data->digest_secure_calc); } free(key); free(search_all); free(search_secure); return data; } const char *rsc_printable_id(resource_t *rsc) { if (is_not_set(rsc->flags, pe_rsc_unique)) { return ID(rsc->xml); } return rsc->id; } void clear_bit_recursive(resource_t * rsc, unsigned long long flag) { GListPtr gIter = rsc->children; clear_bit(rsc->flags, flag); for (; gIter != NULL; gIter = gIter->next) { resource_t *child_rsc = (resource_t *) gIter->data; clear_bit_recursive(child_rsc, flag); } } void set_bit_recursive(resource_t * rsc, unsigned long long flag) { GListPtr gIter = rsc->children; set_bit(rsc->flags, flag); for (; gIter != NULL; gIter = gIter->next) { resource_t *child_rsc = (resource_t *) gIter->data; set_bit_recursive(child_rsc, flag); } } static GListPtr find_unfencing_devices(GListPtr candidates, GListPtr matches) { for (GListPtr gIter = candidates; gIter != NULL; gIter = gIter->next) { resource_t *candidate = gIter->data; const char *provides = g_hash_table_lookup(candidate->meta, XML_RSC_ATTR_PROVIDES); const char *requires = g_hash_table_lookup(candidate->meta, XML_RSC_ATTR_REQUIRES); if(candidate->children) { matches = find_unfencing_devices(candidate->children, matches); } else if (is_not_set(candidate->flags, pe_rsc_fence_device)) { continue; } else if (crm_str_eq(provides, "unfencing", FALSE) || crm_str_eq(requires, "unfencing", FALSE)) { matches = g_list_prepend(matches, candidate); } } return matches; } action_t * pe_fence_op(node_t * node, const char *op, bool optional, const char *reason, pe_working_set_t * data_set) { char *op_key = NULL; action_t *stonith_op = NULL; if(op == NULL) { op = data_set->stonith_action; } op_key = crm_strdup_printf("%s-%s-%s", CRM_OP_FENCE, node->details->uname, op); if(data_set->singletons) { stonith_op = g_hash_table_lookup(data_set->singletons, op_key); } if(stonith_op == NULL) { stonith_op = custom_action(NULL, op_key, CRM_OP_FENCE, node, TRUE, TRUE, data_set); add_hash_param(stonith_op->meta, XML_LRM_ATTR_TARGET, node->details->uname); add_hash_param(stonith_op->meta, XML_LRM_ATTR_TARGET_UUID, node->details->id); add_hash_param(stonith_op->meta, "stonith_action", op); if (pe__is_guest_or_remote_node(node) && is_set(data_set->flags, pe_flag_enable_unfencing)) { /* Extra work to detect device changes on remotes * * We may do this for all nodes in the future, but for now * the check_action_definition() based stuff works fine. * * Use "stonith-on" to avoid creating cache entries for * operations check_action_definition() would look for. */ long max = 1024; long digests_all_offset = 0; long digests_secure_offset = 0; char *digests_all = malloc(max); char *digests_secure = malloc(max); GListPtr matches = find_unfencing_devices(data_set->resources, NULL); for (GListPtr gIter = matches; gIter != NULL; gIter = gIter->next) { resource_t *match = gIter->data; op_digest_cache_t *data = fencing_action_digest_cmp(match, node, data_set); if(data->rc == RSC_DIGEST_ALL) { optional = FALSE; crm_notice("Unfencing %s (remote): because the definition of %s changed", node->details->uname, match->id); if (is_set(data_set->flags, pe_flag_stdout)) { fprintf(stdout, " notice: Unfencing %s (remote): because the definition of %s changed\n", node->details->uname, match->id); } } digests_all_offset += snprintf( digests_all+digests_all_offset, max-digests_all_offset, "%s:%s:%s,", match->id, (const char*)g_hash_table_lookup(match->meta, XML_ATTR_TYPE), data->digest_all_calc); digests_secure_offset += snprintf( digests_secure+digests_secure_offset, max-digests_secure_offset, "%s:%s:%s,", match->id, (const char*)g_hash_table_lookup(match->meta, XML_ATTR_TYPE), data->digest_secure_calc); } g_hash_table_insert(stonith_op->meta, strdup(XML_OP_ATTR_DIGESTS_ALL), digests_all); g_hash_table_insert(stonith_op->meta, strdup(XML_OP_ATTR_DIGESTS_SECURE), digests_secure); } } else { free(op_key); } if(optional == FALSE && pe_can_fence(data_set, node)) { pe_action_required(stonith_op, NULL, reason); } else if(reason && stonith_op->reason == NULL) { stonith_op->reason = strdup(reason); } return stonith_op; } void trigger_unfencing( resource_t * rsc, node_t *node, const char *reason, action_t *dependency, pe_working_set_t * data_set) { if(is_not_set(data_set->flags, pe_flag_enable_unfencing)) { /* No resources require it */ return; } else if (rsc != NULL && is_not_set(rsc->flags, pe_rsc_fence_device)) { /* Wasn't a stonith device */ return; } else if(node && node->details->online && node->details->unclean == FALSE && node->details->shutdown == FALSE) { action_t *unfence = pe_fence_op(node, "on", FALSE, reason, data_set); if(dependency) { order_actions(unfence, dependency, pe_order_optional); } } else if(rsc) { GHashTableIter iter; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if(node->details->online && node->details->unclean == FALSE && node->details->shutdown == FALSE) { trigger_unfencing(rsc, node, reason, dependency, data_set); } } } } gboolean add_tag_ref(GHashTable * tags, const char * tag_name, const char * obj_ref) { tag_t *tag = NULL; GListPtr gIter = NULL; gboolean is_existing = FALSE; CRM_CHECK(tags && tag_name && obj_ref, return FALSE); tag = g_hash_table_lookup(tags, tag_name); if (tag == NULL) { tag = calloc(1, sizeof(tag_t)); if (tag == NULL) { return FALSE; } tag->id = strdup(tag_name); tag->refs = NULL; g_hash_table_insert(tags, strdup(tag_name), tag); } for (gIter = tag->refs; gIter != NULL; gIter = gIter->next) { const char *existing_ref = (const char *) gIter->data; if (crm_str_eq(existing_ref, obj_ref, TRUE)){ is_existing = TRUE; break; } } if (is_existing == FALSE) { tag->refs = g_list_append(tag->refs, strdup(obj_ref)); crm_trace("Added: tag=%s ref=%s", tag->id, obj_ref); } return TRUE; } void pe_action_set_flag_reason(const char *function, long line, pe_action_t *action, pe_action_t *reason, const char *text, enum pe_action_flags flags, bool overwrite) { bool unset = FALSE; bool update = FALSE; const char *change = NULL; if(is_set(flags, pe_action_runnable)) { unset = TRUE; change = "unrunnable"; } else if(is_set(flags, pe_action_optional)) { unset = TRUE; change = "required"; } else if(is_set(flags, pe_action_migrate_runnable)) { unset = TRUE; overwrite = TRUE; change = "unrunnable"; } else if(is_set(flags, pe_action_dangle)) { change = "dangling"; } else if(is_set(flags, pe_action_requires_any)) { change = "required"; } else { crm_err("Unknown flag change to %x by %s: 0x%s", flags, action->uuid, (reason? reason->uuid : "0")); } if(unset) { if(is_set(action->flags, flags)) { action->flags = crm_clear_bit(function, line, action->uuid, action->flags, flags); update = TRUE; } } else { if(is_not_set(action->flags, flags)) { action->flags = crm_set_bit(function, line, action->uuid, action->flags, flags); update = TRUE; } } if((change && update) || text) { char *reason_text = NULL; if(reason == NULL) { pe_action_set_reason(action, text, overwrite); } else if(reason->rsc == NULL) { reason_text = crm_strdup_printf("%s %s%c %s", change, reason->task, text?':':0, text?text:""); } else { reason_text = crm_strdup_printf("%s %s %s%c %s", change, reason->rsc->id, reason->task, text?':':0, text?text:"NA"); } if(reason_text && action->rsc != reason->rsc) { pe_action_set_reason(action, reason_text, overwrite); } free(reason_text); } } void pe_action_set_reason(pe_action_t *action, const char *reason, bool overwrite) { if(action->reason && overwrite) { pe_rsc_trace(action->rsc, "Changing %s reason from '%s' to '%s'", action->uuid, action->reason, reason); free(action->reason); action->reason = NULL; } if(action->reason == NULL) { if(reason) { pe_rsc_trace(action->rsc, "Set %s reason to '%s'", action->uuid, reason); action->reason = strdup(reason); } else { action->reason = NULL; } } } diff --git a/lib/services/services_linux.c b/lib/services/services_linux.c index 8686f29472..a04a8f9956 100644 --- a/lib/services/services_linux.c +++ b/lib/services/services_linux.c @@ -1,1077 +1,1077 @@ /* * Copyright 2010-2019 Andrew Beekhof * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #ifndef _GNU_SOURCE # define _GNU_SOURCE #endif #include #include #include #include #include #include #include #include #include #include #ifdef HAVE_SYS_SIGNALFD_H #include #endif #include "crm/crm.h" #include "crm/common/mainloop.h" #include "crm/services.h" #include "services_private.h" #if SUPPORT_CIBSECRETS # include "crm/common/cib_secrets.h" #endif static gboolean svc_read_output(int fd, svc_action_t * op, bool is_stderr) { char *data = NULL; int rc = 0, len = 0; char buf[500]; static const size_t buf_read_len = sizeof(buf) - 1; if (fd < 0) { crm_trace("No fd for %s", op->id); return FALSE; } if (is_stderr && op->stderr_data) { len = strlen(op->stderr_data); data = op->stderr_data; crm_trace("Reading %s stderr into offset %d", op->id, len); } else if (is_stderr == FALSE && op->stdout_data) { len = strlen(op->stdout_data); data = op->stdout_data; crm_trace("Reading %s stdout into offset %d", op->id, len); } else { crm_trace("Reading %s %s into offset %d", op->id, is_stderr?"stderr":"stdout", len); } do { rc = read(fd, buf, buf_read_len); if (rc > 0) { buf[rc] = 0; crm_trace("Got %d chars: %.80s", rc, buf); data = realloc_safe(data, len + rc + 1); len += sprintf(data + len, "%s", buf); } else if (errno != EINTR) { /* error or EOF * Cleanup happens in pipe_done() */ rc = FALSE; break; } } while (rc == buf_read_len || rc < 0); if (is_stderr) { op->stderr_data = data; } else { op->stdout_data = data; } return rc; } static int dispatch_stdout(gpointer userdata) { svc_action_t *op = (svc_action_t *) userdata; return svc_read_output(op->opaque->stdout_fd, op, FALSE); } static int dispatch_stderr(gpointer userdata) { svc_action_t *op = (svc_action_t *) userdata; return svc_read_output(op->opaque->stderr_fd, op, TRUE); } static void pipe_out_done(gpointer user_data) { svc_action_t *op = (svc_action_t *) user_data; crm_trace("%p", op); op->opaque->stdout_gsource = NULL; if (op->opaque->stdout_fd > STDOUT_FILENO) { close(op->opaque->stdout_fd); } op->opaque->stdout_fd = -1; } static void pipe_err_done(gpointer user_data) { svc_action_t *op = (svc_action_t *) user_data; op->opaque->stderr_gsource = NULL; if (op->opaque->stderr_fd > STDERR_FILENO) { close(op->opaque->stderr_fd); } op->opaque->stderr_fd = -1; } static struct mainloop_fd_callbacks stdout_callbacks = { .dispatch = dispatch_stdout, .destroy = pipe_out_done, }; static struct mainloop_fd_callbacks stderr_callbacks = { .dispatch = dispatch_stderr, .destroy = pipe_err_done, }; static void set_ocf_env(const char *key, const char *value, gpointer user_data) { if (setenv(key, value, 1) != 0) { crm_perror(LOG_ERR, "setenv failed for key:%s and value:%s", key, value); } } static void set_ocf_env_with_prefix(gpointer key, gpointer value, gpointer user_data) { char buffer[500]; snprintf(buffer, sizeof(buffer), "OCF_RESKEY_%s", (char *)key); set_ocf_env(buffer, value, user_data); } /*! * \internal * \brief Add environment variables suitable for an action * * \param[in] op Action to use */ static void add_action_env_vars(const svc_action_t *op) { if (safe_str_eq(op->standard, PCMK_RESOURCE_CLASS_OCF) == FALSE) { return; } if (op->params) { g_hash_table_foreach(op->params, set_ocf_env_with_prefix, NULL); } set_ocf_env("OCF_RA_VERSION_MAJOR", "1", NULL); set_ocf_env("OCF_RA_VERSION_MINOR", "0", NULL); set_ocf_env("OCF_ROOT", OCF_ROOT_DIR, NULL); set_ocf_env("OCF_EXIT_REASON_PREFIX", PCMK_OCF_REASON_PREFIX, NULL); if (op->rsc) { set_ocf_env("OCF_RESOURCE_INSTANCE", op->rsc, NULL); } if (op->agent != NULL) { set_ocf_env("OCF_RESOURCE_TYPE", op->agent, NULL); } /* Notes: this is not added to specification yet. Sept 10,2004 */ if (op->provider != NULL) { set_ocf_env("OCF_RESOURCE_PROVIDER", op->provider, NULL); } } static void pipe_in_single_parameter(gpointer key, gpointer value, gpointer user_data) { svc_action_t *op = user_data; char *buffer = crm_strdup_printf("%s=%s\n", (char *)key, (char *) value); int ret, total = 0, len = strlen(buffer); do { errno = 0; ret = write(op->opaque->stdin_fd, buffer + total, len - total); if (ret > 0) { total += ret; } } while ((errno == EINTR) && (total < len)); free(buffer); } /*! * \internal * \brief Pipe parameters in via stdin for action * * \param[in] op Action to use */ static void pipe_in_action_stdin_parameters(const svc_action_t *op) { crm_debug("sending args"); if (op->params) { g_hash_table_foreach(op->params, pipe_in_single_parameter, (gpointer) op); } } gboolean recurring_action_timer(gpointer data) { svc_action_t *op = data; crm_debug("Scheduling another invocation of %s", op->id); /* Clean out the old result */ free(op->stdout_data); op->stdout_data = NULL; free(op->stderr_data); op->stderr_data = NULL; op->opaque->repeat_timer = 0; services_action_async(op, NULL); return FALSE; } /* Returns FALSE if 'op' should be free'd by the caller */ gboolean operation_finalize(svc_action_t * op) { int recurring = 0; if (op->interval_ms) { if (op->cancel) { op->status = PCMK_LRM_OP_CANCELLED; cancel_recurring_action(op); } else { recurring = 1; op->opaque->repeat_timer = g_timeout_add(op->interval_ms, recurring_action_timer, (void *)op); } } if (op->opaque->callback) { op->opaque->callback(op); } op->pid = 0; services_untrack_op(op); if (!recurring && op->synchronous == FALSE) { /* * If this is a recurring action, do not free explicitly. * It will get freed whenever the action gets cancelled. */ services_action_free(op); return TRUE; } services_action_cleanup(op); return FALSE; } static void operation_finished(mainloop_child_t * p, pid_t pid, int core, int signo, int exitcode) { svc_action_t *op = mainloop_child_userdata(p); char *prefix = crm_strdup_printf("%s:%d", op->id, op->pid); mainloop_clear_child_userdata(p); op->status = PCMK_LRM_OP_DONE; CRM_ASSERT(op->pid == pid); crm_trace("%s %p %p", prefix, op->opaque->stderr_gsource, op->opaque->stdout_gsource); if (op->opaque->stderr_gsource) { /* Make sure we have read everything from the buffer. * Depending on the priority mainloop gives the fd, operation_finished * could occur before all the reads are done. Force the read now.*/ crm_trace("%s dispatching stderr", prefix); dispatch_stderr(op); crm_trace("%s: %p", op->id, op->stderr_data); mainloop_del_fd(op->opaque->stderr_gsource); op->opaque->stderr_gsource = NULL; } if (op->opaque->stdout_gsource) { /* Make sure we have read everything from the buffer. * Depending on the priority mainloop gives the fd, operation_finished * could occur before all the reads are done. Force the read now.*/ crm_trace("%s dispatching stdout", prefix); dispatch_stdout(op); crm_trace("%s: %p", op->id, op->stdout_data); mainloop_del_fd(op->opaque->stdout_gsource); op->opaque->stdout_gsource = NULL; } if (op->opaque->stdin_fd >= 0) { close(op->opaque->stdin_fd); } if (signo) { if (mainloop_child_timeout(p)) { crm_warn("%s - timed out after %dms", prefix, op->timeout); op->status = PCMK_LRM_OP_TIMEOUT; op->rc = PCMK_OCF_TIMEOUT; } else if (op->cancel) { /* If an in-flight recurring operation was killed because it was * cancelled, don't treat that as a failure. */ crm_info("%s - terminated with signal %d", prefix, signo); op->status = PCMK_LRM_OP_CANCELLED; op->rc = PCMK_OCF_OK; } else { crm_warn("%s - terminated with signal %d", prefix, signo); op->status = PCMK_LRM_OP_ERROR; op->rc = PCMK_OCF_SIGNAL; } } else { op->rc = exitcode; crm_debug("%s - exited with rc=%d", prefix, exitcode); } free(prefix); prefix = crm_strdup_printf("%s:%d:stderr", op->id, op->pid); crm_log_output(LOG_NOTICE, prefix, op->stderr_data); free(prefix); prefix = crm_strdup_printf("%s:%d:stdout", op->id, op->pid); crm_log_output(LOG_DEBUG, prefix, op->stdout_data); free(prefix); operation_finalize(op); } /*! * \internal * \brief Set operation rc and status per errno from stat(), fork() or execvp() * * \param[in,out] op Operation to set rc and status for * \param[in] error Value of errno after system call * * \return void */ static void services_handle_exec_error(svc_action_t * op, int error) { int rc_not_installed, rc_insufficient_priv, rc_exec_error; /* Mimic the return codes for each standard as that's what we'll convert back from in get_uniform_rc() */ if (safe_str_eq(op->standard, PCMK_RESOURCE_CLASS_LSB) && safe_str_eq(op->action, "status")) { rc_not_installed = PCMK_LSB_STATUS_NOT_INSTALLED; rc_insufficient_priv = PCMK_LSB_STATUS_INSUFFICIENT_PRIV; rc_exec_error = PCMK_LSB_STATUS_UNKNOWN; #if SUPPORT_NAGIOS } else if (safe_str_eq(op->standard, PCMK_RESOURCE_CLASS_NAGIOS)) { rc_not_installed = NAGIOS_NOT_INSTALLED; rc_insufficient_priv = NAGIOS_INSUFFICIENT_PRIV; rc_exec_error = PCMK_OCF_EXEC_ERROR; #endif } else { rc_not_installed = PCMK_OCF_NOT_INSTALLED; rc_insufficient_priv = PCMK_OCF_INSUFFICIENT_PRIV; rc_exec_error = PCMK_OCF_EXEC_ERROR; } switch (error) { /* see execve(2), stat(2) and fork(2) */ case ENOENT: /* No such file or directory */ case EISDIR: /* Is a directory */ case ENOTDIR: /* Path component is not a directory */ case EINVAL: /* Invalid executable format */ case ENOEXEC: /* Invalid executable format */ op->rc = rc_not_installed; op->status = PCMK_LRM_OP_NOT_INSTALLED; break; case EACCES: /* permission denied (various errors) */ case EPERM: /* permission denied (various errors) */ op->rc = rc_insufficient_priv; op->status = PCMK_LRM_OP_ERROR; break; default: op->rc = rc_exec_error; op->status = PCMK_LRM_OP_ERROR; } } static void action_launch_child(svc_action_t *op) { int lpc; /* SIGPIPE is ignored (which is different from signal blocking) by the gnutls library. * Depending on the libqb version in use, libqb may set SIGPIPE to be ignored as well. * We do not want this to be inherited by the child process. By resetting this the signal * to the default behavior, we avoid some potential odd problems that occur during OCF * scripts when SIGPIPE is ignored by the environment. */ signal(SIGPIPE, SIG_DFL); #if defined(HAVE_SCHED_SETSCHEDULER) if (sched_getscheduler(0) != SCHED_OTHER) { struct sched_param sp; memset(&sp, 0, sizeof(sp)); sp.sched_priority = 0; if (sched_setscheduler(0, SCHED_OTHER, &sp) == -1) { crm_perror(LOG_ERR, "Could not reset scheduling policy to SCHED_OTHER for %s", op->id); } } #endif if (setpriority(PRIO_PROCESS, 0, 0) == -1) { crm_perror(LOG_ERR, "Could not reset process priority to 0 for %s", op->id); } /* Man: The call setpgrp() is equivalent to setpgid(0,0) * _and_ compiles on BSD variants too * need to investigate if it works the same too. */ setpgid(0, 0); // Close all file descriptors except stdin/stdout/stderr for (lpc = getdtablesize() - 1; lpc > STDERR_FILENO; lpc--) { close(lpc); } #if SUPPORT_CIBSECRETS if (replace_secret_params(op->rsc, op->params) < 0) { /* replacing secrets failed! */ if (safe_str_eq(op->action,"stop")) { /* don't fail on stop! */ crm_info("proceeding with the stop operation for %s", op->rsc); } else { crm_err("failed to get secrets for %s, " "considering resource not configured", op->rsc); _exit(PCMK_OCF_NOT_CONFIGURED); } } #endif add_action_env_vars(op); /* Become the desired user */ if (op->opaque->uid && (geteuid() == 0)) { // If requested, set effective group if (op->opaque->gid && (setgid(op->opaque->gid) < 0)) { crm_perror(LOG_ERR, "Could not set child group to %d", op->opaque->gid); _exit(PCMK_OCF_NOT_CONFIGURED); } // Erase supplementary group list // (We could do initgroups() if we kept a copy of the username) if (setgroups(0, NULL) < 0) { crm_perror(LOG_ERR, "Could not set child groups"); _exit(PCMK_OCF_NOT_CONFIGURED); } // Set effective user if (setuid(op->opaque->uid) < 0) { crm_perror(LOG_ERR, "setting user to %d", op->opaque->uid); _exit(PCMK_OCF_NOT_CONFIGURED); } } /* execute the RA */ execvp(op->opaque->exec, op->opaque->args); /* Most cases should have been already handled by stat() */ services_handle_exec_error(op, errno); _exit(op->rc); } #ifndef HAVE_SYS_SIGNALFD_H static int sigchld_pipe[2] = { -1, -1 }; static void sigchld_handler() { if ((sigchld_pipe[1] >= 0) && (write(sigchld_pipe[1], "", 1) == -1)) { crm_perror(LOG_TRACE, "Could not poke SIGCHLD self-pipe"); } } #endif static void action_synced_wait(svc_action_t * op, sigset_t *mask) { int status = 0; int timeout = op->timeout; int sfd = -1; time_t start = -1; struct pollfd fds[3]; int wait_rc = 0; #ifdef HAVE_SYS_SIGNALFD_H sfd = signalfd(-1, mask, SFD_NONBLOCK); if (sfd < 0) { crm_perror(LOG_ERR, "signalfd() failed"); } #else sfd = sigchld_pipe[0]; #endif fds[0].fd = op->opaque->stdout_fd; fds[0].events = POLLIN; fds[0].revents = 0; fds[1].fd = op->opaque->stderr_fd; fds[1].events = POLLIN; fds[1].revents = 0; fds[2].fd = sfd; fds[2].events = POLLIN; fds[2].revents = 0; crm_trace("Waiting for %d", op->pid); start = time(NULL); do { int poll_rc = poll(fds, 3, timeout); if (poll_rc > 0) { if (fds[0].revents & POLLIN) { svc_read_output(op->opaque->stdout_fd, op, FALSE); } if (fds[1].revents & POLLIN) { svc_read_output(op->opaque->stderr_fd, op, TRUE); } if (fds[2].revents & POLLIN) { #ifdef HAVE_SYS_SIGNALFD_H struct signalfd_siginfo fdsi; ssize_t s; s = read(sfd, &fdsi, sizeof(struct signalfd_siginfo)); if (s != sizeof(struct signalfd_siginfo)) { crm_perror(LOG_ERR, "Read from signal fd %d failed", sfd); } else if (fdsi.ssi_signo == SIGCHLD) { #else if (1) { /* Clear out the sigchld pipe. */ char ch; while (read(sfd, &ch, 1) == 1) /*omit*/; #endif wait_rc = waitpid(op->pid, &status, WNOHANG); if (wait_rc > 0) { break; } else if (wait_rc < 0){ if (errno == ECHILD) { /* Here, don't dare to kill and bail out... */ break; } else { /* ...otherwise pretend process still runs. */ wait_rc = 0; } crm_perror(LOG_ERR, "waitpid() for %d failed", op->pid); } } } } else if (poll_rc == 0) { timeout = 0; break; } else if (poll_rc < 0) { if (errno != EINTR) { crm_perror(LOG_ERR, "poll() failed"); break; } } timeout = op->timeout - (time(NULL) - start) * 1000; } while ((op->timeout < 0 || timeout > 0)); crm_trace("Child done: %d", op->pid); if (wait_rc <= 0) { op->rc = PCMK_OCF_UNKNOWN_ERROR; if (op->timeout > 0 && timeout <= 0) { op->status = PCMK_LRM_OP_TIMEOUT; crm_warn("%s:%d - timed out after %dms", op->id, op->pid, op->timeout); } else { op->status = PCMK_LRM_OP_ERROR; } /* If only child hasn't been successfully waited for, yet. This is to limit killing wrong target a bit more. */ if (wait_rc == 0 && waitpid(op->pid, &status, WNOHANG) == 0) { if (kill(op->pid, SIGKILL)) { crm_err("kill(%d, KILL) failed: %d", op->pid, errno); } /* Safe to skip WNOHANG here as we sent non-ignorable signal. */ while (waitpid(op->pid, &status, 0) == (pid_t) -1 && errno == EINTR) /*omit*/; } } else if (WIFEXITED(status)) { op->status = PCMK_LRM_OP_DONE; op->rc = WEXITSTATUS(status); crm_info("Managed %s process %d exited with rc=%d", op->id, op->pid, op->rc); } else if (WIFSIGNALED(status)) { int signo = WTERMSIG(status); op->status = PCMK_LRM_OP_ERROR; crm_err("Managed %s process %d exited with signal=%d", op->id, op->pid, signo); } #ifdef WCOREDUMP if (WCOREDUMP(status)) { crm_err("Managed %s process %d dumped core", op->id, op->pid); } #endif svc_read_output(op->opaque->stdout_fd, op, FALSE); svc_read_output(op->opaque->stderr_fd, op, TRUE); close(op->opaque->stdout_fd); close(op->opaque->stderr_fd); if (op->opaque->stdin_fd >= 0) { close(op->opaque->stdin_fd); } #ifdef HAVE_SYS_SIGNALFD_H close(sfd); #endif } /* For an asynchronous 'op', returns FALSE if 'op' should be free'd by the caller */ /* For a synchronous 'op', returns FALSE if 'op' fails */ gboolean services_os_action_execute(svc_action_t * op) { int stdout_fd[2]; int stderr_fd[2]; int stdin_fd[2] = {-1, -1}; int rc; struct stat st; - sigset_t *pmask; + sigset_t *pmask = NULL; #ifdef HAVE_SYS_SIGNALFD_H sigset_t mask; sigset_t old_mask; #define sigchld_cleanup() do { \ if (sigismember(&old_mask, SIGCHLD) == 0) { \ if (sigprocmask(SIG_UNBLOCK, &mask, NULL) < 0) { \ crm_perror(LOG_ERR, "sigprocmask() failed to unblock sigchld"); \ } \ } \ } while (0) #else struct sigaction sa; struct sigaction old_sa; #define sigchld_cleanup() do { \ if (sigaction(SIGCHLD, &old_sa, NULL) < 0) { \ crm_perror(LOG_ERR, "sigaction() failed to remove sigchld handler"); \ } \ close(sigchld_pipe[0]); \ close(sigchld_pipe[1]); \ sigchld_pipe[0] = sigchld_pipe[1] = -1; \ } while(0) #endif /* Fail fast */ if(stat(op->opaque->exec, &st) != 0) { rc = errno; crm_warn("Cannot execute '%s': %s (%d)", op->opaque->exec, pcmk_strerror(rc), rc); services_handle_exec_error(op, rc); if (!op->synchronous) { return operation_finalize(op); } return FALSE; } if (pipe(stdout_fd) < 0) { rc = errno; crm_err("pipe(stdout_fd) failed. '%s': %s (%d)", op->opaque->exec, pcmk_strerror(rc), rc); services_handle_exec_error(op, rc); if (!op->synchronous) { return operation_finalize(op); } return FALSE; } if (pipe(stderr_fd) < 0) { rc = errno; close(stdout_fd[0]); close(stdout_fd[1]); crm_err("pipe(stderr_fd) failed. '%s': %s (%d)", op->opaque->exec, pcmk_strerror(rc), rc); services_handle_exec_error(op, rc); if (!op->synchronous) { return operation_finalize(op); } return FALSE; } if (safe_str_eq(op->standard, PCMK_RESOURCE_CLASS_STONITH)) { if (pipe(stdin_fd) < 0) { rc = errno; close(stdout_fd[0]); close(stdout_fd[1]); close(stderr_fd[0]); close(stderr_fd[1]); crm_err("pipe(stdin_fd) failed. '%s': %s (%d)", op->opaque->exec, pcmk_strerror(rc), rc); services_handle_exec_error(op, rc); if (!op->synchronous) { return operation_finalize(op); } return FALSE; } } if (op->synchronous) { #ifdef HAVE_SYS_SIGNALFD_H sigemptyset(&mask); sigaddset(&mask, SIGCHLD); sigemptyset(&old_mask); if (sigprocmask(SIG_BLOCK, &mask, &old_mask) < 0) { crm_perror(LOG_ERR, "sigprocmask() failed to block sigchld"); } pmask = &mask; #else if(pipe(sigchld_pipe) == -1) { crm_perror(LOG_ERR, "pipe() failed"); } rc = crm_set_nonblocking(sigchld_pipe[0]); if (rc < 0) { crm_warn("Could not set pipe input non-blocking: %s " CRM_XS " rc=%d", pcmk_strerror(rc), rc); } rc = crm_set_nonblocking(sigchld_pipe[1]); if (rc < 0) { crm_warn("Could not set pipe output non-blocking: %s " CRM_XS " rc=%d", pcmk_strerror(rc), rc); } sa.sa_handler = sigchld_handler; sa.sa_flags = 0; sigemptyset(&sa.sa_mask); if (sigaction(SIGCHLD, &sa, &old_sa) < 0) { crm_perror(LOG_ERR, "sigaction() failed to set sigchld handler"); } pmask = NULL; #endif } op->pid = fork(); switch (op->pid) { case -1: rc = errno; close(stdout_fd[0]); close(stdout_fd[1]); close(stderr_fd[0]); close(stderr_fd[1]); if (stdin_fd[0] >= 0) { close(stdin_fd[0]); close(stdin_fd[1]); } crm_err("Could not execute '%s': %s (%d)", op->opaque->exec, pcmk_strerror(rc), rc); services_handle_exec_error(op, rc); if (!op->synchronous) { return operation_finalize(op); } sigchld_cleanup(); return FALSE; case 0: /* Child */ close(stdout_fd[0]); close(stderr_fd[0]); if (stdin_fd[1] >= 0) { close(stdin_fd[1]); } if (STDOUT_FILENO != stdout_fd[1]) { if (dup2(stdout_fd[1], STDOUT_FILENO) != STDOUT_FILENO) { crm_err("dup2() failed (stdout)"); } close(stdout_fd[1]); } if (STDERR_FILENO != stderr_fd[1]) { if (dup2(stderr_fd[1], STDERR_FILENO) != STDERR_FILENO) { crm_err("dup2() failed (stderr)"); } close(stderr_fd[1]); } if ((stdin_fd[0] >= 0) && (STDIN_FILENO != stdin_fd[0])) { if (dup2(stdin_fd[0], STDIN_FILENO) != STDIN_FILENO) { crm_err("dup2() failed (stdin)"); } close(stdin_fd[0]); } if (op->synchronous) { sigchld_cleanup(); } action_launch_child(op); CRM_ASSERT(0); /* action_launch_child is effectively noreturn */ } /* Only the parent reaches here */ close(stdout_fd[1]); close(stderr_fd[1]); if (stdin_fd[0] >= 0) { close(stdin_fd[0]); } op->opaque->stdout_fd = stdout_fd[0]; rc = crm_set_nonblocking(op->opaque->stdout_fd); if (rc < 0) { crm_warn("Could not set child output non-blocking: %s " CRM_XS " rc=%d", pcmk_strerror(rc), rc); } op->opaque->stderr_fd = stderr_fd[0]; rc = crm_set_nonblocking(op->opaque->stderr_fd); if (rc < 0) { crm_warn("Could not set child error output non-blocking: %s " CRM_XS " rc=%d", pcmk_strerror(rc), rc); } op->opaque->stdin_fd = stdin_fd[1]; if (op->opaque->stdin_fd >= 0) { // using buffer behind non-blocking-fd here - that could be improved // as long as no other standard uses stdin_fd assume stonith rc = crm_set_nonblocking(op->opaque->stdin_fd); if (rc < 0) { crm_warn("Could not set child input non-blocking: %s " CRM_XS " fd=%d,rc=%d", pcmk_strerror(rc), op->opaque->stdin_fd, rc); } pipe_in_action_stdin_parameters(op); // as long as we are handling parameters directly in here just close close(op->opaque->stdin_fd); op->opaque->stdin_fd = -1; } // after fds are setup properly and before we plug anything into mainloop if (op->opaque->fork_callback) { op->opaque->fork_callback(op); } if (op->synchronous) { action_synced_wait(op, pmask); sigchld_cleanup(); } else { crm_trace("Async waiting for %d - %s", op->pid, op->opaque->exec); mainloop_child_add_with_flags(op->pid, op->timeout, op->id, op, (op->flags & SVC_ACTION_LEAVE_GROUP) ? mainloop_leave_pid_group : 0, operation_finished); op->opaque->stdout_gsource = mainloop_add_fd(op->id, G_PRIORITY_LOW, op->opaque->stdout_fd, op, &stdout_callbacks); op->opaque->stderr_gsource = mainloop_add_fd(op->id, G_PRIORITY_LOW, op->opaque->stderr_fd, op, &stderr_callbacks); services_add_inflight_op(op); } return TRUE; } GList * services_os_get_directory_list(const char *root, gboolean files, gboolean executable) { GList *list = NULL; struct dirent **namelist; int entries = 0, lpc = 0; char buffer[PATH_MAX]; entries = scandir(root, &namelist, NULL, alphasort); if (entries <= 0) { return list; } for (lpc = 0; lpc < entries; lpc++) { struct stat sb; if ('.' == namelist[lpc]->d_name[0]) { free(namelist[lpc]); continue; } snprintf(buffer, sizeof(buffer), "%s/%s", root, namelist[lpc]->d_name); if (stat(buffer, &sb)) { continue; } if (S_ISDIR(sb.st_mode)) { if (files) { free(namelist[lpc]); continue; } } else if (S_ISREG(sb.st_mode)) { if (files == FALSE) { free(namelist[lpc]); continue; } else if (executable && (sb.st_mode & S_IXUSR) == 0 && (sb.st_mode & S_IXGRP) == 0 && (sb.st_mode & S_IXOTH) == 0) { free(namelist[lpc]); continue; } } list = g_list_append(list, strdup(namelist[lpc]->d_name)); free(namelist[lpc]); } free(namelist); return list; } GList * resources_os_list_ocf_providers(void) { return get_directory_list(OCF_ROOT_DIR "/resource.d", FALSE, TRUE); } GList * resources_os_list_ocf_agents(const char *provider) { GList *gIter = NULL; GList *result = NULL; GList *providers = NULL; if (provider) { char buffer[500]; snprintf(buffer, sizeof(buffer), "%s/resource.d/%s", OCF_ROOT_DIR, provider); return get_directory_list(buffer, TRUE, TRUE); } providers = resources_os_list_ocf_providers(); for (gIter = providers; gIter != NULL; gIter = gIter->next) { GList *tmp1 = result; GList *tmp2 = resources_os_list_ocf_agents(gIter->data); if (tmp2) { result = g_list_concat(tmp1, tmp2); } } g_list_free_full(providers, free); return result; } gboolean services__ocf_agent_exists(const char *provider, const char *agent) { char *buf = NULL; gboolean rc = FALSE; struct stat st; if (provider == NULL || agent == NULL) { return rc; } buf = crm_strdup_printf(OCF_ROOT_DIR "/resource.d/%s/%s", provider, agent); if (stat(buf, &st) == 0) { rc = TRUE; } free(buf); return rc; } #if SUPPORT_NAGIOS GList * resources_os_list_nagios_agents(void) { GList *plugin_list = NULL; GList *result = NULL; GList *gIter = NULL; plugin_list = get_directory_list(NAGIOS_PLUGIN_DIR, TRUE, TRUE); /* Make sure both the plugin and its metadata exist */ for (gIter = plugin_list; gIter != NULL; gIter = gIter->next) { const char *plugin = gIter->data; char *metadata = crm_strdup_printf(NAGIOS_METADATA_DIR "/%s.xml", plugin); struct stat st; if (stat(metadata, &st) == 0) { result = g_list_append(result, strdup(plugin)); } free(metadata); } g_list_free_full(plugin_list, free); return result; } gboolean services__nagios_agent_exists(const char *name) { char *buf = NULL; gboolean rc = FALSE; struct stat st; if (name == NULL) { return rc; } buf = crm_strdup_printf(NAGIOS_PLUGIN_DIR "/%s", name); if (stat(buf, &st) == 0) { rc = TRUE; } free(buf); return rc; } #endif diff --git a/tools/crm_mon.c b/tools/crm_mon.c index 51a2cf8820..cb77ef5880 100644 --- a/tools/crm_mon.c +++ b/tools/crm_mon.c @@ -1,4428 +1,4435 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include +#include #include #include #include #include #include /* crm_ends_with_ext */ #include #include #include #include #include #include #include #include <../lib/pengine/unpack.h> #include #include static void clean_up_connections(void); static crm_exit_t clean_up(crm_exit_t exit_code); static void crm_diff_update(const char *event, xmlNode * msg); static gboolean mon_refresh_display(gpointer user_data); static int cib_connect(gboolean full); static void mon_st_callback_event(stonith_t * st, stonith_event_t * e); static void mon_st_callback_display(stonith_t * st, stonith_event_t * e); static void kick_refresh(gboolean data_updated); static char *get_node_display_name(node_t *node); /* * Definitions indicating which items to print */ #define mon_show_times (0x0001U) #define mon_show_stack (0x0002U) #define mon_show_dc (0x0004U) #define mon_show_count (0x0008U) #define mon_show_nodes (0x0010U) #define mon_show_resources (0x0020U) #define mon_show_attributes (0x0040U) #define mon_show_failcounts (0x0080U) #define mon_show_operations (0x0100U) #define mon_show_tickets (0x0200U) #define mon_show_bans (0x0400U) #define mon_show_fence_history (0x0800U) #define mon_show_headers (mon_show_times | mon_show_stack | mon_show_dc \ | mon_show_count) #define mon_show_default (mon_show_headers | mon_show_nodes \ | mon_show_resources) #define mon_show_all (mon_show_default | mon_show_attributes \ | mon_show_failcounts | mon_show_operations \ | mon_show_tickets | mon_show_bans \ | mon_show_fence_history) static unsigned int show = mon_show_default; /* * Definitions indicating how to output */ enum mon_output_format_e { mon_output_none, mon_output_monitor, mon_output_plain, mon_output_console, mon_output_xml, mon_output_html, mon_output_cgi } output_format = mon_output_console; static char *output_filename = NULL; /* if sending output to a file, its name */ /* other globals */ static char *pid_file = NULL; static gboolean group_by_node = FALSE; static gboolean inactive_resources = FALSE; static int reconnect_msec = 5000; static gboolean daemonize = FALSE; static GMainLoop *mainloop = NULL; static guint timer_id = 0; static mainloop_timer_t *refresh_timer = NULL; static pe_working_set_t *mon_data_set = NULL; static GList *attr_list = NULL; static const char *external_agent = NULL; static const char *external_recipient = NULL; static cib_t *cib = NULL; static stonith_t *st = NULL; static xmlNode *current_cib = NULL; static gboolean one_shot = FALSE; static gboolean has_warnings = FALSE; static gboolean print_timing = FALSE; static gboolean watch_fencing = FALSE; static gboolean fence_history = FALSE; static gboolean fence_full_history = FALSE; static gboolean fence_connect = FALSE; static int fence_history_level = 1; static gboolean print_brief = FALSE; static gboolean print_pending = TRUE; static gboolean print_clone_detail = FALSE; #if CURSES_ENABLED static gboolean curses_console_initialized = FALSE; #endif /* FIXME allow, detect, and correctly interpret glob pattern or regex? */ const char *print_neg_location_prefix = ""; /* Never display node attributes whose name starts with one of these prefixes */ #define FILTER_STR { CRM_FAIL_COUNT_PREFIX, CRM_LAST_FAILURE_PREFIX, \ "shutdown", "terminate", "standby", "probe_complete", \ "#", NULL } long last_refresh = 0; crm_trigger_t *refresh_trigger = NULL; /* Define exit codes for monitoring-compatible output * For nagios plugins, the possibilities are * OK=0, WARN=1, CRIT=2, and UNKNOWN=3 */ #define MON_STATUS_WARN CRM_EX_ERROR #define MON_STATUS_CRIT CRM_EX_INVALID_PARAM #define MON_STATUS_UNKNOWN CRM_EX_UNIMPLEMENT_FEATURE /* Convenience macro for prettifying output (e.g. "node" vs "nodes") */ #define s_if_plural(i) (((i) == 1)? "" : "s") #if CURSES_ENABLED # define print_dot() if (output_format == mon_output_console) { \ printw("."); \ clrtoeol(); \ refresh(); \ } else { \ fprintf(stdout, "."); \ } #else # define print_dot() fprintf(stdout, "."); #endif #if CURSES_ENABLED # define print_as(fmt, args...) if (output_format == mon_output_console) { \ printw(fmt, ##args); \ clrtoeol(); \ refresh(); \ } else { \ fprintf(stdout, fmt, ##args); \ } #else # define print_as(fmt, args...) fprintf(stdout, fmt, ##args); #endif static void blank_screen(void) { #if CURSES_ENABLED int lpc = 0; for (lpc = 0; lpc < LINES; lpc++) { move(lpc, 0); clrtoeol(); } move(0, 0); refresh(); #endif } static gboolean mon_timer_popped(gpointer data) { int rc = pcmk_ok; #if CURSES_ENABLED if (output_format == mon_output_console) { clear(); refresh(); } #endif if (timer_id > 0) { g_source_remove(timer_id); timer_id = 0; } print_as("Reconnecting...\n"); rc = cib_connect(TRUE); if (rc != pcmk_ok) { timer_id = g_timeout_add(reconnect_msec, mon_timer_popped, NULL); } return FALSE; } static void mon_cib_connection_destroy(gpointer user_data) { print_as("Connection to the cluster-daemons terminated\n"); if (refresh_timer != NULL) { /* we'll trigger a refresh after reconnect */ mainloop_timer_stop(refresh_timer); } if (timer_id) { /* we'll trigger a new reconnect-timeout at the end */ g_source_remove(timer_id); timer_id = 0; } if (st) { /* the client API won't properly reconnect notifications * if they are still in the table - so remove them */ st->cmds->remove_notification(st, T_STONITH_NOTIFY_DISCONNECT); st->cmds->remove_notification(st, T_STONITH_NOTIFY_FENCE); st->cmds->remove_notification(st, T_STONITH_NOTIFY_HISTORY); if (st->state != stonith_disconnected) { st->cmds->disconnect(st); } } if (cib) { cib->cmds->signoff(cib); timer_id = g_timeout_add(reconnect_msec, mon_timer_popped, NULL); } return; } /* * Mainloop signal handler. */ static void mon_shutdown(int nsig) { clean_up(CRM_EX_OK); } -#if ON_DARWIN -# define sighandler_t sig_t -#endif - #if CURSES_ENABLED -# ifndef HAVE_SIGHANDLER_T -typedef void (*sighandler_t) (int); -# endif static sighandler_t ncurses_winch_handler; + static void mon_winresize(int nsig) { static int not_done; int lines = 0, cols = 0; if (!not_done++) { if (ncurses_winch_handler) /* the original ncurses WINCH signal handler does the * magic of retrieving the new window size; * otherwise, we'd have to use ioctl or tgetent */ (*ncurses_winch_handler) (SIGWINCH); getmaxyx(stdscr, lines, cols); resizeterm(lines, cols); mainloop_set_trigger(refresh_trigger); } not_done--; } #endif static int cib_connect(gboolean full) { int rc = pcmk_ok; static gboolean need_pass = TRUE; CRM_CHECK(cib != NULL, return -EINVAL); if (getenv("CIB_passwd") != NULL) { need_pass = FALSE; } if ((fence_connect) && (st == NULL)) { st = stonith_api_new(); } if ((fence_connect) && (st->state == stonith_disconnected)) { crm_trace("Connecting to stonith"); rc = st->cmds->connect(st, crm_system_name, NULL); if (rc == pcmk_ok) { crm_trace("Setting up stonith callbacks"); if (watch_fencing) { st->cmds->register_notification(st, T_STONITH_NOTIFY_DISCONNECT, mon_st_callback_event); st->cmds->register_notification(st, T_STONITH_NOTIFY_FENCE, mon_st_callback_event); } else { st->cmds->register_notification(st, T_STONITH_NOTIFY_DISCONNECT, mon_st_callback_display); st->cmds->register_notification(st, T_STONITH_NOTIFY_HISTORY, mon_st_callback_display); } } } if (cib->state != cib_connected_query && cib->state != cib_connected_command) { crm_trace("Connecting to the CIB"); if ((output_format == mon_output_console) && need_pass && (cib->variant == cib_remote)) { need_pass = FALSE; print_as("Password:"); } rc = cib->cmds->signon(cib, crm_system_name, cib_query); if (rc != pcmk_ok) { return rc; } rc = cib->cmds->query(cib, NULL, ¤t_cib, cib_scope_local | cib_sync_call); if (rc == pcmk_ok) { mon_refresh_display(NULL); } if (rc == pcmk_ok && full) { if (rc == pcmk_ok) { rc = cib->cmds->set_connection_dnotify(cib, mon_cib_connection_destroy); if (rc == -EPROTONOSUPPORT) { print_as ("Notification setup not supported, won't be able to reconnect after failure"); if (output_format == mon_output_console) { sleep(2); } rc = pcmk_ok; } } if (rc == pcmk_ok) { cib->cmds->del_notify_callback(cib, T_CIB_DIFF_NOTIFY, crm_diff_update); rc = cib->cmds->add_notify_callback(cib, T_CIB_DIFF_NOTIFY, crm_diff_update); } if (rc != pcmk_ok) { print_as("Notification setup failed, could not monitor CIB actions"); if (output_format == mon_output_console) { sleep(2); } clean_up_connections(); } } } return rc; } /* *INDENT-OFF* */ static struct crm_option long_options[] = { /* Top-level Options */ {"help", 0, 0, '?', "\tThis text"}, {"version", 0, 0, '$', "\tVersion information" }, {"verbose", 0, 0, 'V', "\tIncrease debug output"}, {"quiet", 0, 0, 'Q', "\tDisplay only essential output" }, {"-spacer-", 1, 0, '-', "\nModes (mutually exclusive):"}, {"as-html", 1, 0, 'h', "\tWrite cluster status to the named html file"}, {"as-xml", 0, 0, 'X', "\t\tWrite cluster status as xml to stdout. This will enable one-shot mode."}, {"web-cgi", 0, 0, 'w', "\t\tWeb mode with output suitable for CGI (preselected when run as *.cgi)"}, {"simple-status", 0, 0, 's', "\tDisplay the cluster status once as a simple one line output (suitable for nagios)"}, {"-spacer-", 1, 0, '-', "\nDisplay Options:"}, {"group-by-node", 0, 0, 'n', "\tGroup resources by node" }, {"inactive", 0, 0, 'r', "\t\tDisplay inactive resources" }, {"failcounts", 0, 0, 'f', "\tDisplay resource fail counts"}, {"operations", 0, 0, 'o', "\tDisplay resource operation history" }, {"timing-details", 0, 0, 't', "\tDisplay resource operation history with timing details" }, {"tickets", 0, 0, 'c', "\t\tDisplay cluster tickets"}, {"watch-fencing", 0, 0, 'W', "\tListen for fencing events. For use with --external-agent"}, {"fence-history", 2, 0, 'm', "Show fence history\n" "\t\t\t\t\t0=off, 1=failures and pending (default without option),\n" "\t\t\t\t\t2=add successes (default without value for option),\n" "\t\t\t\t\t3=show full history without reduction to most recent of each flavor"}, {"neg-locations", 2, 0, 'L', "Display negative location constraints [optionally filtered by id prefix]"}, {"show-node-attributes", 0, 0, 'A', "Display node attributes" }, {"hide-headers", 0, 0, 'D', "\tHide all headers" }, {"show-detail", 0, 0, 'R', "\tShow more details (node IDs, individual clone instances)" }, {"brief", 0, 0, 'b', "\t\tBrief output" }, {"pending", 0, 0, 'j', "\t\tDisplay pending state if 'record-pending' is enabled", pcmk_option_hidden}, {"-spacer-", 1, 0, '-', "\nAdditional Options:"}, {"interval", 1, 0, 'i', "\tUpdate frequency in seconds" }, {"one-shot", 0, 0, '1', "\t\tDisplay the cluster status once on the console and exit"}, {"disable-ncurses",0, 0, 'N', "\tDisable the use of ncurses", !CURSES_ENABLED}, {"daemonize", 0, 0, 'd', "\tRun in the background as a daemon"}, {"pid-file", 1, 0, 'p', "\t(Advanced) Daemon pid file location"}, {"external-agent", 1, 0, 'E', "A program to run when resource operations take place."}, {"external-recipient",1, 0, 'e', "A recipient for your program (assuming you want the program to send something to someone)."}, {"xml-file", 1, 0, 'x', NULL, pcmk_option_hidden}, {"-spacer-", 1, 0, '-', "\nExamples:", pcmk_option_paragraph}, {"-spacer-", 1, 0, '-', "Display the cluster status on the console with updates as they occur:", pcmk_option_paragraph}, {"-spacer-", 1, 0, '-', " crm_mon", pcmk_option_example}, {"-spacer-", 1, 0, '-', "Display the cluster status on the console just once then exit:", pcmk_option_paragraph}, {"-spacer-", 1, 0, '-', " crm_mon -1", pcmk_option_example}, {"-spacer-", 1, 0, '-', "Display your cluster status, group resources by node, and include inactive resources in the list:", pcmk_option_paragraph}, {"-spacer-", 1, 0, '-', " crm_mon --group-by-node --inactive", pcmk_option_example}, {"-spacer-", 1, 0, '-', "Start crm_mon as a background daemon and have it write the cluster status to an HTML file:", pcmk_option_paragraph}, {"-spacer-", 1, 0, '-', " crm_mon --daemonize --as-html /path/to/docroot/filename.html", pcmk_option_example}, {"-spacer-", 1, 0, '-', "Start crm_mon and export the current cluster status as xml to stdout, then exit.:", pcmk_option_paragraph}, {"-spacer-", 1, 0, '-', " crm_mon --as-xml", pcmk_option_example}, {NULL, 0, 0, 0} }; /* *INDENT-ON* */ #if CURSES_ENABLED static const char * get_option_desc(char c) { int lpc; for (lpc = 0; long_options[lpc].name != NULL; lpc++) { if (long_options[lpc].name[0] == '-') continue; if (long_options[lpc].val == c) { static char *buf = NULL; const char *rv; char *nl; /* chop off tabs and cut at newline */ free(buf); /* free string from last usage */ buf = strdup(long_options[lpc].desc); rv = buf; /* make a copy to keep buf pointer unaltered for freeing when we come by next time. Like this the result stays valid until the next call. */ while(isspace(rv[0])) { rv++; } nl = strchr(rv, '\n'); if (nl) { *nl = '\0'; } return rv; } } return NULL; } #define print_option_help(option, condition) \ print_as("%c %c: \t%s\n", ((condition)? '*': ' '), option, get_option_desc(option)); static gboolean detect_user_input(GIOChannel *channel, GIOCondition condition, gpointer unused) { int c; gboolean config_mode = FALSE; while (1) { /* Get user input */ c = getchar(); switch (c) { case 'm': if (!fence_history_level) { fence_history = TRUE; fence_connect = TRUE; if (st == NULL) { mon_cib_connection_destroy(NULL); } } show ^= mon_show_fence_history; break; case 'c': show ^= mon_show_tickets; break; case 'f': show ^= mon_show_failcounts; break; case 'n': group_by_node = ! group_by_node; break; case 'o': show ^= mon_show_operations; if ((show & mon_show_operations) == 0) { print_timing = 0; } break; case 'r': inactive_resources = ! inactive_resources; break; case 'R': print_clone_detail = ! print_clone_detail; break; case 't': print_timing = ! print_timing; if (print_timing) { show |= mon_show_operations; } break; case 'A': show ^= mon_show_attributes; break; case 'L': show ^= mon_show_bans; break; case 'D': /* If any header is shown, clear them all, otherwise set them all */ if (show & mon_show_headers) { show &= ~mon_show_headers; } else { show |= mon_show_headers; } break; case 'b': print_brief = ! print_brief; break; case 'j': print_pending = ! print_pending; break; case '?': config_mode = TRUE; break; default: goto refresh; } if (!config_mode) goto refresh; blank_screen(); print_as("Display option change mode\n"); print_as("\n"); print_option_help('c', show & mon_show_tickets); print_option_help('f', show & mon_show_failcounts); print_option_help('n', group_by_node); print_option_help('o', show & mon_show_operations); print_option_help('r', inactive_resources); print_option_help('t', print_timing); print_option_help('A', show & mon_show_attributes); print_option_help('L', show & mon_show_bans); print_option_help('D', (show & mon_show_headers) == 0); print_option_help('R', print_clone_detail); print_option_help('b', print_brief); print_option_help('j', print_pending); print_option_help('m', (show & mon_show_fence_history)); print_as("\n"); print_as("Toggle fields via field letter, type any other key to return"); } refresh: mon_refresh_display(NULL); return TRUE; } #endif +// Basically crm_signal(SIGCHLD, SIG_IGN) plus the SA_NOCLDWAIT flag +static void +avoid_zombies() +{ + struct sigaction sa; + + memset(&sa, 0, sizeof(struct sigaction)); + if (sigemptyset(&sa.sa_mask) < 0) { + crm_warn("Cannot avoid zombies: %s", pcmk_strerror(errno)); + return; + } + sa.sa_handler = SIG_IGN; + sa.sa_flags = SA_RESTART|SA_NOCLDWAIT; + if (sigaction(SIGCHLD, &sa, NULL) < 0) { + crm_warn("Cannot avoid zombies: %s", pcmk_strerror(errno)); + } +} + int main(int argc, char **argv) { int flag; int argerr = 0; int option_index = 0; int rc = pcmk_ok; pid_file = strdup("/tmp/ClusterMon.pid"); crm_log_cli_init("crm_mon"); crm_set_options(NULL, "mode [options]", long_options, "Provides a summary of cluster's current state." "\n\nOutputs varying levels of detail in a number of different formats.\n"); -#if !defined (ON_DARWIN) && !defined (ON_BSD) - /* prevent zombies */ - signal(SIGCLD, SIG_IGN); -#endif + // Avoid needing to wait for subprocesses forked for -E/--external-agent + avoid_zombies(); if (crm_ends_with_ext(argv[0], ".cgi") == TRUE) { output_format = mon_output_cgi; one_shot = TRUE; } /* to enable stonith-connection when called via some application like pcs * set environment-variable FENCE_HISTORY to desired level * so you don't have to modify this application */ /* fence_history_level = crm_atoi(getenv("FENCE_HISTORY"), "0"); */ while (1) { flag = crm_get_option(argc, argv, &option_index); if (flag == -1) break; switch (flag) { case 'V': crm_bump_log_level(argc, argv); break; case 'Q': show &= ~mon_show_times; break; case 'i': reconnect_msec = crm_get_msec(optarg); break; case 'n': group_by_node = TRUE; break; case 'r': inactive_resources = TRUE; break; case 'W': watch_fencing = TRUE; fence_connect = TRUE; break; case 'm': fence_history_level = crm_atoi(optarg, "2"); break; case 'd': daemonize = TRUE; break; case 't': print_timing = TRUE; show |= mon_show_operations; break; case 'o': show |= mon_show_operations; break; case 'f': show |= mon_show_failcounts; break; case 'A': show |= mon_show_attributes; break; case 'L': show |= mon_show_bans; print_neg_location_prefix = optarg? optarg : ""; break; case 'D': show &= ~mon_show_headers; break; case 'b': print_brief = TRUE; break; case 'j': print_pending = TRUE; break; case 'R': print_clone_detail = TRUE; break; case 'c': show |= mon_show_tickets; break; case 'p': free(pid_file); if(optarg == NULL) { crm_help(flag, CRM_EX_USAGE); } pid_file = strdup(optarg); break; case 'x': if(optarg == NULL) { crm_help(flag, CRM_EX_USAGE); } setenv("CIB_file", optarg, 1); one_shot = TRUE; break; case 'h': if(optarg == NULL) { crm_help(flag, CRM_EX_USAGE); } argerr += (output_format != mon_output_console); output_format = mon_output_html; output_filename = strdup(optarg); umask(S_IWGRP | S_IWOTH); break; case 'X': argerr += (output_format != mon_output_console); output_format = mon_output_xml; one_shot = TRUE; break; case 'w': /* do not allow argv[0] and argv[1...] redundancy */ argerr += (output_format != mon_output_console); output_format = mon_output_cgi; one_shot = TRUE; break; case 's': argerr += (output_format != mon_output_console); output_format = mon_output_monitor; one_shot = TRUE; break; case 'E': external_agent = optarg; break; case 'e': external_recipient = optarg; break; case '1': one_shot = TRUE; break; case 'N': if (output_format == mon_output_console) { output_format = mon_output_plain; } break; case '$': case '?': crm_help(flag, CRM_EX_OK); break; default: printf("Argument code 0%o (%c) is not (?yet?) supported\n", flag, flag); ++argerr; break; } } if (watch_fencing) { /* don't moan as fence_history_level == 1 is default */ fence_history_level = 0; } /* create the cib-object early to be able to do further * decisions based on the cib-source */ cib = cib_new(); if (cib == NULL) { rc = -EINVAL; } else { switch (cib->variant) { case cib_native: /* cib & fencing - everything available */ break; case cib_file: /* Don't try to connect to fencing as we * either don't have a running cluster or * the fencing-information would possibly * not match the cib data from a file. * As we don't expect cib-updates coming * in enforce one-shot. */ fence_history_level = 0; one_shot = TRUE; break; case cib_remote: /* updates coming in but no fencing */ fence_history_level = 0; break; case cib_undefined: case cib_database: default: /* something is odd */ rc = -EINVAL; crm_err("Invalid cib-source"); break; } } switch (fence_history_level) { case 3: fence_full_history = TRUE; /* fall through to next lower level */ case 2: show |= mon_show_fence_history; /* fall through to next lower level */ case 1: fence_history = TRUE; fence_connect = TRUE; break; default: break; } /* Extra sanity checks when in CGI mode */ if (output_format == mon_output_cgi) { argerr += (optind < argc); argerr += (output_filename != NULL); argerr += ((cib) && (cib->variant == cib_file)); argerr += (external_agent != NULL); argerr += (daemonize == TRUE); /* paranoia */ } else if (optind < argc) { printf("non-option ARGV-elements: "); while (optind < argc) printf("%s ", argv[optind++]); printf("\n"); } if (argerr) { return clean_up(CRM_EX_USAGE); } /* XML output always prints everything */ if (output_format == mon_output_xml) { show = mon_show_all; print_timing = TRUE; } if (one_shot) { if (output_format == mon_output_console) { output_format = mon_output_plain; } } else if (daemonize) { if ((output_format == mon_output_console) || (output_format == mon_output_plain)) { output_format = mon_output_none; } crm_enable_stderr(FALSE); if ((output_format != mon_output_html) && !external_agent) { printf ("Looks like you forgot to specify one or more of: " "--as-html, --external-agent\n"); return clean_up(CRM_EX_USAGE); } if (cib) { /* to be on the safe side don't have cib-object around * when we are forking */ cib_delete(cib); cib = NULL; crm_make_daemon(crm_system_name, TRUE, pid_file); cib = cib_new(); if (cib == NULL) { rc = -EINVAL; } /* otherwise assume we've got the same cib-object we've just destroyed * in our parent */ } } else if (output_format == mon_output_console) { #if CURSES_ENABLED initscr(); cbreak(); noecho(); crm_enable_stderr(FALSE); curses_console_initialized = TRUE; #else one_shot = TRUE; output_format = mon_output_plain; printf("Defaulting to one-shot mode\n"); printf("You need to have curses available at compile time to enable console mode\n"); #endif } crm_info("Starting %s", crm_system_name); if (cib) { do { if (!one_shot) { print_as("Waiting until cluster is available on this node ...\n"); } rc = cib_connect(!one_shot); if (one_shot) { break; } else if (rc != pcmk_ok) { sleep(reconnect_msec / 1000); #if CURSES_ENABLED if (output_format == mon_output_console) { clear(); refresh(); } #endif } else { if (output_format == mon_output_html) { print_as("Writing html to %s ...\n", output_filename); } } } while (rc == -ENOTCONN); } if (rc != pcmk_ok) { if (output_format == mon_output_monitor) { printf("CLUSTER CRIT: Connection to cluster failed: %s\n", pcmk_strerror(rc)); return clean_up(MON_STATUS_CRIT); } else { if (rc == -ENOTCONN) { print_as("\nError: cluster is not available on this node\n"); } else { print_as("\nConnection to cluster failed: %s\n", pcmk_strerror(rc)); } } if (output_format == mon_output_console) { sleep(2); } return clean_up(crm_errno2exit(rc)); } if (one_shot) { return clean_up(CRM_EX_OK); } mainloop = g_main_loop_new(NULL, FALSE); mainloop_add_signal(SIGTERM, mon_shutdown); mainloop_add_signal(SIGINT, mon_shutdown); #if CURSES_ENABLED if (output_format == mon_output_console) { - ncurses_winch_handler = signal(SIGWINCH, mon_winresize); + ncurses_winch_handler = crm_signal(SIGWINCH, mon_winresize); if (ncurses_winch_handler == SIG_DFL || ncurses_winch_handler == SIG_IGN || ncurses_winch_handler == SIG_ERR) ncurses_winch_handler = NULL; g_io_add_watch(g_io_channel_unix_new(STDIN_FILENO), G_IO_IN, detect_user_input, NULL); } #endif refresh_trigger = mainloop_add_trigger(G_PRIORITY_LOW, mon_refresh_display, NULL); g_main_loop_run(mainloop); g_main_loop_unref(mainloop); crm_info("Exiting %s", crm_system_name); return clean_up(CRM_EX_OK); } #define mon_warn(fmt...) do { \ if (!has_warnings) { \ print_as("CLUSTER WARN:"); \ } else { \ print_as(","); \ } \ print_as(fmt); \ has_warnings = TRUE; \ } while(0) static int count_resources(pe_working_set_t * data_set, resource_t * rsc) { int count = 0; GListPtr gIter = NULL; if (rsc == NULL) { gIter = data_set->resources; } else if (rsc->children) { gIter = rsc->children; } else { return is_not_set(rsc->flags, pe_rsc_orphan); } for (; gIter != NULL; gIter = gIter->next) { count += count_resources(data_set, gIter->data); } return count; } /*! * \internal * \brief Print one-line status suitable for use with monitoring software * * \param[in] data_set Working set of CIB state * \param[in] history List of stonith actions * * \note This function's output (and the return code when the program exits) * should conform to https://www.monitoring-plugins.org/doc/guidelines.html */ static void print_simple_status(pe_working_set_t * data_set, stonith_history_t *history) { GListPtr gIter = NULL; int nodes_online = 0; int nodes_standby = 0; int nodes_maintenance = 0; if (data_set->dc_node == NULL) { mon_warn(" No DC"); } for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; if (node->details->standby && node->details->online) { nodes_standby++; } else if (node->details->maintenance && node->details->online) { nodes_maintenance++; } else if (node->details->online) { nodes_online++; } else { mon_warn(" offline node: %s", node->details->uname); } } if (!has_warnings) { int nresources = count_resources(data_set, NULL); print_as("CLUSTER OK: %d node%s online", nodes_online, s_if_plural(nodes_online)); if (nodes_standby > 0) { print_as(", %d standby node%s", nodes_standby, s_if_plural(nodes_standby)); } if (nodes_maintenance > 0) { print_as(", %d maintenance node%s", nodes_maintenance, s_if_plural(nodes_maintenance)); } print_as(", %d resource%s configured", nresources, s_if_plural(nresources)); } print_as("\n"); } /*! * \internal * \brief Print a [name]=[value][units] pair, optionally using time string * * \param[in] stream File stream to display output to * \param[in] name Name to display * \param[in] value Value to display (or NULL to convert time instead) * \param[in] units Units to display (or NULL for no units) * \param[in] epoch_time Epoch time to convert if value is NULL */ static void print_nvpair(FILE *stream, const char *name, const char *value, const char *units, time_t epoch_time) { /* print name= */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as(" %s=", name); break; case mon_output_html: case mon_output_cgi: case mon_output_xml: fprintf(stream, " %s=", name); break; default: break; } /* If we have a value (and optionally units), print it */ if (value) { switch (output_format) { case mon_output_plain: case mon_output_console: print_as("%s%s", value, (units? units : "")); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "%s%s", value, (units? units : "")); break; case mon_output_xml: fprintf(stream, "\"%s%s\"", value, (units? units : "")); break; default: break; } /* Otherwise print user-friendly time string */ } else { static char empty_str[] = ""; char *c, *date_str = asctime(localtime(&epoch_time)); for (c = (date_str != NULL) ? date_str : empty_str; *c != '\0'; ++c) { if (*c == '\n') { *c = '\0'; break; } } switch (output_format) { case mon_output_plain: case mon_output_console: print_as("'%s'", date_str); break; case mon_output_html: case mon_output_cgi: case mon_output_xml: fprintf(stream, "\"%s\"", date_str); break; default: break; } } } /*! * \internal * \brief Print whatever is needed to start a node section * * \param[in] stream File stream to display output to * \param[in] node Node to print */ static void print_node_start(FILE *stream, node_t *node) { char *node_name; switch (output_format) { case mon_output_plain: case mon_output_console: node_name = get_node_display_name(node); print_as("* Node %s:\n", node_name); free(node_name); break; case mon_output_html: case mon_output_cgi: node_name = get_node_display_name(node); fprintf(stream, "

Node: %s

\n
    \n", node_name); free(node_name); break; case mon_output_xml: fprintf(stream, " \n", node->details->uname); break; default: break; } } /*! * \internal * \brief Print whatever is needed to end a node section * * \param[in] stream File stream to display output to */ static void print_node_end(FILE *stream) { switch (output_format) { case mon_output_html: case mon_output_cgi: fprintf(stream, "
\n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } /*! * \internal * \brief Print resources section heading appropriate to options * * \param[in] stream File stream to display output to */ static void print_resources_heading(FILE *stream) { const char *heading; if (group_by_node) { /* Active resources have already been printed by node */ heading = (inactive_resources? "Inactive resources" : NULL); } else if (inactive_resources) { heading = "Full list of resources"; } else { heading = "Active resources"; } /* Print section heading */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\n%s:\n\n", heading); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
\n

%s

\n", heading); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } /*! * \internal * \brief Print whatever resource section closing is appropriate * * \param[in] stream File stream to display output to */ static void print_resources_closing(FILE *stream, gboolean printed_heading) { const char *heading; /* What type of resources we did or did not display */ if (group_by_node) { heading = "inactive "; } else if (inactive_resources) { heading = ""; } else { heading = "active "; } switch (output_format) { case mon_output_plain: case mon_output_console: if (!printed_heading) { print_as("\nNo %sresources\n\n", heading); } break; case mon_output_html: case mon_output_cgi: if (!printed_heading) { fprintf(stream, "
\n

No %sresources

\n", heading); } break; case mon_output_xml: fprintf(stream, " %s\n", (printed_heading? "" : "")); break; default: break; } } /*! * \internal * \brief Print whatever resource section(s) are appropriate * * \param[in] stream File stream to display output to * \param[in] data_set Cluster state to display * \param[in] print_opts Bitmask of pe_print_options */ static void print_resources(FILE *stream, pe_working_set_t *data_set, int print_opts) { GListPtr rsc_iter; const char *prefix = NULL; gboolean printed_heading = FALSE; gboolean brief_output = print_brief; /* If we already showed active resources by node, and * we're not showing inactive resources, we have nothing to do */ if (group_by_node && !inactive_resources) { return; } /* XML uses an indent, and ignores brief option for resources */ if (output_format == mon_output_xml) { prefix = " "; brief_output = FALSE; } /* If we haven't already printed resources grouped by node, * and brief output was requested, print resource summary */ if (brief_output && !group_by_node) { print_resources_heading(stream); printed_heading = TRUE; print_rscs_brief(data_set->resources, NULL, print_opts, stream, inactive_resources); } /* For each resource, display it if appropriate */ for (rsc_iter = data_set->resources; rsc_iter != NULL; rsc_iter = rsc_iter->next) { resource_t *rsc = (resource_t *) rsc_iter->data; /* Complex resources may have some sub-resources active and some inactive */ gboolean is_active = rsc->fns->active(rsc, TRUE); gboolean partially_active = rsc->fns->active(rsc, FALSE); /* Skip inactive orphans (deleted but still in CIB) */ if (is_set(rsc->flags, pe_rsc_orphan) && !is_active) { continue; /* Skip active resources if we already displayed them by node */ } else if (group_by_node) { if (is_active) { continue; } /* Skip primitives already counted in a brief summary */ } else if (brief_output && (rsc->variant == pe_native)) { continue; /* Skip resources that aren't at least partially active, * unless we're displaying inactive resources */ } else if (!partially_active && !inactive_resources) { continue; } /* Print this resource */ if (printed_heading == FALSE) { print_resources_heading(stream); printed_heading = TRUE; } rsc->fns->print(rsc, prefix, print_opts, stream); } print_resources_closing(stream, printed_heading); } /*! * \internal * \brief Print heading for resource history * * \param[in] stream File stream to display output to * \param[in] data_set Current state of CIB * \param[in] node Node that ran this resource * \param[in] rsc Resource to print * \param[in] rsc_id ID of resource to print * \param[in] all Whether to print every resource or just failed ones */ static void print_rsc_history_start(FILE *stream, pe_working_set_t *data_set, node_t *node, resource_t *rsc, const char *rsc_id, gboolean all) { time_t last_failure = 0; int failcount = rsc? pe_get_failcount(node, rsc, &last_failure, pe_fc_default, NULL, data_set) : 0; if (!all && !failcount && (last_failure <= 0)) { return; } /* Print resource ID */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as(" %s:", rsc_id); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
  • %s:", rsc_id); break; case mon_output_xml: fprintf(stream, " 0)) { /* Print migration threshold */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as(" migration-threshold=%d", rsc->migration_threshold); break; case mon_output_html: case mon_output_cgi: fprintf(stream, " migration-threshold=%d", rsc->migration_threshold); break; case mon_output_xml: fprintf(stream, " orphan=\"false\" migration-threshold=\"%d\"", rsc->migration_threshold); break; default: break; } /* Print fail count if any */ if (failcount > 0) { switch (output_format) { case mon_output_plain: case mon_output_console: print_as(" " CRM_FAIL_COUNT_PREFIX "=%d", failcount); break; case mon_output_html: case mon_output_cgi: fprintf(stream, " " CRM_FAIL_COUNT_PREFIX "=%d", failcount); break; case mon_output_xml: fprintf(stream, " " CRM_FAIL_COUNT_PREFIX "=\"%d\"", failcount); break; default: break; } } /* Print last failure time if any */ if (last_failure > 0) { print_nvpair(stream, CRM_LAST_FAILURE_PREFIX, NULL, NULL, last_failure); } } /* End the heading */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "\n
      \n"); break; case mon_output_xml: fprintf(stream, ">\n"); break; default: break; } } /*! * \internal * \brief Print closing for resource history * * \param[in] stream File stream to display output to */ static void print_rsc_history_end(FILE *stream) { switch (output_format) { case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n
    • \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } /*! * \internal * \brief Print operation history * * \param[in] stream File stream to display output to * \param[in] data_set Current state of CIB * \param[in] node Node this operation is for * \param[in] xml_op Root of XML tree describing this operation * \param[in] task Task parsed from this operation's XML * \param[in] interval_ms_s Interval parsed from this operation's XML * \param[in] rc Return code parsed from this operation's XML */ static void print_op_history(FILE *stream, pe_working_set_t *data_set, node_t *node, xmlNode *xml_op, const char *task, const char *interval_ms_s, int rc) { const char *value = NULL; const char *call = crm_element_value(xml_op, XML_LRM_ATTR_CALLID); /* Begin the operation description */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as(" + (%s) %s:", call, task); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
  • (%s) %s:", call, task); break; case mon_output_xml: fprintf(stream, " 0) { print_nvpair(stream, attr, NULL, NULL, int_value); } } attr = XML_RSC_OP_LAST_RUN; value = crm_element_value(xml_op, attr); if (value) { int_value = crm_parse_int(value, NULL); if (int_value > 0) { print_nvpair(stream, attr, NULL, NULL, int_value); } } attr = XML_RSC_OP_T_EXEC; value = crm_element_value(xml_op, attr); if (value) { print_nvpair(stream, attr, value, "ms", 0); } attr = XML_RSC_OP_T_QUEUE; value = crm_element_value(xml_op, attr); if (value) { print_nvpair(stream, attr, value, "ms", 0); } } /* End the operation description */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as(" rc=%d (%s)\n", rc, services_ocf_exitcode_str(rc)); break; case mon_output_html: case mon_output_cgi: fprintf(stream, " rc=%d (%s)
    • \n", rc, services_ocf_exitcode_str(rc)); break; case mon_output_xml: fprintf(stream, " rc=\"%d\" rc_text=\"%s\" />\n", rc, services_ocf_exitcode_str(rc)); break; default: break; } } /*! * \internal * \brief Print resource operation/failure history * * \param[in] stream File stream to display output to * \param[in] data_set Current state of CIB * \param[in] node Node that ran this resource * \param[in] rsc_entry Root of XML tree describing resource status * \param[in] operations Whether to print operations or just failcounts */ static void print_rsc_history(FILE *stream, pe_working_set_t *data_set, node_t *node, xmlNode *rsc_entry, gboolean operations) { GListPtr gIter = NULL; GListPtr op_list = NULL; gboolean printed = FALSE; const char *rsc_id = crm_element_value(rsc_entry, XML_ATTR_ID); resource_t *rsc = pe_find_resource(data_set->resources, rsc_id); xmlNode *rsc_op = NULL; /* If we're not showing operations, just print the resource failure summary */ if (operations == FALSE) { print_rsc_history_start(stream, data_set, node, rsc, rsc_id, FALSE); print_rsc_history_end(stream); return; } /* Create a list of this resource's operations */ for (rsc_op = __xml_first_child(rsc_entry); rsc_op != NULL; rsc_op = __xml_next(rsc_op)) { if (crm_str_eq((const char *)rsc_op->name, XML_LRM_TAG_RSC_OP, TRUE)) { op_list = g_list_append(op_list, rsc_op); } } op_list = g_list_sort(op_list, sort_op_by_callid); /* Print each operation */ for (gIter = op_list; gIter != NULL; gIter = gIter->next) { xmlNode *xml_op = (xmlNode *) gIter->data; const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); const char *interval_ms_s = crm_element_value(xml_op, XML_LRM_ATTR_INTERVAL_MS); const char *op_rc = crm_element_value(xml_op, XML_LRM_ATTR_RC); int rc = crm_parse_int(op_rc, "0"); /* Display 0-interval monitors as "probe" */ if (safe_str_eq(task, CRMD_ACTION_STATUS) && ((interval_ms_s == NULL) || safe_str_eq(interval_ms_s, "0"))) { task = "probe"; } /* Ignore notifies and some probes */ if (safe_str_eq(task, CRMD_ACTION_NOTIFY) || (safe_str_eq(task, "probe") && (rc == 7))) { continue; } /* If this is the first printed operation, print heading for resource */ if (printed == FALSE) { printed = TRUE; print_rsc_history_start(stream, data_set, node, rsc, rsc_id, TRUE); } /* Print the operation */ print_op_history(stream, data_set, node, xml_op, task, interval_ms_s, rc); } /* Free the list we created (no need to free the individual items) */ g_list_free(op_list); /* If we printed anything, close the resource */ if (printed) { print_rsc_history_end(stream); } } /*! * \internal * \brief Print node operation/failure history * * \param[in] stream File stream to display output to * \param[in] data_set Current state of CIB * \param[in] node_state Root of XML tree describing node status * \param[in] operations Whether to print operations or just failcounts */ static void print_node_history(FILE *stream, pe_working_set_t *data_set, xmlNode *node_state, gboolean operations) { node_t *node = pe_find_node_id(data_set->nodes, ID(node_state)); xmlNode *lrm_rsc = NULL; xmlNode *rsc_entry = NULL; if (node && node->details && node->details->online) { print_node_start(stream, node); lrm_rsc = find_xml_node(node_state, XML_CIB_TAG_LRM, FALSE); lrm_rsc = find_xml_node(lrm_rsc, XML_LRM_TAG_RESOURCES, FALSE); /* Print history of each of the node's resources */ for (rsc_entry = __xml_first_child(lrm_rsc); rsc_entry != NULL; rsc_entry = __xml_next(rsc_entry)) { if (crm_str_eq((const char *)rsc_entry->name, XML_LRM_TAG_RESOURCE, TRUE)) { print_rsc_history(stream, data_set, node, rsc_entry, operations); } } print_node_end(stream); } } /*! * \internal * \brief Print extended information about an attribute if appropriate * * \param[in] data_set Working set of CIB state * * \return TRUE if extended information was printed, FALSE otherwise * \note Currently, extended information is only supported for ping/pingd * resources, for which a message will be printed if connectivity is lost * or degraded. */ static gboolean print_attr_msg(FILE *stream, node_t * node, GListPtr rsc_list, const char *attrname, const char *attrvalue) { GListPtr gIter = NULL; for (gIter = rsc_list; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; const char *type = g_hash_table_lookup(rsc->meta, "type"); if (rsc->children != NULL) { if (print_attr_msg(stream, node, rsc->children, attrname, attrvalue)) { return TRUE; } } if (safe_str_eq(type, "ping") || safe_str_eq(type, "pingd")) { const char *name = g_hash_table_lookup(rsc->parameters, "name"); if (name == NULL) { name = "pingd"; } /* To identify the resource with the attribute name. */ if (safe_str_eq(name, attrname)) { int host_list_num = 0; int expected_score = 0; int value = crm_parse_int(attrvalue, "0"); const char *hosts = g_hash_table_lookup(rsc->parameters, "host_list"); const char *multiplier = g_hash_table_lookup(rsc->parameters, "multiplier"); if(hosts) { char **host_list = g_strsplit(hosts, " ", 0); host_list_num = g_strv_length(host_list); g_strfreev(host_list); } /* pingd multiplier is the same as the default value. */ expected_score = host_list_num * crm_parse_int(multiplier, "1"); switch (output_format) { case mon_output_plain: case mon_output_console: if (value <= 0) { print_as("\t: Connectivity is lost"); } else if (value < expected_score) { print_as("\t: Connectivity is degraded (Expected=%d)", expected_score); } break; case mon_output_html: case mon_output_cgi: if (value <= 0) { fprintf(stream, " (connectivity is lost)"); } else if (value < expected_score) { fprintf(stream, " (connectivity is degraded -- expected %d)", expected_score); } break; case mon_output_xml: fprintf(stream, " expected=\"%d\"", expected_score); break; default: break; } return TRUE; } } } return FALSE; } static int compare_attribute(gconstpointer a, gconstpointer b) { int rc; rc = strcmp((const char *)a, (const char *)b); return rc; } static void create_attr_list(gpointer name, gpointer value, gpointer data) { int i; const char *filt_str[] = FILTER_STR; CRM_CHECK(name != NULL, return); /* filtering automatic attributes */ for (i = 0; filt_str[i] != NULL; i++) { if (g_str_has_prefix(name, filt_str[i])) { return; } } attr_list = g_list_insert_sorted(attr_list, name, compare_attribute); } /* structure for passing multiple user data to g_list_foreach() */ struct mon_attr_data { FILE *stream; node_t *node; }; static void print_node_attribute(gpointer name, gpointer user_data) { const char *value = NULL; struct mon_attr_data *data = (struct mon_attr_data *) user_data; value = pe_node_attribute_raw(data->node, name); /* Print attribute name and value */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as(" + %-32s\t: %-10s", (char *)name, value); break; case mon_output_html: case mon_output_cgi: fprintf(data->stream, "
  • %s: %s", (char *)name, value); break; case mon_output_xml: fprintf(data->stream, " stream, data->node, data->node->details->running_rsc, name, value); /* Close out the attribute */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\n"); break; case mon_output_html: case mon_output_cgi: fprintf(data->stream, "
    • \n"); break; case mon_output_xml: fprintf(data->stream, " />\n"); break; default: break; } } static void print_node_summary(FILE *stream, pe_working_set_t * data_set, gboolean operations) { xmlNode *node_state = NULL; xmlNode *cib_status = get_object_root(XML_CIB_TAG_STATUS, data_set->input); /* Print heading */ switch (output_format) { case mon_output_plain: case mon_output_console: if (operations) { print_as("\nOperations:\n"); } else { print_as("\nMigration Summary:\n"); } break; case mon_output_html: case mon_output_cgi: if (operations) { fprintf(stream, "
    \n

    Operations

    \n"); } else { fprintf(stream, "
    \n

    Migration Summary

    \n"); } break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } /* Print each node in the CIB status */ for (node_state = __xml_first_child(cib_status); node_state != NULL; node_state = __xml_next(node_state)) { if (crm_str_eq((const char *)node_state->name, XML_CIB_TAG_STATE, TRUE)) { print_node_history(stream, data_set, node_state, operations); } } /* Close section */ switch (output_format) { case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } static void print_ticket(gpointer name, gpointer value, gpointer data) { ticket_t *ticket = (ticket_t *) value; FILE *stream = (FILE *) data; switch (output_format) { case mon_output_plain: case mon_output_console: print_as("* %s:\t%s%s", ticket->id, (ticket->granted? "granted" : "revoked"), (ticket->standby? " [standby]" : "")); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
  • %s: %s%s", ticket->id, (ticket->granted? "granted" : "revoked"), (ticket->standby? " [standby]" : "")); break; case mon_output_xml: fprintf(stream, " id, (ticket->granted? "granted" : "revoked"), (ticket->standby? "true" : "false")); break; default: break; } if (ticket->last_granted > -1) { print_nvpair(stdout, "last-granted", NULL, NULL, ticket->last_granted); } switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
    • \n"); break; case mon_output_xml: fprintf(stream, " />\n"); break; default: break; } } static void print_cluster_tickets(FILE *stream, pe_working_set_t * data_set) { /* Print section heading */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\nTickets:\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n

    Tickets

    \n
      \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } /* Print each ticket */ g_hash_table_foreach(data_set->tickets, print_ticket, stream); /* Close section */ switch (output_format) { case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } /*! * \internal * \brief Return human-friendly string representing node name * * The returned string will be in the format * uname[@hostUname] [(nodeID)] * "@hostUname" will be printed if the node is a guest node. * "(nodeID)" will be printed if the node ID is different from the node uname, * and detailed output has been requested. * * \param[in] node Node to represent * \return Newly allocated string with representation of node name * \note It is the caller's responsibility to free the result with free(). */ static char * get_node_display_name(node_t *node) { char *node_name; const char *node_host = NULL; const char *node_id = NULL; int name_len; CRM_ASSERT((node != NULL) && (node->details != NULL) && (node->details->uname != NULL)); /* Host is displayed only if this is a guest node */ if (pe__is_guest_node(node)) { node_t *host_node = pe__current_node(node->details->remote_rsc); if (host_node && host_node->details) { node_host = host_node->details->uname; } if (node_host == NULL) { node_host = ""; /* so we at least get "uname@" to indicate guest */ } } /* Node ID is displayed if different from uname and detail is requested */ if (print_clone_detail && safe_str_neq(node->details->uname, node->details->id)) { node_id = node->details->id; } /* Determine name length */ name_len = strlen(node->details->uname) + 1; if (node_host) { name_len += strlen(node_host) + 1; /* "@node_host" */ } if (node_id) { name_len += strlen(node_id) + 3; /* + " (node_id)" */ } /* Allocate and populate display name */ node_name = malloc(name_len); CRM_ASSERT(node_name != NULL); strcpy(node_name, node->details->uname); if (node_host) { strcat(node_name, "@"); strcat(node_name, node_host); } if (node_id) { strcat(node_name, " ("); strcat(node_name, node_id); strcat(node_name, ")"); } return node_name; } /*! * \internal * \brief Print a negative location constraint * * \param[in] stream File stream to display output to * \param[in] node Node affected by constraint * \param[in] location Constraint to print */ static void print_ban(FILE *stream, pe_node_t *node, pe__location_t *location) { char *node_name = NULL; switch (output_format) { case mon_output_plain: case mon_output_console: node_name = get_node_display_name(node); print_as(" %s\tprevents %s from running %son %s\n", location->id, location->rsc_lh->id, ((location->role_filter == RSC_ROLE_MASTER)? "as Master " : ""), node_name); break; case mon_output_html: case mon_output_cgi: node_name = get_node_display_name(node); fprintf(stream, "
  • %s prevents %s from running %son %s
    • \n", location->id, location->rsc_lh->id, ((location->role_filter == RSC_ROLE_MASTER)? "as Master " : ""), node_name); break; case mon_output_xml: fprintf(stream, " \n", location->id, location->rsc_lh->id, node->details->uname, node->weight, ((location->role_filter == RSC_ROLE_MASTER)? "true" : "false")); break; default: break; } free(node_name); } /*! * \internal * \brief Print section for negative location constraints * * \param[in] stream File stream to display output to * \param[in] data_set Working set corresponding to CIB status to display */ static void print_neg_locations(FILE *stream, pe_working_set_t *data_set) { GListPtr gIter, gIter2; /* Print section heading */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\nNegative Location Constraints:\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n

    Negative Location Constraints

    \n
      \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } /* Print each ban */ for (gIter = data_set->placement_constraints; gIter != NULL; gIter = gIter->next) { pe__location_t *location = gIter->data; if (!g_str_has_prefix(location->id, print_neg_location_prefix)) continue; for (gIter2 = location->node_list_rh; gIter2 != NULL; gIter2 = gIter2->next) { node_t *node = (node_t *) gIter2->data; if (node->weight < 0) { print_ban(stream, node, location); } } } /* Close section */ switch (output_format) { case mon_output_cgi: case mon_output_html: fprintf(stream, "
    \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } static void crm_mon_get_parameters(resource_t *rsc, pe_working_set_t * data_set) { get_rsc_attributes(rsc->parameters, rsc, NULL, data_set); crm_trace("Beekhof: unpacked params for %s (%d)", rsc->id, g_hash_table_size(rsc->parameters)); if(rsc->children) { GListPtr gIter = NULL; for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) { crm_mon_get_parameters(gIter->data, data_set); } } } /*! * \internal * \brief Print node attributes section * * \param[in] stream File stream to display output to * \param[in] data_set Working set of CIB state */ static void print_node_attributes(FILE *stream, pe_working_set_t *data_set) { GListPtr gIter = NULL; /* Print section heading */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\nNode Attributes:\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n

    Node Attributes

    \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } /* Unpack all resource parameters (it would be more efficient to do this * only when needed for the first time in print_attr_msg()) */ for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) { crm_mon_get_parameters(gIter->data, data_set); } /* Display each node's attributes */ for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { struct mon_attr_data data; data.stream = stream; data.node = (node_t *) gIter->data; if (data.node && data.node->details && data.node->details->online) { print_node_start(stream, data.node); g_hash_table_foreach(data.node->details->attrs, create_attr_list, NULL); g_list_foreach(attr_list, print_node_attribute, &data); g_list_free(attr_list); attr_list = NULL; print_node_end(stream); } } /* Print section footer */ switch (output_format) { case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } /*! * \internal * \brief Return resource display options corresponding to command-line choices * * \return Bitmask of pe_print_options suitable for resource print functions */ static int get_resource_display_options(void) { int print_opts; /* Determine basic output format */ switch (output_format) { case mon_output_console: print_opts = pe_print_ncurses; break; case mon_output_html: case mon_output_cgi: print_opts = pe_print_html; break; case mon_output_xml: print_opts = pe_print_xml; break; default: print_opts = pe_print_printf; break; } /* Add optional display elements */ if (print_pending) { print_opts |= pe_print_pending; } if (print_clone_detail) { print_opts |= pe_print_clone_details|pe_print_implicit; } if (!inactive_resources) { print_opts |= pe_print_clone_active; } if (print_brief) { print_opts |= pe_print_brief; } return print_opts; } /*! * \internal * \brief Return human-friendly string representing current time * * \return Current time as string (as by ctime() but without newline) on success * or "Could not determine current time" on error * \note The return value points to a statically allocated string which might be * overwritten by subsequent calls to any of the C library date and time functions. */ static const char * crm_now_string(void) { time_t a_time = time(NULL); char *since_epoch = ctime(&a_time); if ((a_time == (time_t) -1) || (since_epoch == NULL)) { return "Could not determine current time"; } since_epoch[strlen(since_epoch) - 1] = EOS; /* trim newline */ return (since_epoch); } /*! * \internal * \brief Print header for cluster summary if needed * * \param[in] stream File stream to display output to */ static void print_cluster_summary_header(FILE *stream) { switch (output_format) { case mon_output_html: case mon_output_cgi: fprintf(stream, "

    Cluster Summary

    \n

    \n"); break; case mon_output_xml: fprintf(stream, "

    \n"); break; default: break; } } /*! * \internal * \brief Print footer for cluster summary if needed * * \param[in] stream File stream to display output to */ static void print_cluster_summary_footer(FILE *stream) { switch (output_format) { case mon_output_cgi: case mon_output_html: fprintf(stream, "

    \n"); break; case mon_output_xml: fprintf(stream, "     \n"); break; default: break; } } /*! * \internal * \brief Print times the display was last updated and CIB last changed * * \param[in] stream File stream to display output to * \param[in] data_set Working set of CIB state */ static void print_cluster_times(FILE *stream, pe_working_set_t *data_set) { const char *last_written = crm_element_value(data_set->input, XML_CIB_ATTR_WRITTEN); const char *user = crm_element_value(data_set->input, XML_ATTR_UPDATE_USER); const char *client = crm_element_value(data_set->input, XML_ATTR_UPDATE_CLIENT); const char *origin = crm_element_value(data_set->input, XML_ATTR_UPDATE_ORIG); switch (output_format) { case mon_output_plain: case mon_output_console: print_as("Last updated: %s", crm_now_string()); print_as((user || client || origin)? "\n" : "\t\t"); print_as("Last change: %s", last_written ? last_written : ""); if (user) { print_as(" by %s", user); } if (client) { print_as(" via %s", client); } if (origin) { print_as(" on %s", origin); } print_as("\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, " Last updated: %s
    \n", crm_now_string()); fprintf(stream, " Last change: %s", last_written ? last_written : ""); if (user) { fprintf(stream, " by %s", user); } if (client) { fprintf(stream, " via %s", client); } if (origin) { fprintf(stream, " on %s", origin); } fprintf(stream, "
    \n"); break; case mon_output_xml: fprintf(stream, " \n", crm_now_string()); fprintf(stream, " \n", last_written ? last_written : "", user ? user : "", client ? client : "", origin ? origin : ""); break; default: break; } } /*! * \internal * \brief Print cluster stack * * \param[in] stream File stream to display output to * \param[in] stack_s Stack name */ static void print_cluster_stack(FILE *stream, const char *stack_s) { switch (output_format) { case mon_output_plain: case mon_output_console: print_as("Stack: %s\n", stack_s); break; case mon_output_html: case mon_output_cgi: fprintf(stream, " Stack: %s
    \n", stack_s); break; case mon_output_xml: fprintf(stream, " \n", stack_s); break; default: break; } } /*! * \internal * \brief Print current DC and its version * * \param[in] stream File stream to display output to * \param[in] data_set Working set of CIB state */ static void print_cluster_dc(FILE *stream, pe_working_set_t *data_set) { node_t *dc = data_set->dc_node; xmlNode *dc_version = get_xpath_object("//nvpair[@name='dc-version']", data_set->input, LOG_DEBUG); const char *dc_version_s = dc_version? crm_element_value(dc_version, XML_NVPAIR_ATTR_VALUE) : NULL; const char *quorum = crm_element_value(data_set->input, XML_ATTR_HAVE_QUORUM); char *dc_name = dc? get_node_display_name(dc) : NULL; switch (output_format) { case mon_output_plain: case mon_output_console: print_as("Current DC: "); if (dc) { print_as("%s (version %s) - partition %s quorum\n", dc_name, (dc_version_s? dc_version_s : "unknown"), (crm_is_true(quorum) ? "with" : "WITHOUT")); } else { print_as("NONE\n"); } break; case mon_output_html: case mon_output_cgi: fprintf(stream, " Current DC: "); if (dc) { fprintf(stream, "%s (version %s) - partition %s quorum", dc_name, (dc_version_s? dc_version_s : "unknown"), (crm_is_true(quorum)? "with" : "WITHOUT")); } else { fprintf(stream, "NONE"); } fprintf(stream, "
    \n"); break; case mon_output_xml: fprintf(stream, " details->uname, dc->details->id, (crm_is_true(quorum) ? "true" : "false")); } else { fprintf(stream, "present=\"false\""); } fprintf(stream, " />\n"); break; default: break; } free(dc_name); } /*! * \internal * \brief Print counts of configured nodes and resources * * \param[in] stream File stream to display output to * \param[in] data_set Working set of CIB state * \param[in] stack_s Stack name */ static void print_cluster_counts(FILE *stream, pe_working_set_t *data_set, const char *stack_s) { int nnodes = g_list_length(data_set->nodes); int nresources = count_resources(data_set, NULL); switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\n%d node%s configured\n", nnodes, s_if_plural(nnodes)); print_as("%d resource%s configured", nresources, s_if_plural(nresources)); if(data_set->disabled_resources || data_set->blocked_resources) { print_as(" ("); if (data_set->disabled_resources) { print_as("%d DISABLED", data_set->disabled_resources); } if (data_set->disabled_resources && data_set->blocked_resources) { print_as(", "); } if (data_set->blocked_resources) { print_as("%d BLOCKED from starting due to failure", data_set->blocked_resources); } print_as(")"); } print_as("\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, " %d node%s configured
    \n", nnodes, s_if_plural(nnodes)); fprintf(stream, " %d resource%s configured", nresources, s_if_plural(nresources)); if (data_set->disabled_resources || data_set->blocked_resources) { fprintf(stream, " ("); if (data_set->disabled_resources) { fprintf(stream, "%d DISABLED", data_set->disabled_resources); } if (data_set->disabled_resources && data_set->blocked_resources) { fprintf(stream, ", "); } if (data_set->blocked_resources) { fprintf(stream, "%d BLOCKED from starting due to failure", data_set->blocked_resources); } fprintf(stream, ")"); } fprintf(stream, "
    \n"); break; case mon_output_xml: fprintf(stream, " \n", g_list_length(data_set->nodes)); fprintf(stream, " \n", count_resources(data_set, NULL), data_set->disabled_resources, data_set->blocked_resources); break; default: break; } } /*! * \internal * \brief Print cluster-wide options * * \param[in] stream File stream to display output to * \param[in] data_set Working set of CIB state * * \note Currently this is only implemented for HTML and XML output, and * prints only a few options. If there is demand, more could be added. */ static void print_cluster_options(FILE *stream, pe_working_set_t *data_set) { switch (output_format) { case mon_output_plain: case mon_output_console: if (is_set(data_set->flags, pe_flag_maintenance_mode)) { print_as("\n *** Resource management is DISABLED ***"); print_as("\n The cluster will not attempt to start, stop or recover services"); print_as("\n"); } break; case mon_output_html: fprintf(stream, "

    \n

    Config Options

    \n"); fprintf(stream, " \n"); fprintf(stream, " \n", is_set(data_set->flags, pe_flag_stonith_enabled)? "enabled" : "disabled"); fprintf(stream, " \n", is_set(data_set->flags, pe_flag_symmetric_cluster)? "" : "a"); fprintf(stream, " \n"); fprintf(stream, " \n"); fprintf(stream, "
    STONITH of failed nodes%s
    Cluster is%ssymmetric
    No Quorum Policy"); switch (data_set->no_quorum_policy) { case no_quorum_freeze: fprintf(stream, "Freeze resources"); break; case no_quorum_stop: fprintf(stream, "Stop ALL resources"); break; case no_quorum_ignore: fprintf(stream, "Ignore"); break; case no_quorum_suicide: fprintf(stream, "Suicide"); break; } fprintf(stream, "
    Resource management"); if (is_set(data_set->flags, pe_flag_maintenance_mode)) { fprintf(stream, "DISABLED (the cluster will " "not attempt to start, stop or recover services)"); } else { fprintf(stream, "enabled"); } fprintf(stream, "
    \n

    \n"); break; case mon_output_xml: fprintf(stream, " flags, pe_flag_stonith_enabled)? "true" : "false"); fprintf(stream, " symmetric-cluster=\"%s\"", is_set(data_set->flags, pe_flag_symmetric_cluster)? "true" : "false"); fprintf(stream, " no-quorum-policy=\""); switch (data_set->no_quorum_policy) { case no_quorum_freeze: fprintf(stream, "freeze"); break; case no_quorum_stop: fprintf(stream, "stop"); break; case no_quorum_ignore: fprintf(stream, "ignore"); break; case no_quorum_suicide: fprintf(stream, "suicide"); break; } fprintf(stream, "\""); fprintf(stream, " maintenance-mode=\"%s\"", is_set(data_set->flags, pe_flag_maintenance_mode)? "true" : "false"); fprintf(stream, " />\n"); break; default: break; } } /*! * \internal * \brief Get the name of the stack in use (or "unknown" if not available) * * \param[in] data_set Working set of CIB state * * \return String representing stack name */ static const char * get_cluster_stack(pe_working_set_t *data_set) { xmlNode *stack = get_xpath_object("//nvpair[@name='cluster-infrastructure']", data_set->input, LOG_DEBUG); return stack? crm_element_value(stack, XML_NVPAIR_ATTR_VALUE) : "unknown"; } /*! * \internal * \brief Print a summary of cluster-wide information * * \param[in] stream File stream to display output to * \param[in] data_set Working set of CIB state */ static void print_cluster_summary(FILE *stream, pe_working_set_t *data_set) { const char *stack_s = get_cluster_stack(data_set); gboolean header_printed = FALSE; if (show & mon_show_stack) { if (header_printed == FALSE) { print_cluster_summary_header(stream); header_printed = TRUE; } print_cluster_stack(stream, stack_s); } /* Always print DC if none, even if not requested */ if ((data_set->dc_node == NULL) || (show & mon_show_dc)) { if (header_printed == FALSE) { print_cluster_summary_header(stream); header_printed = TRUE; } print_cluster_dc(stream, data_set); } if (show & mon_show_times) { if (header_printed == FALSE) { print_cluster_summary_header(stream); header_printed = TRUE; } print_cluster_times(stream, data_set); } if (is_set(data_set->flags, pe_flag_maintenance_mode) || data_set->disabled_resources || data_set->blocked_resources || is_set(show, mon_show_count)) { if (header_printed == FALSE) { print_cluster_summary_header(stream); header_printed = TRUE; } print_cluster_counts(stream, data_set, stack_s); } /* There is not a separate option for showing cluster options, so show with * stack for now; a separate option could be added if there is demand */ if (show & mon_show_stack) { print_cluster_options(stream, data_set); } if (header_printed) { print_cluster_summary_footer(stream); } } /*! * \internal * \brief Print a failed action * * \param[in] stream File stream to display output to * \param[in] xml_op Root of XML tree describing failed action */ static void print_failed_action(FILE *stream, xmlNode *xml_op) { const char *op_key = crm_element_value(xml_op, XML_LRM_ATTR_TASK_KEY); const char *op_key_attr = "op_key"; const char *last = crm_element_value(xml_op, XML_RSC_OP_LAST_CHANGE); const char *node = crm_element_value(xml_op, XML_ATTR_UNAME); const char *call = crm_element_value(xml_op, XML_LRM_ATTR_CALLID); const char *exit_reason = crm_element_value(xml_op, XML_LRM_ATTR_EXIT_REASON); int rc = crm_parse_int(crm_element_value(xml_op, XML_LRM_ATTR_RC), "0"); int status = crm_parse_int(crm_element_value(xml_op, XML_LRM_ATTR_OPSTATUS), "0"); char *exit_reason_cleaned; /* If no op_key was given, use id instead */ if (op_key == NULL) { op_key = ID(xml_op); op_key_attr = "id"; } /* If no exit reason was given, use "none" */ if (exit_reason == NULL) { exit_reason = "none"; } /* Print common action information */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("* %s on %s '%s' (%d): call=%s, status=%s, exitreason='%s'", op_key, node, services_ocf_exitcode_str(rc), rc, call, services_lrm_status_str(status), exit_reason); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "

  • %s on %s '%s' (%d): call=%s, status=%s, exitreason='%s'", op_key, node, services_ocf_exitcode_str(rc), rc, call, services_lrm_status_str(status), exit_reason); break; case mon_output_xml: exit_reason_cleaned = crm_xml_escape(exit_reason); fprintf(stream, " \n"); break; case mon_output_xml: fprintf(stream, " />\n"); break; default: break; } } /*! * \internal * \brief Print a section for failed actions * * \param[in] stream File stream to display output to * \param[in] data_set Working set of CIB state */ static void print_failed_actions(FILE *stream, pe_working_set_t *data_set) { xmlNode *xml_op = NULL; /* Print section heading */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\nFailed Resource Actions:\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n

    Failed Resource Actions

    \n
      \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } /* Print each failed action */ for (xml_op = __xml_first_child(data_set->failed); xml_op != NULL; xml_op = __xml_next(xml_op)) { print_failed_action(stream, xml_op); } /* End section */ switch (output_format) { case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } /*! * \internal * \brief Reduce the stonith-history * for successful actions we keep the last of every action-type & target * for failed actions we record as well who had failed * for actions in progress we keep full track * * \param[in] history List of stonith actions * */ static stonith_history_t * reduce_stonith_history(stonith_history_t *history) { stonith_history_t *new = NULL, *hp, *np, *tmp; for (hp = history; hp; ) { for (np = new; np; np = np->next) { if ((hp->state == st_done) || (hp->state == st_failed)) { /* action not in progress */ if (safe_str_eq(hp->target, np->target) && safe_str_eq(hp->action, np->action) && (hp->state == np->state)) { if ((hp->state == st_done) || safe_str_eq(hp->delegate, np->delegate)) { /* replace or purge */ if (hp->completed < np->completed) { /* purge older hp */ tmp = hp->next; hp->next = NULL; stonith_history_free(hp); hp = tmp; break; } /* damn single linked list */ free(hp->target); free(hp->action); free(np->origin); np->origin = hp->origin; free(np->delegate); np->delegate = hp->delegate; free(np->client); np->client = hp->client; np->completed = hp->completed; tmp = hp; hp = hp->next; free(tmp); break; } } if (np->next) { continue; } } np = 0; /* let outer loop progress hp */ break; } /* simply move hp from history to new */ if (np == NULL) { tmp = hp->next; hp->next = new; new = hp; hp = tmp; } } return new; } /*! * \internal * \brief Sort the stonith-history * sort by competed most current on the top * pending actions lacking a completed-stamp are gathered at the top * * \param[in] history List of stonith actions * */ static stonith_history_t * sort_stonith_history(stonith_history_t *history) { stonith_history_t *new = NULL, *pending = NULL, *hp, *np, *tmp; for (hp = history; hp; ) { tmp = hp->next; if ((hp->state == st_done) || (hp->state == st_failed)) { /* sort into new */ if ((!new) || (hp->completed > new->completed)) { hp->next = new; new = hp; } else { np = new; do { if ((!np->next) || (hp->completed > np->next->completed)) { hp->next = np->next; np->next = hp; break; } np = np->next; } while (1); } } else { /* put into pending */ hp->next = pending; pending = hp; } hp = tmp; } /* pending actions don't have a completed-stamp so make them go front */ if (pending) { stonith_history_t *last_pending = pending; while (last_pending->next) { last_pending = last_pending->next; } last_pending->next = new; new = pending; } return new; } /*! * \internal * \brief Print a stonith action * * \param[in] stream File stream to display output to * \param[in] event stonith event */ static void print_stonith_action(FILE *stream, stonith_history_t *event) { const char *action_s = stonith_action_str(event->action); char *run_at_s = ctime(&event->completed); if ((run_at_s) && (run_at_s[0] != 0)) { run_at_s[strlen(run_at_s)-1] = 0; /* Overwrite the newline */ } switch(output_format) { case mon_output_xml: fprintf(stream, " target, event->action); switch(event->state) { case st_done: fprintf(stream, " state=\"success\""); break; case st_failed: fprintf(stream, " state=\"failed\""); break; default: fprintf(stream, " state=\"pending\""); } fprintf(stream, " origin=\"%s\" client=\"%s\"", event->origin, event->client); if (event->delegate) { fprintf(stream, " delegate=\"%s\"", event->delegate); } switch(event->state) { case st_done: case st_failed: fprintf(stream, " completed=\"%s\"", run_at_s?run_at_s:""); break; default: break; } fprintf(stream, " />\n"); break; case mon_output_plain: case mon_output_console: switch(event->state) { case st_done: print_as("* %s of %s successful: delegate=%s, client=%s, origin=%s,\n" " %s='%s'\n", action_s, event->target, event->delegate ? event->delegate : "", event->client, event->origin, fence_full_history?"completed":"last-successful", run_at_s?run_at_s:""); break; case st_failed: print_as("* %s of %s failed: delegate=%s, client=%s, origin=%s,\n" " %s='%s'\n", action_s, event->target, event->delegate ? event->delegate : "", event->client, event->origin, fence_full_history?"completed":"last-failed", run_at_s?run_at_s:""); break; default: print_as("* %s of %s pending: client=%s, origin=%s\n", action_s, event->target, event->client, event->origin); } break; case mon_output_html: case mon_output_cgi: switch(event->state) { case st_done: fprintf(stream, "
  • %s of %s successful: delegate=%s, " "client=%s, origin=%s, %s='%s'
    • \n", action_s, event->target, event->delegate ? event->delegate : "", event->client, event->origin, fence_full_history?"completed":"last-successful", run_at_s?run_at_s:""); break; case st_failed: fprintf(stream, "
  • %s of %s failed: delegate=%s, " "client=%s, origin=%s, %s='%s'
    • \n", action_s, event->target, event->delegate ? event->delegate : "", event->client, event->origin, fence_full_history?"completed":"last-failed", run_at_s?run_at_s:""); break; default: fprintf(stream, "
  • %s of %s pending: client=%s, " "origin=%s
    • \n", action_s, event->target, event->client, event->origin); } break; default: /* no support for fence history for other formats so far */ break; } } /*! * \internal * \brief Print a section for failed stonith actions * * \param[in] stream File stream to display output to * \param[in] history List of stonith actions * */ static void print_failed_stonith_actions(FILE *stream, stonith_history_t *history) { stonith_history_t *hp; for (hp = history; hp; hp = hp->next) { if (hp->state == st_failed) { break; } } if (!hp) { return; } /* Print section heading */ switch (output_format) { /* no need to take care of xml in here as xml gets full * history anyway */ case mon_output_plain: case mon_output_console: print_as("\nFailed Fencing Actions:\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n

    Failed Fencing Actions

    \n
      \n"); break; default: break; } /* Print each failed stonith action */ for (hp = history; hp; hp = hp->next) { if (hp->state == st_failed) { print_stonith_action(stream, hp); } } /* End section */ switch (output_format) { case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n"); break; default: break; } } /*! * \internal * \brief Print pending stonith actions * * \param[in] stream File stream to display output to * \param[in] history List of stonith actions * */ static void print_stonith_pending(FILE *stream, stonith_history_t *history) { /* xml-output always shows the full history * so we'll never have to show pending-actions * separately */ if (history && (history->state != st_failed) && (history->state != st_done)) { stonith_history_t *hp; /* Print section heading */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\nPending Fencing Actions:\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n

    Pending Fencing Actions

    \n
      \n"); break; default: break; } for (hp = history; hp; hp = hp->next) { if ((hp->state == st_failed) || (hp->state == st_done)) { break; } print_stonith_action(stream, hp); } /* End section */ switch (output_format) { case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n"); break; default: break; } } } /*! * \internal * \brief Print a section for stonith-history * * \param[in] stream File stream to display output to * \param[in] history List of stonith actions * */ static void print_stonith_history(FILE *stream, stonith_history_t *history) { stonith_history_t *hp; /* Print section heading */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\nFencing History:\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n

    Fencing History

    \n
      \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } for (hp = history; hp; hp = hp->next) { if ((hp->state != st_failed) || (output_format == mon_output_xml)) { print_stonith_action(stream, hp); } } /* End section */ switch (output_format) { case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } /*! * \internal * \brief Print cluster status to screen * * This uses the global display preferences set by command-line options * to display cluster status in a human-friendly way. * * \param[in] data_set Working set of CIB state * \param[in] stonith_history List of stonith actions */ static void print_status(pe_working_set_t * data_set, stonith_history_t *stonith_history) { GListPtr gIter = NULL; int print_opts = get_resource_display_options(); /* space-separated lists of node names */ char *online_nodes = NULL; char *online_remote_nodes = NULL; char *online_guest_nodes = NULL; char *offline_nodes = NULL; char *offline_remote_nodes = NULL; if (output_format == mon_output_console) { blank_screen(); } print_cluster_summary(stdout, data_set); print_as("\n"); /* Gather node information (and print if in bad state or grouping by node) */ for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; const char *node_mode = NULL; char *node_name = get_node_display_name(node); /* Get node mode */ if (node->details->unclean) { if (node->details->online) { node_mode = "UNCLEAN (online)"; } else if (node->details->pending) { node_mode = "UNCLEAN (pending)"; } else { node_mode = "UNCLEAN (offline)"; } } else if (node->details->pending) { node_mode = "pending"; } else if (node->details->standby_onfail && node->details->online) { node_mode = "standby (on-fail)"; } else if (node->details->standby) { if (node->details->online) { if (node->details->running_rsc) { node_mode = "standby (with active resources)"; } else { node_mode = "standby"; } } else { node_mode = "OFFLINE (standby)"; } } else if (node->details->maintenance) { if (node->details->online) { node_mode = "maintenance"; } else { node_mode = "OFFLINE (maintenance)"; } } else if (node->details->online) { node_mode = "online"; if (group_by_node == FALSE) { if (pe__is_guest_node(node)) { online_guest_nodes = add_list_element(online_guest_nodes, node_name); } else if (pe__is_remote_node(node)) { online_remote_nodes = add_list_element(online_remote_nodes, node_name); } else { online_nodes = add_list_element(online_nodes, node_name); } free(node_name); continue; } } else { node_mode = "OFFLINE"; if (group_by_node == FALSE) { if (pe__is_remote_node(node)) { offline_remote_nodes = add_list_element(offline_remote_nodes, node_name); } else if (pe__is_guest_node(node)) { /* ignore offline guest nodes */ } else { offline_nodes = add_list_element(offline_nodes, node_name); } free(node_name); continue; } } /* If we get here, node is in bad state, or we're grouping by node */ /* Print the node name and status */ if (pe__is_guest_node(node)) { print_as("Guest"); } else if (pe__is_remote_node(node)) { print_as("Remote"); } print_as("Node %s: %s\n", node_name, node_mode); /* If we're grouping by node, print its resources */ if (group_by_node) { if (print_brief) { print_rscs_brief(node->details->running_rsc, "\t", print_opts | pe_print_rsconly, stdout, FALSE); } else { GListPtr gIter2 = NULL; for (gIter2 = node->details->running_rsc; gIter2 != NULL; gIter2 = gIter2->next) { resource_t *rsc = (resource_t *) gIter2->data; rsc->fns->print(rsc, "\t", print_opts | pe_print_rsconly, stdout); } } } free(node_name); } /* If we're not grouping by node, summarize nodes by status */ if (online_nodes) { print_as("Online: [%s ]\n", online_nodes); free(online_nodes); } if (offline_nodes) { print_as("OFFLINE: [%s ]\n", offline_nodes); free(offline_nodes); } if (online_remote_nodes) { print_as("RemoteOnline: [%s ]\n", online_remote_nodes); free(online_remote_nodes); } if (offline_remote_nodes) { print_as("RemoteOFFLINE: [%s ]\n", offline_remote_nodes); free(offline_remote_nodes); } if (online_guest_nodes) { print_as("GuestOnline: [%s ]\n", online_guest_nodes); free(online_guest_nodes); } /* Print resources section, if needed */ print_resources(stdout, data_set, print_opts); /* print Node Attributes section if requested */ if (show & mon_show_attributes) { print_node_attributes(stdout, data_set); } /* If requested, print resource operations (which includes failcounts) * or just failcounts */ if (show & (mon_show_operations | mon_show_failcounts)) { print_node_summary(stdout, data_set, ((show & mon_show_operations)? TRUE : FALSE)); } /* If there were any failed actions, print them */ if (xml_has_children(data_set->failed)) { print_failed_actions(stdout, data_set); } /* Print failed stonith actions */ if (fence_history) { print_failed_stonith_actions(stdout, stonith_history); } /* Print tickets if requested */ if (show & mon_show_tickets) { print_cluster_tickets(stdout, data_set); } /* Print negative location constraints if requested */ if (show & mon_show_bans) { print_neg_locations(stdout, data_set); } /* Print stonith history */ if (fence_history) { if (show & mon_show_fence_history) { print_stonith_history(stdout, stonith_history); } else { print_stonith_pending(stdout, stonith_history); } } #if CURSES_ENABLED if (output_format == mon_output_console) { refresh(); } #endif } /*! * \internal * \brief Print cluster status in XML format * * \param[in] data_set Working set of CIB state */ static void print_xml_status(pe_working_set_t * data_set, stonith_history_t *stonith_history) { FILE *stream = stdout; GListPtr gIter = NULL; int print_opts = get_resource_display_options(); fprintf(stream, "\n"); fprintf(stream, "\n", VERSION); print_cluster_summary(stream, data_set); /*** NODES ***/ fprintf(stream, " \n"); for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; const char *node_type = "unknown"; switch (node->details->type) { case node_member: node_type = "member"; break; case node_remote: node_type = "remote"; break; case node_ping: node_type = "ping"; break; } fprintf(stream, " details->uname); fprintf(stream, "id=\"%s\" ", node->details->id); fprintf(stream, "online=\"%s\" ", node->details->online ? "true" : "false"); fprintf(stream, "standby=\"%s\" ", node->details->standby ? "true" : "false"); fprintf(stream, "standby_onfail=\"%s\" ", node->details->standby_onfail ? "true" : "false"); fprintf(stream, "maintenance=\"%s\" ", node->details->maintenance ? "true" : "false"); fprintf(stream, "pending=\"%s\" ", node->details->pending ? "true" : "false"); fprintf(stream, "unclean=\"%s\" ", node->details->unclean ? "true" : "false"); fprintf(stream, "shutdown=\"%s\" ", node->details->shutdown ? "true" : "false"); fprintf(stream, "expected_up=\"%s\" ", node->details->expected_up ? "true" : "false"); fprintf(stream, "is_dc=\"%s\" ", node->details->is_dc ? "true" : "false"); fprintf(stream, "resources_running=\"%d\" ", g_list_length(node->details->running_rsc)); fprintf(stream, "type=\"%s\" ", node_type); if (pe__is_guest_node(node)) { fprintf(stream, "id_as_resource=\"%s\" ", node->details->remote_rsc->container->id); } if (group_by_node) { GListPtr lpc2 = NULL; fprintf(stream, ">\n"); for (lpc2 = node->details->running_rsc; lpc2 != NULL; lpc2 = lpc2->next) { resource_t *rsc = (resource_t *) lpc2->data; rsc->fns->print(rsc, " ", print_opts | pe_print_rsconly, stream); } fprintf(stream, " \n"); } else { fprintf(stream, "/>\n"); } } fprintf(stream, " \n"); /* Print resources section, if needed */ print_resources(stream, data_set, print_opts); /* print Node Attributes section if requested */ if (show & mon_show_attributes) { print_node_attributes(stream, data_set); } /* If requested, print resource operations (which includes failcounts) * or just failcounts */ if (show & (mon_show_operations | mon_show_failcounts)) { print_node_summary(stream, data_set, ((show & mon_show_operations)? TRUE : FALSE)); } /* If there were any failed actions, print them */ if (xml_has_children(data_set->failed)) { print_failed_actions(stream, data_set); } /* Print stonith history */ if (fence_history) { print_stonith_history(stdout, stonith_history); } /* Print tickets if requested */ if (show & mon_show_tickets) { print_cluster_tickets(stream, data_set); } /* Print negative location constraints if requested */ if (show & mon_show_bans) { print_neg_locations(stream, data_set); } fprintf(stream, "\n"); fflush(stream); fclose(stream); } /*! * \internal * \brief Print cluster status in HTML format (with HTTP headers if CGI) * * \param[in] data_set Working set of CIB state * \param[in] filename Name of file to write HTML to (ignored if CGI) * * \return 0 on success, -1 on error */ static int print_html_status(pe_working_set_t * data_set, const char *filename, stonith_history_t *stonith_history) { FILE *stream; GListPtr gIter = NULL; char *filename_tmp = NULL; int print_opts = get_resource_display_options(); if (output_format == mon_output_cgi) { stream = stdout; fprintf(stream, "Content-Type: text/html\n\n"); } else { filename_tmp = crm_concat(filename, "tmp", '.'); stream = fopen(filename_tmp, "w"); if (stream == NULL) { crm_perror(LOG_ERR, "Cannot open %s for writing", filename_tmp); free(filename_tmp); return -1; } } fprintf(stream, "\n"); fprintf(stream, " \n"); fprintf(stream, " Cluster status\n"); fprintf(stream, " \n", reconnect_msec / 1000); fprintf(stream, " \n"); fprintf(stream, "\n"); print_cluster_summary(stream, data_set); /*** NODE LIST ***/ fprintf(stream, "
    \n

    Node List

    \n"); fprintf(stream, "
      \n"); for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; char *node_name = get_node_display_name(node); fprintf(stream, "
    • Node: %s: ", node_name); if (node->details->standby_onfail && node->details->online) { fprintf(stream, "standby (on-fail)\n"); } else if (node->details->standby && node->details->online) { fprintf(stream, "standby%s\n", node->details->running_rsc?" (with active resources)":""); } else if (node->details->standby) { fprintf(stream, "OFFLINE (standby)\n"); } else if (node->details->maintenance && node->details->online) { fprintf(stream, "maintenance\n"); } else if (node->details->maintenance) { fprintf(stream, "OFFLINE (maintenance)\n"); } else if (node->details->online) { fprintf(stream, "online\n"); } else { fprintf(stream, "OFFLINE\n"); } if (print_brief && group_by_node) { fprintf(stream, "
        \n"); print_rscs_brief(node->details->running_rsc, NULL, print_opts | pe_print_rsconly, stream, FALSE); fprintf(stream, "
      \n"); } else if (group_by_node) { GListPtr lpc2 = NULL; fprintf(stream, "
        \n"); for (lpc2 = node->details->running_rsc; lpc2 != NULL; lpc2 = lpc2->next) { resource_t *rsc = (resource_t *) lpc2->data; fprintf(stream, "
      • "); rsc->fns->print(rsc, NULL, print_opts | pe_print_rsconly, stream); fprintf(stream, "
        • \n"); } fprintf(stream, "
      \n"); } fprintf(stream, "
      • \n"); free(node_name); } fprintf(stream, "
    \n"); /* Print resources section, if needed */ print_resources(stream, data_set, print_opts); /* print Node Attributes section if requested */ if (show & mon_show_attributes) { print_node_attributes(stream, data_set); } /* If requested, print resource operations (which includes failcounts) * or just failcounts */ if (show & (mon_show_operations | mon_show_failcounts)) { print_node_summary(stream, data_set, ((show & mon_show_operations)? TRUE : FALSE)); } /* If there were any failed actions, print them */ if (xml_has_children(data_set->failed)) { print_failed_actions(stream, data_set); } /* Print failed stonith actions */ if (fence_history) { print_failed_stonith_actions(stream, stonith_history); } /* Print stonith history */ if (fence_history) { if (show & mon_show_fence_history) { print_stonith_history(stream, stonith_history); } else { print_stonith_pending(stdout, stonith_history); } } /* Print tickets if requested */ if (show & mon_show_tickets) { print_cluster_tickets(stream, data_set); } /* Print negative location constraints if requested */ if (show & mon_show_bans) { print_neg_locations(stream, data_set); } fprintf(stream, "\n"); fprintf(stream, "\n"); fflush(stream); fclose(stream); if (output_format != mon_output_cgi) { if (rename(filename_tmp, filename) != 0) { crm_perror(LOG_ERR, "Unable to rename %s->%s", filename_tmp, filename); } free(filename_tmp); } return 0; } static int send_custom_trap(const char *node, const char *rsc, const char *task, int target_rc, int rc, int status, const char *desc) { pid_t pid; /*setenv needs chars, these are ints */ char *rc_s = crm_itoa(rc); char *status_s = crm_itoa(status); char *target_rc_s = crm_itoa(target_rc); crm_debug("Sending external notification to '%s' via '%s'", external_recipient, external_agent); if(rsc) { setenv("CRM_notify_rsc", rsc, 1); } if (external_recipient) { setenv("CRM_notify_recipient", external_recipient, 1); } setenv("CRM_notify_node", node, 1); setenv("CRM_notify_task", task, 1); setenv("CRM_notify_desc", desc, 1); setenv("CRM_notify_rc", rc_s, 1); setenv("CRM_notify_target_rc", target_rc_s, 1); setenv("CRM_notify_status", status_s, 1); pid = fork(); if (pid == -1) { crm_perror(LOG_ERR, "notification fork() failed."); } if (pid == 0) { /* crm_debug("notification: I am the child. Executing the nofitication program."); */ execl(external_agent, external_agent, NULL); exit(CRM_EX_ERROR); } crm_trace("Finished running custom notification program '%s'.", external_agent); free(target_rc_s); free(status_s); free(rc_s); return 0; } static void handle_rsc_op(xmlNode * xml, const char *node_id) { int rc = -1; int status = -1; - int action = -1; int target_rc = -1; - int transition_num = -1; gboolean notify = TRUE; char *rsc = NULL; char *task = NULL; const char *desc = NULL; const char *magic = NULL; const char *id = NULL; - char *update_te_uuid = NULL; const char *node = NULL; xmlNode *n = xml; xmlNode * rsc_op = xml; if(strcmp((const char*)xml->name, XML_LRM_TAG_RSC_OP) != 0) { xmlNode *cIter; for(cIter = xml->children; cIter; cIter = cIter->next) { handle_rsc_op(cIter, node_id); } return; } id = crm_element_value(rsc_op, XML_LRM_ATTR_TASK_KEY); if (id == NULL) { /* Compatibility with <= 1.1.5 */ id = ID(rsc_op); } magic = crm_element_value(rsc_op, XML_ATTR_TRANSITION_MAGIC); if (magic == NULL) { /* non-change */ return; } - if (FALSE == decode_transition_magic(magic, &update_te_uuid, &transition_num, &action, - &status, &rc, &target_rc)) { + if (!decode_transition_magic(magic, NULL, NULL, NULL, &status, &rc, + &target_rc)) { crm_err("Invalid event %s detected for %s", magic, id); return; } if (parse_op_key(id, &rsc, &task, NULL) == FALSE) { crm_err("Invalid event detected for %s", id); goto bail; } node = crm_element_value(rsc_op, XML_LRM_ATTR_TARGET); while (n != NULL && safe_str_neq(XML_CIB_TAG_STATE, TYPE(n))) { n = n->parent; } if(node == NULL && n) { node = crm_element_value(n, XML_ATTR_UNAME); } if (node == NULL && n) { node = ID(n); } if (node == NULL) { node = node_id; } if (node == NULL) { crm_err("No node detected for event %s (%s)", magic, id); goto bail; } /* look up where we expected it to be? */ desc = pcmk_strerror(pcmk_ok); if (status == PCMK_LRM_OP_DONE && target_rc == rc) { crm_notice("%s of %s on %s completed: %s", task, rsc, node, desc); if (rc == PCMK_OCF_NOT_RUNNING) { notify = FALSE; } } else if (status == PCMK_LRM_OP_DONE) { desc = services_ocf_exitcode_str(rc); crm_warn("%s of %s on %s failed: %s", task, rsc, node, desc); } else { desc = services_lrm_status_str(status); crm_warn("%s of %s on %s failed: %s", task, rsc, node, desc); } if (notify && external_agent) { send_custom_trap(node, rsc, task, target_rc, rc, status, desc); } bail: - free(update_te_uuid); free(rsc); free(task); } static gboolean mon_trigger_refresh(gpointer user_data) { mainloop_set_trigger(refresh_trigger); return FALSE; } #define NODE_PATT "/lrm[@id=" static char * get_node_from_xpath(const char *xpath) { char *nodeid = NULL; char *tmp = strstr(xpath, NODE_PATT); if(tmp) { tmp += strlen(NODE_PATT); tmp += 1; nodeid = strdup(tmp); tmp = strstr(nodeid, "\'"); CRM_ASSERT(tmp); tmp[0] = 0; } return nodeid; } static void crm_diff_update_v2(const char *event, xmlNode * msg) { xmlNode *change = NULL; xmlNode *diff = get_message_xml(msg, F_CIB_UPDATE_RESULT); for (change = __xml_first_child(diff); change != NULL; change = __xml_next(change)) { const char *name = NULL; const char *op = crm_element_value(change, XML_DIFF_OP); const char *xpath = crm_element_value(change, XML_DIFF_PATH); xmlNode *match = NULL; const char *node = NULL; if(op == NULL) { continue; } else if(strcmp(op, "create") == 0) { match = change->children; } else if(strcmp(op, "move") == 0) { continue; } else if(strcmp(op, "delete") == 0) { continue; } else if(strcmp(op, "modify") == 0) { match = first_named_child(change, XML_DIFF_RESULT); if(match) { match = match->children; } } if(match) { name = (const char *)match->name; } crm_trace("Handling %s operation for %s %p, %s", op, xpath, match, name); if(xpath == NULL) { /* Version field, ignore */ } else if(name == NULL) { crm_debug("No result for %s operation to %s", op, xpath); CRM_ASSERT(strcmp(op, "delete") == 0 || strcmp(op, "move") == 0); } else if(strcmp(name, XML_TAG_CIB) == 0) { xmlNode *state = NULL; xmlNode *status = first_named_child(match, XML_CIB_TAG_STATUS); for (state = __xml_first_child(status); state != NULL; state = __xml_next(state)) { node = crm_element_value(state, XML_ATTR_UNAME); if (node == NULL) { node = ID(state); } handle_rsc_op(state, node); } } else if(strcmp(name, XML_CIB_TAG_STATUS) == 0) { xmlNode *state = NULL; for (state = __xml_first_child(match); state != NULL; state = __xml_next(state)) { node = crm_element_value(state, XML_ATTR_UNAME); if (node == NULL) { node = ID(state); } handle_rsc_op(state, node); } } else if(strcmp(name, XML_CIB_TAG_STATE) == 0) { node = crm_element_value(match, XML_ATTR_UNAME); if (node == NULL) { node = ID(match); } handle_rsc_op(match, node); } else if(strcmp(name, XML_CIB_TAG_LRM) == 0) { node = ID(match); handle_rsc_op(match, node); } else if(strcmp(name, XML_LRM_TAG_RESOURCES) == 0) { char *local_node = get_node_from_xpath(xpath); handle_rsc_op(match, local_node); free(local_node); } else if(strcmp(name, XML_LRM_TAG_RESOURCE) == 0) { char *local_node = get_node_from_xpath(xpath); handle_rsc_op(match, local_node); free(local_node); } else if(strcmp(name, XML_LRM_TAG_RSC_OP) == 0) { char *local_node = get_node_from_xpath(xpath); handle_rsc_op(match, local_node); free(local_node); } else { crm_trace("Ignoring %s operation for %s %p, %s", op, xpath, match, name); } } } static void crm_diff_update_v1(const char *event, xmlNode * msg) { /* Process operation updates */ xmlXPathObject *xpathObj = xpath_search(msg, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_LRM_TAG_RSC_OP); int lpc = 0, max = numXpathResults(xpathObj); for (lpc = 0; lpc < max; lpc++) { xmlNode *rsc_op = getXpathResult(xpathObj, lpc); handle_rsc_op(rsc_op, NULL); } freeXpathObject(xpathObj); } static void crm_diff_update(const char *event, xmlNode * msg) { int rc = -1; static bool stale = FALSE; gboolean cib_updated = FALSE; xmlNode *diff = get_message_xml(msg, F_CIB_UPDATE_RESULT); print_dot(); if (current_cib != NULL) { rc = xml_apply_patchset(current_cib, diff, TRUE); switch (rc) { case -pcmk_err_diff_resync: case -pcmk_err_diff_failed: crm_notice("[%s] Patch aborted: %s (%d)", event, pcmk_strerror(rc), rc); free_xml(current_cib); current_cib = NULL; break; case pcmk_ok: cib_updated = TRUE; break; default: crm_notice("[%s] ABORTED: %s (%d)", event, pcmk_strerror(rc), rc); free_xml(current_cib); current_cib = NULL; } } if (current_cib == NULL) { crm_trace("Re-requesting the full cib"); cib->cmds->query(cib, NULL, ¤t_cib, cib_scope_local | cib_sync_call); } if (external_agent) { int format = 0; crm_element_value_int(diff, "format", &format); switch(format) { case 1: crm_diff_update_v1(event, msg); break; case 2: crm_diff_update_v2(event, msg); break; default: crm_err("Unknown patch format: %d", format); } } if (current_cib == NULL) { if(!stale) { print_as("--- Stale data ---"); } stale = TRUE; return; } stale = FALSE; kick_refresh(cib_updated); } static gboolean mon_refresh_display(gpointer user_data) { xmlNode *cib_copy = copy_xml(current_cib); stonith_history_t *stonith_history = NULL; last_refresh = time(NULL); if (cli_config_update(&cib_copy, NULL, FALSE) == FALSE) { if (cib) { cib->cmds->signoff(cib); } print_as("Upgrade failed: %s", pcmk_strerror(-pcmk_err_schema_validation)); if (output_format == mon_output_console) { sleep(2); } clean_up(CRM_EX_CONFIG); return FALSE; } /* get the stonith-history if there is evidence we need it */ while (fence_history) { if (st != NULL) { if (st->cmds->history(st, st_opt_sync_call, NULL, &stonith_history, 120)) { fprintf(stderr, "Critical: Unable to get stonith-history\n"); mon_cib_connection_destroy(NULL); } else { if ((!fence_full_history) && (output_format != mon_output_xml)) { stonith_history = reduce_stonith_history(stonith_history); } stonith_history = sort_stonith_history(stonith_history); break; /* all other cases are errors */ } } else { fprintf(stderr, "Critical: No stonith-API\n"); } free_xml(cib_copy); print_as("Reading stonith-history failed"); if (output_format == mon_output_console) { sleep(2); } return FALSE; } if (mon_data_set == NULL) { mon_data_set = pe_new_working_set(); CRM_ASSERT(mon_data_set != NULL); } mon_data_set->input = cib_copy; cluster_status(mon_data_set); /* Unpack constraints if any section will need them * (tickets may be referenced in constraints but not granted yet, * and bans need negative location constraints) */ if (show & (mon_show_bans | mon_show_tickets)) { xmlNode *cib_constraints = get_object_root(XML_CIB_TAG_CONSTRAINTS, mon_data_set->input); unpack_constraints(cib_constraints, mon_data_set); } switch (output_format) { case mon_output_html: case mon_output_cgi: if (print_html_status(mon_data_set, output_filename, stonith_history) != 0) { fprintf(stderr, "Critical: Unable to output html file\n"); clean_up(CRM_EX_CANTCREAT); return FALSE; } break; case mon_output_xml: print_xml_status(mon_data_set, stonith_history); break; case mon_output_monitor: print_simple_status(mon_data_set, stonith_history); if (has_warnings) { clean_up(MON_STATUS_WARN); return FALSE; } break; case mon_output_plain: case mon_output_console: print_status(mon_data_set, stonith_history); break; case mon_output_none: break; } stonith_history_free(stonith_history); stonith_history = NULL; pe_reset_working_set(mon_data_set); return TRUE; } static void mon_st_callback_event(stonith_t * st, stonith_event_t * e) { if (st->state == stonith_disconnected) { /* disconnect cib as well and have everything reconnect */ mon_cib_connection_destroy(NULL); } else if (external_agent) { char *desc = crm_strdup_printf("Operation %s requested by %s for peer %s: %s (ref=%s)", e->operation, e->origin, e->target, pcmk_strerror(e->result), e->id); send_custom_trap(e->target, NULL, e->operation, pcmk_ok, e->result, 0, desc); free(desc); } } static void kick_refresh(gboolean data_updated) { static int updates = 0; long now = time(NULL); if (data_updated) { updates++; } if(refresh_timer == NULL) { refresh_timer = mainloop_timer_add("refresh", 2000, FALSE, mon_trigger_refresh, NULL); } /* Refresh * - immediately if the last update was more than 5s ago * - every 10 cib-updates * - at most 2s after the last update */ if ((now - last_refresh) > (reconnect_msec / 1000)) { mainloop_set_trigger(refresh_trigger); mainloop_timer_stop(refresh_timer); updates = 0; } else if(updates >= 10) { mainloop_set_trigger(refresh_trigger); mainloop_timer_stop(refresh_timer); updates = 0; } else { mainloop_timer_start(refresh_timer); } } static void mon_st_callback_display(stonith_t * st, stonith_event_t * e) { if (st->state == stonith_disconnected) { /* disconnect cib as well and have everything reconnect */ mon_cib_connection_destroy(NULL); } else { print_dot(); kick_refresh(TRUE); } } static void clean_up_connections(void) { if (cib != NULL) { cib->cmds->signoff(cib); cib_delete(cib); cib = NULL; } if (st != NULL) { if (st->state != stonith_disconnected) { st->cmds->remove_notification(st, T_STONITH_NOTIFY_DISCONNECT); st->cmds->remove_notification(st, T_STONITH_NOTIFY_FENCE); st->cmds->remove_notification(st, T_STONITH_NOTIFY_HISTORY); st->cmds->disconnect(st); } stonith_api_delete(st); st = NULL; } } /* * De-init ncurses, disconnect from the CIB manager, disconnect fencing, * deallocate memory and show usage-message if requested. * * We don't actually return, but nominally returning crm_exit_t allows a usage * like "return clean_up(exit_code);" which helps static analysis understand the * code flow. */ static crm_exit_t clean_up(crm_exit_t exit_code) { #if CURSES_ENABLED if (curses_console_initialized) { output_format = mon_output_plain; echo(); nocbreak(); endwin(); curses_console_initialized = FALSE; } #endif clean_up_connections(); free(output_filename); free(pid_file); pe_free_working_set(mon_data_set); mon_data_set = NULL; if (exit_code == CRM_EX_USAGE) { if (output_format == mon_output_cgi) { fprintf(stdout, "Content-Type: text/plain\n" "Status: 500\n\n"); } else { crm_help('?', CRM_EX_USAGE); } } crm_exit(exit_code); } diff --git a/tools/crm_resource.h b/tools/crm_resource.h index 487cc28cbf..48236b6695 100644 --- a/tools/crm_resource.h +++ b/tools/crm_resource.h @@ -1,103 +1,103 @@ /* - * Copyright 2004-2018 Andrew Beekhof + * Copyright 2004-2019 the Pacemaker project contributors + * + * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include extern bool print_pending; extern bool scope_master; extern bool do_force; extern bool BE_QUIET; extern int resource_verbose; extern int cib_options; extern int crmd_replies_needed; extern char *move_lifetime; extern const char *attr_set_type; /* ban */ int cli_resource_prefer(const char *rsc_id, const char *host, cib_t * cib_conn); int cli_resource_ban(const char *rsc_id, const char *host, GListPtr allnodes, cib_t * cib_conn); int cli_resource_clear(const char *rsc_id, const char *host, GListPtr allnodes, cib_t * cib_conn, bool clear_ban_constraints); int cli_resource_clear_all_expired(xmlNode *root, cib_t *cib_conn, const char *rsc, const char *node, bool scope_master); /* print */ void cli_resource_print_cts(resource_t * rsc); void cli_resource_print_raw(resource_t * rsc); void cli_resource_print_cts_constraints(pe_working_set_t * data_set); void cli_resource_print_location(resource_t * rsc, const char *prefix); void cli_resource_print_colocation(resource_t * rsc, bool dependents, bool recursive, int offset); int cli_resource_print(resource_t *rsc, pe_working_set_t *data_set, bool expanded); int cli_resource_print_list(pe_working_set_t * data_set, bool raw); int cli_resource_print_attribute(resource_t *rsc, const char *attr, pe_working_set_t *data_set); int cli_resource_print_property(resource_t *rsc, const char *attr, pe_working_set_t *data_set); int cli_resource_print_operations(const char *rsc_id, const char *host_uname, bool active, pe_working_set_t * data_set); /* runtime */ void cli_resource_check(cib_t * cib, resource_t *rsc); int cli_resource_fail(crm_ipc_t * crmd_channel, const char *host_uname, const char *rsc_id, pe_working_set_t * data_set); int cli_resource_search(resource_t *rsc, const char *requested_name, pe_working_set_t *data_set); int cli_resource_delete(crm_ipc_t *crmd_channel, const char *host_uname, resource_t *rsc, const char *operation, const char *interval_spec, bool just_failures, pe_working_set_t *data_set); int cli_cleanup_all(crm_ipc_t *crmd_channel, const char *node_name, const char *operation, const char *interval_spec, pe_working_set_t *data_set); int cli_resource_restart(pe_resource_t *rsc, const char *host, int timeout_ms, cib_t *cib); int cli_resource_move(resource_t *rsc, const char *rsc_id, const char *host_name, cib_t *cib, pe_working_set_t *data_set); int cli_resource_execute_from_params(const char *rsc_name, const char *rsc_class, const char *rsc_prov, const char *rsc_type, const char *rsc_action, GHashTable *params, GHashTable *override_hash, int timeout_ms); int cli_resource_execute(resource_t *rsc, const char *requested_name, const char *rsc_action, GHashTable *override_hash, int timeout_ms, cib_t *cib, pe_working_set_t *data_set); int cli_resource_update_attribute(resource_t *rsc, const char *requested_name, const char *attr_set, const char *attr_id, const char *attr_name, const char *attr_value, bool recursive, cib_t *cib, pe_working_set_t *data_set); int cli_resource_delete_attribute(resource_t *rsc, const char *requested_name, const char *attr_set, const char *attr_id, const char *attr_name, cib_t *cib, pe_working_set_t *data_set); GList* subtract_lists(GList *from, GList *items, GCompareFunc cmp); int update_working_set_xml(pe_working_set_t *data_set, xmlNode **xml); int wait_till_stable(int timeout_ms, cib_t * cib); void cli_resource_why(cib_t *cib_conn, GListPtr resources, resource_t *rsc, node_t *node); - -extern xmlNode *do_calculations(pe_working_set_t * data_set, xmlNode * xml_input, crm_time_t * now); diff --git a/tools/crm_resource_runtime.c b/tools/crm_resource_runtime.c index 5c9fcf74f3..c4ff84c143 100644 --- a/tools/crm_resource_runtime.c +++ b/tools/crm_resource_runtime.c @@ -1,2049 +1,2049 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include int resource_verbose = 0; bool do_force = FALSE; int crmd_replies_needed = 1; /* The welcome message */ const char *attr_set_type = XML_TAG_ATTR_SETS; static int do_find_resource(const char *rsc, resource_t * the_rsc, pe_working_set_t * data_set) { int found = 0; GListPtr lpc = NULL; for (lpc = the_rsc->running_on; lpc != NULL; lpc = lpc->next) { node_t *node = (node_t *) lpc->data; if (BE_QUIET) { fprintf(stdout, "%s\n", node->details->uname); } else { const char *state = ""; if (!pe_rsc_is_clone(the_rsc) && the_rsc->fns->state(the_rsc, TRUE) == RSC_ROLE_MASTER) { state = "Master"; } fprintf(stdout, "resource %s is running on: %s %s\n", rsc, node->details->uname, state); } found++; } if (BE_QUIET == FALSE && found == 0) { fprintf(stderr, "resource %s is NOT running\n", rsc); } return found; } int cli_resource_search(resource_t *rsc, const char *requested_name, pe_working_set_t *data_set) { int found = 0; resource_t *parent = uber_parent(rsc); if (pe_rsc_is_clone(rsc)) { for (GListPtr iter = rsc->children; iter != NULL; iter = iter->next) { found += do_find_resource(requested_name, iter->data, data_set); } /* The anonymous clone children's common ID is supplied */ } else if (pe_rsc_is_clone(parent) && is_not_set(rsc->flags, pe_rsc_unique) && rsc->clone_name && safe_str_eq(requested_name, rsc->clone_name) && safe_str_neq(requested_name, rsc->id)) { for (GListPtr iter = parent->children; iter; iter = iter->next) { found += do_find_resource(requested_name, iter->data, data_set); } } else { found += do_find_resource(requested_name, rsc, data_set); } return found; } #define XPATH_MAX 1024 static int find_resource_attr(cib_t * the_cib, const char *attr, const char *rsc, const char *set_type, const char *set_name, const char *attr_id, const char *attr_name, char **value) { int offset = 0; int rc = pcmk_ok; xmlNode *xml_search = NULL; char *xpath_string = NULL; if(value) { *value = NULL; } if(the_cib == NULL) { return -ENOTCONN; } xpath_string = calloc(1, XPATH_MAX); offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "%s", get_object_path("resources")); offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "//*[@id=\"%s\"]", rsc); if (set_type) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "/%s", set_type); if (set_name) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "[@id=\"%s\"]", set_name); } } offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "//nvpair["); if (attr_id) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "@id=\"%s\"", attr_id); } if (attr_name) { if (attr_id) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, " and "); } offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "@name=\"%s\"", attr_name); } offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "]"); CRM_LOG_ASSERT(offset > 0); rc = the_cib->cmds->query(the_cib, xpath_string, &xml_search, cib_sync_call | cib_scope_local | cib_xpath); if (rc != pcmk_ok) { goto bail; } crm_log_xml_debug(xml_search, "Match"); if (xml_has_children(xml_search)) { xmlNode *child = NULL; rc = -EINVAL; printf("Multiple attributes match name=%s\n", attr_name); for (child = __xml_first_child(xml_search); child != NULL; child = __xml_next(child)) { printf(" Value: %s \t(id=%s)\n", crm_element_value(child, XML_NVPAIR_ATTR_VALUE), ID(child)); } } else if(value) { const char *tmp = crm_element_value(xml_search, attr); if (tmp) { *value = strdup(tmp); } } bail: free(xpath_string); free_xml(xml_search); return rc; } static resource_t * find_matching_attr_resource(resource_t * rsc, const char * rsc_id, const char * attr_set, const char * attr_id, const char * attr_name, cib_t * cib, const char * cmd) { int rc = pcmk_ok; char *lookup_id = NULL; char *local_attr_id = NULL; if(do_force == TRUE) { return rsc; } else if(rsc->parent) { switch(rsc->parent->variant) { case pe_group: if (BE_QUIET == FALSE) { printf("Performing %s of '%s' for '%s' will not apply to its peers in '%s'\n", cmd, attr_name, rsc_id, rsc->parent->id); } break; case pe_clone: rc = find_resource_attr(cib, XML_ATTR_ID, rsc_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); free(local_attr_id); if(rc != pcmk_ok) { rsc = rsc->parent; if (BE_QUIET == FALSE) { printf("Performing %s of '%s' on '%s', the parent of '%s'\n", cmd, attr_name, rsc->id, rsc_id); } } break; default: break; } } else if (rsc->parent && BE_QUIET == FALSE) { printf("Forcing %s of '%s' for '%s' instead of '%s'\n", cmd, attr_name, rsc_id, rsc->parent->id); } else if(rsc->parent == NULL && rsc->children) { resource_t *child = rsc->children->data; if(child->variant == pe_native) { lookup_id = clone_strip(child->id); /* Could be a cloned group! */ rc = find_resource_attr(cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); if(rc == pcmk_ok) { rsc = child; if (BE_QUIET == FALSE) { printf("A value for '%s' already exists in child '%s', performing %s on that instead of '%s'\n", attr_name, lookup_id, cmd, rsc_id); } } free(local_attr_id); free(lookup_id); } } return rsc; } int cli_resource_update_attribute(resource_t *rsc, const char *requested_name, const char *attr_set, const char *attr_id, const char *attr_name, const char *attr_value, bool recursive, cib_t *cib, pe_working_set_t *data_set) { int rc = pcmk_ok; static bool need_init = TRUE; char *lookup_id = NULL; char *local_attr_id = NULL; char *local_attr_set = NULL; xmlNode *xml_top = NULL; xmlNode *xml_obj = NULL; if(attr_id == NULL && do_force == FALSE && find_resource_attr( cib, XML_ATTR_ID, uber_parent(rsc)->id, NULL, NULL, NULL, attr_name, NULL) == -EINVAL) { printf("\n"); } if (safe_str_eq(attr_set_type, XML_TAG_ATTR_SETS)) { if (do_force == FALSE) { rc = find_resource_attr(cib, XML_ATTR_ID, uber_parent(rsc)->id, XML_TAG_META_SETS, attr_set, attr_id, attr_name, &local_attr_id); if (rc == pcmk_ok && BE_QUIET == FALSE) { printf("WARNING: There is already a meta attribute for '%s' called '%s' (id=%s)\n", uber_parent(rsc)->id, attr_name, local_attr_id); printf(" Delete '%s' first or use the force option to override\n", local_attr_id); } free(local_attr_id); if (rc == pcmk_ok) { return -ENOTUNIQ; } } } else { rsc = find_matching_attr_resource(rsc, requested_name, attr_set, attr_id, attr_name, cib, "update"); } lookup_id = clone_strip(rsc->id); /* Could be a cloned group! */ rc = find_resource_attr(cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); if (rc == pcmk_ok) { crm_debug("Found a match for name=%s: id=%s", attr_name, local_attr_id); attr_id = local_attr_id; } else if (rc != -ENXIO) { free(lookup_id); free(local_attr_id); return rc; } else { const char *tag = crm_element_name(rsc->xml); if (attr_set == NULL) { local_attr_set = crm_concat(lookup_id, attr_set_type, '-'); attr_set = local_attr_set; } if (attr_id == NULL) { local_attr_id = crm_concat(attr_set, attr_name, '-'); attr_id = local_attr_id; } xml_top = create_xml_node(NULL, tag); crm_xml_add(xml_top, XML_ATTR_ID, lookup_id); xml_obj = create_xml_node(xml_top, attr_set_type); crm_xml_add(xml_obj, XML_ATTR_ID, attr_set); } xml_obj = crm_create_nvpair_xml(xml_obj, attr_id, attr_name, attr_value); if (xml_top == NULL) { xml_top = xml_obj; } crm_log_xml_debug(xml_top, "Update"); rc = cib->cmds->modify(cib, XML_CIB_TAG_RESOURCES, xml_top, cib_options); if (rc == pcmk_ok && BE_QUIET == FALSE) { printf("Set '%s' option: id=%s%s%s%s%s value=%s\n", lookup_id, local_attr_id, attr_set ? " set=" : "", attr_set ? attr_set : "", attr_name ? " name=" : "", attr_name ? attr_name : "", attr_value); } free_xml(xml_top); free(lookup_id); free(local_attr_id); free(local_attr_set); if(recursive && safe_str_eq(attr_set_type, XML_TAG_META_SETS)) { GListPtr lpc = NULL; if(need_init) { xmlNode *cib_constraints = get_object_root(XML_CIB_TAG_CONSTRAINTS, data_set->input); need_init = FALSE; unpack_constraints(cib_constraints, data_set); for (lpc = data_set->resources; lpc != NULL; lpc = lpc->next) { resource_t *r = (resource_t *) lpc->data; clear_bit(r->flags, pe_rsc_allocating); } } crm_debug("Looking for dependencies %p", rsc->rsc_cons_lhs); set_bit(rsc->flags, pe_rsc_allocating); for (lpc = rsc->rsc_cons_lhs; lpc != NULL; lpc = lpc->next) { rsc_colocation_t *cons = (rsc_colocation_t *) lpc->data; resource_t *peer = cons->rsc_lh; crm_debug("Checking %s %d", cons->id, cons->score); if (cons->score > 0 && is_not_set(peer->flags, pe_rsc_allocating)) { /* Don't get into colocation loops */ crm_debug("Setting %s=%s for dependent resource %s", attr_name, attr_value, peer->id); cli_resource_update_attribute(peer, peer->id, NULL, NULL, attr_name, attr_value, recursive, cib, data_set); } } } return rc; } int cli_resource_delete_attribute(resource_t *rsc, const char *requested_name, const char *attr_set, const char *attr_id, const char *attr_name, cib_t *cib, pe_working_set_t *data_set) { xmlNode *xml_obj = NULL; int rc = pcmk_ok; char *lookup_id = NULL; char *local_attr_id = NULL; if(attr_id == NULL && do_force == FALSE && find_resource_attr( cib, XML_ATTR_ID, uber_parent(rsc)->id, NULL, NULL, NULL, attr_name, NULL) == -EINVAL) { printf("\n"); } if(safe_str_eq(attr_set_type, XML_TAG_META_SETS)) { rsc = find_matching_attr_resource(rsc, requested_name, attr_set, attr_id, attr_name, cib, "delete"); } lookup_id = clone_strip(rsc->id); rc = find_resource_attr(cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); if (rc == -ENXIO) { free(lookup_id); return pcmk_ok; } else if (rc != pcmk_ok) { free(lookup_id); return rc; } if (attr_id == NULL) { attr_id = local_attr_id; } xml_obj = crm_create_nvpair_xml(NULL, attr_id, attr_name, NULL); crm_log_xml_debug(xml_obj, "Delete"); CRM_ASSERT(cib); rc = cib->cmds->remove(cib, XML_CIB_TAG_RESOURCES, xml_obj, cib_options); if (rc == pcmk_ok && BE_QUIET == FALSE) { printf("Deleted '%s' option: id=%s%s%s%s%s\n", lookup_id, local_attr_id, attr_set ? " set=" : "", attr_set ? attr_set : "", attr_name ? " name=" : "", attr_name ? attr_name : ""); } free(lookup_id); free_xml(xml_obj); free(local_attr_id); return rc; } static int send_lrm_rsc_op(crm_ipc_t * crmd_channel, const char *op, const char *host_uname, const char *rsc_id, bool only_failed, pe_working_set_t * data_set) { char *our_pid = NULL; char *key = NULL; int rc = -ECOMM; xmlNode *cmd = NULL; xmlNode *xml_rsc = NULL; const char *value = NULL; const char *router_node = host_uname; xmlNode *params = NULL; xmlNode *msg_data = NULL; resource_t *rsc = pe_find_resource(data_set->resources, rsc_id); if (rsc == NULL) { CMD_ERR("Resource %s not found", rsc_id); return -ENXIO; } else if (rsc->variant != pe_native) { CMD_ERR("We can only process primitive resources, not %s", rsc_id); return -EINVAL; } else if (host_uname == NULL) { CMD_ERR("Please specify a node name"); return -EINVAL; } else { node_t *node = pe_find_node(data_set->nodes, host_uname); if (pe__is_guest_or_remote_node(node)) { node = pe__current_node(node->details->remote_rsc); if (node == NULL) { CMD_ERR("No cluster connection to Pacemaker Remote node %s detected", host_uname); return -ENXIO; } router_node = node->details->uname; } } key = generate_transition_key(0, getpid(), 0, "xxxxxxxx-xrsc-opxx-xcrm-resourcexxxx"); msg_data = create_xml_node(NULL, XML_GRAPH_TAG_RSC_OP); crm_xml_add(msg_data, XML_ATTR_TRANSITION_KEY, key); free(key); crm_xml_add(msg_data, XML_LRM_ATTR_TARGET, host_uname); if (safe_str_neq(router_node, host_uname)) { crm_xml_add(msg_data, XML_LRM_ATTR_ROUTER_NODE, router_node); } xml_rsc = create_xml_node(msg_data, XML_CIB_TAG_RESOURCE); if (rsc->clone_name) { crm_xml_add(xml_rsc, XML_ATTR_ID, rsc->clone_name); crm_xml_add(xml_rsc, XML_ATTR_ID_LONG, rsc->id); } else { crm_xml_add(xml_rsc, XML_ATTR_ID, rsc->id); } value = crm_copy_xml_element(rsc->xml, xml_rsc, XML_ATTR_TYPE); if (value == NULL) { CMD_ERR("%s has no type! Aborting...", rsc_id); return -ENXIO; } value = crm_copy_xml_element(rsc->xml, xml_rsc, XML_AGENT_ATTR_CLASS); if (value == NULL) { CMD_ERR("%s has no class! Aborting...", rsc_id); return -ENXIO; } crm_copy_xml_element(rsc->xml, xml_rsc, XML_AGENT_ATTR_PROVIDER); params = create_xml_node(msg_data, XML_TAG_ATTRS); crm_xml_add(params, XML_ATTR_CRM_VERSION, CRM_FEATURE_SET); key = crm_meta_name(XML_LRM_ATTR_INTERVAL_MS); crm_xml_add(params, key, "60000"); /* 1 minute */ free(key); our_pid = crm_getpid_s(); cmd = create_request(op, msg_data, router_node, CRM_SYSTEM_CRMD, crm_system_name, our_pid); /* crm_log_xml_warn(cmd, "send_lrm_rsc_op"); */ free_xml(msg_data); if (crm_ipc_send(crmd_channel, cmd, 0, 0, NULL) > 0) { rc = 0; } else { crm_debug("Could not send %s op to the controller", op); rc = -ENOTCONN; } free_xml(cmd); return rc; } /*! * \internal * \brief Get resource name as used in failure-related node attributes * * \param[in] rsc Resource to check * * \return Newly allocated string containing resource's fail name * \note The caller is responsible for freeing the result. */ static inline char * rsc_fail_name(resource_t *rsc) { const char *name = (rsc->clone_name? rsc->clone_name : rsc->id); return is_set(rsc->flags, pe_rsc_unique)? strdup(name) : clone_strip(name); } static int clear_rsc_history(crm_ipc_t *crmd_channel, const char *host_uname, const char *rsc_id, pe_working_set_t *data_set) { int rc = pcmk_ok; /* Erase the resource's entire LRM history in the CIB, even if we're only * clearing a single operation's fail count. If we erased only entries for a * single operation, we might wind up with a wrong idea of the current * resource state, and we might not re-probe the resource. */ rc = send_lrm_rsc_op(crmd_channel, CRM_OP_LRM_DELETE, host_uname, rsc_id, TRUE, data_set); if (rc != pcmk_ok) { return rc; } crmd_replies_needed++; crm_trace("Processing %d mainloop inputs", crmd_replies_needed); while (g_main_context_iteration(NULL, FALSE)) { crm_trace("Processed mainloop input, %d still remaining", crmd_replies_needed); } if (crmd_replies_needed < 0) { crmd_replies_needed = 0; } return rc; } static int clear_rsc_failures(crm_ipc_t *crmd_channel, const char *node_name, const char *rsc_id, const char *operation, const char *interval_spec, pe_working_set_t *data_set) { int rc = pcmk_ok; const char *failed_value = NULL; const char *failed_id = NULL; const char *interval_ms_s = NULL; GHashTable *rscs = NULL; GHashTableIter iter; /* Create a hash table to use as a set of resources to clean. This lets us * clean each resource only once (per node) regardless of how many failed * operations it has. */ rscs = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, NULL); // Normalize interval to milliseconds for comparison to history entry if (operation) { interval_ms_s = crm_strdup_printf("%u", crm_parse_interval_spec(interval_spec)); } for (xmlNode *xml_op = __xml_first_child(data_set->failed); xml_op != NULL; xml_op = __xml_next(xml_op)) { failed_id = crm_element_value(xml_op, XML_LRM_ATTR_RSCID); if (failed_id == NULL) { // Malformed history entry, should never happen continue; } // No resource specified means all resources match if (rsc_id) { resource_t *fail_rsc = pe_find_resource_with_flags(data_set->resources, failed_id, pe_find_renamed|pe_find_anon); if (!fail_rsc || safe_str_neq(rsc_id, fail_rsc->id)) { continue; } } // Host name should always have been provided by this point failed_value = crm_element_value(xml_op, XML_ATTR_UNAME); if (safe_str_neq(node_name, failed_value)) { continue; } // No operation specified means all operations match if (operation) { failed_value = crm_element_value(xml_op, XML_LRM_ATTR_TASK); if (safe_str_neq(operation, failed_value)) { continue; } // Interval (if operation was specified) defaults to 0 (not all) failed_value = crm_element_value(xml_op, XML_LRM_ATTR_INTERVAL_MS); if (safe_str_neq(interval_ms_s, failed_value)) { continue; } } /* not available until glib 2.32 g_hash_table_add(rscs, (gpointer) failed_id); */ g_hash_table_insert(rscs, (gpointer) failed_id, (gpointer) failed_id); } g_hash_table_iter_init(&iter, rscs); while (g_hash_table_iter_next(&iter, (gpointer *) &failed_id, NULL)) { crm_debug("Erasing failures of %s on %s", failed_id, node_name); rc = clear_rsc_history(crmd_channel, node_name, failed_id, data_set); if (rc != pcmk_ok) { return rc; } } g_hash_table_destroy(rscs); return rc; } static int clear_rsc_fail_attrs(resource_t *rsc, const char *operation, const char *interval_spec, node_t *node) { int rc = pcmk_ok; int attr_options = attrd_opt_none; char *rsc_name = rsc_fail_name(rsc); if (pe__is_guest_or_remote_node(node)) { attr_options |= attrd_opt_remote; } rc = attrd_clear_delegate(NULL, node->details->uname, rsc_name, operation, interval_spec, NULL, attr_options); free(rsc_name); return rc; } int cli_resource_delete(crm_ipc_t *crmd_channel, const char *host_uname, resource_t *rsc, const char *operation, const char *interval_spec, bool just_failures, pe_working_set_t *data_set) { int rc = pcmk_ok; node_t *node = NULL; if (rsc == NULL) { return -ENXIO; } else if (rsc->children) { GListPtr lpc = NULL; for (lpc = rsc->children; lpc != NULL; lpc = lpc->next) { resource_t *child = (resource_t *) lpc->data; rc = cli_resource_delete(crmd_channel, host_uname, child, operation, interval_spec, just_failures, data_set); if (rc != pcmk_ok) { return rc; } } return pcmk_ok; } else if (host_uname == NULL) { GListPtr lpc = NULL; GListPtr nodes = g_hash_table_get_values(rsc->known_on); if(nodes == NULL && do_force) { nodes = node_list_dup(data_set->nodes, FALSE, FALSE); } else if(nodes == NULL && rsc->exclusive_discover) { GHashTableIter iter; pe_node_t *node = NULL; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void**)&node)) { if(node->weight >= 0) { nodes = g_list_prepend(nodes, node); } } } else if(nodes == NULL) { nodes = g_hash_table_get_values(rsc->allowed_nodes); } for (lpc = nodes; lpc != NULL; lpc = lpc->next) { node = (node_t *) lpc->data; if (node->details->online) { rc = cli_resource_delete(crmd_channel, node->details->uname, rsc, operation, interval_spec, just_failures, data_set); } if (rc != pcmk_ok) { g_list_free(nodes); return rc; } } g_list_free(nodes); return pcmk_ok; } node = pe_find_node(data_set->nodes, host_uname); if (node == NULL) { printf("Unable to clean up %s because node %s not found\n", rsc->id, host_uname); return -ENODEV; } if (!node->details->rsc_discovery_enabled) { printf("Unable to clean up %s because resource discovery disabled on %s\n", rsc->id, host_uname); return -EOPNOTSUPP; } if (crmd_channel == NULL) { printf("Dry run: skipping clean-up of %s on %s due to CIB_file\n", rsc->id, host_uname); return pcmk_ok; } rc = clear_rsc_fail_attrs(rsc, operation, interval_spec, node); if (rc != pcmk_ok) { printf("Unable to clean up %s failures on %s: %s\n", rsc->id, host_uname, pcmk_strerror(rc)); return rc; } if (just_failures) { rc = clear_rsc_failures(crmd_channel, host_uname, rsc->id, operation, interval_spec, data_set); } else { rc = clear_rsc_history(crmd_channel, host_uname, rsc->id, data_set); } if (rc != pcmk_ok) { printf("Cleaned %s failures on %s, but unable to clean history: %s\n", rsc->id, host_uname, pcmk_strerror(rc)); } else { printf("Cleaned up %s on %s\n", rsc->id, host_uname); } return rc; } int cli_cleanup_all(crm_ipc_t *crmd_channel, const char *node_name, const char *operation, const char *interval_spec, pe_working_set_t *data_set) { int rc = pcmk_ok; int attr_options = attrd_opt_none; const char *display_name = node_name? node_name : "all nodes"; if (crmd_channel == NULL) { printf("Dry run: skipping clean-up of %s due to CIB_file\n", display_name); return pcmk_ok; } crmd_replies_needed = 0; if (node_name) { node_t *node = pe_find_node(data_set->nodes, node_name); if (node == NULL) { CMD_ERR("Unknown node: %s", node_name); return -ENXIO; } if (pe__is_guest_or_remote_node(node)) { attr_options |= attrd_opt_remote; } } rc = attrd_clear_delegate(NULL, node_name, NULL, operation, interval_spec, NULL, attr_options); if (rc != pcmk_ok) { printf("Unable to clean up all failures on %s: %s\n", display_name, pcmk_strerror(rc)); return rc; } if (node_name) { rc = clear_rsc_failures(crmd_channel, node_name, NULL, operation, interval_spec, data_set); if (rc != pcmk_ok) { printf("Cleaned all resource failures on %s, but unable to clean history: %s\n", node_name, pcmk_strerror(rc)); return rc; } } else { for (GList *iter = data_set->nodes; iter; iter = iter->next) { pe_node_t *node = (pe_node_t *) iter->data; rc = clear_rsc_failures(crmd_channel, node->details->uname, NULL, operation, interval_spec, data_set); if (rc != pcmk_ok) { printf("Cleaned all resource failures on all nodes, but unable to clean history: %s\n", pcmk_strerror(rc)); return rc; } } } printf("Cleaned up all resources on %s\n", display_name); return pcmk_ok; } void cli_resource_check(cib_t * cib_conn, resource_t *rsc) { int need_nl = 0; char *role_s = NULL; char *managed = NULL; resource_t *parent = uber_parent(rsc); find_resource_attr(cib_conn, XML_NVPAIR_ATTR_VALUE, parent->id, NULL, NULL, NULL, XML_RSC_ATTR_MANAGED, &managed); find_resource_attr(cib_conn, XML_NVPAIR_ATTR_VALUE, parent->id, NULL, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, &role_s); if(role_s) { enum rsc_role_e role = text2role(role_s); free(role_s); if(role == RSC_ROLE_UNKNOWN) { // Treated as if unset } else if(role == RSC_ROLE_STOPPED) { printf("\n * The configuration specifies that '%s' should remain stopped\n", parent->id); need_nl++; } else if (is_set(parent->flags, pe_rsc_promotable) && (role == RSC_ROLE_SLAVE)) { printf("\n * The configuration specifies that '%s' should not be promoted\n", parent->id); need_nl++; } } if(managed && crm_is_true(managed) == FALSE) { printf("%s * The configuration prevents the cluster from stopping or starting '%s' (unmanaged)\n", need_nl == 0?"\n":"", parent->id); need_nl++; } free(managed); if(need_nl) { printf("\n"); } } int cli_resource_fail(crm_ipc_t * crmd_channel, const char *host_uname, const char *rsc_id, pe_working_set_t * data_set) { crm_warn("Failing: %s", rsc_id); return send_lrm_rsc_op(crmd_channel, CRM_OP_LRM_FAIL, host_uname, rsc_id, FALSE, data_set); } static GHashTable * generate_resource_params(resource_t * rsc, pe_working_set_t * data_set) { GHashTable *params = NULL; GHashTable *meta = NULL; GHashTable *combined = NULL; GHashTableIter iter; if (!rsc) { crm_err("Resource does not exist in config"); return NULL; } params = crm_str_table_new(); meta = crm_str_table_new(); combined = crm_str_table_new(); get_rsc_attributes(params, rsc, NULL /* TODO: Pass in local node */ , data_set); get_meta_attributes(meta, rsc, NULL /* TODO: Pass in local node */ , data_set); if (params) { char *key = NULL; char *value = NULL; g_hash_table_iter_init(&iter, params); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { g_hash_table_insert(combined, strdup(key), strdup(value)); } g_hash_table_destroy(params); } if (meta) { char *key = NULL; char *value = NULL; g_hash_table_iter_init(&iter, meta); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { char *crm_name = crm_meta_name(key); g_hash_table_insert(combined, crm_name, strdup(value)); } g_hash_table_destroy(meta); } return combined; } static bool resource_is_running_on(resource_t *rsc, const char *host) { bool found = TRUE; GListPtr hIter = NULL; GListPtr hosts = NULL; if(rsc == NULL) { return FALSE; } rsc->fns->location(rsc, &hosts, TRUE); for (hIter = hosts; host != NULL && hIter != NULL; hIter = hIter->next) { pe_node_t *node = (pe_node_t *) hIter->data; if(strcmp(host, node->details->uname) == 0) { crm_trace("Resource %s is running on %s\n", rsc->id, host); goto done; } else if(strcmp(host, node->details->id) == 0) { crm_trace("Resource %s is running on %s\n", rsc->id, host); goto done; } } if(host != NULL) { crm_trace("Resource %s is not running on: %s\n", rsc->id, host); found = FALSE; } else if(host == NULL && hosts == NULL) { crm_trace("Resource %s is not running\n", rsc->id); found = FALSE; } done: g_list_free(hosts); return found; } /*! * \internal * \brief Create a list of all resources active on host from a given list * * \param[in] host Name of host to check whether resources are active * \param[in] rsc_list List of resources to check * * \return New list of resources from list that are active on host */ static GList * get_active_resources(const char *host, GList *rsc_list) { GList *rIter = NULL; GList *active = NULL; for (rIter = rsc_list; rIter != NULL; rIter = rIter->next) { resource_t *rsc = (resource_t *) rIter->data; /* Expand groups to their members, because if we're restarting a member * other than the first, we can't otherwise tell which resources are * stopping and starting. */ if (rsc->variant == pe_group) { active = g_list_concat(active, get_active_resources(host, rsc->children)); } else if (resource_is_running_on(rsc, host)) { active = g_list_append(active, strdup(rsc->id)); } } return active; } GList* subtract_lists(GList *from, GList *items, GCompareFunc cmp) { GList *item = NULL; GList *result = g_list_copy(from); for (item = items; item != NULL; item = item->next) { GList *candidate = NULL; for (candidate = from; candidate != NULL; candidate = candidate->next) { crm_info("Comparing %s with %s", (const char *) candidate->data, (const char *) item->data); if(cmp(candidate->data, item->data) == 0) { result = g_list_remove(result, candidate->data); break; } } } return result; } static void dump_list(GList *items, const char *tag) { int lpc = 0; GList *item = NULL; for (item = items; item != NULL; item = item->next) { crm_trace("%s[%d]: %s", tag, lpc, (char*)item->data); lpc++; } } static void display_list(GList *items, const char *tag) { GList *item = NULL; for (item = items; item != NULL; item = item->next) { fprintf(stdout, "%s%s\n", tag, (const char *)item->data); } } /*! * \internal * \brief Upgrade XML to latest schema version and use it as working set input * * This also updates the working set timestamp to the current time. * * \param[in] data_set Working set instance to update * \param[in] xml XML to use as input * * \return pcmk_ok on success, -ENOKEY if unable to upgrade XML * \note On success, caller is responsible for freeing memory allocated for * data_set->now. * \todo This follows the example of other callers of cli_config_update() * and returns -ENOKEY ("Required key not available") if that fails, * but perhaps -pcmk_err_schema_validation would be better in that case. */ int update_working_set_xml(pe_working_set_t *data_set, xmlNode **xml) { if (cli_config_update(xml, NULL, FALSE) == FALSE) { return -ENOKEY; } data_set->input = *xml; data_set->now = crm_time_new(NULL); return pcmk_ok; } /*! * \internal * \brief Update a working set's XML input based on a CIB query * * \param[in] data_set Data set instance to initialize * \param[in] cib Connection to the CIB manager * * \return pcmk_ok on success, -errno on failure * \note On success, caller is responsible for freeing memory allocated for * data_set->input and data_set->now. */ static int update_working_set_from_cib(pe_working_set_t * data_set, cib_t *cib) { xmlNode *cib_xml_copy = NULL; int rc; rc = cib->cmds->query(cib, NULL, &cib_xml_copy, cib_scope_local | cib_sync_call); if (rc != pcmk_ok) { fprintf(stderr, "Could not obtain the current CIB: %s (%d)\n", pcmk_strerror(rc), rc); return rc; } rc = update_working_set_xml(data_set, &cib_xml_copy); if (rc != pcmk_ok) { fprintf(stderr, "Could not upgrade the current CIB XML\n"); free_xml(cib_xml_copy); return rc; } return pcmk_ok; } static int update_dataset(cib_t *cib, pe_working_set_t * data_set, bool simulate) { char *pid = NULL; char *shadow_file = NULL; cib_t *shadow_cib = NULL; int rc; pe_reset_working_set(data_set); rc = update_working_set_from_cib(data_set, cib); if (rc != pcmk_ok) { return rc; } if(simulate) { pid = crm_getpid_s(); shadow_cib = cib_shadow_new(pid); shadow_file = get_shadow_file(pid); if (shadow_cib == NULL) { fprintf(stderr, "Could not create shadow cib: '%s'\n", pid); rc = -ENXIO; goto cleanup; } rc = write_xml_file(data_set->input, shadow_file, FALSE); if (rc < 0) { fprintf(stderr, "Could not populate shadow cib: %s (%d)\n", pcmk_strerror(rc), rc); goto cleanup; } rc = shadow_cib->cmds->signon(shadow_cib, crm_system_name, cib_command); if(rc != pcmk_ok) { fprintf(stderr, "Could not connect to shadow cib: %s (%d)\n", pcmk_strerror(rc), rc); goto cleanup; } - do_calculations(data_set, data_set->input, NULL); + pcmk__schedule_actions(data_set, data_set->input, NULL); run_simulation(data_set, shadow_cib, NULL, TRUE); rc = update_dataset(shadow_cib, data_set, FALSE); } else { cluster_status(data_set); } cleanup: /* Do not free data_set->input here, we need rsc->xml to be valid later on */ cib_delete(shadow_cib); free(pid); if(shadow_file) { unlink(shadow_file); free(shadow_file); } return rc; } static int max_delay_for_resource(pe_working_set_t * data_set, resource_t *rsc) { int delay = 0; int max_delay = 0; if(rsc && rsc->children) { GList *iter = NULL; for(iter = rsc->children; iter; iter = iter->next) { resource_t *child = (resource_t *)iter->data; delay = max_delay_for_resource(data_set, child); if(delay > max_delay) { double seconds = delay / 1000.0; crm_trace("Calculated new delay of %.1fs due to %s", seconds, child->id); max_delay = delay; } } } else if(rsc) { char *key = crm_strdup_printf("%s_%s_0", rsc->id, RSC_STOP); action_t *stop = custom_action(rsc, key, RSC_STOP, NULL, TRUE, FALSE, data_set); const char *value = g_hash_table_lookup(stop->meta, XML_ATTR_TIMEOUT); max_delay = crm_int_helper(value, NULL); pe_free_action(stop); } return max_delay; } static int max_delay_in(pe_working_set_t * data_set, GList *resources) { int max_delay = 0; GList *item = NULL; for (item = resources; item != NULL; item = item->next) { int delay = 0; resource_t *rsc = pe_find_resource(data_set->resources, (const char *)item->data); delay = max_delay_for_resource(data_set, rsc); if(delay > max_delay) { double seconds = delay / 1000.0; crm_trace("Calculated new delay of %.1fs due to %s", seconds, rsc->id); max_delay = delay; } } return 5 + (max_delay / 1000); } #define waiting_for_starts(d, r, h) ((g_list_length(d) > 0) || \ (resource_is_running_on((r), (h)) == FALSE)) /*! * \internal * \brief Restart a resource (on a particular host if requested). * * \param[in] rsc The resource to restart * \param[in] host The host to restart the resource on (or NULL for all) * \param[in] timeout_ms Consider failed if actions do not complete in this time * (specified in milliseconds, but a two-second * granularity is actually used; if 0, a timeout will be * calculated based on the resource timeout) * \param[in] cib Connection to the CIB manager * * \return pcmk_ok on success, -errno on failure (exits on certain failures) */ int cli_resource_restart(pe_resource_t *rsc, const char *host, int timeout_ms, cib_t *cib) { int rc = 0; int lpc = 0; int before = 0; int step_timeout_s = 0; int sleep_interval = 2; int timeout = timeout_ms / 1000; bool stop_via_ban = FALSE; char *rsc_id = NULL; char *orig_target_role = NULL; GList *list_delta = NULL; GList *target_active = NULL; GList *current_active = NULL; GList *restart_target_active = NULL; pe_working_set_t *data_set = NULL; if(resource_is_running_on(rsc, host) == FALSE) { const char *id = rsc->clone_name?rsc->clone_name:rsc->id; if(host) { printf("%s is not running on %s and so cannot be restarted\n", id, host); } else { printf("%s is not running anywhere and so cannot be restarted\n", id); } return -ENXIO; } /* We might set the target-role meta-attribute */ attr_set_type = XML_TAG_META_SETS; rsc_id = strdup(rsc->id); if ((pe_rsc_is_clone(rsc) || pe_bundle_replicas(rsc)) && host) { stop_via_ban = TRUE; } /* grab full cib determine originally active resources disable or ban poll cib and watch for affected resources to get stopped without --timeout, calculate the stop timeout for each step and wait for that if we hit --timeout or the service timeout, re-enable or un-ban, report failure and indicate which resources we couldn't take down if everything stopped, re-enable or un-ban poll cib and watch for affected resources to get started without --timeout, calculate the start timeout for each step and wait for that if we hit --timeout or the service timeout, report (different) failure and indicate which resources we couldn't bring back up report success Optimizations: - use constraints to determine ordered list of affected resources - Allow a --no-deps option (aka. --force-restart) */ data_set = pe_new_working_set(); if (data_set == NULL) { crm_perror(LOG_ERR, "Could not allocate working set"); rc = -ENOMEM; goto done; } rc = update_dataset(cib, data_set, FALSE); if(rc != pcmk_ok) { fprintf(stdout, "Could not get new resource list: %s (%d)\n", pcmk_strerror(rc), rc); goto done; } restart_target_active = get_active_resources(host, data_set->resources); current_active = get_active_resources(host, data_set->resources); dump_list(current_active, "Origin"); if (stop_via_ban) { /* Stop the clone or bundle instance by banning it from the host */ BE_QUIET = TRUE; rc = cli_resource_ban(rsc_id, host, NULL, cib); } else { /* Stop the resource by setting target-role to Stopped. * Remember any existing target-role so we can restore it later * (though it only makes any difference if it's Slave). */ char *lookup_id = clone_strip(rsc->id); find_resource_attr(cib, XML_NVPAIR_ATTR_VALUE, lookup_id, NULL, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, &orig_target_role); free(lookup_id); rc = cli_resource_update_attribute(rsc, rsc_id, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, RSC_STOPPED, FALSE, cib, data_set); } if(rc != pcmk_ok) { fprintf(stderr, "Could not set target-role for %s: %s (%d)\n", rsc_id, pcmk_strerror(rc), rc); if (current_active) { g_list_free_full(current_active, free); } if (restart_target_active) { g_list_free_full(restart_target_active, free); } goto done; } rc = update_dataset(cib, data_set, TRUE); if(rc != pcmk_ok) { fprintf(stderr, "Could not determine which resources would be stopped\n"); goto failure; } target_active = get_active_resources(host, data_set->resources); dump_list(target_active, "Target"); list_delta = subtract_lists(current_active, target_active, (GCompareFunc) strcmp); fprintf(stdout, "Waiting for %d resources to stop:\n", g_list_length(list_delta)); display_list(list_delta, " * "); step_timeout_s = timeout / sleep_interval; while(g_list_length(list_delta) > 0) { before = g_list_length(list_delta); if(timeout_ms == 0) { step_timeout_s = max_delay_in(data_set, list_delta) / sleep_interval; } /* We probably don't need the entire step timeout */ for(lpc = 0; lpc < step_timeout_s && g_list_length(list_delta) > 0; lpc++) { sleep(sleep_interval); if(timeout) { timeout -= sleep_interval; crm_trace("%ds remaining", timeout); } rc = update_dataset(cib, data_set, FALSE); if(rc != pcmk_ok) { fprintf(stderr, "Could not determine which resources were stopped\n"); goto failure; } if (current_active) { g_list_free_full(current_active, free); } current_active = get_active_resources(host, data_set->resources); g_list_free(list_delta); list_delta = subtract_lists(current_active, target_active, (GCompareFunc) strcmp); dump_list(current_active, "Current"); dump_list(list_delta, "Delta"); } crm_trace("%d (was %d) resources remaining", g_list_length(list_delta), before); if(before == g_list_length(list_delta)) { /* aborted during stop phase, print the contents of list_delta */ fprintf(stderr, "Could not complete shutdown of %s, %d resources remaining\n", rsc_id, g_list_length(list_delta)); display_list(list_delta, " * "); rc = -ETIME; goto failure; } } if (stop_via_ban) { rc = cli_resource_clear(rsc_id, host, NULL, cib, TRUE); } else if (orig_target_role) { rc = cli_resource_update_attribute(rsc, rsc_id, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, orig_target_role, FALSE, cib, data_set); free(orig_target_role); orig_target_role = NULL; } else { rc = cli_resource_delete_attribute(rsc, rsc_id, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, cib, data_set); } if(rc != pcmk_ok) { fprintf(stderr, "Could not unset target-role for %s: %s (%d)\n", rsc_id, pcmk_strerror(rc), rc); goto done; } if (target_active) { g_list_free_full(target_active, free); } target_active = restart_target_active; if (list_delta) { g_list_free(list_delta); } list_delta = subtract_lists(target_active, current_active, (GCompareFunc) strcmp); fprintf(stdout, "Waiting for %d resources to start again:\n", g_list_length(list_delta)); display_list(list_delta, " * "); step_timeout_s = timeout / sleep_interval; while (waiting_for_starts(list_delta, rsc, host)) { before = g_list_length(list_delta); if(timeout_ms == 0) { step_timeout_s = max_delay_in(data_set, list_delta) / sleep_interval; } /* We probably don't need the entire step timeout */ for (lpc = 0; (lpc < step_timeout_s) && waiting_for_starts(list_delta, rsc, host); lpc++) { sleep(sleep_interval); if(timeout) { timeout -= sleep_interval; crm_trace("%ds remaining", timeout); } rc = update_dataset(cib, data_set, FALSE); if(rc != pcmk_ok) { fprintf(stderr, "Could not determine which resources were started\n"); goto failure; } if (current_active) { g_list_free_full(current_active, free); } /* It's OK if dependent resources moved to a different node, * so we check active resources on all nodes. */ current_active = get_active_resources(NULL, data_set->resources); g_list_free(list_delta); list_delta = subtract_lists(target_active, current_active, (GCompareFunc) strcmp); dump_list(current_active, "Current"); dump_list(list_delta, "Delta"); } if(before == g_list_length(list_delta)) { /* aborted during start phase, print the contents of list_delta */ fprintf(stdout, "Could not complete restart of %s, %d resources remaining\n", rsc_id, g_list_length(list_delta)); display_list(list_delta, " * "); rc = -ETIME; goto failure; } } rc = pcmk_ok; goto done; failure: if (stop_via_ban) { cli_resource_clear(rsc_id, host, NULL, cib, TRUE); } else if (orig_target_role) { cli_resource_update_attribute(rsc, rsc_id, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, orig_target_role, FALSE, cib, data_set); free(orig_target_role); } else { cli_resource_delete_attribute(rsc, rsc_id, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, cib, data_set); } done: if (list_delta) { g_list_free(list_delta); } if (current_active) { g_list_free_full(current_active, free); } if (target_active && (target_active != restart_target_active)) { g_list_free_full(target_active, free); } if (restart_target_active) { g_list_free_full(restart_target_active, free); } free(rsc_id); pe_free_working_set(data_set); return rc; } static inline int action_is_pending(action_t *action) { if(is_set(action->flags, pe_action_optional)) { return FALSE; } else if(is_set(action->flags, pe_action_runnable) == FALSE) { return FALSE; } else if(is_set(action->flags, pe_action_pseudo)) { return FALSE; } else if(safe_str_eq("notify", action->task)) { return FALSE; } return TRUE; } /*! * \internal * \brief Return TRUE if any actions in a list are pending * * \param[in] actions List of actions to check * * \return TRUE if any actions in the list are pending, FALSE otherwise */ static bool actions_are_pending(GListPtr actions) { GListPtr action; for (action = actions; action != NULL; action = action->next) { action_t *a = (action_t *)action->data; if (action_is_pending(a)) { crm_notice("Waiting for %s (flags=0x%.8x)", a->uuid, a->flags); return TRUE; } } return FALSE; } /*! * \internal * \brief Print pending actions to stderr * * \param[in] actions List of actions to check * * \return void */ static void print_pending_actions(GListPtr actions) { GListPtr action; fprintf(stderr, "Pending actions:\n"); for (action = actions; action != NULL; action = action->next) { action_t *a = (action_t *) action->data; if (action_is_pending(a)) { fprintf(stderr, "\tAction %d: %s", a->id, a->uuid); if (a->node) { fprintf(stderr, "\ton %s", a->node->details->uname); } fprintf(stderr, "\n"); } } } /* For --wait, timeout (in seconds) to use if caller doesn't specify one */ #define WAIT_DEFAULT_TIMEOUT_S (60 * 60) /* For --wait, how long to sleep between cluster state checks */ #define WAIT_SLEEP_S (2) /*! * \internal * \brief Wait until all pending cluster actions are complete * * This waits until either the CIB's transition graph is idle or a timeout is * reached. * * \param[in] timeout_ms Consider failed if actions do not complete in this time * (specified in milliseconds, but one-second granularity * is actually used; if 0, a default will be used) * \param[in] cib Connection to the CIB manager * * \return pcmk_ok on success, -errno on failure */ int wait_till_stable(int timeout_ms, cib_t * cib) { pe_working_set_t *data_set = NULL; int rc = -1; int timeout_s = timeout_ms? ((timeout_ms + 999) / 1000) : WAIT_DEFAULT_TIMEOUT_S; time_t expire_time = time(NULL) + timeout_s; time_t time_diff; bool printed_version_warning = BE_QUIET; // i.e. don't print if quiet data_set = pe_new_working_set(); if (data_set == NULL) { return -ENOMEM; } do { /* Abort if timeout is reached */ time_diff = expire_time - time(NULL); if (time_diff > 0) { crm_info("Waiting up to %ld seconds for cluster actions to complete", time_diff); } else { print_pending_actions(data_set->actions); pe_free_working_set(data_set); return -ETIME; } if (rc == pcmk_ok) { /* this avoids sleep on first loop iteration */ sleep(WAIT_SLEEP_S); } /* Get latest transition graph */ pe_reset_working_set(data_set); rc = update_working_set_from_cib(data_set, cib); if (rc != pcmk_ok) { pe_free_working_set(data_set); return rc; } - do_calculations(data_set, data_set->input, NULL); + pcmk__schedule_actions(data_set, data_set->input, NULL); if (!printed_version_warning) { /* If the DC has a different version than the local node, the two * could come to different conclusions about what actions need to be * done. Warn the user in this case. * * @TODO A possible long-term solution would be to reimplement the * wait as a new controller operation that would be forwarded to the * DC. However, that would have potential problems of its own. */ const char *dc_version = g_hash_table_lookup(data_set->config_hash, "dc-version"); if (safe_str_neq(dc_version, PACEMAKER_VERSION "-" BUILD_VERSION)) { printf("warning: wait option may not work properly in " "mixed-version cluster\n"); printed_version_warning = TRUE; } } } while (actions_are_pending(data_set->actions)); pe_free_working_set(data_set); return pcmk_ok; } int cli_resource_execute_from_params(const char *rsc_name, const char *rsc_class, const char *rsc_prov, const char *rsc_type, const char *action, GHashTable *params, GHashTable *override_hash, int timeout_ms) { GHashTable *params_copy = NULL; int rc = pcmk_ok; svc_action_t *op = NULL; if (safe_str_eq(rsc_class, PCMK_RESOURCE_CLASS_STONITH)) { CMD_ERR("Sorry, the %s option doesn't support %s resources yet", action, rsc_class); crm_exit(CRM_EX_UNIMPLEMENT_FEATURE); } /* If no timeout was provided, grab the default. */ if (timeout_ms == 0) { timeout_ms = crm_get_msec(CRM_DEFAULT_OP_TIMEOUT_S); } /* add meta_timeout env needed by some resource agents */ g_hash_table_insert(params, strdup("CRM_meta_timeout"), crm_strdup_printf("%d", timeout_ms)); /* add crm_feature_set env needed by some resource agents */ g_hash_table_insert(params, strdup(XML_ATTR_CRM_VERSION), strdup(CRM_FEATURE_SET)); /* resources_action_create frees the params hash table it's passed, but we * may need to reuse it in a second call to resources_action_create. Thus * we'll make a copy here so that gets freed and the original remains for * reuse. */ params_copy = crm_str_table_dup(params); op = resources_action_create(rsc_name, rsc_class, rsc_prov, rsc_type, action, 0, timeout_ms, params_copy, 0); if (op == NULL) { /* Re-run with stderr enabled so we can display a sane error message */ crm_enable_stderr(TRUE); params_copy = crm_str_table_dup(params); op = resources_action_create(rsc_name, rsc_class, rsc_prov, rsc_type, action, 0, timeout_ms, params_copy, 0); /* Callers of cli_resource_execute expect that the params hash table will * be freed. That function uses this one, so for that reason and for * making the two act the same, we should free the hash table here too. */ g_hash_table_destroy(params); /* We know op will be NULL, but this makes static analysis happy */ services_action_free(op); crm_exit(CRM_EX_DATAERR); } setenv("HA_debug", resource_verbose > 0 ? "1" : "0", 1); if(resource_verbose > 1) { setenv("OCF_TRACE_RA", "1", 1); } /* A resource agent using the standard ocf-shellfuncs library will not print * messages to stderr if it doesn't have a controlling terminal (e.g. if * crm_resource is called via script or ssh). This forces it to do so. */ setenv("OCF_TRACE_FILE", "/dev/stderr", 0); if (override_hash) { GHashTableIter iter; char *name = NULL; char *value = NULL; g_hash_table_iter_init(&iter, override_hash); while (g_hash_table_iter_next(&iter, (gpointer *) & name, (gpointer *) & value)) { printf("Overriding the cluster configuration for '%s' with '%s' = '%s'\n", rsc_name, name, value); g_hash_table_replace(op->params, strdup(name), strdup(value)); } } if (services_action_sync(op)) { int more, lpc, last; char *local_copy = NULL; rc = op->rc; if (op->status == PCMK_LRM_OP_DONE) { printf("Operation %s for %s (%s:%s:%s) returned: '%s' (%d)\n", action, rsc_name, rsc_class, rsc_prov ? rsc_prov : "", rsc_type, services_ocf_exitcode_str(op->rc), op->rc); } else { printf("Operation %s for %s (%s:%s:%s) failed: '%s' (%d)\n", action, rsc_name, rsc_class, rsc_prov ? rsc_prov : "", rsc_type, services_lrm_status_str(op->status), op->status); } /* hide output for validate-all if not in verbose */ if (resource_verbose == 0 && safe_str_eq(action, "validate-all")) goto done; if (op->stdout_data) { local_copy = strdup(op->stdout_data); more = strlen(local_copy); last = 0; for (lpc = 0; lpc < more; lpc++) { if (local_copy[lpc] == '\n' || local_copy[lpc] == 0) { local_copy[lpc] = 0; printf(" > stdout: %s\n", local_copy + last); last = lpc + 1; } } free(local_copy); } if (op->stderr_data) { local_copy = strdup(op->stderr_data); more = strlen(local_copy); last = 0; for (lpc = 0; lpc < more; lpc++) { if (local_copy[lpc] == '\n' || local_copy[lpc] == 0) { local_copy[lpc] = 0; printf(" > stderr: %s\n", local_copy + last); last = lpc + 1; } } free(local_copy); } } else { rc = op->rc == 0 ? pcmk_err_generic : op->rc; } done: services_action_free(op); /* See comment above about why we free params here. */ g_hash_table_destroy(params); return rc; } int cli_resource_execute(resource_t *rsc, const char *requested_name, const char *rsc_action, GHashTable *override_hash, int timeout_ms, cib_t * cib, pe_working_set_t *data_set) { int rc = pcmk_ok; const char *rid = NULL; const char *rtype = NULL; const char *rprov = NULL; const char *rclass = NULL; const char *action = NULL; GHashTable *params = NULL; if (safe_str_eq(rsc_action, "validate")) { action = "validate-all"; } else if (safe_str_eq(rsc_action, "force-check")) { action = "monitor"; } else if (safe_str_eq(rsc_action, "force-stop")) { action = rsc_action+6; } else if (safe_str_eq(rsc_action, "force-start") || safe_str_eq(rsc_action, "force-demote") || safe_str_eq(rsc_action, "force-promote")) { action = rsc_action+6; if(pe_rsc_is_clone(rsc)) { rc = cli_resource_search(rsc, requested_name, data_set); if(rc > 0 && do_force == FALSE) { CMD_ERR("It is not safe to %s %s here: the cluster claims it is already active", action, rsc->id); CMD_ERR("Try setting target-role=Stopped first or specifying " "the force option"); crm_exit(CRM_EX_UNSAFE); } } } else { action = rsc_action; } if(pe_rsc_is_clone(rsc)) { /* Grab the first child resource in the hope it's not a group */ rsc = rsc->children->data; } if(rsc->variant == pe_group) { CMD_ERR("Sorry, the %s option doesn't support group resources", rsc_action); crm_exit(CRM_EX_UNIMPLEMENT_FEATURE); } rclass = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS); rprov = crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER); rtype = crm_element_value(rsc->xml, XML_ATTR_TYPE); params = generate_resource_params(rsc, data_set); if (timeout_ms == 0) { timeout_ms = pe_get_configured_timeout(rsc, action, data_set); } rid = pe_rsc_is_anon_clone(rsc->parent)? requested_name : rsc->id; rc = cli_resource_execute_from_params(rid, rclass, rprov, rtype, action, params, override_hash, timeout_ms); return rc; } int cli_resource_move(resource_t *rsc, const char *rsc_id, const char *host_name, cib_t *cib, pe_working_set_t *data_set) { int rc = pcmk_ok; unsigned int count = 0; node_t *current = NULL; node_t *dest = pe_find_node(data_set->nodes, host_name); bool cur_is_dest = FALSE; if (dest == NULL) { return -pcmk_err_node_unknown; } if (scope_master && is_not_set(rsc->flags, pe_rsc_promotable)) { resource_t *p = uber_parent(rsc); if (is_set(p->flags, pe_rsc_promotable)) { CMD_ERR("Using parent '%s' for move instead of '%s'.", rsc->id, rsc_id); rsc_id = p->id; rsc = p; } else { CMD_ERR("Ignoring master option: %s is not promotable", rsc_id); scope_master = FALSE; } } current = pe__find_active_requires(rsc, &count); if (is_set(rsc->flags, pe_rsc_promotable)) { GListPtr iter = NULL; unsigned int master_count = 0; pe_node_t *master_node = NULL; for(iter = rsc->children; iter; iter = iter->next) { resource_t *child = (resource_t *)iter->data; enum rsc_role_e child_role = child->fns->state(child, TRUE); if(child_role == RSC_ROLE_MASTER) { rsc = child; master_node = pe__current_node(child); master_count++; } } if (scope_master || master_count) { count = master_count; current = master_node; } } if (count > 1) { if (pe_rsc_is_clone(rsc)) { current = NULL; } else { return -pcmk_err_multiple; } } if (current && (current->details == dest->details)) { cur_is_dest = TRUE; if (do_force) { crm_info("%s is already %s on %s, reinforcing placement with location constraint.", rsc_id, scope_master?"promoted":"active", dest->details->uname); } else { return -pcmk_err_already; } } /* Clear any previous prefer constraints across all nodes. */ cli_resource_clear(rsc_id, NULL, data_set->nodes, cib, FALSE); /* Clear any previous ban constraints on 'dest'. */ cli_resource_clear(rsc_id, dest->details->uname, data_set->nodes, cib, TRUE); /* Record an explicit preference for 'dest' */ rc = cli_resource_prefer(rsc_id, dest->details->uname, cib); crm_trace("%s%s now prefers node %s%s", rsc->id, scope_master?" (master)":"", dest->details->uname, do_force?"(forced)":""); /* only ban the previous location if current location != destination location. * it is possible to use -M to enforce a location without regard of where the * resource is currently located */ if(do_force && (cur_is_dest == FALSE)) { /* Ban the original location if possible */ if(current) { (void)cli_resource_ban(rsc_id, current->details->uname, NULL, cib); } else if(count > 1) { CMD_ERR("Resource '%s' is currently %s in %d locations. " "One may now move to %s", rsc_id, (scope_master? "promoted" : "active"), count, dest->details->uname); CMD_ERR("To prevent '%s' from being %s at a specific location, " "specify a node.", rsc_id, (scope_master? "promoted" : "active")); } else { crm_trace("Not banning %s from its current location: not active", rsc_id); } } return rc; } static void cli_resource_why_without_rsc_and_host(cib_t *cib_conn,GListPtr resources) { GListPtr lpc = NULL; GListPtr hosts = NULL; for (lpc = resources; lpc != NULL; lpc = lpc->next) { resource_t *rsc = (resource_t *) lpc->data; rsc->fns->location(rsc, &hosts, TRUE); if (hosts == NULL) { printf("Resource %s is not running\n", rsc->id); } else { printf("Resource %s is running\n", rsc->id); } cli_resource_check(cib_conn, rsc); g_list_free(hosts); hosts = NULL; } } static void cli_resource_why_with_rsc_and_host(cib_t *cib_conn, GListPtr resources, resource_t *rsc, const char *host_uname) { if (resource_is_running_on(rsc, host_uname)) { printf("Resource %s is running on host %s\n",rsc->id,host_uname); } else { printf("Resource %s is not running on host %s\n", rsc->id, host_uname); } cli_resource_check(cib_conn, rsc); } static void cli_resource_why_without_rsc_with_host(cib_t *cib_conn,GListPtr resources,node_t *node) { const char* host_uname = node->details->uname; GListPtr allResources = node->details->allocated_rsc; GListPtr activeResources = node->details->running_rsc; GListPtr unactiveResources = subtract_lists(allResources,activeResources,(GCompareFunc) strcmp); GListPtr lpc = NULL; for (lpc = activeResources; lpc != NULL; lpc = lpc->next) { resource_t *rsc = (resource_t *) lpc->data; printf("Resource %s is running on host %s\n",rsc->id,host_uname); cli_resource_check(cib_conn,rsc); } for(lpc = unactiveResources; lpc != NULL; lpc = lpc->next) { resource_t *rsc = (resource_t *) lpc->data; printf("Resource %s is assigned to host %s but not running\n", rsc->id, host_uname); cli_resource_check(cib_conn,rsc); } g_list_free(allResources); g_list_free(activeResources); g_list_free(unactiveResources); } static void cli_resource_why_with_rsc_without_host(cib_t *cib_conn, GListPtr resources, resource_t *rsc) { GListPtr hosts = NULL; rsc->fns->location(rsc, &hosts, TRUE); printf("Resource %s is %srunning\n", rsc->id, (hosts? "" : "not ")); cli_resource_check(cib_conn, rsc); g_list_free(hosts); } void cli_resource_why(cib_t *cib_conn, GListPtr resources, resource_t *rsc, node_t *node) { const char *host_uname = (node == NULL)? NULL : node->details->uname; if ((rsc == NULL) && (host_uname == NULL)) { cli_resource_why_without_rsc_and_host(cib_conn, resources); } else if ((rsc != NULL) && (host_uname != NULL)) { cli_resource_why_with_rsc_and_host(cib_conn, resources, rsc, host_uname); } else if ((rsc == NULL) && (host_uname != NULL)) { cli_resource_why_without_rsc_with_host(cib_conn, resources, node); } else if ((rsc != NULL) && (host_uname == NULL)) { cli_resource_why_with_rsc_without_host(cib_conn, resources, rsc); } } diff --git a/tools/crm_simulate.c b/tools/crm_simulate.c index 005b0fa98a..1921ee448b 100644 --- a/tools/crm_simulate.c +++ b/tools/crm_simulate.c @@ -1,931 +1,929 @@ /* * Copyright 2009-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include cib_t *global_cib = NULL; GListPtr op_fail = NULL; bool action_numbers = FALSE; gboolean quiet = FALSE; gboolean print_pending = TRUE; char *temp_shadow = NULL; extern gboolean bringing_nodes_online; #define quiet_log(fmt, args...) do { \ if(quiet == FALSE) { \ printf(fmt , ##args); \ } \ } while(0) -extern xmlNode *do_calculations(pe_working_set_t * data_set, xmlNode * xml_input, crm_time_t * now); - char *use_date = NULL; static void get_date(pe_working_set_t * data_set) { int value = 0; time_t original_date = 0; crm_element_value_int(data_set->input, "execution-date", &value); original_date = value; if (use_date) { data_set->now = crm_time_new(use_date); quiet_log(" + Setting effective cluster time: %s", use_date); crm_time_log(LOG_NOTICE, "Pretending 'now' is", data_set->now, crm_time_log_date | crm_time_log_timeofday); } else if(original_date) { char *when = NULL; data_set->now = crm_time_new(NULL); crm_time_set_timet(data_set->now, &original_date); when = crm_time_as_string(data_set->now, crm_time_log_date|crm_time_log_timeofday); printf("Using the original execution date of: %s\n", when); free(when); } } static void print_cluster_status(pe_working_set_t * data_set, long options) { char *online_nodes = NULL; char *online_remote_nodes = NULL; char *online_guest_nodes = NULL; char *offline_nodes = NULL; char *offline_remote_nodes = NULL; GListPtr gIter = NULL; for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; const char *node_mode = NULL; char *node_name = NULL; if (pe__is_guest_node(node)) { node_name = crm_strdup_printf("%s:%s", node->details->uname, node->details->remote_rsc->container->id); } else { node_name = crm_strdup_printf("%s", node->details->uname); } if (node->details->unclean) { if (node->details->online && node->details->unclean) { node_mode = "UNCLEAN (online)"; } else if (node->details->pending) { node_mode = "UNCLEAN (pending)"; } else { node_mode = "UNCLEAN (offline)"; } } else if (node->details->pending) { node_mode = "pending"; } else if (node->details->standby_onfail && node->details->online) { node_mode = "standby (on-fail)"; } else if (node->details->standby) { if (node->details->online) { node_mode = "standby"; } else { node_mode = "OFFLINE (standby)"; } } else if (node->details->maintenance) { if (node->details->online) { node_mode = "maintenance"; } else { node_mode = "OFFLINE (maintenance)"; } } else if (node->details->online) { if (pe__is_guest_node(node)) { online_guest_nodes = add_list_element(online_guest_nodes, node_name); } else if (pe__is_remote_node(node)) { online_remote_nodes = add_list_element(online_remote_nodes, node_name); } else { online_nodes = add_list_element(online_nodes, node_name); } free(node_name); continue; } else { if (pe__is_remote_node(node)) { offline_remote_nodes = add_list_element(offline_remote_nodes, node_name); } else if (pe__is_guest_node(node)) { /* ignore offline container nodes */ } else { offline_nodes = add_list_element(offline_nodes, node_name); } free(node_name); continue; } if (pe__is_guest_node(node)) { printf("GuestNode %s: %s\n", node_name, node_mode); } else if (pe__is_remote_node(node)) { printf("RemoteNode %s: %s\n", node_name, node_mode); } else if (safe_str_eq(node->details->uname, node->details->id)) { printf("Node %s: %s\n", node_name, node_mode); } else { printf("Node %s (%s): %s\n", node_name, node->details->id, node_mode); } free(node_name); } if (online_nodes) { printf("Online: [%s ]\n", online_nodes); free(online_nodes); } if (offline_nodes) { printf("OFFLINE: [%s ]\n", offline_nodes); free(offline_nodes); } if (online_remote_nodes) { printf("RemoteOnline: [%s ]\n", online_remote_nodes); free(online_remote_nodes); } if (offline_remote_nodes) { printf("RemoteOFFLINE: [%s ]\n", offline_remote_nodes); free(offline_remote_nodes); } if (online_guest_nodes) { printf("GuestOnline: [%s ]\n", online_guest_nodes); free(online_guest_nodes); } fprintf(stdout, "\n"); for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; if (is_set(rsc->flags, pe_rsc_orphan) && rsc->role == RSC_ROLE_STOPPED) { continue; } rsc->fns->print(rsc, NULL, pe_print_printf | options, stdout); } fprintf(stdout, "\n"); } static char * create_action_name(action_t * action) { char *action_name = NULL; const char *prefix = NULL; const char *action_host = NULL; const char *task = action->task; if (action->node) { action_host = action->node->details->uname; } else if (is_not_set(action->flags, pe_action_pseudo)) { action_host = ""; } if (safe_str_eq(action->task, RSC_CANCEL)) { prefix = "Cancel "; task = "monitor"; /* TO-DO: Hack! */ } if (action->rsc && action->rsc->clone_name) { char *key = NULL; const char *name = action->rsc->clone_name; const char *interval_ms_s = NULL; guint interval_ms = 0; interval_ms_s = g_hash_table_lookup(action->meta, XML_LRM_ATTR_INTERVAL_MS); interval_ms = crm_parse_ms(interval_ms_s); if (safe_str_eq(action->task, RSC_NOTIFY) || safe_str_eq(action->task, RSC_NOTIFIED)) { const char *n_type = g_hash_table_lookup(action->meta, "notify_key_type"); const char *n_task = g_hash_table_lookup(action->meta, "notify_key_operation"); CRM_ASSERT(n_type != NULL); CRM_ASSERT(n_task != NULL); key = generate_notify_key(name, n_type, n_task); } else { key = generate_op_key(name, task, interval_ms); } if (action_host) { action_name = crm_strdup_printf("%s%s %s", prefix ? prefix : "", key, action_host); } else { action_name = crm_strdup_printf("%s%s", prefix ? prefix : "", key); } free(key); } else if (safe_str_eq(action->task, CRM_OP_FENCE)) { const char *op = g_hash_table_lookup(action->meta, "stonith_action"); action_name = crm_strdup_printf("%s%s '%s' %s", prefix ? prefix : "", action->task, op, action_host); } else if (action->rsc && action_host) { action_name = crm_strdup_printf("%s%s %s", prefix ? prefix : "", action->uuid, action_host); } else if (action_host) { action_name = crm_strdup_printf("%s%s %s", prefix ? prefix : "", action->task, action_host); } else { action_name = crm_strdup_printf("%s", action->uuid); } if(action_numbers) { char *with_id = crm_strdup_printf("%s (%d)", action_name, action->id); free(action_name); action_name = with_id; } return action_name; } static void create_dotfile(pe_working_set_t * data_set, const char *dot_file, gboolean all_actions) { GListPtr gIter = NULL; FILE *dot_strm = fopen(dot_file, "w"); if (dot_strm == NULL) { crm_perror(LOG_ERR, "Could not open %s for writing", dot_file); return; } fprintf(dot_strm, " digraph \"g\" {\n"); for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) { action_t *action = (action_t *) gIter->data; const char *style = "dashed"; const char *font = "black"; const char *color = "black"; char *action_name = create_action_name(action); crm_trace("Action %d: %s %s %p", action->id, action_name, action->uuid, action); if (is_set(action->flags, pe_action_pseudo)) { font = "orange"; } if (is_set(action->flags, pe_action_dumped)) { style = "bold"; color = "green"; } else if (action->rsc != NULL && is_not_set(action->rsc->flags, pe_rsc_managed)) { color = "red"; font = "purple"; if (all_actions == FALSE) { goto dont_write; } } else if (is_set(action->flags, pe_action_optional)) { color = "blue"; if (all_actions == FALSE) { goto dont_write; } } else { color = "red"; CRM_CHECK(is_set(action->flags, pe_action_runnable) == FALSE,; ); } set_bit(action->flags, pe_action_dumped); crm_trace("\"%s\" [ style=%s color=\"%s\" fontcolor=\"%s\"]", action_name, style, color, font); fprintf(dot_strm, "\"%s\" [ style=%s color=\"%s\" fontcolor=\"%s\"]\n", action_name, style, color, font); dont_write: free(action_name); } for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) { action_t *action = (action_t *) gIter->data; GListPtr gIter2 = NULL; for (gIter2 = action->actions_before; gIter2 != NULL; gIter2 = gIter2->next) { action_wrapper_t *before = (action_wrapper_t *) gIter2->data; char *before_name = NULL; char *after_name = NULL; const char *style = "dashed"; gboolean optional = TRUE; if (before->state == pe_link_dumped) { optional = FALSE; style = "bold"; } else if (is_set(action->flags, pe_action_pseudo) && (before->type & pe_order_stonith_stop)) { continue; } else if (before->state == pe_link_dup) { continue; } else if (before->type == pe_order_none) { continue; } else if (is_set(before->action->flags, pe_action_dumped) && is_set(action->flags, pe_action_dumped) && before->type != pe_order_load) { optional = FALSE; } if (all_actions || optional == FALSE) { before_name = create_action_name(before->action); after_name = create_action_name(action); crm_trace("\"%s\" -> \"%s\" [ style = %s]", before_name, after_name, style); fprintf(dot_strm, "\"%s\" -> \"%s\" [ style = %s]\n", before_name, after_name, style); free(before_name); free(after_name); } } } fprintf(dot_strm, "}\n"); fflush(dot_strm); fclose(dot_strm); } static void setup_input(const char *input, const char *output) { int rc = pcmk_ok; cib_t *cib_conn = NULL; xmlNode *cib_object = NULL; char *local_output = NULL; if (input == NULL) { /* Use live CIB */ cib_conn = cib_new(); rc = cib_conn->cmds->signon(cib_conn, crm_system_name, cib_command); if (rc == pcmk_ok) { rc = cib_conn->cmds->query(cib_conn, NULL, &cib_object, cib_scope_local | cib_sync_call); } cib_conn->cmds->signoff(cib_conn); cib_delete(cib_conn); cib_conn = NULL; if (rc != pcmk_ok) { fprintf(stderr, "Live CIB query failed: %s (%d)\n", pcmk_strerror(rc), rc); crm_exit(crm_errno2exit(rc)); } else if (cib_object == NULL) { fprintf(stderr, "Live CIB query failed: empty result\n"); crm_exit(CRM_EX_NOINPUT); } } else if (safe_str_eq(input, "-")) { cib_object = filename2xml(NULL); } else { cib_object = filename2xml(input); } if (get_object_root(XML_CIB_TAG_STATUS, cib_object) == NULL) { create_xml_node(cib_object, XML_CIB_TAG_STATUS); } if (cli_config_update(&cib_object, NULL, FALSE) == FALSE) { free_xml(cib_object); crm_exit(CRM_EX_CONFIG); } if (validate_xml(cib_object, NULL, FALSE) != TRUE) { free_xml(cib_object); crm_exit(CRM_EX_CONFIG); } if (output == NULL) { char *pid = crm_getpid_s(); local_output = get_shadow_file(pid); temp_shadow = strdup(local_output); output = local_output; free(pid); } rc = write_xml_file(cib_object, output, FALSE); free_xml(cib_object); cib_object = NULL; if (rc < 0) { fprintf(stderr, "Could not create '%s': %s\n", output, pcmk_strerror(rc)); crm_exit(CRM_EX_CANTCREAT); } setenv("CIB_file", output, 1); free(local_output); } /* *INDENT-OFF* */ static struct crm_option long_options[] = { /* Top-level Options */ {"help", 0, 0, '?', "\tThis text"}, {"version", 0, 0, '$', "\tVersion information" }, {"quiet", 0, 0, 'Q', "\tDisplay only essentialoutput"}, {"verbose", 0, 0, 'V', "\tIncrease debug output"}, {"-spacer-", 0, 0, '-', "\nOperations:"}, {"run", 0, 0, 'R', "\tDetermine the cluster's response to the given configuration and status"}, {"simulate", 0, 0, 'S', "Simulate the transition's execution and display the resulting cluster status"}, {"in-place", 0, 0, 'X', "Simulate the transition's execution and store the result back to the input file"}, {"show-scores", 0, 0, 's', "Show allocation scores"}, {"show-utilization", 0, 0, 'U', "Show utilization information"}, {"profile", 1, 0, 'P', "Run all tests in the named directory to create profiling data"}, {"pending", 0, 0, 'j', "\tDisplay pending state if 'record-pending' is enabled", pcmk_option_hidden}, {"-spacer-", 0, 0, '-', "\nSynthetic Cluster Events:"}, {"node-up", 1, 0, 'u', "\tBring a node online"}, {"node-down", 1, 0, 'd', "\tTake a node offline"}, {"node-fail", 1, 0, 'f', "\tMark a node as failed"}, {"op-inject", 1, 0, 'i', "\tGenerate a failure for the cluster to react to in the simulation"}, {"-spacer-", 0, 0, '-', "\t\tValue is of the form ${resource}_${task}_${interval_in_ms}@${node}=${rc}."}, {"-spacer-", 0, 0, '-', "\t\tEg. memcached_monitor_20000@bart.example.com=7"}, {"-spacer-", 0, 0, '-', "\t\tFor more information on OCF return codes, refer to: https://clusterlabs.org/pacemaker/doc/en-US/Pacemaker/2.0/html/Pacemaker_Administration/s-ocf-return-codes.html"}, {"op-fail", 1, 0, 'F', "\tIf the specified task occurs during the simulation, have it fail with return code ${rc}"}, {"-spacer-", 0, 0, '-', "\t\tValue is of the form ${resource}_${task}_${interval_in_ms}@${node}=${rc}."}, {"-spacer-", 0, 0, '-', "\t\tEg. memcached_stop_0@bart.example.com=1\n"}, {"-spacer-", 0, 0, '-', "\t\tThe transition will normally stop at the failed action. Save the result with --save-output and re-run with --xml-file"}, { "set-datetime", required_argument, NULL, 't', "Set date/time (ISO 8601 format, see https://en.wikipedia.org/wiki/ISO_8601)" }, {"quorum", 1, 0, 'q', "\tSpecify a value for quorum"}, {"watchdog", 1, 0, 'w', "\tAssume a watchdog device is active"}, {"ticket-grant", 1, 0, 'g', "Grant a ticket"}, {"ticket-revoke", 1, 0, 'r', "Revoke a ticket"}, {"ticket-standby", 1, 0, 'b', "Make a ticket standby"}, {"ticket-activate", 1, 0, 'e', "Activate a ticket"}, {"-spacer-", 0, 0, '-', "\nOutput Options:"}, {"save-input", 1, 0, 'I', "\tSave the input configuration to the named file"}, {"save-output", 1, 0, 'O', "Save the output configuration to the named file"}, {"save-graph", 1, 0, 'G', "\tSave the transition graph (XML format) to the named file"}, {"save-dotfile", 1, 0, 'D', "Save the transition graph (DOT format) to the named file"}, {"all-actions", 0, 0, 'a', "\tDisplay all possible actions in the DOT graph - even ones not part of the transition"}, {"-spacer-", 0, 0, '-', "\nData Source:"}, {"live-check", 0, 0, 'L', "\tConnect to the CIB mamager and use the current CIB contents as input"}, {"xml-file", 1, 0, 'x', "\tRetrieve XML from the named file"}, {"xml-pipe", 0, 0, 'p', "\tRetrieve XML from stdin"}, {"-spacer-", 0, 0, '-', "\nExamples:\n"}, {"-spacer-", 0, 0, '-', "Pretend a recurring monitor action found memcached stopped on node fred.example.com and, during recovery, that the memcached stop action failed", pcmk_option_paragraph}, {"-spacer-", 0, 0, '-', " crm_simulate -LS --op-inject memcached:0_monitor_20000@bart.example.com=7 --op-fail memcached:0_stop_0@fred.example.com=1 --save-output /tmp/memcached-test.xml", pcmk_option_example}, {"-spacer-", 0, 0, '-', "Now see what the reaction to the stop failure would be", pcmk_option_paragraph}, {"-spacer-", 0, 0, '-', " crm_simulate -S --xml-file /tmp/memcached-test.xml", pcmk_option_example}, {0, 0, 0, 0} }; /* *INDENT-ON* */ static void profile_one(const char *xml_file, pe_working_set_t *data_set) { xmlNode *cib_object = NULL; printf("* Testing %s\n", xml_file); cib_object = filename2xml(xml_file); if (get_object_root(XML_CIB_TAG_STATUS, cib_object) == NULL) { create_xml_node(cib_object, XML_CIB_TAG_STATUS); } if (cli_config_update(&cib_object, NULL, FALSE) == FALSE) { free_xml(cib_object); return; } if (validate_xml(cib_object, NULL, FALSE) != TRUE) { free_xml(cib_object); return; } data_set->input = cib_object; get_date(data_set); - do_calculations(data_set, cib_object, NULL); + pcmk__schedule_actions(data_set, cib_object, NULL); pe_reset_working_set(data_set); } #ifndef FILENAME_MAX # define FILENAME_MAX 512 #endif static void profile_all(const char *dir, pe_working_set_t *data_set) { struct dirent **namelist; int file_num = scandir(dir, &namelist, 0, alphasort); if (file_num > 0) { struct stat prop; char buffer[FILENAME_MAX]; while (file_num--) { if ('.' == namelist[file_num]->d_name[0]) { free(namelist[file_num]); continue; } else if (!crm_ends_with_ext(namelist[file_num]->d_name, ".xml")) { free(namelist[file_num]); continue; } snprintf(buffer, sizeof(buffer), "%s/%s", dir, namelist[file_num]->d_name); if (stat(buffer, &prop) == 0 && S_ISREG(prop.st_mode)) { profile_one(buffer, data_set); } free(namelist[file_num]); } free(namelist); } } static int count_resources(pe_working_set_t * data_set, resource_t * rsc) { int count = 0; GListPtr gIter = NULL; if (rsc == NULL) { gIter = data_set->resources; } else if (rsc->children) { gIter = rsc->children; } else { return is_not_set(rsc->flags, pe_rsc_orphan); } for (; gIter != NULL; gIter = gIter->next) { count += count_resources(data_set, gIter->data); } return count; } int main(int argc, char **argv) { int rc = pcmk_ok; guint modified = 0; gboolean store = FALSE; gboolean process = FALSE; gboolean simulate = FALSE; gboolean all_actions = FALSE; gboolean have_stdout = FALSE; pe_working_set_t *data_set = NULL; const char *xml_file = "-"; const char *quorum = NULL; const char *watchdog = NULL; const char *test_dir = NULL; const char *dot_file = NULL; const char *graph_file = NULL; const char *input_file = NULL; const char *output_file = NULL; int flag = 0; int index = 0; int argerr = 0; GListPtr node_up = NULL; GListPtr node_down = NULL; GListPtr node_fail = NULL; GListPtr op_inject = NULL; GListPtr ticket_grant = NULL; GListPtr ticket_revoke = NULL; GListPtr ticket_standby = NULL; GListPtr ticket_activate = NULL; xmlNode *input = NULL; crm_log_cli_init("crm_simulate"); crm_set_options(NULL, "datasource operation [additional options]", long_options, "Tool for simulating the cluster's response to events"); if (argc < 2) { crm_help('?', CRM_EX_USAGE); } while (1) { flag = crm_get_option(argc, argv, &index); if (flag == -1) break; switch (flag) { case 'V': if (have_stdout == FALSE) { /* Redirect stderr to stdout so we can grep the output */ have_stdout = TRUE; close(STDERR_FILENO); dup2(STDOUT_FILENO, STDERR_FILENO); } crm_bump_log_level(argc, argv); action_numbers = TRUE; break; case '?': case '$': crm_help(flag, CRM_EX_OK); break; case 'p': xml_file = "-"; break; case 'Q': quiet = TRUE; break; case 'L': xml_file = NULL; break; case 'x': xml_file = optarg; break; case 'u': modified++; bringing_nodes_online = TRUE; node_up = g_list_append(node_up, optarg); break; case 'd': modified++; node_down = g_list_append(node_down, optarg); break; case 'f': modified++; node_fail = g_list_append(node_fail, optarg); break; case 't': use_date = strdup(optarg); break; case 'i': modified++; op_inject = g_list_append(op_inject, optarg); break; case 'F': process = TRUE; simulate = TRUE; op_fail = g_list_append(op_fail, optarg); break; case 'w': modified++; watchdog = optarg; break; case 'q': modified++; quorum = optarg; break; case 'g': modified++; ticket_grant = g_list_append(ticket_grant, optarg); break; case 'r': modified++; ticket_revoke = g_list_append(ticket_revoke, optarg); break; case 'b': modified++; ticket_standby = g_list_append(ticket_standby, optarg); break; case 'e': modified++; ticket_activate = g_list_append(ticket_activate, optarg); break; case 'a': all_actions = TRUE; break; case 's': process = TRUE; show_scores = TRUE; break; case 'U': process = TRUE; show_utilization = TRUE; break; case 'j': print_pending = TRUE; break; case 'S': process = TRUE; simulate = TRUE; break; case 'X': store = TRUE; process = TRUE; simulate = TRUE; break; case 'R': process = TRUE; break; case 'D': process = TRUE; dot_file = optarg; break; case 'G': process = TRUE; graph_file = optarg; break; case 'I': input_file = optarg; break; case 'O': output_file = optarg; break; case 'P': test_dir = optarg; break; default: ++argerr; break; } } if (optind > argc) { ++argerr; } if (argerr) { crm_help('?', CRM_EX_USAGE); } data_set = pe_new_working_set(); if (data_set == NULL) { crm_perror(LOG_ERR, "Could not allocate working set"); rc = -ENOMEM; goto done; } if (test_dir != NULL) { profile_all(test_dir, data_set); return CRM_EX_OK; } setup_input(xml_file, store ? xml_file : output_file); global_cib = cib_new(); rc = global_cib->cmds->signon(global_cib, crm_system_name, cib_command); if (rc != pcmk_ok) { fprintf(stderr, "Could not connect to the CIB manager: %s\n", pcmk_strerror(rc)); goto done; } rc = global_cib->cmds->query(global_cib, NULL, &input, cib_sync_call | cib_scope_local); if (rc != pcmk_ok) { fprintf(stderr, "Could not get local CIB: %s\n", pcmk_strerror(rc)); goto done; } data_set->input = input; get_date(data_set); if(xml_file) { set_bit(data_set->flags, pe_flag_sanitized); } set_bit(data_set->flags, pe_flag_stdout); cluster_status(data_set); if (quiet == FALSE) { int options = print_pending ? pe_print_pending : 0; if (is_set(data_set->flags, pe_flag_maintenance_mode)) { quiet_log("\n *** Resource management is DISABLED ***"); quiet_log("\n The cluster will not attempt to start, stop or recover services"); quiet_log("\n"); } if (data_set->disabled_resources || data_set->blocked_resources) { quiet_log("%d of %d resources DISABLED and %d BLOCKED from being started due to failures\n", data_set->disabled_resources, count_resources(data_set, NULL), data_set->blocked_resources); } quiet_log("\nCurrent cluster status:\n"); print_cluster_status(data_set, options); } if (modified) { quiet_log("Performing requested modifications\n"); modify_configuration(data_set, global_cib, quorum, watchdog, node_up, node_down, node_fail, op_inject, ticket_grant, ticket_revoke, ticket_standby, ticket_activate); rc = global_cib->cmds->query(global_cib, NULL, &input, cib_sync_call); if (rc != pcmk_ok) { fprintf(stderr, "Could not get modified CIB: %s\n", pcmk_strerror(rc)); goto done; } cleanup_calculations(data_set); data_set->input = input; get_date(data_set); if(xml_file) { set_bit(data_set->flags, pe_flag_sanitized); } set_bit(data_set->flags, pe_flag_stdout); cluster_status(data_set); } if (input_file != NULL) { rc = write_xml_file(input, input_file, FALSE); if (rc < 0) { fprintf(stderr, "Could not create '%s': %s\n", input_file, pcmk_strerror(rc)); goto done; } } if (process || simulate) { crm_time_t *local_date = NULL; if (show_scores && show_utilization) { printf("Allocation scores and utilization information:\n"); } else if (show_scores) { fprintf(stdout, "Allocation scores:\n"); } else if (show_utilization) { printf("Utilization information:\n"); } - do_calculations(data_set, input, local_date); + pcmk__schedule_actions(data_set, input, local_date); input = NULL; /* Don't try and free it twice */ if (graph_file != NULL) { write_xml_file(data_set->graph, graph_file, FALSE); } if (dot_file != NULL) { create_dotfile(data_set, dot_file, all_actions); } if (quiet == FALSE) { GListPtr gIter = NULL; quiet_log("%sTransition Summary:\n", show_scores || show_utilization || modified ? "\n" : ""); fflush(stdout); LogNodeActions(data_set, TRUE); for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; LogActions(rsc, data_set, TRUE); } } } rc = pcmk_ok; if (simulate) { if (run_simulation(data_set, global_cib, op_fail, quiet) != pcmk_ok) { rc = pcmk_err_generic; } if(quiet == FALSE) { get_date(data_set); quiet_log("\nRevised cluster status:\n"); set_bit(data_set->flags, pe_flag_stdout); cluster_status(data_set); print_cluster_status(data_set, 0); } } done: pe_free_working_set(data_set); global_cib->cmds->signoff(global_cib); cib_delete(global_cib); free(use_date); fflush(stderr); if (temp_shadow) { unlink(temp_shadow); free(temp_shadow); } crm_exit(crm_errno2exit(rc)); } diff --git a/tools/crm_verify.c b/tools/crm_verify.c index 4ae926fd5f..870a95ca2b 100644 --- a/tools/crm_verify.c +++ b/tools/crm_verify.c @@ -1,270 +1,270 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include gboolean USE_LIVE_CIB = FALSE; char *cib_save = NULL; extern gboolean stage0(pe_working_set_t * data_set); -extern xmlNode *do_calculations(pe_working_set_t * data_set, xmlNode * xml_input, crm_time_t * now); /* *INDENT-OFF* */ static struct crm_option long_options[] = { /* Top-level Options */ {"help", 0, 0, '?', "\tThis text"}, {"version", 0, 0, '$', "\tVersion information" }, {"verbose", 0, 0, 'V', "\tSpecify multiple times to increase debug output\n"}, {"-spacer-", 1, 0, '-', "\nData sources:"}, {"live-check", 0, 0, 'L', "Check the configuration used by the running cluster\n"}, {"xml-file", 1, 0, 'x', "Check the configuration in the named file"}, {"xml-text", 1, 0, 'X', "Check the configuration in the supplied string"}, {"xml-pipe", 0, 0, 'p', "Check the configuration piped in via stdin"}, {"-spacer-", 1, 0, '-', "\nAdditional Options:"}, {"save-xml", 1, 0, 'S', "Save the verified XML to the named file. Most useful with -L"}, {"-spacer-", 1, 0, '-', "\nExamples:", pcmk_option_paragraph}, {"-spacer-", 1, 0, '-', "Check the consistency of the configuration in the running cluster:", pcmk_option_paragraph}, {"-spacer-", 1, 0, '-', " crm_verify --live-check", pcmk_option_example}, {"-spacer-", 1, 0, '-', "Check the consistency of the configuration in a given file and produce verbose output:", pcmk_option_paragraph}, {"-spacer-", 1, 0, '-', " crm_verify --xml-file file.xml --verbose", pcmk_option_example}, {0, 0, 0, 0} }; /* *INDENT-ON* */ int main(int argc, char **argv) { xmlNode *cib_object = NULL; xmlNode *status = NULL; int argerr = 0; int flag; int option_index = 0; pe_working_set_t *data_set = NULL; cib_t *cib_conn = NULL; int rc = pcmk_ok; bool verbose = FALSE; gboolean xml_stdin = FALSE; const char *xml_tag = NULL; const char *xml_file = NULL; const char *xml_string = NULL; crm_log_cli_init("crm_verify"); crm_set_options(NULL, "[modifiers] data_source", long_options, "check a Pacemaker configuration for errors" "\n\nCheck the well-formedness of a complete Pacemaker XML configuration," "\n\nits conformance to the configured schema, and the presence of common" "\n\nmisconfigurations. Problems reported as errors must be fixed before the" "\n\ncluster will work properly. It is left to the administrator to decide" "\n\nwhether to fix problems reported as warnings."); while (1) { flag = crm_get_option(argc, argv, &option_index); if (flag == -1) break; switch (flag) { case 'X': crm_trace("Option %c => %s", flag, optarg); xml_string = optarg; break; case 'x': crm_trace("Option %c => %s", flag, optarg); xml_file = optarg; break; case 'p': xml_stdin = TRUE; break; case 'S': cib_save = optarg; break; case 'V': verbose = TRUE; crm_bump_log_level(argc, argv); break; case 'L': USE_LIVE_CIB = TRUE; break; case '$': case '?': crm_help(flag, CRM_EX_OK); break; default: fprintf(stderr, "Option -%c is not yet supported\n", flag); ++argerr; break; } } if (optind < argc) { printf("non-option ARGV-elements: "); while (optind < argc) { printf("%s ", argv[optind++]); } printf("\n"); } if (optind > argc) { ++argerr; } if (argerr) { crm_err("%d errors in option parsing", argerr); crm_help(flag, CRM_EX_USAGE); } crm_info("=#=#=#=#= Getting XML =#=#=#=#="); if (USE_LIVE_CIB) { cib_conn = cib_new(); rc = cib_conn->cmds->signon(cib_conn, crm_system_name, cib_command); } if (USE_LIVE_CIB) { if (rc == pcmk_ok) { int options = cib_scope_local | cib_sync_call; crm_info("Reading XML from: live cluster"); rc = cib_conn->cmds->query(cib_conn, NULL, &cib_object, options); } if (rc != pcmk_ok) { fprintf(stderr, "Live CIB query failed: %s\n", pcmk_strerror(rc)); goto done; } if (cib_object == NULL) { fprintf(stderr, "Live CIB query failed: empty result\n"); rc = -ENOMSG; goto done; } } else if (xml_file != NULL) { cib_object = filename2xml(xml_file); if (cib_object == NULL) { fprintf(stderr, "Couldn't parse input file: %s\n", xml_file); rc = -ENODATA; goto done; } } else if (xml_string != NULL) { cib_object = string2xml(xml_string); if (cib_object == NULL) { fprintf(stderr, "Couldn't parse input string: %s\n", xml_string); rc = -ENODATA; goto done; } } else if (xml_stdin) { cib_object = stdin2xml(); if (cib_object == NULL) { fprintf(stderr, "Couldn't parse input from STDIN.\n"); rc = -ENODATA; goto done; } } else { fprintf(stderr, "No configuration source specified." " Use --help for usage information.\n"); rc = -ENODATA; goto done; } xml_tag = crm_element_name(cib_object); if (safe_str_neq(xml_tag, XML_TAG_CIB)) { fprintf(stderr, "This tool can only check complete configurations (i.e. those starting with ).\n"); rc = -EBADMSG; goto done; } if (cib_save != NULL) { write_xml_file(cib_object, cib_save, FALSE); } status = get_object_root(XML_CIB_TAG_STATUS, cib_object); if (status == NULL) { create_xml_node(cib_object, XML_CIB_TAG_STATUS); } if (validate_xml(cib_object, NULL, FALSE) == FALSE) { crm_config_err("CIB did not pass schema validation"); free_xml(cib_object); cib_object = NULL; } else if (cli_config_update(&cib_object, NULL, FALSE) == FALSE) { crm_config_error = TRUE; free_xml(cib_object); cib_object = NULL; fprintf(stderr, "The cluster will NOT be able to use this configuration.\n"); fprintf(stderr, "Please manually update the configuration to conform to the %s syntax.\n", xml_latest_schema()); } data_set = pe_new_working_set(); if (data_set == NULL) { rc = -errno; crm_perror(LOG_CRIT, "Unable to allocate working set"); goto done; } if (cib_object == NULL) { } else if (status != NULL || USE_LIVE_CIB) { /* live queries will always have a status section and can do a full simulation */ - do_calculations(data_set, cib_object, NULL); + pcmk__schedule_actions(data_set, cib_object, NULL); } else { data_set->now = crm_time_new(NULL); data_set->input = cib_object; stage0(data_set); } pe_free_working_set(data_set); if (crm_config_error) { fprintf(stderr, "Errors found during check: config not valid\n"); if (verbose == FALSE) { fprintf(stderr, " -V may provide more details\n"); } rc = -pcmk_err_schema_validation; } else if (crm_config_warning) { fprintf(stderr, "Warnings found during check: config may not be valid\n"); if (verbose == FALSE) { fprintf(stderr, " Use -V -V for more details\n"); } rc = -pcmk_err_schema_validation; } if (USE_LIVE_CIB && cib_conn) { cib_conn->cmds->signoff(cib_conn); cib_delete(cib_conn); } done: crm_exit(crm_errno2exit(rc)); }