diff --git a/.travis.yml b/.travis.yml
index f88060e16d..f5ed4d0f0e 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -1,114 +1,114 @@
# Copyright 2012-2019 the Pacemaker project contributors
#
# The version control history for this file may have further details.
# Control file for the Travis autobuilder
# https://docs.travis-ci.com/user/customizing-the-build/
language: c
# We build with both gcc and clang. If MAINT_EXTRA=1 (gcc only), the
# schema regression tests will additionally be run.
matrix:
include:
- compiler: gcc
env: MAINT_EXTRA=1
arch: amd64
- compiler: clang
env: MAINT_EXTRA=0
arch: amd64
- compiler: gcc
env: MAINT_EXTRA=0
arch: ppc64le
- compiler: gcc
env: MAINT_EXTRA=0
arch: arm64
cache:
directories:
- xml/.relaxng.org
# sudo add-apt-repository ppa:hotot-team
before_install:
- if [ "$TRAVIS_ARCH" == "ppc64le" -o "$TRAVIS_ARCH" == "aarch64" ]; then sudo add-apt-repository "deb http://ports.ubuntu.com/ubuntu-ports/ trusty main"; sudo apt-get update -qq; fi
- if [ "$TRAVIS_ARCH" == "amd64" ]; then sudo add-apt-repository "deb http://archive.ubuntu.com/ubuntu/ trusty main"; sudo apt-get update -qq; fi
# To switch to Travis-CI's containerized (non-sudo) architecture,
# all our dependencies need to be on Travis's whitelist:
# https://github.com/travis-ci/apt-package-whitelist
#
# The only ones that aren't already are:
# - cluster-glue-dev: see open issue:
# https://github.com/travis-ci/apt-package-whitelist/issues/2936
# - resource-agents: see open issue:
# https://github.com/travis-ci/apt-package-whitelist/issues/4261
# - libdbus-1-dev: see multiple open issues:
# https://github.com/travis-ci/apt-package-whitelist/issues?utf8=%E2%9C%93&q=is%3Aissue+libdbus+-1-dev
# (a workaround is to install libdbus-glib-1-dev, which depends on it and is whitelisted)
install:
- sudo apt-get install -qq
automake autoconf libtool
psmisc procps python3 python3-dev
libbz2-dev libdbus-1-dev libglib2.0-dev libgnutls-dev libltdl-dev
libncurses5-dev libpam0g-dev libxml2-dev libxslt1-dev uuid-dev
libqb-dev libcfg-dev libcmap-dev libcorosync-common-dev libcpg-dev
libquorum-dev libsam-dev libtotem-pg-dev libvotequorum-dev
cluster-glue-dev resource-agents
- test $MAINT_EXTRA -eq 0 || sudo apt-get install -qq libxml2-utils xsltproc
before_script:
# some tests (e.g. cts-exec-helper) require actual system-wide credentials
- ./autogen.sh
- ./configure --with-daemon-user=nobody --with-daemon-group=nogroup
--libexecdir=/usr/lib/pacemaker
--with-configdir=/etc/default
script:
# Create directories needed by commands used by regression tests
- sudo make install-exec-local || true
- make
- make check
- ./cts/cts-cli -V
- ./cts/cts-scheduler -V
-- sudo ./cts/cts-exec -V --force-wait
+- if [ "$TRAVIS_ARCH" != "ppc64le" ]; then sudo ./cts/cts-exec -V --force-wait; fi
- test $MAINT_EXTRA -eq 0 ||
{ { echo 'looking for presence of control characters...';
{ git ls-files
| grep -v tap-driver.sh
| xargs grep -Ensv "^([^[:cntrl:]]*|$(printf '\t'))*$"||:; } 2>/dev/null
| { ! grep -Ev '^Binary file' && echo 'ALL OK'; };
} && (
cd xml;
./regression.sh && ./regression.sh -B && ./regression.sh -S && {
schemas=; for schema in *.rng; do
case ${schema} in *cibtr*);; *)schemas="${schemas} ${schema}";; esac;
done;
test -s .relaxng.org/relaxng.rng 2>/dev/null
|| curl --create-dirs -SsLo .relaxng.org/relaxng.rng
'https://raw.githubusercontent.com/relaxng/relaxng.org/master/relaxng.rng';
xmllint --noout --relaxng .relaxng.org/relaxng.rng ${schemas};
}
);
}
#after_script:
#after_success:
after_failure:
- lsb_release -a
- sudo cat /etc/apt/sources.list
- whoami
- env | sort
- cat include/config.h
notifications:
irc: "irc.freenode.org#pcmk"
# email:
# recipients:
# - developers@clusterlabs.org
# whitelist
branches:
only:
- master
- "2.1"
diff --git a/configure.ac b/configure.ac
index 1a501485ec..574308724d 100644
--- a/configure.ac
+++ b/configure.ac
@@ -1,2071 +1,2086 @@
dnl
dnl autoconf for Pacemaker
dnl
dnl Copyright 2009-2021 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)
dnl AC_CONFIG_MACRO_DIR is deprecated as of autoconf 2.70 (2020-12-08).
dnl Once we can require that version, we can simplify this, and no longer
dnl need ACLOCAL_AMFLAGS in Makefile.am.
m4_ifdef([AC_CONFIG_MACRO_DIRS],
[AC_CONFIG_MACRO_DIRS([m4])],
[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([m4/version.m4])
AC_INIT([pacemaker], VERSION_NUMBER, [users@clusterlabs.org], [pacemaker],
PCMK_URL)
PCMK_FEATURES=""
LT_CONFIG_LTDL_DIR([libltdl])
AC_CONFIG_AUX_DIR([libltdl/config])
AC_CANONICAL_HOST
dnl Where #defines that autoconf makes (e.g. HAVE_whatever) go
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])
dnl 1.13: minimum automake version required
dnl foreign: don't require GNU-standard top-level files
dnl tar-ustar: use (older) POSIX variant of generated tar rather than v7
dnl subdir-objects: keep .o's with their .c's (no-op in 2.0+)
AM_INIT_AUTOMAKE([1.13 foreign tar-ustar subdir-objects])
dnl Require minimum version of pkg-config
PKG_PROG_PKG_CONFIG(0.27)
AS_IF([test "x${PKG_CONFIG}" != x], [],
[AC_MSG_FAILURE([Could not find required build tool pkg-config (0.27 or later)])])
PKG_INSTALLDIR
PKG_NOARCH_INSTALLDIR
dnl Example 2.4. Silent Custom Rule to Generate a File
dnl %-bar.pc: %.pc
dnl $(AM_V_GEN)$(LN_S) $(notdir $^) $@
dnl Versioned attributes implementation is not yet production-ready
AC_DEFINE_UNQUOTED(ENABLE_VERSIONED_ATTRS, 0, [Enable versioned attributes])
CC_IN_CONFIGURE=yes
export CC_IN_CONFIGURE
LDD=ldd
GLIB_TESTS
dnl ========================================================================
dnl Compiler characteristics
dnl ========================================================================
dnl A particular compiler can be forced by setting the CC environment variable
AC_PROG_CC
dnl Use at least C99 if possible. This will generate an "obsolete" warning
dnl since autoconf 2.70, but is needed for older versions.
AC_PROG_CC_STDC
dnl C++ is not needed for build, just maintainer utilities
AC_PROG_CXX
dnl We use md5.c from gnulib, which has its own m4 macros. Per its docs:
dnl "The macro gl_EARLY must be called as soon as possible after verifying that
dnl the C compiler is working. ... The core part of the gnulib checks are done
dnl by the macro gl_INIT." In addition, prevent gnulib from introducing OpenSSL
dnl as a dependency.
gl_EARLY
gl_SET_CRYPTO_CHECK_DEFAULT([no])
gl_INIT
# --enable-new-dtags: Use RUNPATH instead of RPATH.
# It is necessary to have this done before libtool does linker detection.
# See also: https://github.com/kronosnet/kronosnet/issues/107
AX_CHECK_LINK_FLAG([-Wl,--enable-new-dtags],
[AM_LDFLAGS=-Wl,--enable-new-dtags],
[AC_MSG_ERROR(["Linker support for --enable-new-dtags is required"])])
AC_SUBST([AM_LDFLAGS])
saved_LDFLAGS="$LDFLAGS"
LDFLAGS="$AM_LDFLAGS $LDFLAGS"
LT_INIT([dlopen])
LDFLAGS="$saved_LDFLAGS"
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
}
# yes_no_try $user_response $default
DISABLED=0
REQUIRED=1
OPTIONAL=2
yes_no_try() {
local value
AS_IF([test x"$1" = x""], [value="$2"], [value="$1"])
AS_CASE(["`echo "$value" | tr '[A-Z]' '[a-z]'`"],
[0|no|false|disable], [return $DISABLED],
[1|yes|true|enable], [return $REQUIRED],
[try|check], [return $OPTIONAL]
)
AC_MSG_ERROR([Invalid option value "$value"])
}
check_systemdsystemunitdir() {
AC_MSG_CHECKING([which system unit file directory to use])
PKG_CHECK_VAR([systemdsystemunitdir], [systemd], [systemdsystemunitdir])
AC_MSG_RESULT([${systemdsystemunitdir}])
test x"$systemdsystemunitdir" != x""
return $?
}
dnl ===============================================
dnl Configure Options
dnl ===============================================
dnl Actual library checks come later, but pkg-config can be used here to grab
dnl external values to use as defaults for configure options
dnl --enable-* options: build process
AC_ARG_ENABLE([quiet],
[AS_HELP_STRING([--enable-quiet],
[suppress make output unless there is an error @<:@no@:>@])]
)
yes_no_try "$enable_quiet" "no"
enable_quiet=$?
AC_ARG_ENABLE([fatal-warnings],
[AS_HELP_STRING([--enable-fatal-warnings],
[enable pedantic and fatal warnings for gcc @<:@try@:>@])],
)
yes_no_try "$enable_fatal_warnings" "try"
enable_fatal_warnings=$?
AC_ARG_ENABLE([hardening],
[AS_HELP_STRING([--enable-hardening],
[harden the resulting executables/libraries @<:@try@:>@])]
)
yes_no_try "$enable_hardening" "try"
enable_hardening=$?
dnl --enable-* options: features
AC_ARG_ENABLE([systemd],
[AS_HELP_STRING([--enable-systemd],
[enable support for managing resources via systemd @<:@try@:>@])]
)
yes_no_try "$enable_systemd" "try"
enable_systemd=$?
AC_ARG_ENABLE([upstart],
[AS_HELP_STRING([--enable-upstart],
[enable support for managing resources via Upstart (deprecated) @<:@try@:>@])]
)
yes_no_try "$enable_upstart" "try"
enable_upstart=$?
dnl --enable-* options: compatibility
AC_ARG_ENABLE([compat-2.0],
[AS_HELP_STRING([--enable-compat-2.0], m4_normalize([
- preserve certain output as it was in 2.0; this option will be
- available only for the lifetime of the 2.1 series @<:@no@:>@]))]
+ preserve certain output as it was in 2.0; this option will be
+ available only for the lifetime of the 2.1 series @<:@no@:>@]))]
)
yes_no_try "$enable_compat_2_0" "no"
enable_compat_2_0=$?
AS_IF([test $enable_compat_2_0 -ne $DISABLED],
[
AC_DEFINE_UNQUOTED([PCMK__COMPAT_2_0], [1],
[Keep certain output compatible with 2.0 release series])
PCMK_FEATURES="$PCMK_FEATURES compat-2.0"
]
)
# Add an option to create symlinks at the pre-2.0.0 daemon name locations, so
# that users and tools can continue to invoke those names directly (e.g., for
# meta-data). This option will be removed in a future release.
AC_ARG_ENABLE([legacy-links],
[AS_HELP_STRING([--enable-legacy-links],
[add symlinks for old daemon names (deprecated) @<:@no@:>@])]
)
yes_no_try "$enable_legacy_links" "no"
enable_legacy_links=$?
AM_CONDITIONAL([BUILD_LEGACY_LINKS], [test $enable_legacy_links -ne $DISABLED])
dnl --with-* options: basic parameters
dnl This argument is defined via an M4 macro so default can be a variable
AC_DEFUN([VERSION_ARG],
[AC_ARG_WITH([version],
[AS_HELP_STRING([--with-version=VERSION],
[override package version @<:@$1@:>@])],
[ PACEMAKER_VERSION="$withval" ],
[ PACEMAKER_VERSION="$PACKAGE_VERSION" ])]
)
VERSION_ARG(VERSION_NUMBER)
# Redefine PACKAGE_VERSION and VERSION according to PACEMAKER_VERSION in case
# the user used --with-version. Unfortunately, this can only affect the
# substitution variables and later uses in this file, not the config.h
# constants, so we have to be careful to use only PACEMAKER_VERSION in C code.
PACKAGE_VERSION=$PACEMAKER_VERSION
VERSION=$PACEMAKER_VERSION
CRM_DAEMON_USER=""
AC_ARG_WITH([daemon-user],
[AS_HELP_STRING([--with-daemon-user=USER],
[user to run unprivileged Pacemaker daemons as (advanced option: changing this may break other cluster components unless similarly configured) @<:@hacluster@:>@])],
[ CRM_DAEMON_USER="$withval" ]
)
CRM_DAEMON_GROUP=""
AC_ARG_WITH([daemon-group],
[AS_HELP_STRING([--with-daemon-group=GROUP],
[group to run unprivileged Pacemaker daemons as (advanced option: changing this may break other cluster components unless similarly configured) @<:@haclient@:>@])],
[ CRM_DAEMON_GROUP="$withval" ]
)
BUG_URL=""
AC_ARG_WITH([bug-url],
[AS_HELP_STRING([--with-bug-url=DIR], m4_normalize([
- address where users should submit bug reports
- @<:@https://bugs.clusterlabs.org/enter_bug.cgi?product=Pacemaker@:>@]))],
+ address where users should submit bug reports
+ @<:@https://bugs.clusterlabs.org/enter_bug.cgi?product=Pacemaker@:>@]))],
[ BUG_URL="$withval" ]
)
dnl --with-* options: features
AC_ARG_WITH([cibsecrets],
[AS_HELP_STRING([--with-cibsecrets],
[support separate file for CIB secrets @<:@no@:>@])]
)
yes_no_try "$with_cibsecrets" "no"
with_cibsecrets=$?
AC_ARG_WITH([gnutls],
[AS_HELP_STRING([--with-gnutls],
[support Pacemaker Remote and remote-tls-port using GnuTLS @<:@try@:>@])]
)
yes_no_try "$with_gnutls" "try"
with_gnutls=$?
PCMK_GNUTLS_PRIORITIES="NORMAL"
AC_ARG_WITH([gnutls-priorities],
[AS_HELP_STRING([--with-gnutls-priorities],
[default GnuTLS cipher priorities @<:@NORMAL@:>@])],
[ test x"$withval" = x"no" || PCMK_GNUTLS_PRIORITIES="$withval" ]
)
AC_ARG_WITH([concurrent-fencing-default],
[AS_HELP_STRING([--with-concurrent-fencing-default],
[default value for concurrent-fencing cluster option @<:@false@:>@])],
)
AS_CASE([$with_concurrent_fencing_default],
- [""], [with_concurrent_fencing_default="false"],
+ [""], [with_concurrent_fencing_default="false"],
[false], [],
[true], [PCMK_FEATURES="$PCMK_FEATURES default-concurrent-fencing"],
[AC_MSG_ERROR([Invalid value "$with_concurrent_fencing_default" for --with-concurrent-fencing-default])]
)
AC_DEFINE_UNQUOTED([PCMK__CONCURRENT_FENCING_DEFAULT],
["$with_concurrent_fencing_default"],
[Default value for concurrent-fencing cluster option])
+AC_ARG_WITH([sbd-sync-default],
+ [AS_HELP_STRING([--with-sbd-sync-default], m4_normalize([
+ default value used by sbd if SBD_SYNC_RESOURCE_STARTUP
+ environment variable is not set @<:@false@:>@]))],
+)
+AS_CASE([$with_sbd_sync_default],
+ [""], [with_sbd_sync_default=false],
+ [false], [],
+ [true], [PCMK_FEATURES="$PCMK_FEATURES default-sbd-sync"],
+ [AC_MSG_ERROR([Invalid value "$with_sbd_sync_default" for --with-sbd-sync-default])]
+)
+AC_DEFINE_UNQUOTED([PCMK__SBD_SYNC_DEFAULT],
+ [$with_sbd_sync_default],
+ [Default value for SBD_SYNC_RESOURCE_STARTUP environment variable])
+
AC_ARG_WITH([resource-stickiness-default],
[AS_HELP_STRING([--with-resource-stickiness-default],
[If positive, value to add to new CIBs as explicit resource default for resource-stickiness @<:@0@:>@])],
)
errmsg="Invalid value \"$with_resource_stickiness_default\" for --with-resource-stickiness-default"
AS_CASE([$with_resource_stickiness_default],
[0|""], [with_resource_stickiness_default="0"],
[*[[!0-9]]*], [AC_MSG_ERROR([$errmsg])],
[PCMK_FEATURES="$PCMK_FEATURES default-resource-stickiness"]
)
AC_DEFINE_UNQUOTED([PCMK__RESOURCE_STICKINESS_DEFAULT],
[$with_resource_stickiness_default],
[Default value for resource-stickiness resource meta-attribute])
AC_ARG_WITH([corosync],
[AS_HELP_STRING([--with-corosync],
[support the Corosync messaging and membership layer @<:@try@:>@])]
)
yes_no_try "$with_corosync" "try"
with_corosync=$?
AC_ARG_WITH([nagios],
[AS_HELP_STRING([--with-nagios], [support nagios resources])]
)
yes_no_try "$with_nagios" "try"
with_nagios=$?
dnl --with-* options: directory locations
AC_ARG_WITH([nagios-plugin-dir],
[AS_HELP_STRING([--with-nagios-plugin-dir=DIR],
[directory for nagios plugins @<:@LIBEXECDIR/nagios/plugins@:>@])],
[ NAGIOS_PLUGIN_DIR="$withval" ]
)
AC_ARG_WITH([nagios-metadata-dir],
[AS_HELP_STRING([--with-nagios-metadata-dir=DIR],
[directory for nagios plugins metadata @<:@DATADIR/nagios/plugins-metadata@:>@])],
[ NAGIOS_METADATA_DIR="$withval" ]
)
INITDIR=""
AC_ARG_WITH([initdir],
[AS_HELP_STRING([--with-initdir=DIR],
[directory for init (rc) scripts])],
[ INITDIR="$withval" ]
)
systemdsystemunitdir="${systemdsystemunitdir-}"
AC_ARG_WITH([systemdsystemunitdir],
[AS_HELP_STRING([--with-systemdsystemunitdir=DIR],
[directory for systemd unit files (advanced option: must match what systemd uses)])],
[ systemdsystemunitdir="$withval" ]
)
CONFIGDIR=""
AC_ARG_WITH([configdir],
[AS_HELP_STRING([--with-configdir=DIR],
[directory for Pacemaker configuration file @<:@SYSCONFDIR/sysconfig@:>@])],
[ CONFIGDIR="$withval" ]
)
dnl --runstatedir is available as of autoconf 2.70 (2020-12-08). When users
dnl have an older version, they can use our --with-runstatedir.
pcmk_runstatedir=""
AC_ARG_WITH([runstatedir],
[AS_HELP_STRING([--with-runstatedir=DIR],
[modifiable per-process data @<:@LOCALSTATEDIR/run@:>@ (ignored if --runstatedir is available)])],
[ pcmk_runstatedir="$withval" ]
)
CRM_LOG_DIR=""
AC_ARG_WITH([logdir],
[AS_HELP_STRING([--with-logdir=DIR],
[directory for Pacemaker log file @<:@LOCALSTATEDIR/log/pacemaker@:>@])],
[ CRM_LOG_DIR="$withval" ]
)
CRM_BUNDLE_DIR=""
AC_ARG_WITH([bundledir],
[AS_HELP_STRING([--with-bundledir=DIR],
[directory for Pacemaker bundle logs @<:@LOCALSTATEDIR/log/pacemaker/bundles@:>@])],
[ CRM_BUNDLE_DIR="$withval" ]
)
dnl Get default from resource-agents if possible. Otherwise, the default uses
dnl /usr/lib rather than libdir because it's determined by the OCF project and
dnl not Pacemaker. Even if a user wants to install Pacemaker to /usr/local or
dnl such, the OCF agents will be expected in their usual location. However, we
dnl do give the user the option to override it.
PKG_CHECK_VAR([OCF_ROOT_DIR], [resource-agents], [ocfrootdir], [],
[OCF_ROOT_DIR="/usr/lib/ocf"])
AC_ARG_WITH([ocfdir],
[AS_HELP_STRING([--with-ocfdir=DIR], m4_normalize([
OCF resource agent root directory (advanced option: changing this
may break other cluster components unless similarly configured)
@<:@value from resource-agents package if available otherwise
/usr/lib/ocf@:>@]))],
[ OCF_ROOT_DIR="$withval" ]
)
AC_SUBST(OCF_ROOT_DIR)
AC_DEFINE_UNQUOTED([OCF_ROOT_DIR], ["$OCF_ROOT_DIR"],
[OCF root directory for resource agents and libraries])
PKG_CHECK_VAR([OCF_RA_PATH], [resource-agents], [ocfrapath], [],
- [OCF_RA_PATH="$OCF_ROOT_DIR/resource.d"])
+ [OCF_RA_PATH="$OCF_ROOT_DIR/resource.d"])
AC_ARG_WITH([ocfrapath],
[AS_HELP_STRING([--with-ocfrapath=DIR], m4_normalize([
OCF resource agent directories (colon-separated) to search
@<:@value from resource-agents package if available otherwise
- OCFDIR/resource.d@:>@]))],
+ OCFDIR/resource.d@:>@]))],
[ OCF_RA_PATH="$withval" ]
)
AC_SUBST(OCF_RA_PATH)
AC_DEFINE_UNQUOTED([OCF_RA_PATH], ["$OCF_RA_PATH"],
[OCF directories to search for resource agents ])
OCF_RA_INSTALL_DIR="$OCF_ROOT_DIR/resource.d"
AC_ARG_WITH([ocfrainstalldir],
[AS_HELP_STRING([--with-ocfrainstalldir=DIR], m4_normalize([
OCF installation directory for Pacemakers resource agents
@<:@OCFDIR/resource.d@:>@]))],
[ OCF_RA_INSTALL_DIR="$withval" ]
)
AC_SUBST(OCF_RA_INSTALL_DIR)
dnl Get default from fence-agents if available
PKG_CHECK_VAR([FA_PREFIX], [fence-agents], [prefix],
[PCMK__FENCE_BINDIR="${FA_PREFIX}/sbin"],
[PCMK__FENCE_BINDIR="$sbindir"])
AC_ARG_WITH([fence-bindir],
[AS_HELP_STRING([--with-fence-bindir=DIR], m4_normalize([
directory for executable fence agents @<:@value from fence-agents
package if available otherwise SBINDIR@:>@]))],
[ PCMK__FENCE_BINDIR="$withval" ]
)
AC_SUBST(PCMK__FENCE_BINDIR)
dnl --with-* options: non-production testing
AC_ARG_WITH([profiling],
[AS_HELP_STRING([--with-profiling],
[disable optimizations, for effective profiling @<:@no@:>@])]
)
yes_no_try "$with_profiling" "no"
with_profiling=$?
AC_ARG_WITH([coverage],
[AS_HELP_STRING([--with-coverage],
[disable optimizations, for effective profiling and coverage testing @<:@no@:>@])]
)
yes_no_try "$with_coverage" "no"
with_coverage=$?
AC_ARG_WITH([sanitizers],
[AS_HELP_STRING([--with-sanitizers=...,...],
[enable SANitizer build, do *NOT* use for production. Only ASAN/UBSAN/TSAN are currently supported])],
[ SANITIZERS="$withval" ],
[ SANITIZERS="" ])
dnl ===============================================
dnl General Processing
dnl ===============================================
AC_DEFINE_UNQUOTED(PACEMAKER_VERSION, "$VERSION",
[Version number of this Pacemaker build])
PACKAGE_SERIES=`echo $VERSION | awk -F. '{ print $1"."$2 }'`
AC_SUBST(PACKAGE_SERIES)
AC_PROG_LN_S
AC_PROG_MKDIR_P
# Check for fatal warning support
AS_IF([test $enable_fatal_warnings -ne $DISABLED && test "$GCC" = "yes" && cc_supports_flag -Werror],
[WERROR="-Werror"],
[
WERROR=""
AS_CASE([$enable_fatal_warnings],
[$REQUIRED], [AC_MSG_ERROR([Compiler does not support fatal warnings])],
[$OPTIONAL], [
AC_MSG_NOTICE([Compiler does not support fatal warnings])
enable_fatal_warnings=$DISABLED
])
])
AC_MSG_NOTICE([Sanitizing prefix: ${prefix}])
AS_IF([test "$prefix" = "NONE"],
[
prefix=/usr
dnl Fix default variables - "prefix" variable if not specified
AS_IF([test "$localstatedir" = "\${prefix}/var"],
[localstatedir="/var"])
AS_IF([test "$sysconfdir" = "\${prefix}/etc"],
[sysconfdir="/etc"])
])
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 ${mandir}`"
dnl Home-grown variables
if [ test "x${runstatedir}" = "x" ]; then
if [ test "x${pcmk_runstatedir}" = "x" ]; then
runstatedir="${localstatedir}/run"
else
runstatedir="${pcmk_runstatedir}"
fi
fi
eval runstatedir="$(eval echo ${runstatedir})"
AC_DEFINE_UNQUOTED([PCMK_RUN_DIR], ["$runstatedir"],
- [Location for modifiable per-process data])
+ [Location for modifiable per-process data])
AC_SUBST(runstatedir)
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)
eval PCMK__FENCE_BINDIR="`eval echo ${PCMK__FENCE_BINDIR}`"
AC_DEFINE_UNQUOTED(PCMK__FENCE_BINDIR,"$PCMK__FENCE_BINDIR",
[Location for executable fence agents])
AS_IF([test x"${PCMK_GNUTLS_PRIORITIES}" != x""], [],
[AC_MSG_ERROR([--with-gnutls-priorities value must not be empty])])
AC_DEFINE_UNQUOTED([PCMK_GNUTLS_PRIORITIES], ["$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 (yet)])
fi
done
us_auth=
AC_CHECK_HEADER([sys/socket.h], [
AC_CHECK_DECL([SO_PEERCRED], [
# Linux
AC_CHECK_TYPE([struct ucred], [
us_auth=peercred_ucred;
AC_DEFINE([US_AUTH_PEERCRED_UCRED], [1],
[Define if Unix socket auth method is
getsockopt(s, SO_PEERCRED, &ucred, ...)])
], [
# OpenBSD
AC_CHECK_TYPE([struct sockpeercred], [
us_auth=localpeercred_sockepeercred;
AC_DEFINE([US_AUTH_PEERCRED_SOCKPEERCRED], [1],
[Define if Unix socket auth method is
getsockopt(s, SO_PEERCRED, &sockpeercred, ...)])
], [], [[#include <sys/socket.h>]])
], [[#define _GNU_SOURCE
#include <sys/socket.h>]])
], [], [[#include <sys/socket.h>]])
])
AS_IF([test -z "${us_auth}"], [
# FreeBSD
AC_CHECK_DECL([getpeereid], [
us_auth=getpeereid;
AC_DEFINE([US_AUTH_GETPEEREID], [1],
[Define if Unix socket auth method is
getpeereid(s, &uid, &gid)])
], [
# Solaris/OpenIndiana
AC_CHECK_DECL([getpeerucred], [
us_auth=getpeerucred;
AC_DEFINE([US_AUTH_GETPEERUCRED], [1],
[Define if Unix socket auth method is
getpeercred(s, &ucred)])
], [
AC_MSG_FAILURE([No way to authenticate a Unix socket peer])
], [[#include <ucred.h>]])
])
])
dnl OS-based decision-making is poor autotools practice; feature-based
dnl mechanisms are strongly preferred. Keep this section to a bare minimum;
dnl regard as a "necessary evil".
INIT_EXT=""
PROCFS=0
case "$host_os" in
dnl Solaris and some *BSD versions support procfs but not files we need
*bsd*)
INIT_EXT=".sh"
;;
*linux*)
PROCFS=1
;;
darwin*)
LIBS="$LIBS -L${prefix}/lib"
CFLAGS="$CFLAGS -I${prefix}/include"
;;
esac
AC_SUBST(INIT_EXT)
AC_DEFINE_UNQUOTED([SUPPORT_PROCFS], [$PROCFS],
- [Define to 1 if procfs is supported])
+ [Define to 1 if procfs is supported])
case "$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 <inttypes.h>]]
)
]
)
(
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)
AS_IF([test "x${LIBTOOL}" != "x"], [],
[AC_MSG_FAILURE([Could not find required build tool libtool (or equivalent)])])
dnl Pacemaker's executable python scripts will invoke the python specified by
dnl configure's PYTHON variable. If not specified, AM_PATH_PYTHON will check a
dnl built-in list with (unversioned) "python" having precedence. To configure
dnl Pacemaker to use a specific python interpreter version, define PYTHON
dnl when calling configure, for example: ./configure PYTHON=/usr/bin/python3.6
dnl Ensure PYTHON is an absolute path
AS_IF([test x"${PYTHON}" != x""], [AC_PATH_PROG([PYTHON], [$PYTHON])])
dnl Require a minimum Python version
AM_PATH_PYTHON([3.4])
AC_PATH_PROGS([ASCIIDOC_CONV], [asciidoc asciidoctor])
AC_PATH_PROG([HELP2MAN], [help2man])
AC_PATH_PROG([SPHINX], [sphinx-build])
AC_PATH_PROG([INKSCAPE], [inkscape])
AC_PATH_PROG([XSLTPROC], [xsltproc])
AC_PATH_PROG([XMLCATALOG], [xmlcatalog])
dnl Bash is needed for building man pages and running regression tests.
dnl BASH is already an environment variable, so use something else.
AC_PATH_PROG([BASH_PATH], [bash])
AS_IF([test "x${BASH_PATH}" != "x"], [],
[AC_MSG_FAILURE([Could not find required build tool bash])])
AC_PATH_PROGS(VALGRIND_BIN, valgrind, /usr/bin/valgrind)
AC_DEFINE_UNQUOTED(VALGRIND_BIN, "$VALGRIND_BIN", Valgrind command)
AM_CONDITIONAL(BUILD_HELP, test x"${HELP2MAN}" != x"")
if test x"${HELP2MAN}" != x""; then
PCMK_FEATURES="$PCMK_FEATURES generated-manpages"
fi
MANPAGE_XSLT=""
if test x"${XSLTPROC}" != x""; then
AC_MSG_CHECKING([for DocBook-to-manpage transform])
# first try to figure out correct template using xmlcatalog query,
# resort to extensive (semi-deterministic) file search if that fails
DOCBOOK_XSL_URI='http://docbook.sourceforge.net/release/xsl/current'
DOCBOOK_XSL_PATH='manpages/docbook.xsl'
MANPAGE_XSLT=$(${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
AM_CONDITIONAL([BUILD_SPHINX_DOCS],
[test x"${SPHINX}" != x"" && test x"${INKSCAPE}" != x""])
AM_COND_IF([BUILD_SPHINX_DOCS], [PCMK_FEATURES="$PCMK_FEATURES books"])
dnl Pacemaker's shell scripts (and thus man page builders) rely on GNU getopt
AC_MSG_CHECKING([for GNU-compatible getopt])
IFS_orig=$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
AS_IF([test -n "$GETOPT_PATH"], [AC_MSG_RESULT([$GETOPT_PATH])],
[
AC_MSG_RESULT([no])
AC_MSG_ERROR([Could not find required build tool GNU-compatible getopt])
])
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 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 tolerably-fast replacement function for it.
dnl
dnl strndup: is a gnu function similar to strdup, but safer.
dnl We wrote a tolerably-fast replacement function for it.
AC_REPLACE_FUNCS(alphasort NoSuchFunctionName scandir strerror strchrnul 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)
PKG_CHECK_MODULES([UUID], [uuid],
[CPPFLAGS="${CPPFLAGS} ${UUID_CFLAGS}"
LIBS="${LIBS} ${UUID_LIBS}"])
AC_CHECK_FUNCS([sched_setscheduler])
if test "$ac_cv_func_sched_setscheduler" != yes; then
- PC_LIBS_RT=""
+ PC_LIBS_RT=""
else
- PC_LIBS_RT="-lrt"
+ PC_LIBS_RT="-lrt"
fi
AC_SUBST(PC_LIBS_RT)
# Require minimum glib version
PKG_CHECK_MODULES([GLIB], [glib-2.0 >= 2.42.0],
[CPPFLAGS="${CPPFLAGS} ${GLIB_CFLAGS}"
LIBS="${LIBS} ${GLIB_LIBS}"])
# Check whether high-resolution sleep function is available
AC_CHECK_FUNCS([nanosleep usleep])
#
# Where is dlopen?
#
if test "$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
PKG_CHECK_MODULES(LIBXML2, [libxml-2.0],
[CPPFLAGS="${CPPFLAGS} ${LIBXML2_CFLAGS}"
LIBS="${LIBS} ${LIBXML2_LIBS}"])
REQUIRE_LIB([xslt], [xsltApplyStylesheet])
dnl ========================================================================
dnl Headers
dnl ========================================================================
# Some distributions insert #warnings into deprecated headers. If we will
# enable fatal warnings for the build, then enable them for the header checks
# as well, otherwise the build could fail even though the header check
# succeeds. (We should probably be doing this in more places.)
cc_temp_flags "$CFLAGS $WERROR"
# Optional headers (inclusion of these should be conditional in C code)
AC_CHECK_HEADERS([getopt.h])
AC_CHECK_HEADERS([linux/swab.h])
AC_CHECK_HEADERS([stddef.h])
AC_CHECK_HEADERS([sys/signalfd.h])
AC_CHECK_HEADERS([uuid/uuid.h])
AC_CHECK_HEADERS([security/pam_appl.h pam/pam_appl.h])
# Required headers
REQUIRE_HEADER([arpa/inet.h])
REQUIRE_HEADER([ctype.h])
REQUIRE_HEADER([dirent.h])
REQUIRE_HEADER([errno.h])
REQUIRE_HEADER([glib.h])
REQUIRE_HEADER([grp.h])
REQUIRE_HEADER([limits.h])
REQUIRE_HEADER([netdb.h])
REQUIRE_HEADER([netinet/in.h])
REQUIRE_HEADER([netinet/ip.h], [
#include <sys/types.h>
#include <netinet/in.h>
])
REQUIRE_HEADER([pwd.h])
REQUIRE_HEADER([signal.h])
REQUIRE_HEADER([stdio.h])
REQUIRE_HEADER([stdlib.h])
REQUIRE_HEADER([string.h])
REQUIRE_HEADER([strings.h])
REQUIRE_HEADER([sys/ioctl.h])
REQUIRE_HEADER([sys/param.h])
REQUIRE_HEADER([sys/reboot.h])
REQUIRE_HEADER([sys/resource.h])
REQUIRE_HEADER([sys/socket.h])
REQUIRE_HEADER([sys/stat.h])
REQUIRE_HEADER([sys/time.h])
REQUIRE_HEADER([sys/types.h])
REQUIRE_HEADER([sys/utsname.h])
REQUIRE_HEADER([sys/wait.h])
REQUIRE_HEADER([time.h])
REQUIRE_HEADER([unistd.h])
REQUIRE_HEADER([libxml/xpath.h])
REQUIRE_HEADER([libxslt/xslt.h])
cc_restore_flags
AC_CHECK_FUNCS([uuid_unparse], [],
[AC_MSG_FAILURE([Could not find required C function uuid_unparse()])])
AC_CACHE_CHECK([whether __progname and __progname_full are available],
[pf_cv_var_progname], AC_LINK_IFELSE([
AC_LANG_PROGRAM([[extern char *__progname, *__progname_full;]],
[[__progname = "foo"; __progname_full = "foo bar";]],
[pf_cv_var_progname="yes"],
[pf_cv_var_progname="no"])
]))
AS_IF([test "$pf_cv_var_progname" = "yes"], [AC_DEFINE(HAVE___PROGNAME,1,[ ])])
dnl ========================================================================
dnl Generic declarations
dnl ========================================================================
AC_CHECK_DECLS([CLOCK_MONOTONIC], [PCMK_FEATURES="$PCMK_FEATURES monotonic"], [], [[
#include <time.h>
]])
dnl ========================================================================
dnl Structures
dnl ========================================================================
AC_CHECK_MEMBERS([struct tm.tm_gmtoff],,,[[#include <time.h>]])
AC_CHECK_MEMBER([struct dirent.d_type],
AC_DEFINE(HAVE_STRUCT_DIRENT_D_TYPE,1,[Define this if struct dirent has d_type]),,
[#include <dirent.h>])
dnl ========================================================================
dnl Functions
dnl ========================================================================
REQUIRE_FUNC([getopt])
REQUIRE_FUNC([setenv])
REQUIRE_FUNC([unsetenv])
AC_CACHE_CHECK(whether sscanf supports %m,
pf_cv_var_sscanf,
AC_RUN_IFELSE([AC_LANG_SOURCE([[
#include <stdio.h>
const char *s = "some-command-line-arg";
int main(int argc, char **argv) {
char *name = NULL;
int n = sscanf(s, "%ms", &name);
return n == 1 ? 0 : 1;
}
]])],
pf_cv_var_sscanf="yes", pf_cv_var_sscanf="no", pf_cv_var_sscanf="no"))
AS_IF([test "$pf_cv_var_sscanf" = "yes"], [AC_DEFINE(SSCANF_HAS_M, 1, [ ])])
dnl ========================================================================
dnl bzip2
dnl ========================================================================
REQUIRE_HEADER([bzlib.h])
REQUIRE_LIB([bz2], [BZ2_bzBuffToBuffCompress])
dnl ========================================================================
dnl sighandler_t is missing from Illumos, Solaris11 systems
dnl ========================================================================
AC_MSG_CHECKING([for sighandler_t])
AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[#include <signal.h>]], [[sighandler_t *f;]])],
[
AC_MSG_RESULT([yes])
AC_DEFINE([HAVE_SIGHANDLER_T], [1],
[Define to 1 if sighandler_t is available])
],
[AC_MSG_RESULT([no])])
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 ncurses takes precedence.
dnl
AC_CHECK_HEADERS([curses.h curses/curses.h ncurses.h ncurses/ncurses.h])
dnl Although n-library is preferred, only look for it if the n-header was found.
CURSESLIBS=''
PC_NAME_CURSES=""
PC_LIBS_CURSES=""
AS_IF([test "$ac_cv_header_ncurses_h" = "yes"], [
AC_CHECK_LIB(ncurses, printw,
[AC_DEFINE(HAVE_LIBNCURSES,1, have ncurses library)])
CURSESLIBS=`$PKG_CONFIG --libs ncurses` || CURSESLIBS='-lncurses'
PC_NAME_CURSES="ncurses"
])
AS_IF([test "$ac_cv_header_ncurses_ncurses_h" = "yes"], [
AC_CHECK_LIB(ncurses, printw,
[AC_DEFINE(HAVE_LIBNCURSES,1, have ncurses library)])
CURSESLIBS=`$PKG_CONFIG --libs ncurses` || CURSESLIBS='-lncurses'
PC_NAME_CURSES="ncurses"
])
dnl Only look for non-n-library if there was no n-library.
AS_IF([test X"$CURSESLIBS" = X"" && test "$ac_cv_header_curses_h" = "yes"], [
AC_CHECK_LIB(curses, printw,
[CURSESLIBS='-lcurses'; AC_DEFINE(HAVE_LIBCURSES,1, have curses library)])
PC_LIBS_CURSES="$CURSESLIBS"
])
dnl Only look for non-n-library if there was no n-library.
AS_IF([test X"$CURSESLIBS" = X"" && test "$ac_cv_header_curses_curses_h" = "yes"], [
AC_CHECK_LIB(curses, printw,
[CURSESLIBS='-lcurses'; AC_DEFINE(HAVE_LIBCURSES,1, have curses library)])
PC_LIBS_CURSES="$CURSESLIBS"
])
if test "x$CURSESLIBS" != "x"; then
PCMK_FEATURES="$PCMK_FEATURES ncurses"
fi
dnl Check for printw() prototype compatibility
AS_IF([test X"$CURSESLIBS" != X"" && cc_supports_flag -Wcast-qual], [
ac_save_LIBS=$LIBS
LIBS="$CURSESLIBS"
cc_temp_flags "-Wcast-qual $WERROR"
# avoid broken test because of hardened build environment in Fedora 23+
# - https://fedoraproject.org/wiki/Changes/Harden_All_Packages
# - https://bugzilla.redhat.com/1297985
AS_IF([cc_supports_flag -fPIC], [CFLAGS="$CFLAGS -fPIC"])
AC_MSG_CHECKING([whether curses library is compatible])
AC_LINK_IFELSE(
[AC_LANG_PROGRAM([
#if defined(HAVE_NCURSES_H)
# include <ncurses.h>
#elif defined(HAVE_NCURSES_NCURSES_H)
# include <ncurses/ncurses.h>
#elif defined(HAVE_CURSES_H)
# include <curses.h>
#endif
],
[printw((const char *)"Test");]
)],
[AC_MSG_RESULT([yes])],
[
AC_MSG_RESULT([no])
AC_MSG_WARN(m4_normalize([Disabling curses because the printw()
function of your (n)curses library is old.
If you wish to enable curses, update to a
newer version (ncurses 5.4 or later is
recommended, available from
https://invisible-island.net/ncurses/)
]))
AC_DEFINE([HAVE_INCOMPATIBLE_PRINTW], [1],
[Define to 1 if curses library has incompatible printw()])
]
)
LIBS=$ac_save_LIBS
cc_restore_flags
])
AC_SUBST(CURSESLIBS)
AC_SUBST(PC_NAME_CURSES)
AC_SUBST(PC_LIBS_CURSES)
dnl ========================================================================
dnl Profiling and GProf
dnl ========================================================================
CFLAGS_ORIG="$CFLAGS"
AS_IF([test $with_coverage -ne $DISABLED],
[
with_profiling=$REQUIRED
PCMK_FEATURES="$PCMK_FEATURES coverage"
CFLAGS="$CFLAGS -fprofile-arcs -ftest-coverage"
dnl During linking, make sure to specify -lgcov or -coverage
]
)
AS_IF([test $with_profiling -ne $DISABLED],
[
with_profiling=$REQUIRED
PCMK_FEATURES="$PCMK_FEATURES profile"
dnl Disable various compiler optimizations
CFLAGS="$CFLAGS -fno-omit-frame-pointer -fno-inline -fno-builtin"
dnl CFLAGS="$CFLAGS -fno-inline-functions"
dnl CFLAGS="$CFLAGS -fno-default-inline"
dnl CFLAGS="$CFLAGS -fno-inline-functions-called-once"
dnl CFLAGS="$CFLAGS -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"
AC_MSG_NOTICE([CFLAGS before adding profiling options: $CFLAGS_ORIG])
AC_MSG_NOTICE([CFLAGS after: $CFLAGS])
]
)
AC_DEFINE_UNQUOTED([SUPPORT_PROFILING], [$with_profiling], [Support profiling])
dnl ========================================================================
dnl Cluster infrastructure - LibQB
dnl ========================================================================
PKG_CHECK_MODULES(libqb, libqb >= 0.17)
CPPFLAGS="$libqb_CFLAGS $CPPFLAGS"
LIBS="$libqb_LIBS $LIBS"
dnl libqb 2.0.2+ (2020-10)
AC_CHECK_FUNCS(qb_ipcc_auth_get,
AC_DEFINE(HAVE_IPCC_AUTH_GET, 1,
[Have qb_ipcc_auth_get function]))
dnl libqb 2.0.0+ (2020-05)
CHECK_ENUM_VALUE([qb/qblog.h],[qb_log_conf],[QB_LOG_CONF_MAX_LINE_LEN])
CHECK_ENUM_VALUE([qb/qblog.h],[qb_log_conf],[QB_LOG_CONF_ELLIPSIS])
dnl Support Linux-HA fence agents if available
if test "$cross_compiling" != "yes"; then
CPPFLAGS="$CPPFLAGS -I${prefix}/include/heartbeat"
fi
AC_CHECK_HEADERS([stonith/stonith.h],
[
AC_CHECK_LIB([pils], [PILLoadPlugin])
AC_CHECK_LIB([plumb], [G_main_add_IPC_Channel])
PCMK_FEATURES="$PCMK_FEATURES lha"
])
AM_CONDITIONAL([BUILD_LHA_SUPPORT], [test "$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)
if test x"${CRM_DAEMON_USER}" = x""; then
CRM_DAEMON_USER="hacluster"
fi
AC_DEFINE_UNQUOTED(CRM_DAEMON_USER,"$CRM_DAEMON_USER", User to run Pacemaker daemons as)
AC_SUBST(CRM_DAEMON_USER)
if test x"${CRM_DAEMON_GROUP}" = x""; then
CRM_DAEMON_GROUP="haclient"
fi
AC_DEFINE_UNQUOTED(CRM_DAEMON_GROUP,"$CRM_DAEMON_GROUP", Group to run Pacemaker daemons as)
AC_SUBST(CRM_DAEMON_GROUP)
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)
CRM_STATE_DIR="${runstatedir}/crm"
AC_DEFINE_UNQUOTED([CRM_STATE_DIR], ["$CRM_STATE_DIR"],
- [Where to keep state files and sockets])
+ [Where to keep state files and sockets])
AC_SUBST(CRM_STATE_DIR)
CRM_RSCTMP_DIR="${runstatedir}/resource-agents"
AC_DEFINE_UNQUOTED(CRM_RSCTMP_DIR,"$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)
AC_DEFINE_UNQUOTED(SBIN_DIR,"$sbindir",[Location for system binaries])
AC_PATH_PROGS(GIT, git false)
AC_MSG_CHECKING([build version])
BUILD_VERSION=$Format:%h$
if test $BUILD_VERSION != ":%h$"; then
AC_MSG_RESULT([$BUILD_VERSION (archive hash)])
elif test -x $GIT && test -d .git; then
BUILD_VERSION=`$GIT log --pretty="format:%h" -n 1`
AC_MSG_RESULT([$BUILD_VERSION (git hash)])
else
# The current directory name make a reasonable default
# Most generated archives will include the hash or tag
BASE=`basename $PWD`
BUILD_VERSION=`echo $BASE | sed s:.*[[Pp]]acemaker-::`
AC_MSG_RESULT([$BUILD_VERSION (directory name)])
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} ${DBUS_CFLAGS}"],
- [HAVE_dbus=0])
+ [CPPFLAGS="${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 <dbus/dbus.h>]])
if test $HAVE_dbus = 0; then
- PC_NAME_DBUS=""
+ PC_NAME_DBUS=""
else
- PC_NAME_DBUS="dbus-1"
+ PC_NAME_DBUS="dbus-1"
fi
AC_SUBST(PC_NAME_DBUS)
AS_CASE([$enable_systemd],
[$REQUIRED], [
AS_IF([test $HAVE_dbus = 0],
[AC_MSG_FAILURE([Cannot support systemd resources without DBus])])
AS_IF([test "$ac_cv_have_decl_CLOCK_MONOTONIC" = "no"],
[AC_MSG_FAILURE([Cannot support systemd resources without monotonic clock])])
AS_IF([check_systemdsystemunitdir], [],
[AC_MSG_FAILURE([Cannot support systemd resources without systemdsystemunitdir])])
],
[$OPTIONAL], [
AS_IF([test $HAVE_dbus = 0 \
|| test "$ac_cv_have_decl_CLOCK_MONOTONIC" = "no"],
[enable_systemd=$DISABLED],
[
AC_MSG_CHECKING([for systemd version (using dbus-send)])
ret=$({ 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 "version unavailable"; } | 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}])
AS_IF([test "$ret" != "unavailable" \
|| systemctl --version 2>/dev/null | grep -q systemd],
[
AS_IF([check_systemdsystemunitdir],
[enable_systemd=$REQUIRED],
[enable_systemd=$DISABLED])
],
[enable_systemd=$DISABLED]
)
])
],
)
AC_MSG_CHECKING([whether to enable support for managing resources via systemd])
AS_IF([test $enable_systemd -eq $DISABLED], [AC_MSG_RESULT([no])],
[
AC_MSG_RESULT([yes])
PCMK_FEATURES="$PCMK_FEATURES systemd"
]
)
AC_SUBST([systemdsystemunitdir])
AC_DEFINE_UNQUOTED([SUPPORT_SYSTEMD], [$enable_systemd],
[Support systemd resources])
AM_CONDITIONAL([BUILD_SYSTEMD], [test $enable_systemd = $REQUIRED])
AC_SUBST(SUPPORT_SYSTEMD)
AS_CASE([$enable_upstart],
[$REQUIRED], [
AS_IF([test $HAVE_dbus = 0],
[AC_MSG_FAILURE([Cannot support Upstart resources without DBus])])
],
[$OPTIONAL], [
AS_IF([test $HAVE_dbus = 0], [enable_upstart=$DISABLED],
[
AC_MSG_CHECKING([for Upstart version (using dbus-send)])
ret=$({ dbus-send --system --print-reply \
--dest=com.ubuntu.Upstart \
/com/ubuntu/Upstart org.freedesktop.DBus.Properties.Get \
string:com.ubuntu.Upstart0_6 string:version 2>/dev/null \
|| echo "version unavailable"; } | tail -n1)
# sanitize output a bit (interested just in value, not type),
# ret is intentionally unenquoted so as to normalize whitespace
ret=$(echo ${ret} | cut -d' ' -f2-)
AC_MSG_RESULT([${ret}])
AS_IF([test "$ret" != "unavailable" \
|| initctl --version 2>/dev/null | grep -q upstart],
[enable_upstart=$REQUIRED],
[enable_upstart=$DISABLED]
)
])
],
)
AC_MSG_CHECKING([whether to enable support for managing resources via Upstart])
AS_IF([test $enable_upstart -eq $DISABLED], [AC_MSG_RESULT([no])],
[
AC_MSG_RESULT([yes])
PCMK_FEATURES="$PCMK_FEATURES upstart"
]
)
AC_DEFINE_UNQUOTED([SUPPORT_UPSTART], [$enable_upstart],
[Support Upstart resources])
AM_CONDITIONAL([BUILD_UPSTART], [test $enable_upstart -eq $REQUIRED])
AC_SUBST(SUPPORT_UPSTART)
AS_CASE([$with_nagios],
[$REQUIRED], [
AS_IF([test "$ac_cv_have_decl_CLOCK_MONOTONIC" = "no"],
[AC_MSG_FAILURE([Cannot support nagios resources without monotonic clock])])
],
[$OPTIONAL], [
AS_IF([test "$ac_cv_have_decl_CLOCK_MONOTONIC" = "no"],
[with_nagios=$DISABLED], [with_nagios=$REQUIRED])
]
)
AS_IF([test $with_nagios -eq $REQUIRED], [PCMK_FEATURES="$PCMK_FEATURES nagios"])
AC_DEFINE_UNQUOTED([SUPPORT_NAGIOS], [$with_nagios], [Support nagios plugins])
AM_CONDITIONAL([BUILD_NAGIOS], [test $with_nagios -eq $REQUIRED])
if test x"$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, "$NAGIOS_METADATA_DIR", Directory for nagios plugins metadata)
AC_SUBST(NAGIOS_METADATA_DIR)
STACKS=""
CLUSTERLIBS=""
PC_NAME_CLUSTER=""
dnl ========================================================================
dnl Cluster stack - Corosync
dnl ========================================================================
COROSYNC_LIBS=""
AS_CASE([$with_corosync],
[$REQUIRED], [
# These will be fatal if unavailable
PKG_CHECK_MODULES([cpg], [libcpg])
PKG_CHECK_MODULES([cfg], [libcfg])
PKG_CHECK_MODULES([cmap], [libcmap])
PKG_CHECK_MODULES([quorum], [libquorum])
PKG_CHECK_MODULES([libcorosync_common], [libcorosync_common])
]
[$OPTIONAL], [
PKG_CHECK_MODULES([cpg], [libcpg], [], [with_corosync=$DISABLED])
PKG_CHECK_MODULES([cfg], [libcfg], [], [with_corosync=$DISABLED])
PKG_CHECK_MODULES([cmap], [libcmap], [], [with_corosync=$DISABLED])
PKG_CHECK_MODULES([quorum], [libquorum], [], [with_corosync=$DISABLED])
PKG_CHECK_MODULES([libcorosync_common], [libcorosync_common], [], [with_corosync=$DISABLED])
AS_IF([test $with_corosync -ne $DISABLED], [with_corosync=$REQUIRED])
]
)
AS_IF([test $with_corosync -ne $DISABLED],
[
AC_MSG_CHECKING([for Corosync 2 or later])
AC_MSG_RESULT([yes])
CFLAGS="$CFLAGS $libqb_CFLAGS $cpg_CFLAGS $cfg_CFLAGS $cmap_CFLAGS $quorum_CFLAGS $libcorosync_common_CFLAGS"
COROSYNC_LIBS="$COROSYNC_LIBS $cpg_LIBS $cfg_LIBS $cmap_LIBS $quorum_LIBS $libcorosync_common_LIBS"
CLUSTERLIBS="$CLUSTERLIBS $COROSYNC_LIBS"
PC_NAME_CLUSTER="$PC_CLUSTER_NAME libcfg libcmap libcorosync_common libcpg libquorum"
STACKS="$STACKS corosync-ge-2"
dnl Shutdown tracking added (back) to corosync Jan 2021
saved_LIBS="$LIBS"
LIBS="$LIBS $COROSYNC_LIBS"
AC_CHECK_FUNCS(corosync_cfg_trackstart,
AC_DEFINE(HAVE_COROSYNC_CFG_TRACKSTART, 1,
[Have corosync_cfg_trackstart function]))
LIBS="$saved_LIBS"
]
)
AC_DEFINE_UNQUOTED([SUPPORT_COROSYNC], [$with_corosync],
[Support the Corosync messaging and membership layer])
AM_CONDITIONAL([BUILD_CS_SUPPORT], [test $with_corosync -eq $REQUIRED])
AC_SUBST([SUPPORT_COROSYNC])
dnl
dnl Cluster stack - Sanity
dnl
AS_IF([test "x$STACKS" != "x"], [AC_MSG_NOTICE([Supported stacks:${STACKS}])],
[AC_MSG_FAILURE([At least one cluster stack must be supported])])
PCMK_FEATURES="${PCMK_FEATURES}${STACKS}"
AC_SUBST(CLUSTERLIBS)
AC_SUBST(PC_NAME_CLUSTER)
dnl ========================================================================
dnl CIB secrets
dnl ========================================================================
AS_IF([test $with_cibsecrets -ne $DISABLED],
[
with_cibsecrets=$REQUIRED
PCMK_FEATURES="$PCMK_FEATURES cibsecrets"
LRM_CIBSECRETS_DIR="${localstatedir}/lib/pacemaker/lrm/secrets"
AC_DEFINE_UNQUOTED([LRM_CIBSECRETS_DIR], ["$LRM_CIBSECRETS_DIR"],
[Location for CIB secrets])
AC_SUBST([LRM_CIBSECRETS_DIR])
]
)
AC_DEFINE_UNQUOTED([SUPPORT_CIBSECRETS], [$with_cibsecrets], [Support CIB secrets])
AM_CONDITIONAL([BUILD_CIBSECRETS], [test $with_cibsecrets -eq $REQUIRED])
dnl ========================================================================
dnl GnuTLS
dnl ========================================================================
dnl Require GnuTLS >=2.12.0 (2011-03) for Pacemaker Remote support
PC_NAME_GNUTLS=""
AS_CASE([$with_gnutls],
[$REQUIRED], [
REQUIRE_LIB([gnutls], [gnutls_sec_param_to_pk_bits])
REQUIRE_HEADER([gnutls/gnutls.h])
],
[$OPTIONAL], [
AC_CHECK_LIB([gnutls], [gnutls_sec_param_to_pk_bits],
[], [with_gnutls=$DISABLED])
AC_CHECK_HEADERS([gnutls/gnutls.h], [], [with_gnutls=$DISABLED])
]
)
AS_IF([test $with_gnutls -ne $DISABLED],
[
PC_NAME_GNUTLS="gnutls"
PCMK_FEATURES="$PCMK_FEATURES remote"
]
)
AC_SUBST([PC_NAME_GNUTLS])
AM_CONDITIONAL([BUILD_REMOTE], [test $with_gnutls -ne $DISABLED])
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 $SERVICELOG
then
PKG_CHECK_MODULES([SERVICELOG], [servicelog-1])
SERVICELOG_EXISTS="yes"
PCMK_FEATURES="$PCMK_FEATURES servicelog"
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 $SERVICELOG
then
PKG_CHECK_MODULES([OPENIPMI_SERVICELOG],[OpenIPMI OpenIPMIposix])
REQUIRE_HEADER([malloc.h])
OPENIPMI_SERVICELOG_EXISTS="yes"
PCMK_FEATURES="$PCMK_FEATURES ipmiservicelogd"
fi
AC_MSG_RESULT([$OPENIPMI_SERVICELOG_EXISTS])
AM_CONDITIONAL(BUILD_OPENIPMI_SERVICELOG, test "$OPENIPMI_SERVICELOG_EXISTS" = "yes")
# --- ASAN/UBSAN/TSAN (see man gcc) ---
# when using SANitizers, we need to pass the -fsanitize..
# to both CFLAGS and LDFLAGS. The CFLAGS/LDFLAGS must be
# specified as first in the list or there will be runtime
# issues (for example user has to LD_PRELOAD asan for it to work
# properly).
AS_IF([test -n "${SANITIZERS}"], [
SANITIZERS=$(echo $SANITIZERS | sed -e 's/,/ /g')
for SANITIZER in $SANITIZERS
do
AS_CASE([$SANITIZER],
[asan|ASAN], [
SANITIZERS_CFLAGS="$SANITIZERS_CFLAGS -fsanitize=address"
SANITIZERS_LDFLAGS="$SANITIZERS_LDFLAGS -fsanitize=address -lasan"
PCMK_FEATURES="$PCMK_FEATURES asan"
REQUIRE_LIB([asan],[main])
],
[ubsan|UBSAN], [
SANITIZERS_CFLAGS="$SANITIZERS_CFLAGS -fsanitize=undefined"
SANITIZERS_LDFLAGS="$SANITIZERS_LDFLAGS -fsanitize=undefined -lubsan"
PCMK_FEATURES="$PCMK_FEATURES ubsan"
REQUIRE_LIB([ubsan],[main])
],
[tsan|TSAN], [
SANITIZERS_CFLAGS="$SANITIZERS_CFLAGS -fsanitize=thread"
SANITIZERS_LDFLAGS="$SANITIZERS_LDFLAGS -fsanitize=thread -ltsan"
PCMK_FEATURES="$PCMK_FEATURES tsan"
REQUIRE_LIB([tsan],[main])
])
done
])
dnl ========================================================================
dnl Compiler flags
dnl ========================================================================
dnl Make sure that CFLAGS is not exported. If the user did
dnl not have CFLAGS in their environment then this should have
dnl no effect. However if CFLAGS was exported from the user's
dnl environment, then the new CFLAGS will also be exported
dnl to sub processes.
if export | fgrep " CFLAGS=" > /dev/null; then
SAVED_CFLAGS="$CFLAGS"
unset CFLAGS
CFLAGS="$SAVED_CFLAGS"
unset SAVED_CFLAGS
fi
AC_ARG_VAR([CFLAGS_HARDENED_LIB], [extra C compiler flags for hardened libraries])
AC_ARG_VAR([LDFLAGS_HARDENED_LIB], [extra linker flags for hardened libraries])
AC_ARG_VAR([CFLAGS_HARDENED_EXE], [extra C compiler flags for hardened executables])
AC_ARG_VAR([LDFLAGS_HARDENED_EXE], [extra linker flags for hardened executables])
CC_EXTRAS=""
AS_IF([test "$GCC" != yes], [CFLAGS="$CFLAGS -g"], [
CFLAGS="$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 $WERROR"
AC_MSG_CHECKING([for gcc diagnostic push / pull])
AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[
#pragma GCC diagnostic push
#pragma GCC diagnostic pop
]])],
[
AC_MSG_RESULT([yes])
gcc_diagnostic_push_pull=yes
], AC_MSG_RESULT([no]))
cc_restore_flags
AS_IF([cc_supports_flag "-Wformat-nonliteral"],
[gcc_format_nonliteral=yes],
[gcc_format_nonliteral=no])
# We had to eliminate -Wnested-externs because of libtool changes
# Make sure to order options so that the former stand for prerequisites
# of the latter (e.g., -Wformat-nonliteral requires -Wformat).
EXTRA_FLAGS="-fgnu89-inline"
EXTRA_FLAGS="$EXTRA_FLAGS -Wall"
EXTRA_FLAGS="$EXTRA_FLAGS -Waggregate-return"
EXTRA_FLAGS="$EXTRA_FLAGS -Wbad-function-cast"
EXTRA_FLAGS="$EXTRA_FLAGS -Wcast-align"
EXTRA_FLAGS="$EXTRA_FLAGS -Wdeclaration-after-statement"
EXTRA_FLAGS="$EXTRA_FLAGS -Wendif-labels"
EXTRA_FLAGS="$EXTRA_FLAGS -Wfloat-equal"
EXTRA_FLAGS="$EXTRA_FLAGS -Wformat-security"
EXTRA_FLAGS="$EXTRA_FLAGS -Wmissing-prototypes"
EXTRA_FLAGS="$EXTRA_FLAGS -Wmissing-declarations"
EXTRA_FLAGS="$EXTRA_FLAGS -Wnested-externs"
EXTRA_FLAGS="$EXTRA_FLAGS -Wno-long-long"
EXTRA_FLAGS="$EXTRA_FLAGS -Wno-strict-aliasing"
EXTRA_FLAGS="$EXTRA_FLAGS -Wpointer-arith"
EXTRA_FLAGS="$EXTRA_FLAGS -Wstrict-prototypes"
EXTRA_FLAGS="$EXTRA_FLAGS -Wwrite-strings"
EXTRA_FLAGS="$EXTRA_FLAGS -Wunused-but-set-variable"
EXTRA_FLAGS="$EXTRA_FLAGS -Wunsigned-char"
AS_IF([test "x$gcc_diagnostic_push_pull" = "xyes"],
[
AC_DEFINE([GCC_FORMAT_NONLITERAL_CHECKING_ENABLED], [],
[gcc can complain about nonliterals in format])
EXTRA_FLAGS="$EXTRA_FLAGS -Wformat=2 -Wformat-nonliteral"
],
[test "x$gcc_format_nonliteral" = "xyes"],
[EXTRA_FLAGS="$EXTRA_FLAGS -Wformat=2"])
# Additional warnings it might be nice to enable one day
# -Wshadow
# -Wunreachable-code
for j in $EXTRA_FLAGS
do
AS_IF([cc_supports_flag $CC_EXTRAS $j], [CC_EXTRAS="$CC_EXTRAS $j"])
done
AC_MSG_NOTICE([Using additional gcc flags: ${CC_EXTRAS}])
])
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
AS_IF([test $enable_hardening -eq $OPTIONAL],
[
AS_IF([test "$(env | grep -Ec '^(C|LD)FLAGS_HARDENED_(EXE|LIB)=.')" = 0],
[enable_hardening=$REQUIRED],
[AC_MSG_NOTICE([Hardening: using custom flags from environment])]
)
],
[
unset CFLAGS_HARDENED_EXE
unset CFLAGS_HARDENED_LIB
unset LDFLAGS_HARDENED_EXE
unset LDFLAGS_HARDENED_LIB
]
)
AS_CASE([$enable_hardening],
[$DISABLED], [AC_MSG_NOTICE([Hardening: explicitly disabled])],
[$REQUIRED], [
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} ${flag}"
relro=1
])
# daemons: PIE for both CFLAGS and LDFLAGS
AS_IF([cc_supports_flag -fPIE],
[
flag="-pie"
CC_CHECK_LDFLAGS(["${flag}"],
[
CFLAGS_HARDENED_EXE="${CFLAGS_HARDENED_EXE} -fPIE"
LDFLAGS_HARDENED_EXE="${LDFLAGS_HARDENED_EXE} ${flag}"
pie=1
])
]
)
# daemons incl. libs: full RELRO if sensible + as-needed linking
# so as to possibly mitigate startup performance
# hit caused by excessive linking with unneeded
# libraries
AS_IF([test "${relro}" = 1 && test "${pie}" = 1],
[
flag="-Wl,-z,now"
CC_CHECK_LDFLAGS(["${flag}"],
[
LDFLAGS_HARDENED_EXE="${LDFLAGS_HARDENED_EXE} ${flag}"
LDFLAGS_HARDENED_LIB="${LDFLAGS_HARDENED_LIB} ${flag}"
bindnow=1
])
]
)
AS_IF([test "${bindnow}" = 1],
[
flag="-Wl,--as-needed"
CC_CHECK_LDFLAGS(["${flag}"],
[
LDFLAGS_HARDENED_EXE="${LDFLAGS_HARDENED_EXE} ${flag}"
LDFLAGS_HARDENED_LIB="${LDFLAGS_HARDENED_LIB} ${flag}"
])
])
# universal: prefer strong > all > default stack protector if possible
flag=
AS_IF([cc_supports_flag -fstack-protector-strong],
[flag="-fstack-protector-strong"],
[cc_supports_flag -fstack-protector-all],
[flag="-fstack-protector-all"],
[cc_supports_flag -fstack-protector],
[flag="-fstack-protector"]
)
AS_IF([test -n "${flag}"],
[
CC_EXTRAS="${CC_EXTRAS} ${flag}"
stackprot=1
]
)
AS_IF([test "${relro}" = 1 || test "${pie}" = 1 || test "${stackprot}" = 1],
[AC_MSG_NOTICE([Hardening: relro=${relro} pie=${pie} bindnow=${bindnow} stackprot=${flag}])],
[AC_MSG_WARN([Hardening: no suitable features in the toolchain detected])]
)
],
)
CFLAGS="$SANITIZERS_CFLAGS $CFLAGS $CC_EXTRAS"
LDFLAGS="$SANITIZERS_LDFLAGS $LDFLAGS"
CFLAGS_HARDENED_EXE="$SANITIZERS_CFLAGS $CFLAGS_HARDENED_EXE"
LDFLAGS_HARDENED_EXE="$SANITIZERS_LDFLAGS $LDFLAGS_HARDENED_EXE"
NON_FATAL_CFLAGS="$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
AS_IF([test $enable_fatal_warnings -ne $DISABLED], [
AC_MSG_NOTICE([Enabling fatal compiler warnings])
CFLAGS="$CFLAGS $WERROR"
])
AC_SUBST(CFLAGS)
dnl This is useful for use in Makefiles that need to remove one specific flag
CFLAGS_COPY="$CFLAGS"
AC_SUBST(CFLAGS_COPY)
AC_SUBST(LIBADD_DL) dnl extra flags for dynamic linking libraries
AC_SUBST(LOCALE)
dnl Options for cleaning up the compiler output
AS_IF([test $enable_quiet -ne $DISABLED],
[
AC_MSG_NOTICE([Suppressing make details])
QUIET_LIBTOOL_OPTS="--silent"
QUIET_MAKE_OPTS="-s" # POSIX compliant
],
[
QUIET_LIBTOOL_OPTS=""
QUIET_MAKE_OPTS=""
]
)
dnl Put the above variables to use
LIBTOOL="${LIBTOOL} --tag=CC \$(QUIET_LIBTOOL_OPTS)"
MAKEFLAGS="${MAKEFLAGS} ${QUIET_MAKE_OPTS}"
# Make features list available (sorted alphabetically, without leading space)
PCMK_FEATURES=`echo "$PCMK_FEATURES" | sed -e 's/^ //' -e 's/ /\n/g' | sort | xargs`
AC_DEFINE_UNQUOTED(CRM_FEATURES, "$PCMK_FEATURES", Set of enabled features)
AC_SUBST(PCMK_FEATURES)
AC_SUBST(CC)
AC_SUBST(MAKEFLAGS)
AC_SUBST(LIBTOOL)
AC_SUBST(QUIET_LIBTOOL_OPTS)
dnl Files we output that need to be executable
CONFIG_FILES_EXEC([cts/cts-cli],
[cts/cts-coverage],
[cts/cts-exec],
[cts/cts-fencing],
[cts/cts-regression],
[cts/cts-scheduler],
[cts/lxc_autogen.sh],
[cts/benchmark/clubench],
[cts/lab/CTSlab.py],
[cts/lab/OCFIPraTest.py],
[cts/lab/cluster_test],
[cts/lab/cts],
[cts/lab/cts-log-watcher],
[cts/support/LSBDummy],
[cts/support/cts-support],
[cts/support/fence_dummy],
[cts/support/pacemaker-cts-dummyd],
[daemons/fenced/fence_legacy],
[doc/abi-check],
[extra/resources/ClusterMon],
[extra/resources/HealthSMART],
[extra/resources/SysInfo],
[extra/resources/ifspeed],
[extra/resources/o2cb],
[tools/crm_failcount],
[tools/crm_master],
[tools/crm_report],
[tools/crm_standby],
[tools/cibsecret],
[tools/pcmk_simtimes])
dnl Other files we output
AC_CONFIG_FILES(Makefile \
cts/Makefile \
cts/benchmark/Makefile \
cts/lab/Makefile \
cts/lab/CTS.py \
cts/lab/CTSvars.py \
cts/scheduler/Makefile \
cts/scheduler/dot/Makefile \
cts/scheduler/exp/Makefile \
cts/scheduler/scores/Makefile \
cts/scheduler/stderr/Makefile \
cts/scheduler/summary/Makefile \
cts/scheduler/xml/Makefile \
cts/support/Makefile \
cts/support/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.service \
daemons/pacemakerd/pacemaker.upstart \
daemons/pacemakerd/pacemaker.combined.upstart \
daemons/schedulerd/Makefile \
devel/Makefile \
doc/Doxyfile \
doc/Makefile \
doc/sphinx/Makefile \
etc/Makefile \
etc/init.d/pacemaker \
etc/logrotate.d/pacemaker \
extra/Makefile \
extra/alerts/Makefile \
extra/resources/Makefile \
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/common/tests/Makefile \
lib/common/tests/agents/Makefile \
lib/common/tests/cmdline/Makefile \
lib/common/tests/flags/Makefile \
lib/common/tests/operations/Makefile \
lib/common/tests/strings/Makefile \
lib/common/tests/utils/Makefile \
lib/common/tests/xpath/Makefile \
lib/cluster/Makefile \
lib/cib/Makefile \
lib/gnu/Makefile \
lib/pacemaker/Makefile \
lib/pengine/Makefile \
lib/pengine/tests/Makefile \
lib/pengine/tests/rules/Makefile \
lib/fencing/Makefile \
lib/lrmd/Makefile \
lib/services/Makefile \
maint/Makefile \
tests/Makefile \
tools/Makefile \
tools/report.collector \
tools/report.common \
tools/crm_mon.service \
tools/crm_mon.upstart \
xml/Makefile \
xml/pacemaker-schemas.pc \
)
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_NOTICE([])
AC_MSG_NOTICE([$PACKAGE configuration:])
AC_MSG_NOTICE([ Version = ${VERSION} (Build: $BUILD_VERSION)])
AC_MSG_NOTICE([ Features = ${PCMK_FEATURES}])
AC_MSG_NOTICE([])
AC_MSG_NOTICE([ Prefix = ${prefix}])
AC_MSG_NOTICE([ Executables = ${sbindir}])
AC_MSG_NOTICE([ Man pages = ${mandir}])
AC_MSG_NOTICE([ Libraries = ${libdir}])
AC_MSG_NOTICE([ Header files = ${includedir}])
AC_MSG_NOTICE([ Arch-independent files = ${datadir}])
AC_MSG_NOTICE([ State information = ${localstatedir}])
AC_MSG_NOTICE([ System configuration = ${sysconfdir}])
AC_MSG_NOTICE([ OCF agents = ${OCF_ROOT_DIR}])
AC_MSG_NOTICE([])
AC_MSG_NOTICE([ HA group name = ${CRM_DAEMON_GROUP}])
AC_MSG_NOTICE([ HA user name = ${CRM_DAEMON_USER}])
AC_MSG_NOTICE([])
AC_MSG_NOTICE([ CFLAGS = ${CFLAGS}])
AC_MSG_NOTICE([ CFLAGS_HARDENED_EXE = ${CFLAGS_HARDENED_EXE}])
AC_MSG_NOTICE([ CFLAGS_HARDENED_LIB = ${CFLAGS_HARDENED_LIB}])
AC_MSG_NOTICE([ LDFLAGS_HARDENED_EXE = ${LDFLAGS_HARDENED_EXE}])
AC_MSG_NOTICE([ LDFLAGS_HARDENED_LIB = ${LDFLAGS_HARDENED_LIB}])
AC_MSG_NOTICE([ Libraries = ${LIBS}])
AC_MSG_NOTICE([ Stack Libraries = ${CLUSTERLIBS}])
AC_MSG_NOTICE([ Unix socket auth method = ${us_auth}])
diff --git a/cts/lab/CTStests.py b/cts/lab/CTStests.py
index af73ffd5fb..c5c7393bdc 100644
--- a/cts/lab/CTStests.py
+++ b/cts/lab/CTStests.py
@@ -1,3122 +1,3142 @@
""" Test-specific classes for Pacemaker's Cluster Test Suite (CTS)
"""
__copyright__ = "Copyright 2000-2021 the Pacemaker project contributors"
__license__ = "GNU General Public License version 2 or later (GPLv2+) WITHOUT ANY WARRANTY"
#
# SPECIAL NOTE:
#
# Tests may NOT implement any cluster-manager-specific code in them.
# EXTEND the ClusterManager object to provide the base capabilities
# the test needs if you need to do something that the current CM classes
# do not. Otherwise you screw up the whole point of the object structure
# in CTS.
#
# Thank you.
#
import os
import re
import time
import subprocess
import tempfile
from stat import *
from cts import CTS
from cts.CTSaudits import *
from cts.CTSvars import *
from cts.patterns import PatternSelector
from cts.logging import LogFactory
from cts.remote import RemoteFactory, input_wrapper
from cts.watcher import LogWatcher
from cts.environment import EnvFactory
AllTestClasses = [ ]
class CTSTest(object):
'''
A Cluster test.
We implement the basic set of properties and behaviors for a generic
cluster test.
Cluster tests track their own statistics.
We keep each of the kinds of counts we track as separate {name,value}
pairs.
'''
def __init__(self, cm):
#self.name="the unnamed test"
self.Stats = {"calls":0
, "success":0
, "failure":0
, "skipped":0
, "auditfail":0}
# if not issubclass(cm.__class__, ClusterManager):
# raise ValueError("Must be a ClusterManager object")
self.CM = cm
self.Env = EnvFactory().getInstance()
self.rsh = RemoteFactory().getInstance()
self.logger = LogFactory()
self.templates = PatternSelector(cm["Name"])
self.Audits = []
self.timeout = 120
self.passed = 1
self.is_loop = 0
self.is_unsafe = 0
self.is_docker_unsafe = 0
self.is_experimental = 0
self.is_container = 0
self.is_valgrind = 0
self.benchmark = 0 # which tests to benchmark
self.timer = {} # timers
def log(self, args):
self.logger.log(args)
def debug(self, args):
self.logger.debug(args)
def has_key(self, key):
return key in self.Stats
def __setitem__(self, key, value):
self.Stats[key] = value
def __getitem__(self, key):
if str(key) == "0":
raise ValueError("Bad call to 'foo in X', should reference 'foo in X.Stats' instead")
if key in self.Stats:
return self.Stats[key]
return None
def log_mark(self, msg):
self.debug("MARK: test %s %s %d" % (self.name,msg,time.time()))
return
def get_timer(self,key = "test"):
try: return self.timer[key]
except: return 0
def set_timer(self,key = "test"):
self.timer[key] = time.time()
return self.timer[key]
def log_timer(self,key = "test"):
elapsed = 0
if key in self.timer:
elapsed = time.time() - self.timer[key]
s = key == "test" and self.name or "%s:%s" % (self.name,key)
self.debug("%s runtime: %.2f" % (s, elapsed))
del self.timer[key]
return elapsed
def incr(self, name):
'''Increment (or initialize) the value associated with the given name'''
if not name in self.Stats:
self.Stats[name] = 0
self.Stats[name] = self.Stats[name]+1
# Reset the test passed boolean
if name == "calls":
self.passed = 1
def failure(self, reason="none"):
'''Increment the failure count'''
self.passed = 0
self.incr("failure")
self.logger.log(("Test %s" % self.name).ljust(35) + " FAILED: %s" % reason)
return None
def success(self):
'''Increment the success count'''
self.incr("success")
return 1
def skipped(self):
'''Increment the skipped count'''
self.incr("skipped")
return 1
def __call__(self, node):
'''Perform the given test'''
raise ValueError("Abstract Class member (__call__)")
self.incr("calls")
return self.failure()
def audit(self):
passed = 1
if len(self.Audits) > 0:
for audit in self.Audits:
if not audit():
self.logger.log("Internal %s Audit %s FAILED." % (self.name, audit.name()))
self.incr("auditfail")
passed = 0
return passed
def setup(self, node):
'''Setup the given test'''
return self.success()
def teardown(self, node):
'''Tear down the given test'''
return self.success()
def create_watch(self, patterns, timeout, name=None):
if not name:
name = self.name
return LogWatcher(self.Env["LogFileName"], patterns, name, timeout, kind=self.Env["LogWatcher"], hosts=self.Env["nodes"])
def local_badnews(self, prefix, watch, local_ignore=[]):
errcount = 0
if not prefix:
prefix = "LocalBadNews:"
ignorelist = []
ignorelist.append(" CTS: ")
ignorelist.append(prefix)
ignorelist.extend(local_ignore)
while errcount < 100:
match = watch.look(0)
if match:
add_err = 1
for ignore in ignorelist:
if add_err == 1 and re.search(ignore, match):
add_err = 0
if add_err == 1:
self.logger.log(prefix + " " + match)
errcount = errcount + 1
else:
break
else:
self.logger.log("Too many errors!")
watch.end()
return errcount
def is_applicable(self):
return self.is_applicable_common()
def is_applicable_common(self):
'''Return TRUE if we are applicable in the current test configuration'''
#raise ValueError("Abstract Class member (is_applicable)")
if self.is_loop and not self.Env["loop-tests"]:
return 0
elif self.is_unsafe and not self.Env["unsafe-tests"]:
return 0
elif self.is_valgrind and not self.Env["valgrind-tests"]:
return 0
elif self.is_experimental and not self.Env["experimental-tests"]:
return 0
elif self.is_docker_unsafe and self.Env["docker"]:
return 0
elif self.is_container and not self.Env["container-tests"]:
return 0
elif self.Env["benchmark"] and self.benchmark == 0:
return 0
return 1
def find_ocfs2_resources(self, node):
self.r_o2cb = None
self.r_ocfs2 = []
(rc, lines) = self.rsh(node, "crm_resource -c", None)
for line in lines:
if re.search("^Resource", line):
r = AuditResource(self.CM, line)
if r.rtype == "o2cb" and r.parent != "NA":
self.debug("Found o2cb: %s" % self.r_o2cb)
self.r_o2cb = r.parent
if re.search("^Constraint", line):
c = AuditConstraint(self.CM, line)
if c.type == "rsc_colocation" and c.target == self.r_o2cb:
self.r_ocfs2.append(c.rsc)
self.debug("Found ocfs2 filesystems: %s" % repr(self.r_ocfs2))
return len(self.r_ocfs2)
def canrunnow(self, node):
'''Return TRUE if we can meaningfully run right now'''
return 1
def errorstoignore(self):
'''Return list of errors which are 'normal' and should be ignored'''
return []
class StopTest(CTSTest):
'''Stop (deactivate) the cluster manager on a node'''
def __init__(self, cm):
CTSTest.__init__(self, cm)
self.name = "Stop"
def __call__(self, node):
'''Perform the 'stop' test. '''
self.incr("calls")
if self.CM.ShouldBeStatus[node] != "up":
return self.skipped()
patterns = []
# Technically we should always be able to notice ourselves stopping
patterns.append(self.templates["Pat:We_stopped"] % node)
# Any active node needs to notice this one left
# (note that this won't work if we have multiple partitions)
for other in self.Env["nodes"]:
if self.CM.ShouldBeStatus[other] == "up" and other != node:
patterns.append(self.templates["Pat:They_stopped"] %(other, self.CM.key_for_node(node)))
#self.debug("Checking %s will notice %s left"%(other, node))
watch = self.create_watch(patterns, self.Env["DeadTime"])
watch.setwatch()
if node == self.CM.OurNode:
self.incr("us")
else:
if self.CM.upcount() <= 1:
self.incr("all")
else:
self.incr("them")
self.CM.StopaCM(node)
watch_result = watch.lookforall()
failreason = None
UnmatchedList = "||"
if watch.unmatched:
(rc, output) = self.rsh(node, "/bin/ps axf", None)
for line in output:
self.debug(line)
(rc, output) = self.rsh(node, "/usr/sbin/dlm_tool dump 2>/dev/null", None)
for line in output:
self.debug(line)
for regex in watch.unmatched:
self.logger.log ("ERROR: Shutdown pattern not found: %s" % (regex))
UnmatchedList += regex + "||";
failreason = "Missing shutdown pattern"
self.CM.cluster_stable(self.Env["DeadTime"])
if not watch.unmatched or self.CM.upcount() == 0:
return self.success()
if len(watch.unmatched) >= self.CM.upcount():
return self.failure("no match against (%s)" % UnmatchedList)
if failreason == None:
return self.success()
else:
return self.failure(failreason)
#
# We don't register StopTest because it's better when called by
# another test...
#
class StartTest(CTSTest):
'''Start (activate) the cluster manager on a node'''
def __init__(self, cm, debug=None):
CTSTest.__init__(self,cm)
self.name = "start"
self.debug = debug
def __call__(self, node):
'''Perform the 'start' test. '''
self.incr("calls")
if self.CM.upcount() == 0:
self.incr("us")
else:
self.incr("them")
if self.CM.ShouldBeStatus[node] != "down":
return self.skipped()
elif self.CM.StartaCM(node):
return self.success()
else:
return self.failure("Startup %s on node %s failed"
% (self.Env["Name"], node))
#
# We don't register StartTest because it's better when called by
# another test...
#
class FlipTest(CTSTest):
'''If it's running, stop it. If it's stopped start it.
Overthrow the status quo...
'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "Flip"
self.start = StartTest(cm)
self.stop = StopTest(cm)
def __call__(self, node):
'''Perform the 'Flip' test. '''
self.incr("calls")
if self.CM.ShouldBeStatus[node] == "up":
self.incr("stopped")
ret = self.stop(node)
type = "up->down"
# Give the cluster time to recognize it's gone...
time.sleep(self.Env["StableTime"])
elif self.CM.ShouldBeStatus[node] == "down":
self.incr("started")
ret = self.start(node)
type = "down->up"
else:
return self.skipped()
self.incr(type)
if ret:
return self.success()
else:
return self.failure("%s failure" % type)
# Register FlipTest as a good test to run
AllTestClasses.append(FlipTest)
class RestartTest(CTSTest):
'''Stop and restart a node'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "Restart"
self.start = StartTest(cm)
self.stop = StopTest(cm)
self.benchmark = 1
def __call__(self, node):
'''Perform the 'restart' test. '''
self.incr("calls")
self.incr("node:" + node)
ret1 = 1
if self.CM.StataCM(node):
self.incr("WasStopped")
if not self.start(node):
return self.failure("start (setup) failure: "+node)
self.set_timer()
if not self.stop(node):
return self.failure("stop failure: "+node)
if not self.start(node):
return self.failure("start failure: "+node)
return self.success()
# Register RestartTest as a good test to run
AllTestClasses.append(RestartTest)
class StonithdTest(CTSTest):
def __init__(self, cm):
CTSTest.__init__(self, cm)
self.name = "Stonithd"
self.startall = SimulStartLite(cm)
self.benchmark = 1
def __call__(self, node):
self.incr("calls")
if len(self.Env["nodes"]) < 2:
return self.skipped()
ret = self.startall(None)
if not ret:
return self.failure("Setup failed")
is_dc = self.CM.is_node_dc(node)
watchpats = []
watchpats.append(self.templates["Pat:FenceOpOK"] % node)
watchpats.append(self.templates["Pat:NodeFenced"] % node)
if self.Env["at-boot"] == 0:
self.debug("Expecting %s to stay down" % node)
self.CM.ShouldBeStatus[node] = "down"
else:
self.debug("Expecting %s to come up again %d" % (node, self.Env["at-boot"]))
watchpats.append("%s.* S_STARTING -> S_PENDING" % node)
watchpats.append("%s.* S_PENDING -> S_NOT_DC" % node)
watch = self.create_watch(watchpats, 30 + self.Env["DeadTime"] + self.Env["StableTime"] + self.Env["StartTime"])
watch.setwatch()
origin = self.Env.RandomGen.choice(self.Env["nodes"])
rc = self.rsh(origin, "stonith_admin --reboot %s -VVVVVV" % node)
if rc == 194:
# 194 - 256 = -62 = Timer expired
#
# Look for the patterns, usually this means the required
# device was running on the node to be fenced - or that
# the required devices were in the process of being loaded
# and/or moved
#
# Effectively the node committed suicide so there will be
# no confirmation, but pacemaker should be watching and
# fence the node again
self.logger.log("Fencing command on %s to fence %s timed out" % (origin, node))
elif origin != node and rc != 0:
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
self.debug("Waiting for fenced node to come back up")
self.CM.ns.WaitForAllNodesToComeUp(self.Env["nodes"], 600)
self.logger.log("Fencing command on %s failed to fence %s (rc=%d)" % (origin, node, rc))
elif origin == node and rc != 255:
# 255 == broken pipe, ie. the node was fenced as expected
self.logger.log("Locally originated fencing returned %d" % rc)
self.set_timer("fence")
matched = watch.lookforall()
self.log_timer("fence")
self.set_timer("reform")
if watch.unmatched:
self.logger.log("Patterns not found: " + repr(watch.unmatched))
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
self.debug("Waiting for fenced node to come back up")
self.CM.ns.WaitForAllNodesToComeUp(self.Env["nodes"], 600)
self.debug("Waiting for the cluster to re-stabilize with all nodes")
is_stable = self.CM.cluster_stable(self.Env["StartTime"])
if not matched:
return self.failure("Didn't find all expected patterns")
elif not is_stable:
return self.failure("Cluster did not become stable")
self.log_timer("reform")
return self.success()
def errorstoignore(self):
return [
self.templates["Pat:Fencing_start"] % ".*",
self.templates["Pat:Fencing_ok"] % ".*",
self.templates["Pat:Fencing_active"],
r"error.*: Operation 'reboot' targeting .* by .* for stonith_admin.*: Timer expired",
]
def is_applicable(self):
if not self.is_applicable_common():
return 0
if "DoFencing" in list(self.Env.keys()):
return self.Env["DoFencing"]
return 1
AllTestClasses.append(StonithdTest)
class StartOnebyOne(CTSTest):
'''Start all the nodes ~ one by one'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "StartOnebyOne"
self.stopall = SimulStopLite(cm)
self.start = StartTest(cm)
self.ns = CTS.NodeStatus(cm.Env)
def __call__(self, dummy):
'''Perform the 'StartOnebyOne' test. '''
self.incr("calls")
# We ignore the "node" parameter...
# Shut down all the nodes...
ret = self.stopall(None)
if not ret:
return self.failure("Test setup failed")
failed = []
self.set_timer()
for node in self.Env["nodes"]:
if not self.start(node):
failed.append(node)
if len(failed) > 0:
return self.failure("Some node failed to start: " + repr(failed))
return self.success()
# Register StartOnebyOne as a good test to run
AllTestClasses.append(StartOnebyOne)
class SimulStart(CTSTest):
'''Start all the nodes ~ simultaneously'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "SimulStart"
self.stopall = SimulStopLite(cm)
self.startall = SimulStartLite(cm)
def __call__(self, dummy):
'''Perform the 'SimulStart' test. '''
self.incr("calls")
# We ignore the "node" parameter...
# Shut down all the nodes...
ret = self.stopall(None)
if not ret:
return self.failure("Setup failed")
if not self.startall(None):
return self.failure("Startall failed")
return self.success()
# Register SimulStart as a good test to run
AllTestClasses.append(SimulStart)
class SimulStop(CTSTest):
'''Stop all the nodes ~ simultaneously'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "SimulStop"
self.startall = SimulStartLite(cm)
self.stopall = SimulStopLite(cm)
def __call__(self, dummy):
'''Perform the 'SimulStop' test. '''
self.incr("calls")
# We ignore the "node" parameter...
# Start up all the nodes...
ret = self.startall(None)
if not ret:
return self.failure("Setup failed")
if not self.stopall(None):
return self.failure("Stopall failed")
return self.success()
# Register SimulStop as a good test to run
AllTestClasses.append(SimulStop)
class StopOnebyOne(CTSTest):
'''Stop all the nodes in order'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "StopOnebyOne"
self.startall = SimulStartLite(cm)
self.stop = StopTest(cm)
def __call__(self, dummy):
'''Perform the 'StopOnebyOne' test. '''
self.incr("calls")
# We ignore the "node" parameter...
# Start up all the nodes...
ret = self.startall(None)
if not ret:
return self.failure("Setup failed")
failed = []
self.set_timer()
for node in self.Env["nodes"]:
if not self.stop(node):
failed.append(node)
if len(failed) > 0:
return self.failure("Some node failed to stop: " + repr(failed))
return self.success()
# Register StopOnebyOne as a good test to run
AllTestClasses.append(StopOnebyOne)
class RestartOnebyOne(CTSTest):
'''Restart all the nodes in order'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "RestartOnebyOne"
self.startall = SimulStartLite(cm)
def __call__(self, dummy):
'''Perform the 'RestartOnebyOne' test. '''
self.incr("calls")
# We ignore the "node" parameter...
# Start up all the nodes...
ret = self.startall(None)
if not ret:
return self.failure("Setup failed")
did_fail = []
self.set_timer()
self.restart = RestartTest(self.CM)
for node in self.Env["nodes"]:
if not self.restart(node):
did_fail.append(node)
if did_fail:
return self.failure("Could not restart %d nodes: %s"
% (len(did_fail), repr(did_fail)))
return self.success()
# Register StopOnebyOne as a good test to run
AllTestClasses.append(RestartOnebyOne)
class PartialStart(CTSTest):
'''Start a node - but tell it to stop before it finishes starting up'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "PartialStart"
self.startall = SimulStartLite(cm)
self.stopall = SimulStopLite(cm)
self.stop = StopTest(cm)
#self.is_unsafe = 1
def __call__(self, node):
'''Perform the 'PartialStart' test. '''
self.incr("calls")
ret = self.stopall(None)
if not ret:
return self.failure("Setup failed")
watchpats = []
watchpats.append("pacemaker-controld.*Connecting to .* cluster infrastructure")
watch = self.create_watch(watchpats, self.Env["DeadTime"]+10)
watch.setwatch()
self.CM.StartaCMnoBlock(node)
ret = watch.lookforall()
if not ret:
self.logger.log("Patterns not found: " + repr(watch.unmatched))
return self.failure("Setup of %s failed" % node)
ret = self.stop(node)
if not ret:
return self.failure("%s did not stop in time" % node)
return self.success()
def errorstoignore(self):
'''Return list of errors which should be ignored'''
# We might do some fencing in the 2-node case if we make it up far enough
return [
r"Executing reboot fencing operation",
r"Requesting fencing \([^)]+\) of node ",
]
# Register StopOnebyOne as a good test to run
AllTestClasses.append(PartialStart)
class StandbyTest(CTSTest):
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "Standby"
self.benchmark = 1
self.start = StartTest(cm)
self.startall = SimulStartLite(cm)
# make sure the node is active
# set the node to standby mode
# check resources, none resource should be running on the node
# set the node to active mode
# check resouces, resources should have been migrated back (SHOULD THEY?)
def __call__(self, node):
self.incr("calls")
ret = self.startall(None)
if not ret:
return self.failure("Start all nodes failed")
self.debug("Make sure node %s is active" % node)
if self.CM.StandbyStatus(node) != "off":
if not self.CM.SetStandbyMode(node, "off"):
return self.failure("can't set node %s to active mode" % node)
self.CM.cluster_stable()
status = self.CM.StandbyStatus(node)
if status != "off":
return self.failure("standby status of %s is [%s] but we expect [off]" % (node, status))
self.debug("Getting resources running on node %s" % node)
rsc_on_node = self.CM.active_resources(node)
watchpats = []
watchpats.append(r"State transition .* -> S_POLICY_ENGINE")
watch = self.create_watch(watchpats, self.Env["DeadTime"]+10)
watch.setwatch()
self.debug("Setting node %s to standby mode" % node)
if not self.CM.SetStandbyMode(node, "on"):
return self.failure("can't set node %s to standby mode" % node)
self.set_timer("on")
ret = watch.lookforall()
if not ret:
self.logger.log("Patterns not found: " + repr(watch.unmatched))
self.CM.SetStandbyMode(node, "off")
return self.failure("cluster didn't react to standby change on %s" % node)
self.CM.cluster_stable()
status = self.CM.StandbyStatus(node)
if status != "on":
return self.failure("standby status of %s is [%s] but we expect [on]" % (node, status))
self.log_timer("on")
self.debug("Checking resources")
bad_run = self.CM.active_resources(node)
if len(bad_run) > 0:
rc = self.failure("%s set to standby, %s is still running on it" % (node, repr(bad_run)))
self.debug("Setting node %s to active mode" % node)
self.CM.SetStandbyMode(node, "off")
return rc
self.debug("Setting node %s to active mode" % node)
if not self.CM.SetStandbyMode(node, "off"):
return self.failure("can't set node %s to active mode" % node)
self.set_timer("off")
self.CM.cluster_stable()
status = self.CM.StandbyStatus(node)
if status != "off":
return self.failure("standby status of %s is [%s] but we expect [off]" % (node, status))
self.log_timer("off")
return self.success()
AllTestClasses.append(StandbyTest)
class ValgrindTest(CTSTest):
'''Check for memory leaks'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "Valgrind"
self.stopall = SimulStopLite(cm)
self.startall = SimulStartLite(cm)
self.is_valgrind = 1
self.is_loop = 1
def setup(self, node):
self.incr("calls")
ret = self.stopall(None)
if not ret:
return self.failure("Stop all nodes failed")
# @TODO Edit /etc/sysconfig/pacemaker on all nodes to enable valgrind,
# and clear any valgrind logs from previous runs. For now, we rely on
# the user to do this manually.
ret = self.startall(None)
if not ret:
return self.failure("Start all nodes failed")
return self.success()
def teardown(self, node):
# Return all nodes to normal
# @TODO Edit /etc/sysconfig/pacemaker on all nodes to disable valgrind
ret = self.stopall(None)
if not ret:
return self.failure("Stop all nodes failed")
return self.success()
def find_leaks(self):
# Check for leaks
# (no longer used but kept in case feature is restored)
leaked = []
self.stop = StopTest(self.CM)
for node in self.Env["nodes"]:
rc = self.stop(node)
if not rc:
self.failure("Couldn't shut down %s" % node)
rc = self.rsh(node, "grep -e indirectly.*lost:.*[1-9] -e definitely.*lost:.*[1-9] -e (ERROR|error).*SUMMARY:.*[1-9].*errors %s" % self.logger.logPat, 0)
if rc != 1:
leaked.append(node)
self.failure("Valgrind errors detected on %s" % node)
(rc, output) = self.rsh(node, "grep -e lost: -e SUMMARY: %s" % self.logger.logPat, None)
for line in output:
self.logger.log(line)
(rc, output) = self.rsh(node, "cat %s" % self.logger.logPat, None)
for line in output:
self.debug(line)
self.rsh(node, "rm -f %s" % self.logger.logPat, None)
return leaked
def __call__(self, node):
#leaked = self.find_leaks()
#if len(leaked) > 0:
# return self.failure("Nodes %s leaked" % repr(leaked))
return self.success()
def errorstoignore(self):
'''Return list of errors which should be ignored'''
return [
r"pacemaker-based.*: \*\*\*\*\*\*\*\*\*\*\*\*\*",
r"pacemaker-based.*: .* avoid confusing Valgrind",
r"HA_VALGRIND_ENABLED",
]
class StandbyLoopTest(ValgrindTest):
'''Check for memory leaks by putting a node in and out of standby for an hour'''
# @TODO This is not a useful test for memory leaks
def __init__(self, cm):
ValgrindTest.__init__(self,cm)
self.name = "StandbyLoop"
def __call__(self, node):
lpc = 0
delay = 2
failed = 0
done = time.time() + self.Env["loop-minutes"] * 60
while time.time() <= done and not failed:
lpc = lpc + 1
time.sleep(delay)
if not self.CM.SetStandbyMode(node, "on"):
self.failure("can't set node %s to standby mode" % node)
failed = lpc
time.sleep(delay)
if not self.CM.SetStandbyMode(node, "off"):
self.failure("can't set node %s to active mode" % node)
failed = lpc
leaked = self.find_leaks()
if failed:
return self.failure("Iteration %d failed" % failed)
elif len(leaked) > 0:
return self.failure("Nodes %s leaked" % repr(leaked))
return self.success()
#AllTestClasses.append(StandbyLoopTest)
class BandwidthTest(CTSTest):
# Tests should not be cluster-manager-specific
# If you need to find out cluster manager configuration to do this, then
# it should be added to the generic cluster manager API.
'''Test the bandwidth which the cluster uses'''
def __init__(self, cm):
CTSTest.__init__(self, cm)
self.name = "Bandwidth"
self.start = StartTest(cm)
self.__setitem__("min",0)
self.__setitem__("max",0)
self.__setitem__("totalbandwidth",0)
(handle, self.tempfile) = tempfile.mkstemp(".cts")
os.close(handle)
self.startall = SimulStartLite(cm)
def __call__(self, node):
'''Perform the Bandwidth test'''
self.incr("calls")
if self.CM.upcount() < 1:
return self.skipped()
Path = self.CM.InternalCommConfig()
if "ip" not in Path["mediatype"]:
return self.skipped()
port = Path["port"][0]
port = int(port)
ret = self.startall(None)
if not ret:
return self.failure("Test setup failed")
time.sleep(5) # We get extra messages right after startup.
fstmpfile = "/var/run/band_estimate"
dumpcmd = "tcpdump -p -n -c 102 -i any udp port %d > %s 2>&1" \
% (port, fstmpfile)
rc = self.rsh(node, dumpcmd)
if rc == 0:
farfile = "root@%s:%s" % (node, fstmpfile)
self.rsh.cp(farfile, self.tempfile)
Bandwidth = self.countbandwidth(self.tempfile)
if not Bandwidth:
self.logger.log("Could not compute bandwidth.")
return self.success()
intband = int(Bandwidth + 0.5)
self.logger.log("...bandwidth: %d bits/sec" % intband)
self.Stats["totalbandwidth"] = self.Stats["totalbandwidth"] + Bandwidth
if self.Stats["min"] == 0:
self.Stats["min"] = Bandwidth
if Bandwidth > self.Stats["max"]:
self.Stats["max"] = Bandwidth
if Bandwidth < self.Stats["min"]:
self.Stats["min"] = Bandwidth
self.rsh(node, "rm -f %s" % fstmpfile)
os.unlink(self.tempfile)
return self.success()
else:
return self.failure("no response from tcpdump command [%d]!" % rc)
def countbandwidth(self, file):
fp = open(file, "r")
fp.seek(0)
count = 0
sum = 0
while 1:
line = fp.readline()
if not line:
return None
if re.search("udp",line) or re.search("UDP,", line):
count = count + 1
linesplit = line.split(" ")
for j in range(len(linesplit)-1):
if linesplit[j] == "udp": break
if linesplit[j] == "length:": break
try:
sum = sum + int(linesplit[j+1])
except ValueError:
self.logger.log("Invalid tcpdump line: %s" % line)
return None
T1 = linesplit[0]
timesplit = T1.split(":")
time2split = timesplit[2].split(".")
time1 = (int(timesplit[0])*60+int(timesplit[1]))*60+int(time2split[0])+int(time2split[1])*0.000001
break
while count < 100:
line = fp.readline()
if not line:
return None
if re.search("udp",line) or re.search("UDP,", line):
count = count+1
linessplit = line.split(" ")
for j in range(len(linessplit)-1):
if linessplit[j] == "udp": break
if linessplit[j] == "length:": break
try:
sum = int(linessplit[j+1]) + sum
except ValueError:
self.logger.log("Invalid tcpdump line: %s" % line)
return None
T2 = linessplit[0]
timesplit = T2.split(":")
time2split = timesplit[2].split(".")
time2 = (int(timesplit[0])*60+int(timesplit[1]))*60+int(time2split[0])+int(time2split[1])*0.000001
time = time2-time1
if (time <= 0):
return 0
return int((sum*8)/time)
def is_applicable(self):
'''BandwidthTest never applicable'''
return 0
AllTestClasses.append(BandwidthTest)
###################################################################
class MaintenanceMode(CTSTest):
###################################################################
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "MaintenanceMode"
self.start = StartTest(cm)
self.startall = SimulStartLite(cm)
self.max = 30
#self.is_unsafe = 1
self.benchmark = 1
self.action = "asyncmon"
self.interval = 0
self.rid = "maintenanceDummy"
def toggleMaintenanceMode(self, node, action):
pats = []
pats.append(self.templates["Pat:DC_IDLE"])
# fail the resource right after turning Maintenance mode on
# verify it is not recovered until maintenance mode is turned off
if action == "On":
pats.append(self.templates["Pat:RscOpFail"] % (self.action, self.rid))
else:
pats.append(self.templates["Pat:RscOpOK"] % ("stop", self.rid))
pats.append(self.templates["Pat:RscOpOK"] % ("start", self.rid))
watch = self.create_watch(pats, 60)
watch.setwatch()
self.debug("Turning maintenance mode %s" % action)
self.rsh(node, self.templates["MaintenanceMode%s" % (action)])
if (action == "On"):
self.rsh(node, "crm_resource -V -F -r %s -H %s &>/dev/null" % (self.rid, node))
self.set_timer("recover%s" % (action))
watch.lookforall()
self.log_timer("recover%s" % (action))
if watch.unmatched:
self.debug("Failed to find patterns when turning maintenance mode %s" % action)
return repr(watch.unmatched)
return ""
def insertMaintenanceDummy(self, node):
pats = []
pats.append(("%s.*" % node) + (self.templates["Pat:RscOpOK"] % ("start", self.rid)))
watch = self.create_watch(pats, 60)
watch.setwatch()
self.CM.AddDummyRsc(node, self.rid)
self.set_timer("addDummy")
watch.lookforall()
self.log_timer("addDummy")
if watch.unmatched:
self.debug("Failed to find patterns when adding maintenance dummy resource")
return repr(watch.unmatched)
return ""
def removeMaintenanceDummy(self, node):
pats = []
pats.append(self.templates["Pat:RscOpOK"] % ("stop", self.rid))
watch = self.create_watch(pats, 60)
watch.setwatch()
self.CM.RemoveDummyRsc(node, self.rid)
self.set_timer("removeDummy")
watch.lookforall()
self.log_timer("removeDummy")
if watch.unmatched:
self.debug("Failed to find patterns when removing maintenance dummy resource")
return repr(watch.unmatched)
return ""
def managedRscList(self, node):
rscList = []
(rc, lines) = self.rsh(node, "crm_resource -c", None)
for line in lines:
if re.search("^Resource", line):
tmp = AuditResource(self.CM, line)
if tmp.managed():
rscList.append(tmp.id)
return rscList
def verifyResources(self, node, rscList, managed):
managedList = list(rscList)
managed_str = "managed"
if not managed:
managed_str = "unmanaged"
(rc, lines) = self.rsh(node, "crm_resource -c", None)
for line in lines:
if re.search("^Resource", line):
tmp = AuditResource(self.CM, line)
if managed and not tmp.managed():
continue
elif not managed and tmp.managed():
continue
elif managedList.count(tmp.id):
managedList.remove(tmp.id)
if len(managedList) == 0:
self.debug("Found all %s resources on %s" % (managed_str, node))
return True
self.logger.log("Could not find all %s resources on %s. %s" % (managed_str, node, managedList))
return False
def __call__(self, node):
'''Perform the 'MaintenanceMode' test. '''
self.incr("calls")
verify_managed = False
verify_unmanaged = False
failPat = ""
ret = self.startall(None)
if not ret:
return self.failure("Setup failed")
# get a list of all the managed resources. We use this list
# after enabling maintenance mode to verify all managed resources
# become un-managed. After maintenance mode is turned off, we use
# this list to verify all the resources become managed again.
managedResources = self.managedRscList(node)
if len(managedResources) == 0:
self.logger.log("No managed resources on %s" % node)
return self.skipped()
# insert a fake resource we can fail during maintenance mode
# so we can verify recovery does not take place until after maintenance
# mode is disabled.
failPat = failPat + self.insertMaintenanceDummy(node)
# toggle maintenance mode ON, then fail dummy resource.
failPat = failPat + self.toggleMaintenanceMode(node, "On")
# verify all the resources are now unmanaged
if self.verifyResources(node, managedResources, False):
verify_unmanaged = True
# Toggle maintenance mode OFF, verify dummy is recovered.
failPat = failPat + self.toggleMaintenanceMode(node, "Off")
# verify all the resources are now managed again
if self.verifyResources(node, managedResources, True):
verify_managed = True
# Remove our maintenance dummy resource.
failPat = failPat + self.removeMaintenanceDummy(node)
self.CM.cluster_stable()
if failPat != "":
return self.failure("Unmatched patterns: %s" % (failPat))
elif verify_unmanaged is False:
return self.failure("Failed to verify resources became unmanaged during maintenance mode")
elif verify_managed is False:
return self.failure("Failed to verify resources switched back to managed after disabling maintenance mode")
return self.success()
def errorstoignore(self):
'''Return list of errors which should be ignored'''
return [
r"Updating failcount for %s" % self.rid,
r"schedulerd.*: Recover %s\s*\(.*\)" % self.rid,
r"Unknown operation: fail",
self.templates["Pat:RscOpOK"] % (self.action, self.rid),
r"(ERROR|error).*: Action %s_%s_%d .* initiated outside of a transition" % (self.rid, self.action, self.interval),
]
AllTestClasses.append(MaintenanceMode)
class ResourceRecover(CTSTest):
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "ResourceRecover"
self.start = StartTest(cm)
self.startall = SimulStartLite(cm)
self.max = 30
self.rid = None
self.rid_alt = None
#self.is_unsafe = 1
self.benchmark = 1
# these are the values used for the new LRM API call
self.action = "asyncmon"
self.interval = 0
def __call__(self, node):
'''Perform the 'ResourceRecover' test. '''
self.incr("calls")
ret = self.startall(None)
if not ret:
return self.failure("Setup failed")
resourcelist = self.CM.active_resources(node)
# if there are no resourcelist, return directly
if len(resourcelist) == 0:
self.logger.log("No active resources on %s" % node)
return self.skipped()
self.rid = self.Env.RandomGen.choice(resourcelist)
self.rid_alt = self.rid
rsc = None
(rc, lines) = self.rsh(node, "crm_resource -c", None)
for line in lines:
if re.search("^Resource", line):
tmp = AuditResource(self.CM, line)
if tmp.id == self.rid:
rsc = tmp
# Handle anonymous clones that get renamed
self.rid = rsc.clone_id
break
if not rsc:
return self.failure("Could not find %s in the resource list" % self.rid)
self.debug("Shooting %s aka. %s" % (rsc.clone_id, rsc.id))
pats = []
pats.append(self.templates["Pat:CloneOpFail"] % (self.action, rsc.id, rsc.clone_id))
if rsc.managed():
pats.append(self.templates["Pat:RscOpOK"] % ("stop", self.rid))
if rsc.unique():
pats.append(self.templates["Pat:RscOpOK"] % ("start", self.rid))
else:
# Anonymous clones may get restarted with a different clone number
pats.append(self.templates["Pat:RscOpOK"] % ("start", ".*"))
watch = self.create_watch(pats, 60)
watch.setwatch()
self.rsh(node, "crm_resource -V -F -r %s -H %s &>/dev/null" % (self.rid, node))
self.set_timer("recover")
watch.lookforall()
self.log_timer("recover")
self.CM.cluster_stable()
recovered = self.CM.ResourceLocation(self.rid)
if watch.unmatched:
return self.failure("Patterns not found: %s" % repr(watch.unmatched))
elif rsc.unique() and len(recovered) > 1:
return self.failure("%s is now active on more than one node: %s"%(self.rid, repr(recovered)))
elif len(recovered) > 0:
self.debug("%s is running on: %s" % (self.rid, repr(recovered)))
elif rsc.managed():
return self.failure("%s was not recovered and is inactive" % self.rid)
return self.success()
def errorstoignore(self):
'''Return list of errors which should be ignored'''
return [
r"Updating failcount for %s" % self.rid,
r"schedulerd.*: Recover (%s|%s)\s*\(.*\)" % (self.rid, self.rid_alt),
r"Unknown operation: fail",
self.templates["Pat:RscOpOK"] % (self.action, self.rid),
r"(ERROR|error).*: Action %s_%s_%d .* initiated outside of a transition" % (self.rid, self.action, self.interval),
]
AllTestClasses.append(ResourceRecover)
class ComponentFail(CTSTest):
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "ComponentFail"
# TODO make this work correctly in docker.
self.is_docker_unsafe = 1
self.startall = SimulStartLite(cm)
self.complist = cm.Components()
self.patterns = []
self.okerrpatterns = []
self.is_unsafe = 1
def __call__(self, node):
'''Perform the 'ComponentFail' test. '''
self.incr("calls")
self.patterns = []
self.okerrpatterns = []
# start all nodes
ret = self.startall(None)
if not ret:
return self.failure("Setup failed")
if not self.CM.cluster_stable(self.Env["StableTime"]):
return self.failure("Setup failed - unstable")
node_is_dc = self.CM.is_node_dc(node, None)
# select a component to kill
chosen = self.Env.RandomGen.choice(self.complist)
while chosen.dc_only == 1 and node_is_dc == 0:
chosen = self.Env.RandomGen.choice(self.complist)
self.debug("...component %s (dc=%d,boot=%d)" % (chosen.name, node_is_dc,chosen.triggersreboot))
self.incr(chosen.name)
if chosen.name != "corosync":
self.patterns.append(self.templates["Pat:ChildKilled"] %(node, chosen.name))
self.patterns.append(self.templates["Pat:ChildRespawn"] %(node, chosen.name))
self.patterns.extend(chosen.pats)
if node_is_dc:
self.patterns.extend(chosen.dc_pats)
# @TODO this should be a flag in the Component
if chosen.name in [ "corosync", "pacemaker-based", "pacemaker-fenced" ]:
# Ignore actions for fence devices if fencer will respawn
# (their registration will be lost, and probes will fail)
self.okerrpatterns = [ self.templates["Pat:Fencing_active"] ]
(rc, lines) = self.rsh(node, "crm_resource -c", None)
for line in lines:
if re.search("^Resource", line):
r = AuditResource(self.CM, line)
if r.rclass == "stonith":
self.okerrpatterns.append(self.templates["Pat:Fencing_recover"] % r.id)
self.okerrpatterns.append(self.templates["Pat:Fencing_probe"] % r.id)
# supply a copy so self.patterns doesn't end up empty
tmpPats = []
tmpPats.extend(self.patterns)
self.patterns.extend(chosen.badnews_ignore)
# Look for STONITH ops, depending on Env["at-boot"] we might need to change the nodes status
stonithPats = []
stonithPats.append(self.templates["Pat:Fencing_ok"] % node)
stonith = self.create_watch(stonithPats, 0)
stonith.setwatch()
# set the watch for stable
watch = self.create_watch(
tmpPats, self.Env["DeadTime"] + self.Env["StableTime"] + self.Env["StartTime"])
watch.setwatch()
# kill the component
chosen.kill(node)
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
self.debug("Waiting for any fenced node to come back up")
self.CM.ns.WaitForAllNodesToComeUp(self.Env["nodes"], 600)
self.debug("Waiting for the cluster to re-stabilize with all nodes")
self.CM.cluster_stable(self.Env["StartTime"])
self.debug("Checking if %s was shot" % node)
shot = stonith.look(60)
if shot:
self.debug("Found: " + repr(shot))
self.okerrpatterns.append(self.templates["Pat:Fencing_start"] % node)
if self.Env["at-boot"] == 0:
self.CM.ShouldBeStatus[node] = "down"
# If fencing occurred, chances are many (if not all) the expected logs
# will not be sent - or will be lost when the node reboots
return self.success()
# check for logs indicating a graceful recovery
matched = watch.lookforall(allow_multiple_matches=1)
if watch.unmatched:
self.logger.log("Patterns not found: " + repr(watch.unmatched))
self.debug("Waiting for the cluster to re-stabilize with all nodes")
is_stable = self.CM.cluster_stable(self.Env["StartTime"])
if not matched:
return self.failure("Didn't find all expected %s patterns" % chosen.name)
elif not is_stable:
return self.failure("Cluster did not become stable after killing %s" % chosen.name)
return self.success()
def errorstoignore(self):
'''Return list of errors which should be ignored'''
# Note that okerrpatterns refers to the last time we ran this test
# The good news is that this works fine for us...
self.okerrpatterns.extend(self.patterns)
return self.okerrpatterns
AllTestClasses.append(ComponentFail)
class SplitBrainTest(CTSTest):
'''It is used to test split-brain. when the path between the two nodes break
check the two nodes both take over the resource'''
def __init__(self,cm):
CTSTest.__init__(self,cm)
self.name = "SplitBrain"
self.start = StartTest(cm)
self.startall = SimulStartLite(cm)
self.is_experimental = 1
def isolate_partition(self, partition):
other_nodes = []
other_nodes.extend(self.Env["nodes"])
for node in partition:
try:
other_nodes.remove(node)
except ValueError:
self.logger.log("Node "+node+" not in " + repr(self.Env["nodes"]) + " from " +repr(partition))
if len(other_nodes) == 0:
return 1
self.debug("Creating partition: " + repr(partition))
self.debug("Everyone else: " + repr(other_nodes))
for node in partition:
if not self.CM.isolate_node(node, other_nodes):
self.logger.log("Could not isolate %s" % node)
return 0
return 1
def heal_partition(self, partition):
other_nodes = []
other_nodes.extend(self.Env["nodes"])
for node in partition:
try:
other_nodes.remove(node)
except ValueError:
self.logger.log("Node "+node+" not in " + repr(self.Env["nodes"]))
if len(other_nodes) == 0:
return 1
self.debug("Healing partition: " + repr(partition))
self.debug("Everyone else: " + repr(other_nodes))
for node in partition:
self.CM.unisolate_node(node, other_nodes)
def __call__(self, node):
'''Perform split-brain test'''
self.incr("calls")
self.passed = 1
partitions = {}
ret = self.startall(None)
if not ret:
return self.failure("Setup failed")
while 1:
# Retry until we get multiple partitions
partitions = {}
p_max = len(self.Env["nodes"])
for node in self.Env["nodes"]:
p = self.Env.RandomGen.randint(1, p_max)
if not p in partitions:
partitions[p] = []
partitions[p].append(node)
p_max = len(list(partitions.keys()))
if p_max > 1:
break
# else, try again
self.debug("Created %d partitions" % p_max)
for key in list(partitions.keys()):
self.debug("Partition["+str(key)+"]:\t"+repr(partitions[key]))
# Disabling STONITH to reduce test complexity for now
self.rsh(node, "crm_attribute -V -n stonith-enabled -v false")
for key in list(partitions.keys()):
self.isolate_partition(partitions[key])
count = 30
while count > 0:
if len(self.CM.find_partitions()) != p_max:
time.sleep(10)
else:
break
else:
self.failure("Expected partitions were not created")
# Target number of partitions formed - wait for stability
if not self.CM.cluster_stable():
self.failure("Partitioned cluster not stable")
# Now audit the cluster state
self.CM.partitions_expected = p_max
if not self.audit():
self.failure("Audits failed")
self.CM.partitions_expected = 1
# And heal them again
for key in list(partitions.keys()):
self.heal_partition(partitions[key])
# Wait for a single partition to form
count = 30
while count > 0:
if len(self.CM.find_partitions()) != 1:
time.sleep(10)
count -= 1
else:
break
else:
self.failure("Cluster did not reform")
# Wait for it to have the right number of members
count = 30
while count > 0:
members = []
partitions = self.CM.find_partitions()
if len(partitions) > 0:
members = partitions[0].split()
if len(members) != len(self.Env["nodes"]):
time.sleep(10)
count -= 1
else:
break
else:
self.failure("Cluster did not completely reform")
# Wait up to 20 minutes - the delay is more preferable than
# trying to continue with in a messed up state
if not self.CM.cluster_stable(1200):
self.failure("Reformed cluster not stable")
if self.Env["continue"] == 1:
answer = "Y"
else:
try:
answer = input_wrapper('Continue? [nY]')
except EOFError as e:
answer = "n"
if answer and answer == "n":
raise ValueError("Reformed cluster not stable")
# Turn fencing back on
if self.Env["DoFencing"]:
self.rsh(node, "crm_attribute -V -D -n stonith-enabled")
self.CM.cluster_stable()
if self.passed:
return self.success()
return self.failure("See previous errors")
def errorstoignore(self):
'''Return list of errors which are 'normal' and should be ignored'''
return [
r"Another DC detected:",
r"(ERROR|error).*: .*Application of an update diff failed",
r"pacemaker-controld.*:.*not in our membership list",
r"CRIT:.*node.*returning after partition",
]
def is_applicable(self):
if not self.is_applicable_common():
return 0
return len(self.Env["nodes"]) > 2
AllTestClasses.append(SplitBrainTest)
class Reattach(CTSTest):
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "Reattach"
self.startall = SimulStartLite(cm)
self.restart1 = RestartTest(cm)
self.stopall = SimulStopLite(cm)
self.is_unsafe = 0 # Handled by canrunnow()
def _is_managed(self, node):
is_managed = self.rsh(node, "crm_attribute -t rsc_defaults -n is-managed -q -G -d true", 1)
is_managed = is_managed[:-1] # Strip off the newline
return is_managed == "true"
def _set_unmanaged(self, node):
self.debug("Disable resource management")
self.rsh(node, "crm_attribute -t rsc_defaults -n is-managed -v false")
def _set_managed(self, node):
self.debug("Re-enable resource management")
self.rsh(node, "crm_attribute -t rsc_defaults -n is-managed -D")
def setup(self, node):
attempt = 0
if not self.startall(None):
return None
# Make sure we are really _really_ stable and that all
# resources, including those that depend on transient node
# attributes, are started
while not self.CM.cluster_stable(double_check=True):
if attempt < 5:
attempt += 1
self.debug("Not stable yet, re-testing")
else:
self.logger.log("Cluster is not stable")
return None
return 1
def teardown(self, node):
# Make sure 'node' is up
start = StartTest(self.CM)
start(node)
if not self._is_managed(node):
self.logger.log("Attempting to re-enable resource management on %s" % node)
self._set_managed(node)
self.CM.cluster_stable()
if not self._is_managed(node):
self.logger.log("Could not re-enable resource management")
return 0
return 1
def canrunnow(self, node):
'''Return TRUE if we can meaningfully run right now'''
if self.find_ocfs2_resources(node):
self.logger.log("Detach/Reattach scenarios are not possible with OCFS2 services present")
return 0
return 1
def __call__(self, node):
self.incr("calls")
pats = []
# Conveniently, the scheduler will display this message when disabling
# management, even if fencing is not enabled, so we can rely on it.
managed = self.create_watch(["Delaying fencing operations"], 60)
managed.setwatch()
self._set_unmanaged(node)
if not managed.lookforall():
self.logger.log("Patterns not found: " + repr(managed.unmatched))
return self.failure("Resource management not disabled")
pats = []
pats.append(self.templates["Pat:RscOpOK"] % ("start", ".*"))
pats.append(self.templates["Pat:RscOpOK"] % ("stop", ".*"))
pats.append(self.templates["Pat:RscOpOK"] % ("promote", ".*"))
pats.append(self.templates["Pat:RscOpOK"] % ("demote", ".*"))
pats.append(self.templates["Pat:RscOpOK"] % ("migrate", ".*"))
watch = self.create_watch(pats, 60, "ShutdownActivity")
watch.setwatch()
self.debug("Shutting down the cluster")
ret = self.stopall(None)
if not ret:
self._set_managed(node)
return self.failure("Couldn't shut down the cluster")
self.debug("Bringing the cluster back up")
ret = self.startall(None)
time.sleep(5) # allow ping to update the CIB
if not ret:
self._set_managed(node)
return self.failure("Couldn't restart the cluster")
if self.local_badnews("ResourceActivity:", watch):
self._set_managed(node)
return self.failure("Resources stopped or started during cluster restart")
watch = self.create_watch(pats, 60, "StartupActivity")
watch.setwatch()
# Re-enable resource management (and verify it happened).
self._set_managed(node)
self.CM.cluster_stable()
if not self._is_managed(node):
return self.failure("Could not re-enable resource management")
# Ignore actions for STONITH resources
ignore = []
(rc, lines) = self.rsh(node, "crm_resource -c", None)
for line in lines:
if re.search("^Resource", line):
r = AuditResource(self.CM, line)
if r.rclass == "stonith":
self.debug("Ignoring start actions for %s" % r.id)
ignore.append(self.templates["Pat:RscOpOK"] % ("start", r.id))
if self.local_badnews("ResourceActivity:", watch, ignore):
return self.failure("Resources stopped or started after resource management was re-enabled")
return ret
def errorstoignore(self):
'''Return list of errors which should be ignored'''
return [
r"resource( was|s were) active at shutdown",
]
def is_applicable(self):
return 1
AllTestClasses.append(Reattach)
class SpecialTest1(CTSTest):
'''Set up a custom test to cause quorum failure issues for Andrew'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "SpecialTest1"
self.startall = SimulStartLite(cm)
self.restart1 = RestartTest(cm)
self.stopall = SimulStopLite(cm)
def __call__(self, node):
'''Perform the 'SpecialTest1' test for Andrew. '''
self.incr("calls")
# Shut down all the nodes...
ret = self.stopall(None)
if not ret:
return self.failure("Could not stop all nodes")
# Test config recovery when the other nodes come up
self.rsh(node, "rm -f "+CTSvars.CRM_CONFIG_DIR+"/cib*")
# Start the selected node
ret = self.restart1(node)
if not ret:
return self.failure("Could not start "+node)
# Start all remaining nodes
ret = self.startall(None)
if not ret:
return self.failure("Could not start the remaining nodes")
return self.success()
def errorstoignore(self):
'''Return list of errors which should be ignored'''
# Errors that occur as a result of the CIB being wiped
return [
r"error.*: v1 patchset error, patch failed to apply: Application of an update diff failed",
r"error.*: Resource start-up disabled since no STONITH resources have been defined",
r"error.*: Either configure some or disable STONITH with the stonith-enabled option",
r"error.*: NOTE: Clusters with shared data need STONITH to ensure data integrity",
]
AllTestClasses.append(SpecialTest1)
class HAETest(CTSTest):
'''Set up a custom test to cause quorum failure issues for Andrew'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "HAETest"
self.stopall = SimulStopLite(cm)
self.startall = SimulStartLite(cm)
self.is_loop = 1
def setup(self, node):
# Start all remaining nodes
ret = self.startall(None)
if not ret:
return self.failure("Couldn't start all nodes")
return self.success()
def teardown(self, node):
# Stop everything
ret = self.stopall(None)
if not ret:
return self.failure("Couldn't stop all nodes")
return self.success()
def wait_on_state(self, node, resource, expected_clones, attempts=240):
while attempts > 0:
active = 0
(rc, lines) = self.rsh(node, "crm_resource -r %s -W -Q" % resource, stdout=None)
# Hack until crm_resource does the right thing
if rc == 0 and lines:
active = len(lines)
if len(lines) == expected_clones:
return 1
elif rc == 1:
self.debug("Resource %s is still inactive" % resource)
elif rc == 234:
self.logger.log("Unknown resource %s" % resource)
return 0
elif rc == 246:
self.logger.log("Cluster is inactive")
return 0
elif rc != 0:
self.logger.log("Call to crm_resource failed, rc=%d" % rc)
return 0
else:
self.debug("Resource %s is active on %d times instead of %d" % (resource, active, expected_clones))
attempts -= 1
time.sleep(1)
return 0
def find_dlm(self, node):
self.r_dlm = None
(rc, lines) = self.rsh(node, "crm_resource -c", None)
for line in lines:
if re.search("^Resource", line):
r = AuditResource(self.CM, line)
if r.rtype == "controld" and r.parent != "NA":
self.debug("Found dlm: %s" % self.r_dlm)
self.r_dlm = r.parent
return 1
return 0
def find_hae_resources(self, node):
self.r_dlm = None
self.r_o2cb = None
self.r_ocfs2 = []
if self.find_dlm(node):
self.find_ocfs2_resources(node)
def is_applicable(self):
if not self.is_applicable_common():
return 0
if self.Env["Schema"] == "hae":
return 1
return None
class HAERoleTest(HAETest):
def __init__(self, cm):
'''Lars' mount/unmount test for the HA extension. '''
HAETest.__init__(self,cm)
self.name = "HAERoleTest"
def change_state(self, node, resource, target):
rc = self.rsh(node, "crm_resource -V -r %s -p target-role -v %s --meta" % (resource, target))
return rc
def __call__(self, node):
self.incr("calls")
lpc = 0
failed = 0
delay = 2
done = time.time() + self.Env["loop-minutes"]*60
self.find_hae_resources(node)
clone_max = len(self.Env["nodes"])
while time.time() <= done and not failed:
lpc = lpc + 1
self.change_state(node, self.r_dlm, "Stopped")
if not self.wait_on_state(node, self.r_dlm, 0):
self.failure("%s did not go down correctly" % self.r_dlm)
failed = lpc
self.change_state(node, self.r_dlm, "Started")
if not self.wait_on_state(node, self.r_dlm, clone_max):
self.failure("%s did not come up correctly" % self.r_dlm)
failed = lpc
if not self.wait_on_state(node, self.r_o2cb, clone_max):
self.failure("%s did not come up correctly" % self.r_o2cb)
failed = lpc
for fs in self.r_ocfs2:
if not self.wait_on_state(node, fs, clone_max):
self.failure("%s did not come up correctly" % fs)
failed = lpc
if failed:
return self.failure("iteration %d failed" % failed)
return self.success()
AllTestClasses.append(HAERoleTest)
class HAEStandbyTest(HAETest):
'''Set up a custom test to cause quorum failure issues for Andrew'''
def __init__(self, cm):
HAETest.__init__(self,cm)
self.name = "HAEStandbyTest"
def change_state(self, node, resource, target):
rc = self.rsh(node, "crm_standby -V -l reboot -v %s" % (target))
return rc
def __call__(self, node):
self.incr("calls")
lpc = 0
failed = 0
done = time.time() + self.Env["loop-minutes"]*60
self.find_hae_resources(node)
clone_max = len(self.Env["nodes"])
while time.time() <= done and not failed:
lpc = lpc + 1
self.change_state(node, self.r_dlm, "true")
if not self.wait_on_state(node, self.r_dlm, clone_max-1):
self.failure("%s did not go down correctly" % self.r_dlm)
failed = lpc
self.change_state(node, self.r_dlm, "false")
if not self.wait_on_state(node, self.r_dlm, clone_max):
self.failure("%s did not come up correctly" % self.r_dlm)
failed = lpc
if not self.wait_on_state(node, self.r_o2cb, clone_max):
self.failure("%s did not come up correctly" % self.r_o2cb)
failed = lpc
for fs in self.r_ocfs2:
if not self.wait_on_state(node, fs, clone_max):
self.failure("%s did not come up correctly" % fs)
failed = lpc
if failed:
return self.failure("iteration %d failed" % failed)
return self.success()
AllTestClasses.append(HAEStandbyTest)
class NearQuorumPointTest(CTSTest):
'''
This test brings larger clusters near the quorum point (50%).
In addition, it will test doing starts and stops at the same time.
Here is how I think it should work:
- loop over the nodes and decide randomly which will be up and which
will be down Use a 50% probability for each of up/down.
- figure out what to do to get into that state from the current state
- in parallel, bring up those going up and bring those going down.
'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "NearQuorumPoint"
def __call__(self, dummy):
'''Perform the 'NearQuorumPoint' test. '''
self.incr("calls")
startset = []
stopset = []
stonith = self.CM.prepare_fencing_watcher("NearQuorumPoint")
#decide what to do with each node
for node in self.Env["nodes"]:
action = self.Env.RandomGen.choice(["start","stop"])
#action = self.Env.RandomGen.choice(["start","stop","no change"])
if action == "start" :
startset.append(node)
elif action == "stop" :
stopset.append(node)
self.debug("start nodes:" + repr(startset))
self.debug("stop nodes:" + repr(stopset))
#add search patterns
watchpats = [ ]
for node in stopset:
if self.CM.ShouldBeStatus[node] == "up":
watchpats.append(self.templates["Pat:We_stopped"] % node)
for node in startset:
if self.CM.ShouldBeStatus[node] == "down":
#watchpats.append(self.templates["Pat:NonDC_started"] % node)
watchpats.append(self.templates["Pat:Local_started"] % node)
else:
for stopping in stopset:
if self.CM.ShouldBeStatus[stopping] == "up":
watchpats.append(self.templates["Pat:They_stopped"] % (node, self.CM.key_for_node(stopping)))
if len(watchpats) == 0:
return self.skipped()
if len(startset) != 0:
watchpats.append(self.templates["Pat:DC_IDLE"])
watch = self.create_watch(watchpats, self.Env["DeadTime"]+10)
watch.setwatch()
#begin actions
for node in stopset:
if self.CM.ShouldBeStatus[node] == "up":
self.CM.StopaCMnoBlock(node)
for node in startset:
if self.CM.ShouldBeStatus[node] == "down":
self.CM.StartaCMnoBlock(node)
#get the result
if watch.lookforall():
self.CM.cluster_stable()
self.CM.fencing_cleanup("NearQuorumPoint", stonith)
return self.success()
self.logger.log("Warn: Patterns not found: " + repr(watch.unmatched))
#get the "bad" nodes
upnodes = []
for node in stopset:
if self.CM.StataCM(node) == 1:
upnodes.append(node)
downnodes = []
for node in startset:
if self.CM.StataCM(node) == 0:
downnodes.append(node)
self.CM.fencing_cleanup("NearQuorumPoint", stonith)
if upnodes == [] and downnodes == []:
self.CM.cluster_stable()
# Make sure they're completely down with no residule
for node in stopset:
self.rsh(node, self.templates["StopCmd"])
return self.success()
if len(upnodes) > 0:
self.logger.log("Warn: Unstoppable nodes: " + repr(upnodes))
if len(downnodes) > 0:
self.logger.log("Warn: Unstartable nodes: " + repr(downnodes))
return self.failure()
def is_applicable(self):
return 1
AllTestClasses.append(NearQuorumPointTest)
class RollingUpgradeTest(CTSTest):
'''Perform a rolling upgrade of the cluster'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "RollingUpgrade"
self.start = StartTest(cm)
self.stop = StopTest(cm)
self.stopall = SimulStopLite(cm)
self.startall = SimulStartLite(cm)
def setup(self, node):
# Start all remaining nodes
ret = self.stopall(None)
if not ret:
return self.failure("Couldn't stop all nodes")
for node in self.Env["nodes"]:
if not self.downgrade(node, None):
return self.failure("Couldn't downgrade %s" % node)
ret = self.startall(None)
if not ret:
return self.failure("Couldn't start all nodes")
return self.success()
def teardown(self, node):
# Stop everything
ret = self.stopall(None)
if not ret:
return self.failure("Couldn't stop all nodes")
for node in self.Env["nodes"]:
if not self.upgrade(node, None):
return self.failure("Couldn't upgrade %s" % node)
return self.success()
def install(self, node, version, start=1, flags="--force"):
target_dir = "/tmp/rpm-%s" % version
src_dir = "%s/%s" % (self.Env["rpm-dir"], version)
self.logger.log("Installing %s on %s with %s" % (version, node, flags))
if not self.stop(node):
return self.failure("stop failure: "+node)
rc = self.rsh(node, "mkdir -p %s" % target_dir)
rc = self.rsh(node, "rm -f %s/*.rpm" % target_dir)
(rc, lines) = self.rsh(node, "ls -1 %s/*.rpm" % src_dir, None)
for line in lines:
line = line[:-1]
rc = self.rsh.cp("%s" % (line), "%s:%s/" % (node, target_dir))
rc = self.rsh(node, "rpm -Uvh %s %s/*.rpm" % (flags, target_dir))
if start and not self.start(node):
return self.failure("start failure: "+node)
return self.success()
def upgrade(self, node, start=1):
return self.install(node, self.Env["current-version"], start)
def downgrade(self, node, start=1):
return self.install(node, self.Env["previous-version"], start, "--force --nodeps")
def __call__(self, node):
'''Perform the 'Rolling Upgrade' test. '''
self.incr("calls")
for node in self.Env["nodes"]:
if self.upgrade(node):
return self.failure("Couldn't upgrade %s" % node)
self.CM.cluster_stable()
return self.success()
def is_applicable(self):
if not self.is_applicable_common():
return None
if not "rpm-dir" in list(self.Env.keys()):
return None
if not "current-version" in list(self.Env.keys()):
return None
if not "previous-version" in list(self.Env.keys()):
return None
return 1
# Register RestartTest as a good test to run
AllTestClasses.append(RollingUpgradeTest)
class BSC_AddResource(CTSTest):
'''Add a resource to the cluster'''
def __init__(self, cm):
CTSTest.__init__(self, cm)
self.name = "AddResource"
self.resource_offset = 0
self.cib_cmd = """cibadmin -C -o %s -X '%s' """
def __call__(self, node):
self.incr("calls")
self.resource_offset = self.resource_offset + 1
r_id = "bsc-rsc-%s-%d" % (node, self.resource_offset)
start_pat = "pacemaker-controld.*%s_start_0.*confirmed.*ok"
patterns = []
patterns.append(start_pat % r_id)
watch = self.create_watch(patterns, self.Env["DeadTime"])
watch.setwatch()
ip = self.NextIP()
if not self.make_ip_resource(node, r_id, "ocf", "IPaddr", ip):
return self.failure("Make resource %s failed" % r_id)
failed = 0
watch_result = watch.lookforall()
if watch.unmatched:
for regex in watch.unmatched:
self.logger.log ("Warn: Pattern not found: %s" % (regex))
failed = 1
if failed:
return self.failure("Resource pattern(s) not found")
if not self.CM.cluster_stable(self.Env["DeadTime"]):
return self.failure("Unstable cluster")
return self.success()
def NextIP(self):
ip = self.Env["IPBase"]
if ":" in ip:
fields = ip.rpartition(":")
fields[2] = str(hex(int(fields[2], 16)+1))
print(str(hex(int(f[2], 16)+1)))
else:
fields = ip.rpartition('.')
fields[2] = str(int(fields[2])+1)
ip = fields[0] + fields[1] + fields[3];
self.Env["IPBase"] = ip
return ip.strip()
def make_ip_resource(self, node, id, rclass, type, ip):
self.logger.log("Creating %s:%s:%s (%s) on %s" % (rclass,type,id,ip,node))
rsc_xml="""
<primitive id="%s" class="%s" type="%s" provider="heartbeat">
<instance_attributes id="%s"><attributes>
<nvpair id="%s" name="ip" value="%s"/>
</attributes></instance_attributes>
</primitive>""" % (id, rclass, type, id, id, ip)
node_constraint = """
<rsc_location id="run_%s" rsc="%s">
<rule id="pref_run_%s" score="100">
<expression id="%s_loc_expr" attribute="#uname" operation="eq" value="%s"/>
</rule>
</rsc_location>""" % (id, id, id, id, node)
rc = 0
(rc, lines) = self.rsh(node, self.cib_cmd % ("constraints", node_constraint), None)
if rc != 0:
self.logger.log("Constraint creation failed: %d" % rc)
return None
(rc, lines) = self.rsh(node, self.cib_cmd % ("resources", rsc_xml), None)
if rc != 0:
self.logger.log("Resource creation failed: %d" % rc)
return None
return 1
def is_applicable(self):
if self.Env["DoBSC"]:
return 1
return None
AllTestClasses.append(BSC_AddResource)
class SimulStopLite(CTSTest):
'''Stop any active nodes ~ simultaneously'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "SimulStopLite"
def __call__(self, dummy):
'''Perform the 'SimulStopLite' setup work. '''
self.incr("calls")
self.debug("Setup: " + self.name)
# We ignore the "node" parameter...
watchpats = [ ]
for node in self.Env["nodes"]:
if self.CM.ShouldBeStatus[node] == "up":
self.incr("WasStarted")
watchpats.append(self.templates["Pat:We_stopped"] % node)
if len(watchpats) == 0:
return self.success()
# Stop all the nodes - at about the same time...
watch = self.create_watch(watchpats, self.Env["DeadTime"]+10)
watch.setwatch()
self.set_timer()
for node in self.Env["nodes"]:
if self.CM.ShouldBeStatus[node] == "up":
self.CM.StopaCMnoBlock(node)
if watch.lookforall():
# Make sure they're completely down with no residule
for node in self.Env["nodes"]:
self.rsh(node, self.templates["StopCmd"])
return self.success()
did_fail = 0
up_nodes = []
for node in self.Env["nodes"]:
if self.CM.StataCM(node) == 1:
did_fail = 1
up_nodes.append(node)
if did_fail:
return self.failure("Active nodes exist: " + repr(up_nodes))
self.logger.log("Warn: All nodes stopped but CTS didnt detect: "
+ repr(watch.unmatched))
return self.failure("Missing log message: "+repr(watch.unmatched))
def is_applicable(self):
'''SimulStopLite is a setup test and never applicable'''
return 0
class SimulStartLite(CTSTest):
'''Start any stopped nodes ~ simultaneously'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "SimulStartLite"
def __call__(self, dummy):
'''Perform the 'SimulStartList' setup work. '''
self.incr("calls")
self.debug("Setup: " + self.name)
# We ignore the "node" parameter...
node_list = []
for node in self.Env["nodes"]:
if self.CM.ShouldBeStatus[node] == "down":
self.incr("WasStopped")
node_list.append(node)
self.set_timer()
while len(node_list) > 0:
# Repeat until all nodes come up
watchpats = [ ]
uppat = self.templates["Pat:NonDC_started"]
if self.CM.upcount() == 0:
uppat = self.templates["Pat:Local_started"]
watchpats.append(self.templates["Pat:DC_IDLE"])
for node in node_list:
watchpats.append(uppat % node)
watchpats.append(self.templates["Pat:InfraUp"] % node)
watchpats.append(self.templates["Pat:PacemakerUp"] % node)
# Start all the nodes - at about the same time...
watch = self.create_watch(watchpats, self.Env["DeadTime"]+10)
watch.setwatch()
stonith = self.CM.prepare_fencing_watcher(self.name)
for node in node_list:
self.CM.StartaCMnoBlock(node)
watch.lookforall()
node_list = self.CM.fencing_cleanup(self.name, stonith)
if node_list == None:
return self.failure("Cluster did not stabilize")
# Remove node_list messages from watch.unmatched
for node in node_list:
self.logger.debug("Dealing with stonith operations for %s" % repr(node_list))
if watch.unmatched:
try:
watch.unmatched.remove(uppat % node)
except:
self.debug("Already matched: %s" % (uppat % node))
try:
watch.unmatched.remove(self.templates["Pat:InfraUp"] % node)
except:
self.debug("Already matched: %s" % (self.templates["Pat:InfraUp"] % node))
try:
watch.unmatched.remove(self.templates["Pat:PacemakerUp"] % node)
except:
self.debug("Already matched: %s" % (self.templates["Pat:PacemakerUp"] % node))
if watch.unmatched:
for regex in watch.unmatched:
self.logger.log ("Warn: Startup pattern not found: %s" %(regex))
if not self.CM.cluster_stable():
return self.failure("Cluster did not stabilize")
did_fail = 0
unstable = []
for node in self.Env["nodes"]:
if self.CM.StataCM(node) == 0:
did_fail = 1
unstable.append(node)
if did_fail:
return self.failure("Unstarted nodes exist: " + repr(unstable))
unstable = []
for node in self.Env["nodes"]:
if not self.CM.node_stable(node):
did_fail = 1
unstable.append(node)
if did_fail:
return self.failure("Unstable cluster nodes exist: " + repr(unstable))
return self.success()
def is_applicable(self):
'''SimulStartLite is a setup test and never applicable'''
return 0
def TestList(cm, audits):
result = []
for testclass in AllTestClasses:
bound_test = testclass(cm)
if bound_test.is_applicable():
bound_test.Audits = audits
result.append(bound_test)
return result
class RemoteLXC(CTSTest):
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "RemoteLXC"
self.start = StartTest(cm)
self.startall = SimulStartLite(cm)
self.num_containers = 2
self.is_container = 1
self.is_docker_unsafe = 1
self.failed = 0
self.fail_string = ""
def start_lxc_simple(self, node):
# restore any artifacts laying around from a previous test.
self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -s -R &>/dev/null")
# generate the containers, put them in the config, add some resources to them
pats = [ ]
watch = self.create_watch(pats, 120)
watch.setwatch()
pats.append(self.templates["Pat:RscOpOK"] % ("start", "lxc1"))
pats.append(self.templates["Pat:RscOpOK"] % ("start", "lxc2"))
pats.append(self.templates["Pat:RscOpOK"] % ("start", "lxc-ms"))
pats.append(self.templates["Pat:RscOpOK"] % ("promote", "lxc-ms"))
self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -g -a -m -s -c %d &>/dev/null" % self.num_containers)
self.set_timer("remoteSimpleInit")
watch.lookforall()
self.log_timer("remoteSimpleInit")
if watch.unmatched:
self.fail_string = "Unmatched patterns: %s" % (repr(watch.unmatched))
self.failed = 1
def cleanup_lxc_simple(self, node):
pats = [ ]
# if the test failed, attempt to clean up the cib and libvirt environment
# as best as possible
if self.failed == 1:
# restore libvirt and cib
self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -s -R &>/dev/null")
return
watch = self.create_watch(pats, 120)
watch.setwatch()
pats.append(self.templates["Pat:RscOpOK"] % ("stop", "container1"))
pats.append(self.templates["Pat:RscOpOK"] % ("stop", "container2"))
self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -p &>/dev/null")
self.set_timer("remoteSimpleCleanup")
watch.lookforall()
self.log_timer("remoteSimpleCleanup")
if watch.unmatched:
self.fail_string = "Unmatched patterns: %s" % (repr(watch.unmatched))
self.failed = 1
# cleanup libvirt
self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -s -R &>/dev/null")
def __call__(self, node):
'''Perform the 'RemoteLXC' test. '''
self.incr("calls")
ret = self.startall(None)
if not ret:
return self.failure("Setup failed, start all nodes failed.")
rc = self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -v &>/dev/null")
if rc == 1:
self.log("Environment test for lxc support failed.")
return self.skipped()
self.start_lxc_simple(node)
self.cleanup_lxc_simple(node)
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
if self.failed == 1:
return self.failure(self.fail_string)
return self.success()
def errorstoignore(self):
'''Return list of errors which should be ignored'''
return [
r"Updating failcount for ping",
r"schedulerd.*: Recover (ping|lxc-ms|container)\s*\(.*\)",
# The orphaned lxc-ms resource causes an expected transition error
# that is a result of the scheduler not having knowledge that the
# promotable resource used to be a clone. As a result, it looks like that
# resource is running in multiple locations when it shouldn't... But in
# this instance we know why this error is occurring and that it is expected.
r"Calculated [Tt]ransition .*pe-error",
r"Resource lxc-ms .* is active on 2 nodes attempting recovery",
r"Unknown operation: fail",
r"VirtualDomain.*ERROR: Unable to determine emulator",
]
AllTestClasses.append(RemoteLXC)
class RemoteDriver(CTSTest):
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = self.__class__.__name__
self.is_docker_unsafe = 1
self.start = StartTest(cm)
self.startall = SimulStartLite(cm)
self.stop = StopTest(cm)
self.remote_rsc = "remote-rsc"
self.cib_cmd = """cibadmin -C -o %s -X '%s' """
self.reset()
def reset(self):
self.pcmk_started = 0
self.failed = False
self.fail_string = ""
self.remote_node_added = 0
self.remote_rsc_added = 0
self.remote_use_reconnect_interval = self.Env.RandomGen.choice([True,False])
def fail(self, msg):
""" Mark test as failed. """
self.failed = True
# Always log the failure.
self.logger.log(msg)
# Use first failure as test status, as it's likely to be most useful.
if not self.fail_string:
self.fail_string = msg
def get_othernode(self, node):
for othernode in self.Env["nodes"]:
if othernode == node:
# we don't want to try and use the cib that we just shutdown.
# find a cluster node that is not our soon to be remote-node.
continue
else:
return othernode
def del_rsc(self, node, rsc):
othernode = self.get_othernode(node)
rc = self.rsh(othernode, "crm_resource -D -r %s -t primitive" % (rsc))
if rc != 0:
self.fail("Removal of resource '%s' failed" % rsc)
def add_rsc(self, node, rsc_xml):
othernode = self.get_othernode(node)
rc = self.rsh(othernode, self.cib_cmd % ("resources", rsc_xml))
if rc != 0:
self.fail("resource creation failed")
def add_primitive_rsc(self, node):
rsc_xml = """
<primitive class="ocf" id="%(node)s" provider="heartbeat" type="Dummy">
<meta_attributes id="%(node)s-meta_attributes"/>
<operations>
<op id="%(node)s-monitor-interval-20s" interval="20s" name="monitor"/>
</operations>
</primitive>""" % { "node": self.remote_rsc }
self.add_rsc(node, rsc_xml)
if not self.failed:
self.remote_rsc_added = 1
def add_connection_rsc(self, node):
rsc_xml = """
<primitive class="ocf" id="%(node)s" provider="pacemaker" type="remote">
<instance_attributes id="%(node)s-instance_attributes">
<nvpair id="%(node)s-instance_attributes-server" name="server" value="%(server)s"/>
""" % { "node": self.remote_node, "server": node }
if self.remote_use_reconnect_interval:
# Set reconnect interval on resource
rsc_xml = rsc_xml + """
<nvpair id="%s-instance_attributes-reconnect_interval" name="reconnect_interval" value="60s"/>
""" % (self.remote_node)
rsc_xml = rsc_xml + """
</instance_attributes>
<operations>
<op id="%(node)s-start" name="start" interval="0" timeout="120s"/>
<op id="%(node)s-monitor-20s" name="monitor" interval="20s" timeout="45s"/>
</operations>
</primitive>
""" % { "node": self.remote_node }
self.add_rsc(node, rsc_xml)
if not self.failed:
self.remote_node_added = 1
def disable_services(self, node):
self.corosync_enabled = self.Env.service_is_enabled(node, "corosync")
if self.corosync_enabled:
self.Env.disable_service(node, "corosync")
self.pacemaker_enabled = self.Env.service_is_enabled(node, "pacemaker")
if self.pacemaker_enabled:
self.Env.disable_service(node, "pacemaker")
def restore_services(self, node):
if self.corosync_enabled:
self.Env.enable_service(node, "corosync")
if self.pacemaker_enabled:
self.Env.enable_service(node, "pacemaker")
def stop_pcmk_remote(self, node):
# disable pcmk remote
for i in range(10):
rc = self.rsh(node, "service pacemaker_remote stop")
if rc != 0:
time.sleep(6)
else:
break
def start_pcmk_remote(self, node):
for i in range(10):
rc = self.rsh(node, "service pacemaker_remote start")
if rc != 0:
time.sleep(6)
else:
self.pcmk_started = 1
break
def freeze_pcmk_remote(self, node):
""" Simulate a Pacemaker Remote daemon failure. """
# We freeze the process.
self.rsh(node, "killall -STOP pacemaker-remoted")
def resume_pcmk_remote(self, node):
# We resume the process.
self.rsh(node, "killall -CONT pacemaker-remoted")
def start_metal(self, node):
# Cluster nodes are reused as remote nodes in remote tests. If cluster
# services were enabled at boot, in case the remote node got fenced, the
# cluster node would join instead of the expected remote one. Meanwhile
# pacemaker_remote would not be able to start. Depending on the chances,
# the situations might not be able to be orchestrated gracefully any more.
#
# Temporarily disable any enabled cluster serivces.
self.disable_services(node)
pcmk_started = 0
# make sure the resource doesn't already exist for some reason
self.rsh(node, "crm_resource -D -r %s -t primitive" % (self.remote_rsc))
self.rsh(node, "crm_resource -D -r %s -t primitive" % (self.remote_node))
if not self.stop(node):
self.fail("Failed to shutdown cluster node %s" % node)
return
self.start_pcmk_remote(node)
if self.pcmk_started == 0:
self.fail("Failed to start pacemaker_remote on node %s" % node)
return
# Convert node to baremetal now that it has shutdown the cluster stack
pats = [ ]
watch = self.create_watch(pats, 120)
watch.setwatch()
pats.append(self.templates["Pat:RscOpOK"] % ("start", self.remote_node))
pats.append(self.templates["Pat:DC_IDLE"])
self.add_connection_rsc(node)
self.set_timer("remoteMetalInit")
watch.lookforall()
self.log_timer("remoteMetalInit")
if watch.unmatched:
self.fail("Unmatched patterns: %s" % watch.unmatched)
def migrate_connection(self, node):
if self.failed:
return
pats = [ ]
pats.append(self.templates["Pat:RscOpOK"] % ("migrate_to", self.remote_node))
pats.append(self.templates["Pat:RscOpOK"] % ("migrate_from", self.remote_node))
pats.append(self.templates["Pat:DC_IDLE"])
watch = self.create_watch(pats, 120)
watch.setwatch()
(rc, lines) = self.rsh(node, "crm_resource -M -r %s" % (self.remote_node), None)
if rc != 0:
self.fail("failed to move remote node connection resource")
return
self.set_timer("remoteMetalMigrate")
watch.lookforall()
self.log_timer("remoteMetalMigrate")
if watch.unmatched:
self.fail("Unmatched patterns: %s" % watch.unmatched)
return
def fail_rsc(self, node):
if self.failed:
return
watchpats = [ ]
watchpats.append(self.templates["Pat:RscRemoteOpOK"] % ("stop", self.remote_rsc, self.remote_node))
watchpats.append(self.templates["Pat:RscRemoteOpOK"] % ("start", self.remote_rsc, self.remote_node))
watchpats.append(self.templates["Pat:DC_IDLE"])
watch = self.create_watch(watchpats, 120)
watch.setwatch()
self.debug("causing dummy rsc to fail.")
rc = self.rsh(node, "rm -f /var/run/resource-agents/Dummy*")
self.set_timer("remoteRscFail")
watch.lookforall()
self.log_timer("remoteRscFail")
if watch.unmatched:
self.fail("Unmatched patterns during rsc fail: %s" % watch.unmatched)
def fail_connection(self, node):
if self.failed:
return
watchpats = [ ]
watchpats.append(self.templates["Pat:FenceOpOK"] % self.remote_node)
watchpats.append(self.templates["Pat:NodeFenced"] % self.remote_node)
watch = self.create_watch(watchpats, 120)
watch.setwatch()
# freeze the pcmk remote daemon. this will result in fencing
self.debug("Force stopped active remote node")
self.freeze_pcmk_remote(node)
self.debug("Waiting for remote node to be fenced.")
self.set_timer("remoteMetalFence")
watch.lookforall()
self.log_timer("remoteMetalFence")
if watch.unmatched:
self.fail("Unmatched patterns: %s" % watch.unmatched)
return
self.debug("Waiting for the remote node to come back up")
self.CM.ns.WaitForNodeToComeUp(node, 120);
pats = [ ]
watch = self.create_watch(pats, 240)
watch.setwatch()
pats.append(self.templates["Pat:RscOpOK"] % ("start", self.remote_node))
if self.remote_rsc_added == 1:
pats.append(self.templates["Pat:RscRemoteOpOK"] % ("start", self.remote_rsc, self.remote_node))
# start the remote node again watch it integrate back into cluster.
self.start_pcmk_remote(node)
if self.pcmk_started == 0:
self.fail("Failed to start pacemaker_remote on node %s" % node)
return
self.debug("Waiting for remote node to rejoin cluster after being fenced.")
self.set_timer("remoteMetalRestart")
watch.lookforall()
self.log_timer("remoteMetalRestart")
if watch.unmatched:
self.fail("Unmatched patterns: %s" % watch.unmatched)
return
def add_dummy_rsc(self, node):
if self.failed:
return
# verify we can put a resource on the remote node
pats = [ ]
watch = self.create_watch(pats, 120)
watch.setwatch()
pats.append(self.templates["Pat:RscRemoteOpOK"] % ("start", self.remote_rsc, self.remote_node))
pats.append(self.templates["Pat:DC_IDLE"])
# Add a resource that must live on remote-node
self.add_primitive_rsc(node)
# force that rsc to prefer the remote node.
(rc, line) = self.CM.rsh(node, "crm_resource -M -r %s -N %s -f" % (self.remote_rsc, self.remote_node), None)
if rc != 0:
self.fail("Failed to place remote resource on remote node.")
return
self.set_timer("remoteMetalRsc")
watch.lookforall()
self.log_timer("remoteMetalRsc")
if watch.unmatched:
self.fail("Unmatched patterns: %s" % watch.unmatched)
def test_attributes(self, node):
if self.failed:
return
# This verifies permanent attributes can be set on a remote-node. It also
# verifies the remote-node can edit its own cib node section remotely.
(rc, line) = self.CM.rsh(node, "crm_attribute -l forever -n testattr -v testval -N %s" % (self.remote_node), None)
if rc != 0:
self.fail("Failed to set remote-node attribute. rc:%s output:%s" % (rc, line))
return
(rc, line) = self.CM.rsh(node, "crm_attribute -l forever -n testattr -q -N %s" % (self.remote_node), None)
if rc != 0:
self.fail("Failed to get remote-node attribute")
return
(rc, line) = self.CM.rsh(node, "crm_attribute -l forever -n testattr -D -N %s" % (self.remote_node), None)
if rc != 0:
self.fail("Failed to delete remote-node attribute")
return
def cleanup_metal(self, node):
self.restore_services(node)
if self.pcmk_started == 0:
return
pats = [ ]
watch = self.create_watch(pats, 120)
watch.setwatch()
if self.remote_rsc_added == 1:
pats.append(self.templates["Pat:RscOpOK"] % ("stop", self.remote_rsc))
if self.remote_node_added == 1:
pats.append(self.templates["Pat:RscOpOK"] % ("stop", self.remote_node))
self.set_timer("remoteMetalCleanup")
self.resume_pcmk_remote(node)
if self.remote_rsc_added == 1:
# Remove dummy resource added for remote node tests
self.debug("Cleaning up dummy rsc put on remote node")
self.rsh(self.get_othernode(node), "crm_resource -U -r %s" % self.remote_rsc)
self.del_rsc(node, self.remote_rsc)
if self.remote_node_added == 1:
# Remove remote node's connection resource
self.debug("Cleaning up remote node connection resource")
self.rsh(self.get_othernode(node), "crm_resource -U -r %s" % (self.remote_node))
self.del_rsc(node, self.remote_node)
watch.lookforall()
self.log_timer("remoteMetalCleanup")
if watch.unmatched:
self.fail("Unmatched patterns: %s" % watch.unmatched)
self.stop_pcmk_remote(node)
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
if self.remote_node_added == 1:
# Remove remote node itself
self.debug("Cleaning up node entry for remote node")
self.rsh(self.get_othernode(node), "crm_node --force --remove %s" % self.remote_node)
def setup_env(self, node):
self.remote_node = "remote-%s" % (node)
# we are assuming if all nodes have a key, that it is
# the right key... If any node doesn't have a remote
# key, we regenerate it everywhere.
if self.rsh.exists_on_all("/etc/pacemaker/authkey", self.Env["nodes"]):
return
# create key locally
(handle, keyfile) = tempfile.mkstemp(".cts")
os.close(handle)
subprocess.check_call(["dd", "if=/dev/urandom", "of=%s" % keyfile, "bs=4096", "count=1"],
stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
# sync key throughout the cluster
for node in self.Env["nodes"]:
self.rsh(node, "mkdir -p --mode=0750 /etc/pacemaker")
self.rsh.cp(keyfile, "root@%s:/etc/pacemaker/authkey" % node)
self.rsh(node, "chgrp haclient /etc/pacemaker /etc/pacemaker/authkey")
self.rsh(node, "chmod 0640 /etc/pacemaker/authkey")
os.unlink(keyfile)
def is_applicable(self):
if not self.is_applicable_common():
return False
for node in self.Env["nodes"]:
rc = self.rsh(node, "which pacemaker-remoted >/dev/null 2>&1")
if rc != 0:
return False
return True
def start_new_test(self, node):
self.incr("calls")
self.reset()
ret = self.startall(None)
if not ret:
return self.failure("setup failed: could not start all nodes")
self.setup_env(node)
self.start_metal(node)
self.add_dummy_rsc(node)
return True
def __call__(self, node):
return self.failure("This base class is not meant to be called directly.")
def errorstoignore(self):
'''Return list of errors which should be ignored'''
return [ r"""is running on remote.*which isn't allowed""",
r"""Connection terminated""",
r"""Could not send remote""",
]
# RemoteDriver is just a base class for other tests, so it is not added to AllTestClasses
class RemoteBasic(RemoteDriver):
def __call__(self, node):
'''Perform the 'RemoteBaremetal' test. '''
if not self.start_new_test(node):
return self.failure(self.fail_string)
self.test_attributes(node)
self.cleanup_metal(node)
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
if self.failed:
return self.failure(self.fail_string)
return self.success()
AllTestClasses.append(RemoteBasic)
class RemoteStonithd(RemoteDriver):
def __call__(self, node):
'''Perform the 'RemoteStonithd' test. '''
if not self.start_new_test(node):
return self.failure(self.fail_string)
self.fail_connection(node)
self.cleanup_metal(node)
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
if self.failed:
return self.failure(self.fail_string)
return self.success()
def is_applicable(self):
if not RemoteDriver.is_applicable(self):
return False
if "DoFencing" in list(self.Env.keys()):
return self.Env["DoFencing"]
return True
def errorstoignore(self):
ignore_pats = [
r"Lost connection to Pacemaker Remote node",
r"Software caused connection abort",
r"pacemaker-controld.*:\s+error.*: Operation remote-.*_monitor",
r"pacemaker-controld.*:\s+error.*: Result of monitor operation for remote-.*",
r"schedulerd.*:\s+Recover remote-.*\s*\(.*\)",
r"error: Result of monitor operation for .* on remote-.*: No executor connection",
]
ignore_pats.extend(RemoteDriver.errorstoignore(self))
return ignore_pats
AllTestClasses.append(RemoteStonithd)
class RemoteMigrate(RemoteDriver):
def __call__(self, node):
'''Perform the 'RemoteMigrate' test. '''
if not self.start_new_test(node):
return self.failure(self.fail_string)
self.migrate_connection(node)
self.cleanup_metal(node)
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
if self.failed:
return self.failure(self.fail_string)
return self.success()
+ def is_applicable(self):
+ if not RemoteDriver.is_applicable(self):
+ return 0
+ # This test requires at least three nodes: one to convert to a
+ # remote node, one to host the connection originally, and one
+ # to migrate the connection to.
+ if len(self.Env["nodes"]) < 3:
+ return 0
+ return 1
+
AllTestClasses.append(RemoteMigrate)
class RemoteRscFailure(RemoteDriver):
def __call__(self, node):
'''Perform the 'RemoteRscFailure' test. '''
if not self.start_new_test(node):
return self.failure(self.fail_string)
# This is an important step. We are migrating the connection
# before failing the resource. This verifies that the migration
# has properly maintained control over the remote-node.
self.migrate_connection(node)
self.fail_rsc(node)
self.cleanup_metal(node)
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
if self.failed:
return self.failure(self.fail_string)
return self.success()
def errorstoignore(self):
ignore_pats = [
r"schedulerd.*: Recover remote-rsc\s*\(.*\)",
r"Dummy.*: No process state file found",
]
ignore_pats.extend(RemoteDriver.errorstoignore(self))
return ignore_pats
+ def is_applicable(self):
+ if not RemoteDriver.is_applicable(self):
+ return 0
+ # This test requires at least three nodes: one to convert to a
+ # remote node, one to host the connection originally, and one
+ # to migrate the connection to.
+ if len(self.Env["nodes"]) < 3:
+ return 0
+ return 1
+
AllTestClasses.append(RemoteRscFailure)
# vim:ts=4:sw=4:et:
diff --git a/daemons/pacemakerd/pcmkd_corosync.c b/daemons/pacemakerd/pcmkd_corosync.c
index 4aaa383077..40e19a45dd 100644
--- a/daemons/pacemakerd/pcmkd_corosync.c
+++ b/daemons/pacemakerd/pcmkd_corosync.c
@@ -1,347 +1,352 @@
/*
* Copyright 2010-2021 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 <crm_internal.h>
#include "pacemakerd.h"
#include <sys/utsname.h>
#include <sys/stat.h> /* for calls to stat() */
#include <libgen.h> /* For basename() and dirname() */
#include <sys/types.h>
#include <pwd.h> /* For getpwname() */
#include <corosync/hdb.h>
#include <corosync/cfg.h>
#include <corosync/cpg.h>
#include <corosync/cmap.h>
#include <crm/cluster/internal.h>
#include <crm/common/ipc.h> /* for crm_ipc_is_authentic_process */
#include <crm/common/mainloop.h>
#include <crm/common/ipc_internal.h> /* PCMK__SPECIAL_PID* */
static corosync_cfg_handle_t cfg_handle = 0;
static mainloop_timer_t *reconnect_timer = NULL;
/* =::=::=::= CFG - Shutdown stuff =::=::=::= */
static void
cfg_shutdown_callback(corosync_cfg_handle_t h, corosync_cfg_shutdown_flags_t flags)
{
crm_info("Corosync wants to shut down: %s",
(flags == COROSYNC_CFG_SHUTDOWN_FLAG_IMMEDIATE) ? "immediate" :
(flags == COROSYNC_CFG_SHUTDOWN_FLAG_REGARDLESS) ? "forced" : "optional");
/* Never allow corosync to shut down while we're running */
corosync_cfg_replyto_shutdown(h, COROSYNC_CFG_SHUTDOWN_FLAG_NO);
}
static corosync_cfg_callbacks_t cfg_callbacks = {
.corosync_cfg_shutdown_callback = cfg_shutdown_callback,
};
static int
pcmk_cfg_dispatch(gpointer user_data)
{
corosync_cfg_handle_t *handle = (corosync_cfg_handle_t *) user_data;
cs_error_t rc = corosync_cfg_dispatch(*handle, CS_DISPATCH_ALL);
if (rc != CS_OK) {
return -1;
}
return 0;
}
+static void
+close_cfg(void)
+{
+ if (cfg_handle != 0) {
+#ifdef HAVE_COROSYNC_CFG_TRACKSTART
+ /* Ideally, we would call corosync_cfg_trackstop(cfg_handle) here, but a
+ * bug in corosync 3.1.1 and 3.1.2 makes it hang forever. Thankfully,
+ * it's not necessary since we exit immediately after this.
+ */
+#endif
+ corosync_cfg_finalize(cfg_handle);
+ cfg_handle = 0;
+ }
+}
+
static gboolean
cluster_reconnect_cb(gpointer data)
{
if (cluster_connect_cfg()) {
mainloop_timer_del(reconnect_timer);
reconnect_timer = NULL;
crm_notice("Cluster reconnect succeeded");
} else {
crm_info("Cluster reconnect failed"
"(connection will be reattempted once per second)");
}
/*
* In theory this will continue forever. In practice the CIB connection from
* attrd will timeout and shut down Pacemaker when it gets bored.
*/
return TRUE;
}
static void
cfg_connection_destroy(gpointer user_data)
{
crm_warn("Lost connection to cluster layer "
"(connection will be reattempted once per second)");
corosync_cfg_finalize(cfg_handle);
cfg_handle = 0;
reconnect_timer = mainloop_timer_add("corosync reconnect", 1000, TRUE, cluster_reconnect_cb, NULL);
mainloop_timer_start(reconnect_timer);
}
void
cluster_disconnect_cfg(void)
{
- if (cfg_handle) {
-#ifdef HAVE_COROSYNC_CFG_TRACKSTART
- corosync_cfg_trackstop(cfg_handle);
-#endif
- corosync_cfg_finalize(cfg_handle);
- cfg_handle = 0;
- }
+ close_cfg();
if (reconnect_timer != NULL) {
/* The mainloop should be gone by this point, so this isn't necessary,
* but cleaning up memory should make valgrind happier.
*/
mainloop_timer_del(reconnect_timer);
reconnect_timer = NULL;
}
}
#define cs_repeat(counter, max, code) do { \
code; \
if(rc == CS_ERR_TRY_AGAIN || rc == CS_ERR_QUEUE_FULL) { \
counter++; \
crm_debug("Retrying Corosync operation after %ds", counter); \
sleep(counter); \
} else { \
break; \
} \
} while(counter < max)
gboolean
cluster_connect_cfg(void)
{
cs_error_t rc;
int fd = -1, retries = 0, rv;
uid_t found_uid = 0;
gid_t found_gid = 0;
pid_t found_pid = 0;
uint32_t nodeid;
static struct mainloop_fd_callbacks cfg_fd_callbacks = {
.dispatch = pcmk_cfg_dispatch,
.destroy = cfg_connection_destroy,
};
cs_repeat(retries, 30, rc = corosync_cfg_initialize(&cfg_handle, &cfg_callbacks));
if (rc != CS_OK) {
crm_crit("Could not connect to Corosync CFG: %s " CRM_XS " rc=%d",
cs_strerror(rc), rc);
return FALSE;
}
rc = corosync_cfg_fd_get(cfg_handle, &fd);
if (rc != CS_OK) {
crm_crit("Could not get Corosync CFG descriptor: %s " CRM_XS " rc=%d",
cs_strerror(rc), rc);
goto bail;
}
/* CFG provider run as root (in given user namespace, anyway)? */
if (!(rv = crm_ipc_is_authentic_process(fd, (uid_t) 0,(gid_t) 0, &found_pid,
&found_uid, &found_gid))) {
crm_crit("Rejecting Corosync CFG provider because process %lld "
"is running as uid %lld gid %lld, not root",
(long long) PCMK__SPECIAL_PID_AS_0(found_pid),
(long long) found_uid, (long long) found_gid);
goto bail;
} else if (rv < 0) {
crm_crit("Could not authenticate Corosync CFG provider: %s "
CRM_XS " rc=%d", strerror(-rv), -rv);
goto bail;
}
retries = 0;
cs_repeat(retries, 30, rc = corosync_cfg_local_get(cfg_handle, &nodeid));
if (rc != CS_OK) {
crm_crit("Could not get local node ID from Corosync: %s "
CRM_XS " rc=%d", cs_strerror(rc), rc);
goto bail;
}
crm_debug("Corosync reports local node ID is %lu", (unsigned long) nodeid);
#ifdef HAVE_COROSYNC_CFG_TRACKSTART
rc = corosync_cfg_trackstart(cfg_handle, 0);
if (rc != CS_OK) {
crm_crit("Could not enable Corosync CFG shutdown tracker: %s " CRM_XS " rc=%d",
cs_strerror(rc), rc);
goto bail;
}
#endif
mainloop_add_fd("corosync-cfg", G_PRIORITY_DEFAULT, fd, &cfg_handle, &cfg_fd_callbacks);
return TRUE;
bail:
corosync_cfg_finalize(cfg_handle);
return FALSE;
}
void
pcmkd_shutdown_corosync(void)
{
cs_error_t rc;
if (cfg_handle == 0) {
crm_warn("Unable to shut down Corosync: No connection");
return;
}
crm_info("Asking Corosync to shut down");
rc = corosync_cfg_try_shutdown(cfg_handle,
COROSYNC_CFG_SHUTDOWN_FLAG_IMMEDIATE);
if (rc == CS_OK) {
-#ifdef HAVE_COROSYNC_CFG_TRACKSTART
- corosync_cfg_trackstop(cfg_handle);
-#endif
- corosync_cfg_finalize(cfg_handle);
- cfg_handle = 0;
+ close_cfg();
} else {
crm_warn("Corosync shutdown failed: %s " CRM_XS " rc=%d",
cs_strerror(rc), rc);
}
}
/* =::=::=::= Configuration =::=::=::= */
static int
get_config_opt(uint64_t unused, cmap_handle_t object_handle, const char *key, char **value,
const char *fallback)
{
int rc = 0, retries = 0;
cs_repeat(retries, 5, rc = cmap_get_string(object_handle, key, value));
if (rc != CS_OK) {
crm_trace("Search for %s failed %d, defaulting to %s", key, rc, fallback);
if (fallback) {
*value = strdup(fallback);
} else {
*value = NULL;
}
}
crm_trace("%s: %s", key, *value);
return rc;
}
gboolean
mcp_read_config(void)
{
cs_error_t rc = CS_OK;
int retries = 0;
cmap_handle_t local_handle;
uint64_t config = 0;
int fd = -1;
uid_t found_uid = 0;
gid_t found_gid = 0;
pid_t found_pid = 0;
int rv;
enum cluster_type_e stack;
// There can be only one possibility
do {
rc = cmap_initialize(&local_handle);
if (rc != CS_OK) {
retries++;
crm_info("Could not connect to Corosync CMAP: %s (retrying in %ds) "
CRM_XS " rc=%d", cs_strerror(rc), retries, rc);
sleep(retries);
} else {
break;
}
} while (retries < 5);
if (rc != CS_OK) {
crm_crit("Could not connect to Corosync CMAP: %s "
CRM_XS " rc=%d", cs_strerror(rc), rc);
return FALSE;
}
rc = cmap_fd_get(local_handle, &fd);
if (rc != CS_OK) {
crm_crit("Could not get Corosync CMAP descriptor: %s " CRM_XS " rc=%d",
cs_strerror(rc), rc);
cmap_finalize(local_handle);
return FALSE;
}
/* CMAP provider run as root (in given user namespace, anyway)? */
if (!(rv = crm_ipc_is_authentic_process(fd, (uid_t) 0,(gid_t) 0, &found_pid,
&found_uid, &found_gid))) {
crm_crit("Rejecting Corosync CMAP provider because process %lld "
"is running as uid %lld gid %lld, not root",
(long long) PCMK__SPECIAL_PID_AS_0(found_pid),
(long long) found_uid, (long long) found_gid);
cmap_finalize(local_handle);
return FALSE;
} else if (rv < 0) {
crm_crit("Could not authenticate Corosync CMAP provider: %s "
CRM_XS " rc=%d", strerror(-rv), -rv);
cmap_finalize(local_handle);
return FALSE;
}
stack = get_cluster_type();
if (stack != pcmk_cluster_corosync) {
crm_crit("Expected corosync stack but detected %s " CRM_XS " stack=%d",
name_for_cluster_type(stack), stack);
return FALSE;
}
crm_info("Reading configuration for %s stack",
name_for_cluster_type(stack));
pcmk__set_env_option("cluster_type", "corosync");
pcmk__set_env_option("quorum_type", "corosync");
// If debug logging is not configured, check whether corosync has it
if (pcmk__env_option("debug") == NULL) {
char *debug_enabled = NULL;
get_config_opt(config, local_handle, "logging.debug", &debug_enabled, "off");
if (crm_is_true(debug_enabled)) {
pcmk__set_env_option("debug", "1");
if (get_crm_log_level() < LOG_DEBUG) {
set_crm_log_level(LOG_DEBUG);
}
} else {
pcmk__set_env_option("debug", "0");
}
free(debug_enabled);
}
if(local_handle){
gid_t gid = 0;
if (pcmk_daemon_user(NULL, &gid) < 0) {
crm_warn("Could not authorize group with Corosync " CRM_XS
" No group found for user %s", CRM_DAEMON_USER);
} else {
char key[PATH_MAX];
snprintf(key, PATH_MAX, "uidgid.gid.%u", gid);
rc = cmap_set_uint8(local_handle, key, 1);
if (rc != CS_OK) {
crm_warn("Could not authorize group with Corosync: %s " CRM_XS
" group=%u rc=%d", pcmk__cs_err_str(rc), gid, rc);
}
}
}
cmap_finalize(local_handle);
return TRUE;
}
diff --git a/lib/common/watchdog.c b/lib/common/watchdog.c
index 2d1f401ba1..ea4b62c819 100644
--- a/lib/common/watchdog.c
+++ b/lib/common/watchdog.c
@@ -1,307 +1,315 @@
/*
* Copyright 2013-2020 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 <crm_internal.h>
#include <sched.h>
#include <sys/ioctl.h>
#include <sys/reboot.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <unistd.h>
#include <ctype.h>
#include <dirent.h>
#include <signal.h>
#ifdef _POSIX_MEMLOCK
# include <sys/mman.h>
#endif
static pid_t sbd_pid = 0;
static void
sysrq_trigger(char t)
{
#if SUPPORT_PROCFS
FILE *procf;
// Root can always write here, regardless of kernel.sysrq value
procf = fopen("/proc/sysrq-trigger", "a");
if (!procf) {
crm_perror(LOG_WARNING, "Opening sysrq-trigger failed");
return;
}
crm_info("sysrq-trigger: %c", t);
fprintf(procf, "%c\n", t);
fclose(procf);
#endif // SUPPORT_PROCFS
return;
}
/*!
* \internal
* \brief Panic the local host (if root) or tell pacemakerd to do so
*/
static void
panic_local(void)
{
int rc = pcmk_ok;
uid_t uid = geteuid();
pid_t ppid = getppid();
if(uid != 0 && ppid > 1) {
/* We're a non-root pacemaker daemon (pacemaker-based,
* pacemaker-controld, pacemaker-schedulerd, pacemaker-attrd, etc.) with
* the original pacemakerd parent.
*
* Of these, only the controller is likely to be initiating resets.
*/
crm_emerg("Signaling parent %lld to panic", (long long) ppid);
crm_exit(CRM_EX_PANIC);
return;
} else if (uid != 0) {
#if SUPPORT_PROCFS
/*
* No permissions, and no pacemakerd parent to escalate to.
* Track down the new pacemakerd process and send a signal instead.
*/
union sigval signal_value;
memset(&signal_value, 0, sizeof(signal_value));
ppid = pcmk__procfs_pid_of("pacemakerd");
crm_emerg("Signaling pacemakerd[%lld] to panic", (long long) ppid);
if(ppid > 1 && sigqueue(ppid, SIGQUIT, signal_value) < 0) {
crm_perror(LOG_EMERG, "Cannot signal pacemakerd[%lld] to panic",
(long long) ppid);
}
#endif // SUPPORT_PROCFS
/* The best we can do now is die */
crm_exit(CRM_EX_PANIC);
return;
}
/* We're either pacemakerd, or a pacemaker daemon running as root */
if (pcmk__str_eq("crash", getenv("PCMK_panic_action"), pcmk__str_casei)) {
sysrq_trigger('c');
} else if (pcmk__str_eq("sync-crash", getenv("PCMK_panic_action"), pcmk__str_casei)) {
sync();
sysrq_trigger('c');
} else {
if (pcmk__str_eq("sync-reboot", getenv("PCMK_panic_action"), pcmk__str_casei)) {
sync();
}
sysrq_trigger('b');
}
/* reboot(RB_HALT_SYSTEM); rc = errno; */
reboot(RB_AUTOBOOT);
rc = errno;
crm_emerg("Reboot failed, escalating to parent %lld: %s " CRM_XS " rc=%d",
(long long) ppid, pcmk_rc_str(rc), rc);
if(ppid > 1) {
/* child daemon */
exit(CRM_EX_PANIC);
} else {
/* pacemakerd or orphan child */
exit(CRM_EX_FATAL);
}
}
/*!
* \internal
* \brief Tell sbd to kill the local host, then exit
*/
static void
panic_sbd(void)
{
union sigval signal_value;
pid_t ppid = getppid();
crm_emerg("Signaling sbd[%lld] to panic", (long long) sbd_pid);
memset(&signal_value, 0, sizeof(signal_value));
/* TODO: Arrange for a slightly less brutal option? */
if(sigqueue(sbd_pid, SIGKILL, signal_value) < 0) {
crm_perror(LOG_EMERG, "Cannot signal sbd[%lld] to terminate",
(long long) sbd_pid);
panic_local();
}
if(ppid > 1) {
/* child daemon */
exit(CRM_EX_PANIC);
} else {
/* pacemakerd or orphan child */
exit(CRM_EX_FATAL);
}
}
/*!
* \internal
* \brief Panic the local host
*
* Panic the local host either by sbd (if running), directly, or by asking
* pacemakerd. If trace logging this function, exit instead.
*
* \param[in] origin Function caller (for logging only)
*/
void
pcmk__panic(const char *origin)
{
static struct qb_log_callsite *panic_cs = NULL;
if (panic_cs == NULL) {
panic_cs = qb_log_callsite_get(__func__, __FILE__, "panic-delay",
LOG_TRACE, __LINE__, crm_trace_nonlog);
}
/* Ensure sbd_pid is set */
(void) pcmk__locate_sbd();
if (panic_cs && panic_cs->targets) {
/* getppid() == 1 means our original parent no longer exists */
crm_emerg("Shutting down instead of panicking the node "
CRM_XS " origin=%s sbd=%lld parent=%d",
origin, (long long) sbd_pid, getppid());
crm_exit(CRM_EX_FATAL);
return;
}
if(sbd_pid > 1) {
crm_emerg("Signaling sbd[%lld] to panic the system: %s",
(long long) sbd_pid, origin);
panic_sbd();
} else {
crm_emerg("Panicking the system directly: %s", origin);
panic_local();
}
}
/*!
* \internal
* \brief Return the process ID of sbd (or 0 if it is not running)
*/
pid_t
pcmk__locate_sbd(void)
{
char *pidfile = NULL;
char *sbd_path = NULL;
int rc;
if(sbd_pid > 1) {
return sbd_pid;
}
/* Look for the pid file */
pidfile = crm_strdup_printf(PCMK_RUN_DIR "/sbd.pid");
sbd_path = crm_strdup_printf("%s/sbd", SBIN_DIR);
/* Read the pid file */
rc = pcmk__pidfile_matches(pidfile, 0, sbd_path, &sbd_pid);
if (rc == pcmk_rc_ok) {
crm_trace("SBD detected at pid %lld (via PID file %s)",
(long long) sbd_pid, pidfile);
#if SUPPORT_PROCFS
} else {
/* Fall back to /proc for systems that support it */
sbd_pid = pcmk__procfs_pid_of("sbd");
crm_trace("SBD detected at pid %lld (via procfs)",
(long long) sbd_pid);
#endif // SUPPORT_PROCFS
}
if(sbd_pid < 0) {
sbd_pid = 0;
crm_trace("SBD not detected");
}
free(pidfile);
free(sbd_path);
return sbd_pid;
}
long
pcmk__get_sbd_timeout(void)
{
static long sbd_timeout = -2;
if (sbd_timeout == -2) {
sbd_timeout = crm_get_msec(getenv("SBD_WATCHDOG_TIMEOUT"));
}
return sbd_timeout;
}
bool
pcmk__get_sbd_sync_resource_startup(void)
{
- static bool sync_resource_startup = false;
+ static int sync_resource_startup = PCMK__SBD_SYNC_DEFAULT;
static bool checked_sync_resource_startup = false;
if (!checked_sync_resource_startup) {
- sync_resource_startup =
- crm_is_true(getenv("SBD_SYNC_RESOURCE_STARTUP"));
+ const char *sync_env = getenv("SBD_SYNC_RESOURCE_STARTUP");
+
+ if (sync_env == NULL) {
+ crm_trace("Defaulting to %sstart-up synchronization with sbd",
+ (PCMK__SBD_SYNC_DEFAULT? "" : "no "));
+
+ } else if (crm_str_to_boolean(sync_env, &sync_resource_startup) < 0) {
+ crm_warn("Defaulting to %sstart-up synchronization with sbd "
+ "because environment value '%s' is invalid",
+ (PCMK__SBD_SYNC_DEFAULT? "" : "no "), sync_env);
+ }
checked_sync_resource_startup = true;
}
-
- return sync_resource_startup;
+ return sync_resource_startup != 0;
}
long
pcmk__auto_watchdog_timeout()
{
long sbd_timeout = pcmk__get_sbd_timeout();
return (sbd_timeout <= 0)? 0 : (2 * sbd_timeout);
}
bool
pcmk__valid_sbd_timeout(const char *value)
{
long st_timeout = value? crm_get_msec(value) : 0;
if (st_timeout < 0) {
st_timeout = pcmk__auto_watchdog_timeout();
crm_debug("Using calculated value %ld for stonith-watchdog-timeout (%s)",
st_timeout, value);
}
if (st_timeout == 0) {
crm_debug("Watchdog may be enabled but stonith-watchdog-timeout is disabled (%s)",
value? value : "default");
} else if (pcmk__locate_sbd() == 0) {
crm_emerg("Shutting down: stonith-watchdog-timeout configured (%s) "
"but SBD not active", (value? value : "auto"));
crm_exit(CRM_EX_FATAL);
return false;
} else {
long sbd_timeout = pcmk__get_sbd_timeout();
if (st_timeout < sbd_timeout) {
crm_emerg("Shutting down: stonith-watchdog-timeout (%s) too short "
"(must be >%ldms)", value, sbd_timeout);
crm_exit(CRM_EX_FATAL);
return false;
}
crm_info("Watchdog configured with stonith-watchdog-timeout %s and SBD timeout %ldms",
value, sbd_timeout);
}
return true;
}
diff --git a/rpm/pacemaker.spec.in b/rpm/pacemaker.spec.in
index 1e7e4b9eb3..f7bdc208c2 100644
--- a/rpm/pacemaker.spec.in
+++ b/rpm/pacemaker.spec.in
@@ -1,872 +1,886 @@
# User-configurable globals and defines to control package behavior
# (these should not test {with X} values, which are declared later)
## User and group to use for nonprivileged services
%global uname hacluster
%global gname haclient
## Where to install Pacemaker documentation
%if 0%{?suse_version} > 0
%global pcmk_docdir %{_docdir}/%{name}-%{version}
%else
%if 0%{?rhel} > 7
%global pcmk_docdir %{_docdir}/%{name}-doc
%else
%global pcmk_docdir %{_docdir}/%{name}
%endif
%endif
## GitHub entity that distributes source (for ease of using a fork)
%global github_owner ClusterLabs
## Where bug reports should be submitted
## Leave bug_url undefined to use ClusterLabs default, others define it here
## Upstream pacemaker version, and its package version (specversion
## can be incremented to build packages reliably considered "newer"
## than previously built packages with the same pcmkversion)
%global pcmkversion X.Y.Z
%global specversion 1
## Upstream commit (full commit ID, abbreviated commit ID, or tag) to build
%global commit HEAD
## Since git v2.11, the extent of abbreviation is autoscaled by default
## (used to be constant of 7), so we need to convey it for non-tags, too.
%global commit_abbrev 7
# Define conditionals so that "rpmbuild --with <feature>" and
# "rpmbuild --without <feature>" can enable and disable specific features
## Add option to enable support for stonith/external fencing agents
%bcond_with stonithd
## Add option to enable support for storing sensitive information outside CIB
%bcond_with cibsecrets
## Add option to create binaries suitable for use with profiling tools
%bcond_with profiling
## Add option to create binaries with coverage analysis
%bcond_with coverage
## Add option to skip (or enable, on RHEL) generating documentation
## (the build tools aren't available everywhere)
%if 0%{?rhel}
%bcond_with doc
%else
%bcond_without doc
%endif
+## Add option to default to start-up synchronization with SBD.
+##
+## If enabled, SBD *MUST* be built to default similarly, otherwise data
+## corruption could occur. Building both Pacemaker and SBD to default
+## to synchronization improves safety, without requiring higher-level tools
+## to be aware of the setting or requiring users to modify configurations
+## after upgrading to versions that support synchronization.
+%if 0%{?rhel} && 0%{?rhel} > 8
+%bcond_without sbd_sync
+%else
+%bcond_with sbd_sync
+%endif
+
## Add option to prefix package version with "0."
## (so later "official" packages will be considered updates)
%bcond_with pre_release
## Add option to ship Upstart job files
%bcond_with upstart_job
## Add option to turn off hardening of libraries and daemon executables
%bcond_without hardening
## Add option to enable (or disable, on RHEL 8) links for legacy daemon names
%if 0%{?rhel} && 0%{?rhel} <= 8
%bcond_without legacy_links
%else
%bcond_with legacy_links
%endif
# Define globals for convenient use later
## Workaround to use parentheses in other globals
%global lparen (
%global rparen )
## Whether this is a tagged release (final or release candidate)
%define tag_release %(c=%{commit}; case ${c} in Pacemaker-*%{rparen} echo 1 ;;
*%{rparen} echo 0 ;; esac)
## Portion of export/dist tarball name after "pacemaker-", and release version
%if 0%{tag_release}
%define archive_version %(c=%{commit}; echo ${c:10})
%define archive_github_url %{commit}#/%{name}-%{archive_version}.tar.gz
%define pcmk_release %(c=%{commit}; case $c in *-rc[[:digit:]]*%{rparen}
echo 0.%{specversion}.${c: -3} ;;
*%{rparen} echo %{specversion} ;; esac)
%else
%define archive_version %(c=%{commit}; echo ${c:0:%{commit_abbrev}})
%define archive_github_url %{archive_version}#/%{name}-%{archive_version}.tar.gz
%if %{with pre_release}
%define pcmk_release 0.%{specversion}.%{archive_version}.git
%else
%define pcmk_release %{specversion}.%{archive_version}.git
%endif
%endif
## Whether this platform defaults to using systemd as an init system
## (needs to be evaluated prior to BuildRequires being enumerated and
## installed as it's intended to conditionally select some of these, and
## for that there are only few indicators with varying reliability:
## - presence of systemd-defined macros (when building in a full-fledged
## environment, which is not the case with ordinary mock-based builds)
## - systemd-aware rpm as manifested with the presence of particular
## macro (rpm itself will trivially always be present when building)
## - existence of /usr/lib/os-release file, which is something heavily
## propagated by systemd project
## - when not good enough, there's always a possibility to check
## particular distro-specific macros (incl. version comparison)
%define systemd_native (%{?_unitdir:1}%{!?_unitdir:0}%{nil \
} || %{?__transaction_systemd_inhibit:1}%{!?__transaction_systemd_inhibit:0}%{nil \
} || %(test -f /usr/lib/os-release; test $? -ne 0; echo $?))
%if 0%{?fedora} > 20 || 0%{?rhel} > 7
## Base GnuTLS cipher priorities (presumably only the initial, required keyword)
## overridable with "rpmbuild --define 'pcmk_gnutls_priorities PRIORITY-SPEC'"
%define gnutls_priorities %{?pcmk_gnutls_priorities}%{!?pcmk_gnutls_priorities:@SYSTEM}
%endif
%if !%{defined _rundir}
%if 0%{?fedora} >= 15 || 0%{?rhel} >= 7 || 0%{?suse_version} >= 1200
%define _rundir /run
%else
%define _rundir /var/run
%endif
%endif
%if 0%{?fedora} > 22 || 0%{?rhel} > 7
%global supports_recommends 1
%endif
## Different distros name certain packages differently
## (note: corosync libraries also differ, but all provide corosync-devel)
%if 0%{?suse_version} > 0
%global pkgname_bzip2_devel libbz2-devel
%global pkgname_docbook_xsl docbook-xsl-stylesheets
%global pkgname_gnutls_devel libgnutls-devel
%global pkgname_shadow_utils shadow
%global pkgname_procps procps
%global pkgname_glue_libs libglue
%global pkgname_pcmk_libs lib%{name}3
%global hacluster_id 90
%else
%global pkgname_libtool_devel libtool-ltdl-devel
%global pkgname_libtool_devel_arch libtool-ltdl-devel%{?_isa}
%global pkgname_bzip2_devel bzip2-devel
%global pkgname_docbook_xsl docbook-style-xsl
%global pkgname_gnutls_devel gnutls-devel
%global pkgname_shadow_utils shadow-utils
%global pkgname_procps procps-ng
%global pkgname_glue_libs cluster-glue-libs
%global pkgname_pcmk_libs %{name}-libs
%global hacluster_id 189
%endif
## Distro-specific configuration choices
### Use 2.0-style output when other distro packages don't support current output
%if 0%{?fedora} || ( 0%{?rhel} && 0%{?rhel} <= 8 )
%global compat20 --enable-compat-2.0
%endif
### Default concurrent-fencing to true when distro prefers that
%if 0%{?rhel} >= 7
%global concurrent_fencing --with-concurrent-fencing-default=true
%endif
### Default resource-stickiness to 1 when distro prefers that
%if 0%{?fedora} >= 35 || 0%{?rhel} >= 9
%global resource_stickiness --with-resource-stickiness-default=1
%endif
# Python-related definitions
## Turn off auto-compilation of Python files outside Python specific paths,
## so there's no risk that unexpected "__python" macro gets picked to do the
## RPM-native byte-compiling there (only "{_datadir}/pacemaker/tests" affected)
## -- distro-dependent tricks or automake's fallback to be applied there
%if %{defined _python_bytecompile_extra}
%global _python_bytecompile_extra 0
%else
### the statement effectively means no RPM-native byte-compiling will occur at
### all, so distro-dependent tricks for Python-specific packages to be applied
%global __os_install_post %(echo '%{__os_install_post}' | {
sed -e 's!/usr/lib[^[:space:]]*/brp-python-bytecompile[[:space:]].*$!!g'; })
%endif
## Prefer Python 3 definitions explicitly, in case 2 is also available
%if %{defined __python3}
%global python_name python3
%global python_path %{__python3}
%define python_site %{?python3_sitelib}%{!?python3_sitelib:%(
%{python_path} -c 'from distutils.sysconfig import get_python_lib as gpl; print(gpl(1))' 2>/dev/null)}
%else
%if %{defined python_version}
%global python_name python%(echo %{python_version} | cut -d'.' -f1)
%define python_path %{?__python}%{!?__python:/usr/bin/%{python_name}}
%else
%global python_name python
%global python_path %{?__python}%{!?__python:/usr/bin/python%{?python_pkgversion}}
%endif
%define python_site %{?python_sitelib}%{!?python_sitelib:%(
%{python_name} -c 'from distutils.sysconfig import get_python_lib as gpl; print(gpl(1))' 2>/dev/null)}
%endif
# Keep sane profiling data if requested
%if %{with profiling}
## Disable -debuginfo package and stripping binaries/libraries
%define debug_package %{nil}
%endif
Name: pacemaker
Summary: Scalable High-Availability cluster resource manager
Version: %{pcmkversion}
Release: %{pcmk_release}%{?dist}
%if %{defined _unitdir}
License: GPLv2+ and LGPLv2+
%else
# initscript is Revised BSD
License: GPLv2+ and LGPLv2+ and BSD
%endif
Url: https://www.clusterlabs.org/
# Example: https://codeload.github.com/ClusterLabs/pacemaker/tar.gz/e91769e
# will download pacemaker-e91769e.tar.gz
#
# The ending part starting with '#' is ignored by github but necessary for
# rpmbuild to know what the tar archive name is. (The downloaded file will be
# named correctly only for commit IDs, not tagged releases.)
#
# You can use "spectool -s 0 pacemaker.spec" (rpmdevtools) to show final URL.
Source0: https://codeload.github.com/%{github_owner}/%{name}/tar.gz/%{archive_github_url}
Requires: resource-agents
Requires: %{pkgname_pcmk_libs}%{?_isa} = %{version}-%{release}
Requires: %{name}-cluster-libs%{?_isa} = %{version}-%{release}
Requires: %{name}-cli = %{version}-%{release}
%if !%{defined _unitdir}
Requires: %{pkgname_procps}
Requires: psmisc
%endif
%{?systemd_requires}
Requires: %{python_path}
BuildRequires: %{python_name}-devel
# Pacemaker requires a minimum libqb functionality
Requires: libqb >= 0.17.0
BuildRequires: libqb-devel >= 0.17.0
# Required basic build tools
BuildRequires: coreutils findutils grep sed
BuildRequires: autoconf automake gcc make pkgconfig
BuildRequires: libtool %{?pkgname_libtool_devel}
# Required for core functionality
BuildRequires: pkgconfig(glib-2.0) >= 2.42
BuildRequires: libxml2-devel libxslt-devel libuuid-devel
BuildRequires: %{pkgname_bzip2_devel}
# Enables optional functionality
BuildRequires: ncurses-devel %{pkgname_docbook_xsl}
BuildRequires: help2man %{pkgname_gnutls_devel} pam-devel pkgconfig(dbus-1)
%if %{systemd_native}
BuildRequires: pkgconfig(systemd)
%endif
Requires: corosync >= 2.0.0
BuildRequires: corosync-devel >= 2.0.0
%if %{with stonithd}
BuildRequires: %{pkgname_glue_libs}-devel
%endif
%if %{with doc}
BuildRequires: asciidoc inkscape %{python_name}-sphinx
%endif
Provides: pcmk-cluster-manager = %{version}-%{release}
Provides: pcmk-cluster-manager%{?_isa} = %{version}-%{release}
# Bundled bits
## Pacemaker uses the crypto/md5-buffer module from gnulib
%if 0%{?fedora} || 0%{?rhel}
Provides: bundled(gnulib)
%endif
%description
Pacemaker is an advanced, scalable High-Availability cluster resource
manager.
It supports more than 16 node clusters with significant capabilities
for managing resources and dependencies.
It will run scripts at initialization, when machines go up or down,
when related resources fail and can be configured to periodically check
resource health.
Available rpmbuild rebuild options:
--with(out) : cibsecrets coverage doc hardening pre_release profiling stonithd
upstart_job
%package cli
License: GPLv2+ and LGPLv2+
Summary: Command line tools for controlling Pacemaker clusters
Requires: %{pkgname_pcmk_libs}%{?_isa} = %{version}-%{release}
%if 0%{?supports_recommends}
Recommends: pcmk-cluster-manager = %{version}-%{release}
# For crm_report
Recommends: tar
Recommends: bzip2
%endif
Requires: perl-TimeDate
Requires: %{pkgname_procps}
Requires: psmisc
Requires(post):coreutils
%description cli
Pacemaker is an advanced, scalable High-Availability cluster resource
manager.
The %{name}-cli package contains command line tools that can be used
to query and control the cluster from machines that may, or may not,
be part of the cluster.
%package -n %{pkgname_pcmk_libs}
License: GPLv2+ and LGPLv2+
Summary: Core Pacemaker libraries
Requires(pre): %{pkgname_shadow_utils}
Requires: %{name}-schemas = %{version}-%{release}
# sbd 1.4.0+ supports the libpe_status API for pe_working_set_t
Conflicts: sbd < 1.4.0
%description -n %{pkgname_pcmk_libs}
Pacemaker is an advanced, scalable High-Availability cluster resource
manager.
The %{pkgname_pcmk_libs} package contains shared libraries needed for cluster
nodes and those just running the CLI tools.
%package cluster-libs
License: GPLv2+ and LGPLv2+
Summary: Cluster Libraries used by Pacemaker
Requires: %{pkgname_pcmk_libs}%{?_isa} = %{version}-%{release}
%description cluster-libs
Pacemaker is an advanced, scalable High-Availability cluster resource
manager.
The %{name}-cluster-libs package contains cluster-aware shared
libraries needed for nodes that will form part of the cluster nodes.
%package remote
%if %{defined _unitdir}
License: GPLv2+ and LGPLv2+
%else
# initscript is Revised BSD
License: GPLv2+ and LGPLv2+ and BSD
%endif
Summary: Pacemaker remote executor daemon for non-cluster nodes
Requires: %{pkgname_pcmk_libs}%{?_isa} = %{version}-%{release}
Requires: %{name}-cli = %{version}-%{release}
Requires: resource-agents
%if !%{defined _unitdir}
Requires: %{pkgname_procps}
%endif
# -remote can be fully independent of systemd
%{?systemd_ordering}%{!?systemd_ordering:%{?systemd_requires}}
Provides: pcmk-cluster-manager = %{version}-%{release}
Provides: pcmk-cluster-manager%{?_isa} = %{version}-%{release}
%description remote
Pacemaker is an advanced, scalable High-Availability cluster resource
manager.
The %{name}-remote package contains the Pacemaker Remote daemon
which is capable of extending pacemaker functionality to remote
nodes not running the full corosync/cluster stack.
%package -n %{pkgname_pcmk_libs}-devel
License: GPLv2+ and LGPLv2+
Summary: Pacemaker development package
Requires: %{pkgname_pcmk_libs}%{?_isa} = %{version}-%{release}
Requires: %{name}-cluster-libs%{?_isa} = %{version}-%{release}
Requires: %{?pkgname_libtool_devel_arch} libuuid-devel%{?_isa}
Requires: libxml2-devel%{?_isa} libxslt-devel%{?_isa}
Requires: %{pkgname_bzip2_devel}%{?_isa} glib2-devel%{?_isa}
Requires: libqb-devel%{?_isa}
Requires: corosync-devel >= 2.0.0
%description -n %{pkgname_pcmk_libs}-devel
Pacemaker is an advanced, scalable High-Availability cluster resource
manager.
The %{pkgname_pcmk_libs}-devel package contains headers and shared libraries
for developing tools for Pacemaker.
%package cts
License: GPLv2+ and LGPLv2+
Summary: Test framework for cluster-related technologies like Pacemaker
Requires: %{python_path}
Requires: %{pkgname_pcmk_libs} = %{version}-%{release}
Requires: %{name}-cli = %{version}-%{release}
Requires: %{pkgname_procps}
Requires: psmisc
BuildArch: noarch
# systemd Python bindings are a separate package in some distros
%if %{defined systemd_requires}
%if 0%{?fedora} > 22 || 0%{?rhel} > 7
Requires: %{python_name}-systemd
%endif
%endif
%description cts
Test framework for cluster-related technologies like Pacemaker
%package doc
License: CC-BY-SA-4.0
Summary: Documentation for Pacemaker
BuildArch: noarch
%description doc
Documentation for Pacemaker.
Pacemaker is an advanced, scalable High-Availability cluster resource
manager.
%package schemas
License: GPLv2+
Summary: Schemas and upgrade stylesheets for Pacemaker
BuildArch: noarch
%description schemas
Schemas and upgrade stylesheets for Pacemaker
Pacemaker is an advanced, scalable High-Availability cluster resource
manager.
%prep
%setup -q -n %{name}-%{archive_version}
%build
export systemdsystemunitdir=%{?_unitdir}%{!?_unitdir:no}
%if %{with hardening}
# prefer distro-provided hardening flags in case they are defined
# through _hardening_{c,ld}flags macros, configure script will
# use its own defaults otherwise; if such hardenings are completely
# undesired, rpmbuild using "--without hardening"
# (or "--define '_without_hardening 1'")
export CFLAGS_HARDENED_EXE="%{?_hardening_cflags}"
export CFLAGS_HARDENED_LIB="%{?_hardening_cflags}"
export LDFLAGS_HARDENED_EXE="%{?_hardening_ldflags}"
export LDFLAGS_HARDENED_LIB="%{?_hardening_ldflags}"
%endif
./autogen.sh
%{configure} \
PYTHON=%{python_path} \
%{!?with_hardening: --disable-hardening} \
%{?with_legacy_links: --enable-legacy-links} \
%{?with_profiling: --with-profiling} \
%{?with_coverage: --with-coverage} \
%{?with_cibsecrets: --with-cibsecrets} \
+ %{?with_sbd_sync: --with-sbd-sync-default="true"} \
%{?gnutls_priorities: --with-gnutls-priorities="%{gnutls_priorities}"} \
%{?bug_url: --with-bug-url=%{bug_url}} \
%{?concurrent_fencing} \
%{?resource_stickiness} \
%{?compat20} \
--with-initdir=%{_initrddir} \
--with-runstatedir=%{_rundir} \
--localstatedir=%{_var} \
--with-version=%{version}-%{release}
%if 0%{?suse_version} >= 1200
# Fedora handles rpath removal automagically
sed -i 's|^hardcode_libdir_flag_spec=.*|hardcode_libdir_flag_spec=""|g' libtool
sed -i 's|^runpath_var=LD_RUN_PATH|runpath_var=DIE_RPATH_DIE|g' libtool
%endif
make %{_smp_mflags} V=1
%check
make %{_smp_mflags} check
{ cts/cts-scheduler --run load-stopped-loop \
&& cts/cts-cli \
&& touch .CHECKED
} 2>&1 | sed 's/[fF]ail/faiil/g' # prevent false positives in rpmlint
[ -f .CHECKED ] && rm -f -- .CHECKED
exit $? # TODO remove when rpm<4.14 compatibility irrelevant
%install
# skip automake-native Python byte-compilation, since RPM-native one (possibly
# distro-confined to Python-specific directories, which is currently the only
# relevant place, anyway) assures proper intrinsic alignment with wider system
# (such as with py_byte_compile macro, which is concurrent Fedora/EL specific)
make install \
DESTDIR=%{buildroot} V=1 docdir=%{pcmk_docdir} \
%{?_python_bytecompile_extra:%{?py_byte_compile:am__py_compile=true}}
%if %{with upstart_job}
mkdir -p ${RPM_BUILD_ROOT}%{_sysconfdir}/init
install -m 644 pacemakerd/pacemaker.upstart ${RPM_BUILD_ROOT}%{_sysconfdir}/init/pacemaker.conf
install -m 644 pacemakerd/pacemaker.combined.upstart ${RPM_BUILD_ROOT}%{_sysconfdir}/init/pacemaker.combined.conf
install -m 644 tools/crm_mon.upstart ${RPM_BUILD_ROOT}%{_sysconfdir}/init/crm_mon.conf
%endif
%if %{defined _unitdir}
mkdir -p ${RPM_BUILD_ROOT}%{_localstatedir}/lib/rpm-state/%{name}
%endif
# Don't package static libs
find %{buildroot} -name '*.a' -type f -print0 | xargs -0 rm -f
find %{buildroot} -name '*.la' -type f -print0 | xargs -0 rm -f
# For now, don't package the servicelog-related binaries built only for
# ppc64le when certain dependencies are installed. If they get more exercise by
# advanced users, we can reconsider.
rm -f %{buildroot}/%{_sbindir}/notifyServicelogEvent
rm -f %{buildroot}/%{_sbindir}/ipmiservicelogd
# Don't ship init scripts for systemd based platforms
%if %{defined _unitdir}
rm -f %{buildroot}/%{_initrddir}/pacemaker
rm -f %{buildroot}/%{_initrddir}/pacemaker_remote
%endif
# Byte-compile Python sources where suitable and the distro procedures known
%if %{defined py_byte_compile}
%{py_byte_compile %{python_path} %{buildroot}%{_datadir}/pacemaker/tests}
%if !%{defined _python_bytecompile_extra}
%{py_byte_compile %{python_path} %{buildroot}%{python_site}/cts}
%endif
%endif
%if %{with coverage}
GCOV_BASE=%{buildroot}/%{_var}/lib/pacemaker/gcov
mkdir -p $GCOV_BASE
find . -name '*.gcno' -type f | while read F ; do
D=`dirname $F`
mkdir -p ${GCOV_BASE}/$D
cp $F ${GCOV_BASE}/$D
done
%endif
%post
%if %{defined _unitdir}
%systemd_post pacemaker.service
%else
/sbin/chkconfig --add pacemaker || :
%endif
%preun
%if %{defined _unitdir}
%systemd_preun pacemaker.service
%else
/sbin/service pacemaker stop >/dev/null 2>&1 || :
if [ "$1" -eq 0 ]; then
# Package removal, not upgrade
/sbin/chkconfig --del pacemaker || :
fi
%endif
%postun
%if %{defined _unitdir}
%systemd_postun_with_restart pacemaker.service
%endif
%pre remote
%if %{defined _unitdir}
# Stop the service before anything is touched, and remember to restart
# it as one of the last actions (compared to using systemd_postun_with_restart,
# this avoids suicide when sbd is in use)
systemctl --quiet is-active pacemaker_remote
if [ $? -eq 0 ] ; then
mkdir -p %{_localstatedir}/lib/rpm-state/%{name}
touch %{_localstatedir}/lib/rpm-state/%{name}/restart_pacemaker_remote
systemctl stop pacemaker_remote >/dev/null 2>&1
else
rm -f %{_localstatedir}/lib/rpm-state/%{name}/restart_pacemaker_remote
fi
%endif
%post remote
%if %{defined _unitdir}
%systemd_post pacemaker_remote.service
%else
/sbin/chkconfig --add pacemaker_remote || :
%endif
%preun remote
%if %{defined _unitdir}
%systemd_preun pacemaker_remote.service
%else
/sbin/service pacemaker_remote stop >/dev/null 2>&1 || :
if [ "$1" -eq 0 ]; then
# Package removal, not upgrade
/sbin/chkconfig --del pacemaker_remote || :
fi
%endif
%postun remote
%if %{defined _unitdir}
# This next line is a no-op, because we stopped the service earlier, but
# we leave it here because it allows us to revert to the standard behavior
# in the future if desired
%systemd_postun_with_restart pacemaker_remote.service
# Explicitly take care of removing the flag-file(s) upon final removal
if [ "$1" -eq 0 ] ; then
rm -f %{_localstatedir}/lib/rpm-state/%{name}/restart_pacemaker_remote
fi
%endif
%posttrans remote
%if %{defined _unitdir}
if [ -e %{_localstatedir}/lib/rpm-state/%{name}/restart_pacemaker_remote ] ; then
systemctl start pacemaker_remote >/dev/null 2>&1
rm -f %{_localstatedir}/lib/rpm-state/%{name}/restart_pacemaker_remote
fi
%endif
%post cli
%if %{defined _unitdir}
%systemd_post crm_mon.service
%endif
if [ "$1" -eq 2 ]; then
# Package upgrade, not initial install:
# Move any pre-2.0 logs to new location to ensure they get rotated
{ mv -fbS.rpmsave %{_var}/log/pacemaker.log* %{_var}/log/pacemaker \
|| mv -f %{_var}/log/pacemaker.log* %{_var}/log/pacemaker
} >/dev/null 2>/dev/null || :
fi
%preun cli
%if %{defined _unitdir}
%systemd_preun crm_mon.service
%endif
%postun cli
%if %{defined _unitdir}
%systemd_postun_with_restart crm_mon.service
%endif
%pre -n %{pkgname_pcmk_libs}
getent group %{gname} >/dev/null || groupadd -r %{gname} -g %{hacluster_id}
getent passwd %{uname} >/dev/null || useradd -r -g %{gname} -u %{hacluster_id} -s /sbin/nologin -c "cluster user" %{uname}
exit 0
%if %{defined ldconfig_scriptlets}
%ldconfig_scriptlets -n %{pkgname_pcmk_libs}
%ldconfig_scriptlets cluster-libs
%else
%post -n %{pkgname_pcmk_libs} -p /sbin/ldconfig
%postun -n %{pkgname_pcmk_libs} -p /sbin/ldconfig
%post cluster-libs -p /sbin/ldconfig
%postun cluster-libs -p /sbin/ldconfig
%endif
%files
###########################################################
%config(noreplace) %{_sysconfdir}/sysconfig/pacemaker
%{_sbindir}/pacemakerd
%if %{defined _unitdir}
%{_unitdir}/pacemaker.service
%else
%{_initrddir}/pacemaker
%endif
%exclude %{_libexecdir}/pacemaker/cts-log-watcher
%exclude %{_libexecdir}/pacemaker/cts-support
%exclude %{_sbindir}/pacemaker-remoted
%exclude %{_sbindir}/pacemaker_remoted
%{_libexecdir}/pacemaker/*
%{_sbindir}/crm_attribute
%{_sbindir}/crm_master
%{_sbindir}/fence_legacy
%doc %{_mandir}/man7/pacemaker-controld.*
%doc %{_mandir}/man7/pacemaker-schedulerd.*
%doc %{_mandir}/man7/pacemaker-fenced.*
%doc %{_mandir}/man7/ocf_pacemaker_controld.*
%doc %{_mandir}/man7/ocf_pacemaker_o2cb.*
%doc %{_mandir}/man7/ocf_pacemaker_remote.*
%doc %{_mandir}/man8/crm_attribute.*
%doc %{_mandir}/man8/crm_master.*
%doc %{_mandir}/man8/fence_legacy.*
%doc %{_mandir}/man8/pacemakerd.*
%doc %{_datadir}/pacemaker/alerts
%license licenses/GPLv2
%doc COPYING
%doc ChangeLog
%dir %attr (750, %{uname}, %{gname}) %{_var}/lib/pacemaker/cib
%dir %attr (750, %{uname}, %{gname}) %{_var}/lib/pacemaker/pengine
/usr/lib/ocf/resource.d/pacemaker/controld
/usr/lib/ocf/resource.d/pacemaker/o2cb
/usr/lib/ocf/resource.d/pacemaker/remote
%if %{with upstart_job}
%config(noreplace) %{_sysconfdir}/init/pacemaker.conf
%config(noreplace) %{_sysconfdir}/init/pacemaker.combined.conf
%endif
%files cli
%dir %attr (750, root, %{gname}) %{_sysconfdir}/pacemaker
%config(noreplace) %{_sysconfdir}/logrotate.d/pacemaker
%config(noreplace) %{_sysconfdir}/sysconfig/crm_mon
%if %{defined _unitdir}
%{_unitdir}/crm_mon.service
%endif
%if %{with upstart_job}
%config(noreplace) %{_sysconfdir}/init/crm_mon.conf
%endif
%{_sbindir}/attrd_updater
%{_sbindir}/cibadmin
%if %{with cibsecrets}
%{_sbindir}/cibsecret
%endif
%{_sbindir}/crm_diff
%{_sbindir}/crm_error
%{_sbindir}/crm_failcount
%{_sbindir}/crm_mon
%{_sbindir}/crm_node
%{_sbindir}/crm_resource
%{_sbindir}/crm_rule
%{_sbindir}/crm_standby
%{_sbindir}/crm_verify
%{_sbindir}/crmadmin
%{_sbindir}/iso8601
%{_sbindir}/crm_shadow
%{_sbindir}/crm_simulate
%{_sbindir}/crm_report
%{_sbindir}/crm_ticket
%{_sbindir}/stonith_admin
# "dirname" is owned by -schemas, which is a prerequisite
%{_datadir}/pacemaker/report.collector
%{_datadir}/pacemaker/report.common
# XXX "dirname" is not owned by any prerequisite
%{_datadir}/snmp/mibs/PCMK-MIB.txt
%exclude /usr/lib/ocf/resource.d/pacemaker/controld
%exclude /usr/lib/ocf/resource.d/pacemaker/o2cb
%exclude /usr/lib/ocf/resource.d/pacemaker/remote
%dir /usr/lib/ocf
%dir /usr/lib/ocf/resource.d
/usr/lib/ocf/resource.d/pacemaker
%doc %{_mandir}/man7/*
%exclude %{_mandir}/man7/pacemaker-controld.*
%exclude %{_mandir}/man7/pacemaker-schedulerd.*
%exclude %{_mandir}/man7/pacemaker-fenced.*
%exclude %{_mandir}/man7/ocf_pacemaker_controld.*
%exclude %{_mandir}/man7/ocf_pacemaker_o2cb.*
%exclude %{_mandir}/man7/ocf_pacemaker_remote.*
%doc %{_mandir}/man8/*
%exclude %{_mandir}/man8/crm_attribute.*
%exclude %{_mandir}/man8/crm_master.*
%exclude %{_mandir}/man8/fence_legacy.*
%exclude %{_mandir}/man8/pacemakerd.*
%exclude %{_mandir}/man8/pacemaker-remoted.*
%license licenses/GPLv2
%doc COPYING
%doc ChangeLog
%dir %attr (750, %{uname}, %{gname}) %{_var}/lib/pacemaker
%dir %attr (750, %{uname}, %{gname}) %{_var}/lib/pacemaker/blackbox
%dir %attr (750, %{uname}, %{gname}) %{_var}/lib/pacemaker/cores
%dir %attr (770, %{uname}, %{gname}) %{_var}/log/pacemaker
%dir %attr (770, %{uname}, %{gname}) %{_var}/log/pacemaker/bundles
%files -n %{pkgname_pcmk_libs}
%{_libdir}/libcib.so.*
%{_libdir}/liblrmd.so.*
%{_libdir}/libcrmservice.so.*
%{_libdir}/libcrmcommon.so.*
%{_libdir}/libpe_status.so.*
%{_libdir}/libpe_rules.so.*
%{_libdir}/libpacemaker.so.*
%{_libdir}/libstonithd.so.*
%license licenses/LGPLv2.1
%doc COPYING
%doc ChangeLog
%files cluster-libs
%{_libdir}/libcrmcluster.so.*
%license licenses/LGPLv2.1
%doc COPYING
%doc ChangeLog
%files remote
%config(noreplace) %{_sysconfdir}/sysconfig/pacemaker
%if %{defined _unitdir}
# state directory is shared between the subpackets
# let rpm take care of removing it once it isn't
# referenced anymore and empty
%ghost %dir %{_localstatedir}/lib/rpm-state/%{name}
%{_unitdir}/pacemaker_remote.service
%else
%{_initrddir}/pacemaker_remote
%endif
%{_sbindir}/pacemaker-remoted
%{_sbindir}/pacemaker_remoted
%{_mandir}/man8/pacemaker-remoted.*
%license licenses/GPLv2
%doc COPYING
%doc ChangeLog
%files doc
%doc %{pcmk_docdir}
%license licenses/CC-BY-SA-4.0
%files cts
%{python_site}/cts
%{_datadir}/pacemaker/tests
%{_libexecdir}/pacemaker/cts-log-watcher
%{_libexecdir}/pacemaker/cts-support
%license licenses/GPLv2
%doc COPYING
%doc ChangeLog
%files -n %{pkgname_pcmk_libs}-devel
%{_includedir}/pacemaker
%{_libdir}/*.so
%if %{with coverage}
%{_var}/lib/pacemaker/gcov
%endif
%{_libdir}/pkgconfig/*.pc
%license licenses/LGPLv2.1
%doc COPYING
%doc ChangeLog
%files schemas
%license licenses/GPLv2
%dir %{_datadir}/pacemaker
%{_datadir}/pacemaker/*.rng
%{_datadir}/pacemaker/*.xsl
%{_datadir}/pacemaker/api
%{_datadir}/pkgconfig/pacemaker-schemas.pc
%changelog
* PACKAGE_DATE ClusterLabs <admin@clusterlabs.org> PACKAGE_VERSION-1
- See included ChangeLog file for details