Page MenuHomeClusterLabs Projects
Files F4624545

No OneTemporary
Actions

View Options

This file is larger than 256 KB, so syntax highlighting was skipped.
diff --git a/GNUmakefile b/GNUmakefile
index 9a472d96df..408f7c12a9 100644
--- a/GNUmakefile
+++ b/GNUmakefile
@@ -1,420 +1,420 @@
#
# Copyright 2008-2019 the Pacemaker project contributors
#
# The version control history for this file may have further details.
#
# This source code is licensed under the GNU General Public License version 2
# or later (GPLv2+) WITHOUT ANY WARRANTY.
#
default: build
.PHONY: default
# The toplevel "clean" targets are generated from Makefile.am, not this file.
# We can't use autotools' CLEANFILES, clean-local, etc. here. Instead, we
# define this target, which Makefile.am can use as a dependency of clean-local.
EXTRA_CLEAN_TARGETS = ancillary-clean
-include Makefile
# The main purpose of this GNUmakefile is that its targets can be invoked
# without having to call autogen.sh and configure first. That means automake
# variables may or may not be defined. Here, we use the current working
# directory if a relevant variable hasn't been defined.
#
# The idea is to keep generated artifacts in the build tree, in case a VPATH
# build is in use, but in practice it would be difficult to make the targets
# here usable from a different location than the source tree.
abs_srcdir ?= $(shell pwd)
abs_builddir ?= $(shell pwd)
PACKAGE ?= pacemaker
# Definitions that specify what various targets will apply to
COMMIT ?= HEAD
TAG ?= $(shell T=$$(git describe --tags --exact-match '$(COMMIT)' 2>/dev/null); \
test -n "$${T}" && echo "$${T}" \
|| git log --pretty=format:%H -n 1 '$(COMMIT)' 2>/dev/null || echo DIST)
lparen = (
rparen = )
SHORTTAG ?= $(shell case $(TAG) in Pacemaker-*|DIST$(rparen) echo '$(TAG)' | cut -c11-;; \
*$(rparen) git log --pretty=format:%h -n 1 '$(TAG)';; esac)
SHORTTAG_ABBREV = $(shell printf %s '$(SHORTTAG)' | wc -c)
LAST_RC ?= $(shell test -e /Volumes || git tag -l | grep Pacemaker | sort -Vr | grep rc | head -n 1)
ifneq ($(origin VERSION), undefined)
LAST_RELEASE ?= Pacemaker-$(VERSION)
else
LAST_RELEASE ?= $(shell git tag -l | grep Pacemaker | sort -Vr | grep -v rc | head -n 1)
endif
NEXT_RELEASE ?= $(shell echo $(LAST_RELEASE) | awk -F. '/[0-9]+\./{$$3+=1;OFS=".";print $$1,$$2,$$3}')
# This Makefile can create 2 types of distributions:
#
# - "make dist" is automake's native functionality, based on the various
# dist/nodist make variables; it always uses the current sources
#
# - "make export" is a custom target based on git archive and relevant entries
# from .gitattributes; it defaults to current sources but can use any git tag
#
# Both types use the TARFILE name for the result, though they generate
# different contents.
distdir = $(PACKAGE)-$(SHORTTAG)
TARFILE = $(abs_builddir)/$(PACKAGE)-$(SHORTTAG).tar.gz
.PHONY: init
init:
test -e $(top_srcdir)/configure || ./autogen.sh
test -e $(abs_builddir)/Makefile || $(abs_builddir)/configure
.PHONY: build
build: init
$(MAKE) $(AM_MAKEFLAGS) core
export:
if [ ! -f "$(TARFILE)" ]; then \
if [ $(TAG) = dirty ]; then \
git commit -m "DO-NOT-PUSH" -a; \
git archive --prefix=$(distdir)/ -o "$(TARFILE)" HEAD^{tree}; \
git reset --mixed HEAD^; \
else \
git archive --prefix=$(distdir)/ -o "$(TARFILE)" $(TAG)^{tree}; \
fi; \
echo "`date`: Rebuilt $(TARFILE)"; \
else \
echo "`date`: Using existing tarball: $(TARFILE)"; \
fi
## RPM-related targets
# Where to put RPM artifacts; possible values:
#
# - toplevel (default): RPM sources, spec, and source rpm in top-level build
# directory (everything else uses the usual defaults)
#
# - subtree: RPM sources (i.e. TARFILE) in top-level build directory,
# everything else in dedicated "rpm" subdirectory of build tree
RPMDEST ?= toplevel
RPM_SPEC_DIR_toplevel = $(abs_builddir)
RPM_SRCRPM_DIR_toplevel = $(abs_builddir)
RPM_OPTS_toplevel = --define "_sourcedir $(abs_builddir)" \
--define "_specdir $(RPM_SPEC_DIR_toplevel)" \
--define "_srcrpmdir $(RPM_SRCRPM_DIR_toplevel)"
RPM_SPEC_DIR_subtree = $(abs_builddir)/rpm/SPECS
RPM_SRCRPM_DIR_subtree = $(abs_builddir)/rpm/SRPMS
RPM_OPTS_subtree = --define "_sourcedir $(abs_builddir)" \
--define "_topdir $(abs_builddir)/rpm"
RPM_SPEC_DIR = $(RPM_SPEC_DIR_$(RPMDEST))
RPM_SRCRPM_DIR = $(RPM_SRCRPM_DIR_$(RPMDEST))
RPM_OPTS = $(RPM_OPTS_$(RPMDEST))
WITH ?= --without doc
BUILD_COUNTER ?= build.counter
LAST_COUNT = $(shell test ! -e $(BUILD_COUNTER) && echo 0; test -e $(BUILD_COUNTER) && cat $(BUILD_COUNTER))
COUNT = $(shell expr 1 + $(LAST_COUNT))
SPECVERSION ?= $(COUNT)
MOCK_DIR = $(abs_builddir)/mock
MOCK_OPTIONS ?= --resultdir=$(MOCK_DIR) --no-cleanup-after
F ?= $(shell test ! -e /etc/fedora-release && echo 0; test -e /etc/fedora-release && rpm --eval %{fedora})
ARCH ?= $(shell test ! -e /etc/fedora-release && uname -m; test -e /etc/fedora-release && rpm --eval %{_arch})
MOCK_CFG ?= $(shell test -e /etc/fedora-release && echo fedora-$(F)-$(ARCH))
# rpmbuild wrapper that translates "--with[out] FEATURE" into RPM macros
#
# Unfortunately, at least recent versions of rpm do not support mentioned
# switch. To work this around, we can emulate mechanism that rpm uses
# internally: unfold the flags into respective macro definitions:
#
# --with[out] FOO -> --define "_with[out]_FOO --with[out]-FOO"
#
# $(1) ... WITH string (e.g., --with pre_release --without doc)
# $(2) ... options following the initial "rpmbuild" in the command
# $(3) ... final arguments determined with $2 (e.g., pacemaker.spec)
#
# Note that if $(3) is a specfile, extra case is taken so as to reflect
# pcmkversion correctly (using in-place modification).
#
# Also note that both ways to specify long option with an argument
# (i.e., what getopt and, importantly, rpm itself support) can be used:
#
# --with FOO
# --with=FOO
rpmbuild-with = \
WITH=$$(getopt -o "" -l with:,without: -- $(1)) || exit 1; \
CMD='rpmbuild $(2)'; PREREL=0; \
eval set -- "$${WITH}"; \
while true; do \
case "$$1" in \
--with) CMD="$${CMD} --define \"_with_$$2 --with-$$2\""; \
[ "$$2" != pre_release ] || PREREL=1; shift 2;; \
--without) CMD="$${CMD} --define \"_without_$$2 --without-$$2\""; \
[ "$$2" != pre_release ] || PREREL=0; shift 2;; \
--) shift ; break ;; \
*) echo "cannot parse WITH: $$1"; exit 1;; \
esac; \
done; \
case "$(3)" in \
*.spec) { [ $${PREREL} -eq 0 ] || [ $(LAST_RELEASE) = $(TAG) ]; } \
&& sed -i "s/^\(%global pcmkversion \).*/\1$$(echo $(LAST_RELEASE) | sed -e s:Pacemaker-:: -e s:-.*::)/" $(3) \
|| sed -i "s/^\(%global pcmkversion \).*/\1$$(echo $(NEXT_RELEASE) | sed -e s:Pacemaker-:: -e s:-.*::)/" $(3);; \
esac; \
CMD="$${CMD} $(3)"; \
eval "$${CMD}"
# Depend on spec-clean so it gets rebuilt every time
$(RPM_SPEC_DIR)/$(PACKAGE).spec: spec-clean rpm/pacemaker.spec.in
$(AM_V_at)$(MKDIR_P) $(RPM_SPEC_DIR) # might not exist in VPATH build
$(AM_V_GEN)if [ x != x"`git ls-files -m rpm/pacemaker.spec.in 2>/dev/null`" ]; then \
cat $(abs_srcdir)/rpm/pacemaker.spec.in; \
elif git cat-file -e $(TAG):rpm/pacemaker.spec.in 2>/dev/null; then \
git show $(TAG):rpm/pacemaker.spec.in; \
elif git cat-file -e $(TAG):pacemaker.spec.in 2>/dev/null; then \
git show $(TAG):pacemaker.spec.in; \
else \
cat $(abs_srcdir)/rpm/pacemaker.spec.in; \
fi | sed \
-e "s/^\(%global pcmkversion \).*/\1$$(echo $(LAST_RELEASE) | sed -e s:Pacemaker-:: -e s:-.*::)/" \
-e 's/global\ specversion\ .*/global\ specversion\ $(SPECVERSION)/' \
-e 's/global\ commit\ .*/global\ commit\ $(SHORTTAG)/' \
-e 's/global\ commit_abbrev\ .*/global\ commit_abbrev\ $(SHORTTAG_ABBREV)/' \
-e "s/PACKAGE_DATE/$$(date +'%a %b %d %Y')/" \
-e "s/PACKAGE_VERSION/$$(git describe --tags $(TAG) | sed -e s:Pacemaker-:: -e s:-.*::)/" \
> "$@"
.PHONY: $(PACKAGE).spec
$(PACKAGE).spec: $(RPM_SPEC_DIR)/$(PACKAGE).spec
.PHONY: spec-clean
spec-clean:
-rm -f $(RPM_SPEC_DIR)/$(PACKAGE).spec
.PHONY: srpm
srpm: export srpm-clean $(RPM_SPEC_DIR)/$(PACKAGE).spec
if [ -e $(BUILD_COUNTER) ]; then \
echo $(COUNT) > $(BUILD_COUNTER); \
fi
$(call rpmbuild-with,$(WITH),-bs $(RPM_OPTS),$(RPM_SPEC_DIR)/$(PACKAGE).spec)
.PHONY: srpm-clean
srpm-clean:
-rm -f $(RPM_SRCRPM_DIR)/*.src.rpm
.PHONY: chroot
chroot: mock-$(MOCK_CFG) mock-install-$(MOCK_CFG) mock-sh-$(MOCK_CFG)
@echo "Done"
.PHONY: mock-next
mock-next:
$(MAKE) $(AM_MAKEFLAGS) F=$(shell expr 1 + $(F)) mock
.PHONY: mock-rawhide
mock-rawhide:
$(MAKE) $(AM_MAKEFLAGS) F=rawhide mock
mock-install-%:
@echo "Installing packages"
mock --root=$* $(MOCK_OPTIONS) --install $(MOCK_DIR)/*.rpm \
vi sudo valgrind lcov gdb fence-agents psmisc
.PHONY: mock-install
mock-install: mock-install-$(MOCK_CFG)
@echo "Done"
.PHONY: mock-sh
mock-sh: mock-sh-$(MOCK_CFG)
@echo "Done"
mock-sh-%:
@echo "Connecting"
mock --root=$* $(MOCK_OPTIONS) --shell
@echo "Done"
mock-%: srpm mock-clean
mock $(MOCK_OPTIONS) --root=$* --no-cleanup-after --rebuild \
$(WITH) $(RPM_SRCRPM_DIR)/*.src.rpm
.PHONY: mock
mock: mock-$(MOCK_CFG)
@echo "Done"
.PHONY: dirty
dirty:
$(MAKE) $(AM_MAKEFLAGS) TAG=dirty mock
.PHONY: mock-clean
mock-clean:
-rm -rf $(MOCK_DIR)
.PHONY: rpm-dep
rpm-dep: $(RPM_SPEC_DIR)/$(PACKAGE).spec
sudo yum-builddep $(PACKAGE).spec
# e.g. make WITH="--with pre_release" rpm
.PHONY: rpm
rpm: srpm
@echo To create custom builds, edit the flags and options in $(PACKAGE).spec first
$(call rpmbuild-with,$(WITH),$(RPM_OPTS),--rebuild $(RPM_SRCRPM_DIR)/*.src.rpm)
.PHONY: rpmlint
rpmlint: $(RPM_SPEC_DIR)/$(PACKAGE).spec
rpmlint -f rpm/rpmlintrc "$<"
.PHONY: release
release:
$(MAKE) $(AM_MAKEFLAGS) TAG=$(LAST_RELEASE) rpm
.PHONY: rc
rc:
$(MAKE) $(AM_MAKEFLAGS) TAG=$(LAST_RC) rpm
## Static analysis via coverity
# Aggressiveness (low, medium, or high)
COVLEVEL ?= low
# Generated outputs
COVERITY_DIR = $(abs_builddir)/coverity-$(TAG)
COVTAR = $(abs_builddir)/$(PACKAGE)-coverity-$(TAG).tgz
COVEMACS = $(abs_builddir)/$(TAG).coverity
COVHTML = $(COVERITY_DIR)/output/errors
# Coverity outputs are phony so they get rebuilt every invocation
.PHONY: $(COVERITY_DIR)
$(COVERITY_DIR): init core-clean coverity-clean
$(AM_V_GEN)cov-build --dir "$@" $(MAKE) $(AM_MAKEFLAGS) core
# Public coverity instance
.PHONY: $(COVTAR)
$(COVTAR): $(COVERITY_DIR)
$(AM_V_GEN)tar czf "$@" --transform="s@.*$(TAG)@cov-int@" "$<"
.PHONY: coverity
coverity: $(COVTAR)
@echo "Now go to https://scan.coverity.com/users/sign_in and upload:"
@echo " $(COVTAR)"
@echo "then make core-clean coverity-clean"
# Licensed coverity instance
#
# The prerequisites are a little hacky; rather than actually required, some
# of them are designed so that things execute in the proper order (which is
# not the same as GNU make's order-only prerequisites).
.PHONY: coverity-analyze
coverity-analyze: $(COVERITY_DIR)
@echo ""
@echo "Analyzing (waiting for coverity license if necessary) ..."
cov-analyze --dir "$<" --wait-for-license --security \
--aggressiveness-level "$(COVLEVEL)"
.PHONY: $(COVEMACS)
$(COVEMACS): coverity-analyze
$(AM_V_GEN)cov-format-errors --dir "$(COVERITY_DIR)" --emacs-style > "$@"
.PHONY: $(COVHTML)
$(COVHTML): $(COVEMACS)
$(AM_V_GEN)cov-format-errors --dir "$(COVERITY_DIR)" --html-output "$@"
.PHONY: coverity-corp
coverity-corp: $(COVHTML)
$(MAKE) $(AM_MAKEFLAGS) core-clean
@echo "Done. See:"
@echo " file://$(COVHTML)/index.html"
@echo "When no longer needed, make coverity-clean"
# Remove all outputs regardless of tag
.PHONY: coverity-clean
coverity-clean:
-rm -rf "$(abs_builddir)"/coverity-* \
"$(abs_builddir)"/$(PACKAGE)-coverity-*.tgz \
"$(abs_builddir)"/*.coverity
## Change log generation
summary:
@printf "\n* `date +"%a %b %d %Y"` `git config user.name` <`git config user.email`> $(NEXT_RELEASE)"
@printf "\n- Changesets: `git log --pretty=oneline --no-merges $(LAST_RELEASE)..HEAD | wc -l`"
@printf "\n- Diff:\n"
@git diff $(LAST_RELEASE)..HEAD --shortstat include lib daemons tools xml
rc-changes:
@$(MAKE) $(AM_MAKEFLAGS) NEXT_RELEASE=$(shell echo $(LAST_RC) | sed s:-rc.*::) LAST_RELEASE=$(LAST_RC) changes
changes: summary
@printf "\n- Features added since $(LAST_RELEASE)\n"
@git log --pretty=format:' +%s' --abbrev-commit $(LAST_RELEASE)..HEAD | grep -e Feature: | sed -e 's@Feature:@@' | sort -uf
@printf "\n- Changes since $(LAST_RELEASE)\n"
@git log --pretty=format:' +%s' --no-merges --abbrev-commit $(LAST_RELEASE)..HEAD \
| grep -e High: -e Fix: -e Bug | sed \
-e 's@\(Fix\|High\|Bug\):@@' \
-e 's@\(cib\|pacemaker-based\|based\):@CIB:@' \
-e 's@\(crmd\|pacemaker-controld\|controld\):@controller:@' \
-e 's@\(lrmd\|pacemaker-execd\|execd\):@executor:@' \
-e 's@\(Fencing\|stonithd\|stonith\|pacemaker-fenced\|fenced\):@fencing:@' \
-e 's@\(PE\|pengine\|pacemaker-schedulerd\|schedulerd\):@scheduler:@' \
| sort -uf
authors:
git log $(LAST_RELEASE)..$(COMMIT) --format='%an' | sort -u
changelog:
@$(MAKE) $(AM_MAKEFLAGS) changes > ChangeLog
@printf "\n">> ChangeLog
git show $(LAST_RELEASE):ChangeLog >> ChangeLog
DO_NOT_INDENT = lib/gnu daemons/controld/controld_fsa.h
indent:
find . -name "*.[ch]" -exec ./p-indent \{\} \;
git co HEAD $(DO_NOT_INDENT)
rel-tags: tags
find . -name TAGS -exec sed -i 's:\(.*\)/\(.*\)/TAGS:\2/TAGS:g' \{\} \;
CLANG_analyzer = $(shell which scan-build)
CLANG_checkers =
# Use CPPCHECK_ARGS to pass extra cppcheck options, e.g.:
# --enable={warning,style,performance,portability,information,all}
# --inconclusive --std=posix
CPPCHECK_ARGS ?=
cppcheck:
- cppcheck $(CPPCHECK_ARGS) -I include --max-configs=25 \
+ cppcheck $(CPPCHECK_ARGS) -I include --max-configs=30 \
--library=posix --library=gnu --library=gtk \
--inline-suppr -q replace lib daemons tools
clang:
test -e $(CLANG_analyzer)
scan-build $(CLANG_checkers:%=-enable-checker %) $(MAKE) $(AM_MAKEFLAGS) clean all
# V3 = scandir unsetenv alphasort xalloc
# V2 = setenv strerror strchrnul strndup
# https://www.gnu.org/software/gnulib/manual/html_node/Initial-import.html#Initial-import
# previously, this was crypto/md5, but got spoiled with streams/kernel crypto
GNU_MODS = crypto/md5-buffer
# stdint appears to be surrogate only for C99-lacking environments
GNU_MODS_AVOID = stdint
# only for plain crypto/md5: we make do without kernel-assisted crypto
# GNU_MODS_AVOID += crypto/af_alg
gnulib-update:
-test -e maint/gnulib \
|| git clone https://git.savannah.gnu.org/git/gnulib.git maint/gnulib
cd maint/gnulib && git pull
maint/gnulib/gnulib-tool \
--source-base=lib/gnu --lgpl=2 --no-vc-files --no-conditional-dependencies \
$(GNU_MODS_AVOID:%=--avoid %) --import $(GNU_MODS)
ancillary-clean: spec-clean srpm-clean mock-clean coverity-clean
-rm -f $(TARFILE)
diff --git a/daemons/based/based_callbacks.c b/daemons/based/based_callbacks.c
index 67597fb863..4a3a1fb056 100644
--- a/daemons/based/based_callbacks.c
+++ b/daemons/based/based_callbacks.c
@@ -1,1550 +1,1550 @@
/*
* Copyright 2004-2020 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 <sys/param.h>
#include <stdio.h>
#include <sys/types.h>
#include <unistd.h>
#include <stdlib.h>
#include <errno.h>
#include <fcntl.h>
#include <inttypes.h> /* U64T ~ PRIu64 */
#include <crm/crm.h>
#include <crm/cib.h>
#include <crm/msg_xml.h>
#include <crm/cluster/internal.h>
#include <crm/common/xml.h>
#include <crm/common/remote_internal.h>
#include <pacemaker-based.h>
#define EXIT_ESCALATION_MS 10000
static unsigned long cib_local_bcast_num = 0;
typedef struct cib_local_notify_s {
xmlNode *notify_src;
char *client_id;
gboolean from_peer;
gboolean sync_reply;
} cib_local_notify_t;
int next_client_id = 0;
gboolean legacy_mode = FALSE;
qb_ipcs_service_t *ipcs_ro = NULL;
qb_ipcs_service_t *ipcs_rw = NULL;
qb_ipcs_service_t *ipcs_shm = NULL;
void send_cib_replace(const xmlNode * sync_request, const char *host);
static void cib_process_request(xmlNode* request, gboolean force_synchronous,
gboolean privileged,
pcmk__client_t *cib_client);
static int cib_process_command(xmlNode *request, xmlNode **reply,
xmlNode **cib_diff, gboolean privileged);
gboolean cib_common_callback(qb_ipcs_connection_t * c, void *data, size_t size,
gboolean privileged);
gboolean cib_legacy_mode(void)
{
return legacy_mode;
}
static int32_t
cib_ipc_accept(qb_ipcs_connection_t * c, uid_t uid, gid_t gid)
{
if (cib_shutdown_flag) {
crm_info("Ignoring new IPC client [%d] during shutdown",
pcmk__client_pid(c));
return -EPERM;
}
if (pcmk__new_client(c, uid, gid) == NULL) {
return -EIO;
}
return 0;
}
static int32_t
cib_ipc_dispatch_rw(qb_ipcs_connection_t * c, void *data, size_t size)
{
pcmk__client_t *client = pcmk__find_client(c);
crm_trace("%p message from %s", c, client->id);
return cib_common_callback(c, data, size, TRUE);
}
static int32_t
cib_ipc_dispatch_ro(qb_ipcs_connection_t * c, void *data, size_t size)
{
pcmk__client_t *client = pcmk__find_client(c);
crm_trace("%p message from %s", c, client->id);
return cib_common_callback(c, data, size, FALSE);
}
/* Error code means? */
static int32_t
cib_ipc_closed(qb_ipcs_connection_t * c)
{
pcmk__client_t *client = pcmk__find_client(c);
if (client == NULL) {
return 0;
}
crm_trace("Connection %p", c);
pcmk__free_client(client);
return 0;
}
static void
cib_ipc_destroy(qb_ipcs_connection_t * c)
{
crm_trace("Connection %p", c);
cib_ipc_closed(c);
if (cib_shutdown_flag) {
cib_shutdown(0);
}
}
struct qb_ipcs_service_handlers ipc_ro_callbacks = {
.connection_accept = cib_ipc_accept,
.connection_created = NULL,
.msg_process = cib_ipc_dispatch_ro,
.connection_closed = cib_ipc_closed,
.connection_destroyed = cib_ipc_destroy
};
struct qb_ipcs_service_handlers ipc_rw_callbacks = {
.connection_accept = cib_ipc_accept,
.connection_created = NULL,
.msg_process = cib_ipc_dispatch_rw,
.connection_closed = cib_ipc_closed,
.connection_destroyed = cib_ipc_destroy
};
void
cib_common_callback_worker(uint32_t id, uint32_t flags, xmlNode * op_request,
pcmk__client_t *cib_client, gboolean privileged)
{
const char *op = crm_element_value(op_request, F_CIB_OPERATION);
if (crm_str_eq(op, CRM_OP_REGISTER, TRUE)) {
if (flags & crm_ipc_client_response) {
xmlNode *ack = create_xml_node(NULL, __FUNCTION__);
crm_xml_add(ack, F_CIB_OPERATION, CRM_OP_REGISTER);
crm_xml_add(ack, F_CIB_CLIENTID, cib_client->id);
pcmk__ipc_send_xml(cib_client, id, ack, flags);
cib_client->request_id = 0;
free_xml(ack);
}
return;
} else if (crm_str_eq(op, T_CIB_NOTIFY, TRUE)) {
/* Update the notify filters for this client */
int on_off = 0;
long long bit = 0;
const char *type = crm_element_value(op_request, F_CIB_NOTIFY_TYPE);
crm_element_value_int(op_request, F_CIB_NOTIFY_ACTIVATE, &on_off);
crm_debug("Setting %s callbacks for %s (%s): %s",
type, cib_client->name, cib_client->id, on_off ? "on" : "off");
if (safe_str_eq(type, T_CIB_POST_NOTIFY)) {
bit = cib_notify_post;
} else if (safe_str_eq(type, T_CIB_PRE_NOTIFY)) {
bit = cib_notify_pre;
} else if (safe_str_eq(type, T_CIB_UPDATE_CONFIRM)) {
bit = cib_notify_confirm;
} else if (safe_str_eq(type, T_CIB_DIFF_NOTIFY)) {
bit = cib_notify_diff;
} else if (safe_str_eq(type, T_CIB_REPLACE_NOTIFY)) {
bit = cib_notify_replace;
}
if (on_off) {
set_bit(cib_client->options, bit);
} else {
clear_bit(cib_client->options, bit);
}
if (flags & crm_ipc_client_response) {
/* TODO - include rc */
pcmk__ipc_send_ack(cib_client, id, flags, "ack");
}
return;
}
cib_process_request(op_request, FALSE, privileged, cib_client);
}
int32_t
cib_common_callback(qb_ipcs_connection_t * c, void *data, size_t size, gboolean privileged)
{
uint32_t id = 0;
uint32_t flags = 0;
int call_options = 0;
pcmk__client_t *cib_client = pcmk__find_client(c);
xmlNode *op_request = pcmk__client_data2xml(cib_client, data, &id, &flags);
if (op_request) {
crm_element_value_int(op_request, F_CIB_CALLOPTS, &call_options);
}
if (op_request == NULL) {
crm_trace("Invalid message from %p", c);
pcmk__ipc_send_ack(cib_client, id, flags, "nack");
return 0;
} else if(cib_client == NULL) {
crm_trace("Invalid client %p", c);
return 0;
}
if (is_set(call_options, cib_sync_call)) {
CRM_LOG_ASSERT(flags & crm_ipc_client_response);
CRM_LOG_ASSERT(cib_client->request_id == 0); /* This means the client has two synchronous events in-flight */
cib_client->request_id = id; /* Reply only to the last one */
}
if (cib_client->name == NULL) {
const char *value = crm_element_value(op_request, F_CIB_CLIENTNAME);
if (value == NULL) {
cib_client->name = crm_itoa(cib_client->pid);
} else {
cib_client->name = strdup(value);
if (crm_is_daemon_name(value)) {
set_bit(cib_client->options, cib_is_daemon);
}
}
}
/* Allow cluster daemons more leeway before being evicted */
if (is_set(cib_client->options, cib_is_daemon)) {
const char *qmax = cib_config_lookup("cluster-ipc-limit");
if (pcmk__set_client_queue_max(cib_client, qmax)) {
crm_trace("IPC threshold for %s[%u] is now %u",
cib_client->name, cib_client->pid, cib_client->queue_max);
}
}
crm_xml_add(op_request, F_CIB_CLIENTID, cib_client->id);
crm_xml_add(op_request, F_CIB_CLIENTNAME, cib_client->name);
#if ENABLE_ACL
CRM_LOG_ASSERT(cib_client->user != NULL);
pcmk__update_acl_user(op_request, F_CIB_USER, cib_client->user);
#endif
cib_common_callback_worker(id, flags, op_request, cib_client, privileged);
free_xml(op_request);
return 0;
}
static uint64_t ping_seq = 0;
static char *ping_digest = NULL;
static bool ping_modified_since = FALSE;
int sync_our_cib(xmlNode * request, gboolean all);
static gboolean
cib_digester_cb(gpointer data)
{
if (cib_is_master) {
char buffer[32];
xmlNode *ping = create_xml_node(NULL, "ping");
ping_seq++;
free(ping_digest);
ping_digest = NULL;
ping_modified_since = FALSE;
snprintf(buffer, 32, "%" U64T, ping_seq);
crm_trace("Requesting peer digests (%s)", buffer);
crm_xml_add(ping, F_TYPE, "cib");
crm_xml_add(ping, F_CIB_OPERATION, CRM_OP_PING);
crm_xml_add(ping, F_CIB_PING_ID, buffer);
crm_xml_add(ping, XML_ATTR_CRM_VERSION, CRM_FEATURE_SET);
send_cluster_message(NULL, crm_msg_cib, ping, TRUE);
free_xml(ping);
}
return FALSE;
}
static void
process_ping_reply(xmlNode *reply)
{
uint64_t seq = 0;
const char *host = crm_element_value(reply, F_ORIG);
xmlNode *pong = get_message_xml(reply, F_CIB_CALLDATA);
const char *seq_s = crm_element_value(pong, F_CIB_PING_ID);
const char *digest = crm_element_value(pong, XML_ATTR_DIGEST);
if (seq_s) {
seq = (uint64_t) crm_parse_ll(seq_s, NULL);
}
if(digest == NULL) {
crm_trace("Ignoring ping reply %s from %s with no digest", seq_s, host);
} else if(seq != ping_seq) {
crm_trace("Ignoring out of sequence ping reply %s from %s", seq_s, host);
} else if(ping_modified_since) {
crm_trace("Ignoring ping reply %s from %s: cib updated since", seq_s, host);
} else {
const char *version = crm_element_value(pong, XML_ATTR_CRM_VERSION);
if(ping_digest == NULL) {
crm_trace("Calculating new digest");
ping_digest = calculate_xml_versioned_digest(the_cib, FALSE, TRUE, version);
}
crm_trace("Processing ping reply %s from %s (%s)", seq_s, host, digest);
if(safe_str_eq(ping_digest, digest) == FALSE) {
xmlNode *remote_cib = get_message_xml(pong, F_CIB_CALLDATA);
crm_notice("Local CIB %s.%s.%s.%s differs from %s: %s.%s.%s.%s %p",
crm_element_value(the_cib, XML_ATTR_GENERATION_ADMIN),
crm_element_value(the_cib, XML_ATTR_GENERATION),
crm_element_value(the_cib, XML_ATTR_NUMUPDATES),
ping_digest, host,
remote_cib?crm_element_value(remote_cib, XML_ATTR_GENERATION_ADMIN):"_",
remote_cib?crm_element_value(remote_cib, XML_ATTR_GENERATION):"_",
remote_cib?crm_element_value(remote_cib, XML_ATTR_NUMUPDATES):"_",
digest, remote_cib);
if(remote_cib && remote_cib->children) {
/* Additional debug */
xml_calculate_changes(the_cib, remote_cib);
xml_log_changes(LOG_INFO, __FUNCTION__, remote_cib);
crm_trace("End of differences");
}
free_xml(remote_cib);
sync_our_cib(reply, FALSE);
}
}
}
static void
do_local_notify(xmlNode * notify_src, const char *client_id,
gboolean sync_reply, gboolean from_peer)
{
int rid = 0;
int call_id = 0;
pcmk__client_t *client_obj = NULL;
CRM_ASSERT(notify_src && client_id);
crm_element_value_int(notify_src, F_CIB_CALLID, &call_id);
client_obj = pcmk__find_client_by_id(client_id);
if (client_obj == NULL) {
crm_debug("Could not send response %d: client %s not found",
call_id, client_id);
return;
}
if (sync_reply) {
if (client_obj->ipcs) {
CRM_LOG_ASSERT(client_obj->request_id);
rid = client_obj->request_id;
client_obj->request_id = 0;
crm_trace("Sending response %d to %s %s",
rid, client_obj->name,
from_peer ? "(originator of delegated request)" : "");
} else {
crm_trace("Sending response [call %d] to %s %s",
call_id, client_obj->name, from_peer ? "(originator of delegated request)" : "");
}
} else {
crm_trace("Sending event %d to %s %s",
call_id, client_obj->name, from_peer ? "(originator of delegated request)" : "");
}
switch (client_obj->kind) {
case PCMK__CLIENT_IPC:
{
int rc = pcmk__ipc_send_xml(client_obj, rid, notify_src,
(sync_reply? crm_ipc_flags_none
: crm_ipc_server_event));
if (rc != pcmk_rc_ok) {
crm_warn("%s reply to %s failed: %s " CRM_XS " rc=%d",
(sync_reply? "Synchronous" : "Asynchronous"),
client_obj->name, pcmk_rc_str(rc), rc);
}
}
break;
#ifdef HAVE_GNUTLS_GNUTLS_H
case PCMK__CLIENT_TLS:
#endif
case PCMK__CLIENT_TCP:
pcmk__remote_send_xml(client_obj->remote, notify_src);
break;
default:
crm_err("Unknown transport %d for %s", client_obj->kind, client_obj->name);
}
}
static void
local_notify_destroy_callback(gpointer data)
{
cib_local_notify_t *notify = data;
free_xml(notify->notify_src);
free(notify->client_id);
free(notify);
}
static void
check_local_notify(int bcast_id)
{
cib_local_notify_t *notify = NULL;
if (!local_notify_queue) {
return;
}
notify = g_hash_table_lookup(local_notify_queue, GINT_TO_POINTER(bcast_id));
if (notify) {
do_local_notify(notify->notify_src, notify->client_id, notify->sync_reply,
notify->from_peer);
g_hash_table_remove(local_notify_queue, GINT_TO_POINTER(bcast_id));
}
}
static void
queue_local_notify(xmlNode * notify_src, const char *client_id, gboolean sync_reply,
gboolean from_peer)
{
cib_local_notify_t *notify = calloc(1, sizeof(cib_local_notify_t));
notify->notify_src = notify_src;
notify->client_id = strdup(client_id);
notify->sync_reply = sync_reply;
notify->from_peer = from_peer;
if (!local_notify_queue) {
local_notify_queue = g_hash_table_new_full(g_direct_hash,
g_direct_equal, NULL,
local_notify_destroy_callback);
}
- // notify will get freed when the hash table is destroyed
- // cppcheck-suppress memleak
g_hash_table_insert(local_notify_queue, GINT_TO_POINTER(cib_local_bcast_num), notify);
+ // cppcheck doesn't know notify will get freed when hash table is destroyed
+ // cppcheck-suppress memleak
}
static void
parse_local_options_v1(pcmk__client_t *cib_client, int call_type,
int call_options, const char *host, const char *op,
gboolean *local_notify, gboolean *needs_reply,
gboolean *process, gboolean *needs_forward)
{
if (cib_op_modifies(call_type)
&& !(call_options & cib_inhibit_bcast)) {
/* we need to send an update anyway */
*needs_reply = TRUE;
} else {
*needs_reply = FALSE;
}
if (host == NULL && (call_options & cib_scope_local)) {
crm_trace("Processing locally scoped %s op from %s", op, cib_client->name);
*local_notify = TRUE;
} else if (host == NULL && cib_is_master) {
crm_trace("Processing master %s op locally from %s", op, cib_client->name);
*local_notify = TRUE;
} else if (safe_str_eq(host, cib_our_uname)) {
crm_trace("Processing locally addressed %s op from %s", op, cib_client->name);
*local_notify = TRUE;
} else if (stand_alone) {
*needs_forward = FALSE;
*local_notify = TRUE;
*process = TRUE;
} else {
crm_trace("%s op from %s needs to be forwarded to %s",
op, cib_client->name, host ? host : "the master instance");
*needs_forward = TRUE;
*process = FALSE;
}
}
static void
parse_local_options_v2(pcmk__client_t *cib_client, int call_type,
int call_options, const char *host, const char *op,
gboolean *local_notify, gboolean *needs_reply,
gboolean *process, gboolean *needs_forward)
{
if (cib_op_modifies(call_type)) {
if(safe_str_eq(op, CIB_OP_MASTER) || safe_str_eq(op, CIB_OP_SLAVE)) {
/* Always handle these locally */
*process = TRUE;
*needs_reply = FALSE;
*local_notify = TRUE;
*needs_forward = FALSE;
return;
} else {
/* Redirect all other updates via CPG */
*needs_reply = TRUE;
*needs_forward = TRUE;
*process = FALSE;
crm_trace("%s op from %s needs to be forwarded to %s",
op, cib_client->name, host ? host : "the master instance");
return;
}
}
*process = TRUE;
*needs_reply = FALSE;
*local_notify = TRUE;
*needs_forward = FALSE;
if (stand_alone) {
crm_trace("Processing %s op from %s (stand-alone)", op, cib_client->name);
} else if (host == NULL) {
crm_trace("Processing unaddressed %s op from %s", op, cib_client->name);
} else if (safe_str_eq(host, cib_our_uname)) {
crm_trace("Processing locally addressed %s op from %s", op, cib_client->name);
} else {
crm_trace("%s op from %s needs to be forwarded to %s", op, cib_client->name, host);
*needs_forward = TRUE;
*process = FALSE;
}
}
static void
parse_local_options(pcmk__client_t *cib_client, int call_type,
int call_options, const char *host, const char *op,
gboolean *local_notify, gboolean *needs_reply,
gboolean *process, gboolean *needs_forward)
{
if(cib_legacy_mode()) {
parse_local_options_v1(cib_client, call_type, call_options, host,
op, local_notify, needs_reply, process, needs_forward);
} else {
parse_local_options_v2(cib_client, call_type, call_options, host,
op, local_notify, needs_reply, process, needs_forward);
}
}
static gboolean
parse_peer_options_v1(int call_type, xmlNode * request,
gboolean * local_notify, gboolean * needs_reply, gboolean * process,
gboolean * needs_forward)
{
const char *op = NULL;
const char *host = NULL;
const char *delegated = NULL;
const char *originator = crm_element_value(request, F_ORIG);
const char *reply_to = crm_element_value(request, F_CIB_ISREPLY);
const char *update = crm_element_value(request, F_CIB_GLOBAL_UPDATE);
gboolean is_reply = safe_str_eq(reply_to, cib_our_uname);
if (crm_is_true(update)) {
*needs_reply = FALSE;
if (is_reply) {
*local_notify = TRUE;
crm_trace("Processing global/peer update from %s"
" that originated from us", originator);
} else {
crm_trace("Processing global/peer update from %s", originator);
}
return TRUE;
}
op = crm_element_value(request, F_CIB_OPERATION);
crm_trace("Processing %s request sent by %s", op, originator);
if (safe_str_eq(op, "cib_shutdown_req")) {
/* Always process these */
*local_notify = FALSE;
if (reply_to == NULL || is_reply) {
*process = TRUE;
}
if (is_reply) {
*needs_reply = FALSE;
}
return *process;
}
if (is_reply && safe_str_eq(op, CRM_OP_PING)) {
process_ping_reply(request);
return FALSE;
}
if (is_reply) {
crm_trace("Forward reply sent from %s to local clients", originator);
*process = FALSE;
*needs_reply = FALSE;
*local_notify = TRUE;
return TRUE;
}
host = crm_element_value(request, F_CIB_HOST);
if (host != NULL && safe_str_eq(host, cib_our_uname)) {
crm_trace("Processing %s request sent to us from %s", op, originator);
return TRUE;
} else if(is_reply == FALSE && safe_str_eq(op, CRM_OP_PING)) {
crm_trace("Processing %s request sent to %s by %s", op, host?host:"everyone", originator);
*needs_reply = TRUE;
return TRUE;
} else if (host == NULL && cib_is_master == TRUE) {
crm_trace("Processing %s request sent to master instance from %s", op, originator);
return TRUE;
}
delegated = crm_element_value(request, F_CIB_DELEGATED);
if (delegated != NULL) {
crm_trace("Ignoring msg for master instance");
} else if (host != NULL) {
/* this is for a specific instance and we're not it */
crm_trace("Ignoring msg for instance on %s", crm_str(host));
} else if (reply_to == NULL && cib_is_master == FALSE) {
/* this is for the master instance and we're not it */
crm_trace("Ignoring reply to %s", crm_str(reply_to));
} else if (safe_str_eq(op, "cib_shutdown_req")) {
if (reply_to != NULL) {
crm_debug("Processing %s from %s", op, originator);
*needs_reply = FALSE;
} else {
crm_debug("Processing %s reply from %s", op, originator);
}
return TRUE;
} else {
crm_err("Nothing for us to do?");
crm_log_xml_err(request, "Peer[inbound]");
}
return FALSE;
}
static gboolean
parse_peer_options_v2(int call_type, xmlNode * request,
gboolean * local_notify, gboolean * needs_reply, gboolean * process,
gboolean * needs_forward)
{
const char *host = NULL;
const char *delegated = crm_element_value(request, F_CIB_DELEGATED);
const char *op = crm_element_value(request, F_CIB_OPERATION);
const char *originator = crm_element_value(request, F_ORIG);
const char *reply_to = crm_element_value(request, F_CIB_ISREPLY);
const char *update = crm_element_value(request, F_CIB_GLOBAL_UPDATE);
gboolean is_reply = safe_str_eq(reply_to, cib_our_uname);
if(safe_str_eq(op, CIB_OP_REPLACE)) {
/* sync_our_cib() sets F_CIB_ISREPLY */
if (reply_to) {
delegated = reply_to;
}
goto skip_is_reply;
} else if(safe_str_eq(op, CIB_OP_SYNC)) {
} else if (is_reply && safe_str_eq(op, CRM_OP_PING)) {
process_ping_reply(request);
return FALSE;
} else if (safe_str_eq(op, CIB_OP_UPGRADE)) {
/* Only the DC (node with the oldest software) should process
* this operation if F_CIB_SCHEMA_MAX is unset
*
* If the DC is happy it will then send out another
* CIB_OP_UPGRADE which will tell all nodes to do the actual
* upgrade.
*
* Except this time F_CIB_SCHEMA_MAX will be set which puts a
* limit on how far newer nodes will go
*/
const char *max = crm_element_value(request, F_CIB_SCHEMA_MAX);
const char *upgrade_rc = crm_element_value(request, F_CIB_UPGRADE_RC);
crm_trace("Parsing %s operation%s for %s with max=%s and upgrade_rc=%s",
op, (is_reply? " reply" : ""),
(cib_is_master? "master" : "slave"),
(max? max : "none"), (upgrade_rc? upgrade_rc : "none"));
if (upgrade_rc != NULL) {
// Our upgrade request was rejected by DC, notify clients of result
crm_xml_add(request, F_CIB_RC, upgrade_rc);
} else if ((max == NULL) && cib_is_master) {
/* We are the DC, check if this upgrade is allowed */
goto skip_is_reply;
} else if(max) {
/* Ok, go ahead and upgrade to 'max' */
goto skip_is_reply;
} else {
// Ignore broadcast client requests when we're not DC
return FALSE;
}
} else if (crm_is_true(update)) {
crm_info("Detected legacy %s global update from %s", op, originator);
send_sync_request(NULL);
legacy_mode = TRUE;
return FALSE;
} else if (is_reply && cib_op_modifies(call_type)) {
crm_trace("Ignoring legacy %s reply sent from %s to local clients", op, originator);
return FALSE;
} else if (safe_str_eq(op, "cib_shutdown_req")) {
/* Legacy handling */
crm_debug("Legacy handling of %s message from %s", op, originator);
*local_notify = FALSE;
if (reply_to == NULL) {
*process = TRUE;
}
return *process;
}
if(is_reply) {
crm_trace("Handling %s reply sent from %s to local clients", op, originator);
*process = FALSE;
*needs_reply = FALSE;
*local_notify = TRUE;
return TRUE;
}
skip_is_reply:
*process = TRUE;
*needs_reply = FALSE;
if(safe_str_eq(delegated, cib_our_uname)) {
*local_notify = TRUE;
} else {
*local_notify = FALSE;
}
host = crm_element_value(request, F_CIB_HOST);
if (host != NULL && safe_str_eq(host, cib_our_uname)) {
crm_trace("Processing %s request sent to us from %s", op, originator);
*needs_reply = TRUE;
return TRUE;
} else if (host != NULL) {
/* this is for a specific instance and we're not it */
crm_trace("Ignoring %s operation for instance on %s", op, crm_str(host));
return FALSE;
} else if(is_reply == FALSE && safe_str_eq(op, CRM_OP_PING)) {
*needs_reply = TRUE;
}
crm_trace("Processing %s request sent to everyone by %s/%s on %s %s", op,
crm_element_value(request, F_CIB_CLIENTNAME),
crm_element_value(request, F_CIB_CALLID),
originator, (*local_notify)?"(notify)":"");
return TRUE;
}
static gboolean
parse_peer_options(int call_type, xmlNode * request,
gboolean * local_notify, gboolean * needs_reply, gboolean * process,
gboolean * needs_forward)
{
/* TODO: What happens when an update comes in after node A
* requests the CIB from node B, but before it gets the reply (and
* sends out the replace operation)
*/
if(cib_legacy_mode()) {
return parse_peer_options_v1(
call_type, request, local_notify, needs_reply, process, needs_forward);
} else {
return parse_peer_options_v2(
call_type, request, local_notify, needs_reply, process, needs_forward);
}
}
static void
forward_request(xmlNode * request, pcmk__client_t *cib_client, int call_options)
{
const char *op = crm_element_value(request, F_CIB_OPERATION);
const char *host = crm_element_value(request, F_CIB_HOST);
crm_xml_add(request, F_CIB_DELEGATED, cib_our_uname);
if (host != NULL) {
crm_trace("Forwarding %s op to %s", op, host);
send_cluster_message(crm_get_peer(0, host), crm_msg_cib, request, FALSE);
} else {
crm_trace("Forwarding %s op to master instance", op);
send_cluster_message(NULL, crm_msg_cib, request, FALSE);
}
/* Return the request to its original state */
xml_remove_prop(request, F_CIB_DELEGATED);
if (call_options & cib_discard_reply) {
crm_trace("Client not interested in reply");
}
}
static gboolean
send_peer_reply(xmlNode * msg, xmlNode * result_diff, const char *originator, gboolean broadcast)
{
CRM_ASSERT(msg != NULL);
if (broadcast) {
/* this (successful) call modified the CIB _and_ the
* change needs to be broadcast...
* send via HA to other nodes
*/
int diff_add_updates = 0;
int diff_add_epoch = 0;
int diff_add_admin_epoch = 0;
int diff_del_updates = 0;
int diff_del_epoch = 0;
int diff_del_admin_epoch = 0;
const char *digest = NULL;
int format = 1;
CRM_LOG_ASSERT(result_diff != NULL);
digest = crm_element_value(result_diff, XML_ATTR_DIGEST);
crm_element_value_int(result_diff, "format", &format);
cib_diff_version_details(result_diff,
&diff_add_admin_epoch, &diff_add_epoch, &diff_add_updates,
&diff_del_admin_epoch, &diff_del_epoch, &diff_del_updates);
crm_trace("Sending update diff %d.%d.%d -> %d.%d.%d %s",
diff_del_admin_epoch, diff_del_epoch, diff_del_updates,
diff_add_admin_epoch, diff_add_epoch, diff_add_updates, digest);
crm_xml_add(msg, F_CIB_ISREPLY, originator);
crm_xml_add(msg, F_CIB_GLOBAL_UPDATE, XML_BOOLEAN_TRUE);
crm_xml_add(msg, F_CIB_OPERATION, CIB_OP_APPLY_DIFF);
crm_xml_add(msg, F_CIB_USER, CRM_DAEMON_USER);
if (format == 1) {
CRM_ASSERT(digest != NULL);
}
add_message_xml(msg, F_CIB_UPDATE_DIFF, result_diff);
crm_log_xml_explicit(msg, "copy");
return send_cluster_message(NULL, crm_msg_cib, msg, TRUE);
} else if (originator != NULL) {
/* send reply via HA to originating node */
crm_trace("Sending request result to %s only", originator);
crm_xml_add(msg, F_CIB_ISREPLY, originator);
return send_cluster_message(crm_get_peer(0, originator), crm_msg_cib, msg, FALSE);
}
return FALSE;
}
static void
cib_process_request(xmlNode *request, gboolean force_synchronous,
gboolean privileged, pcmk__client_t *cib_client)
{
int call_type = 0;
int call_options = 0;
gboolean process = TRUE;
gboolean is_update = TRUE;
gboolean from_peer = TRUE;
gboolean needs_reply = TRUE;
gboolean local_notify = FALSE;
gboolean needs_forward = FALSE;
gboolean global_update = crm_is_true(crm_element_value(request, F_CIB_GLOBAL_UPDATE));
xmlNode *op_reply = NULL;
xmlNode *result_diff = NULL;
int rc = pcmk_ok;
const char *op = crm_element_value(request, F_CIB_OPERATION);
const char *originator = crm_element_value(request, F_ORIG);
const char *host = crm_element_value(request, F_CIB_HOST);
const char *target = NULL;
const char *call_id = crm_element_value(request, F_CIB_CALLID);
const char *client_id = crm_element_value(request, F_CIB_CLIENTID);
const char *client_name = crm_element_value(request, F_CIB_CLIENTNAME);
const char *reply_to = crm_element_value(request, F_CIB_ISREPLY);
if (cib_client) {
from_peer = FALSE;
}
crm_element_value_int(request, F_CIB_CALLOPTS, &call_options);
if (force_synchronous) {
call_options |= cib_sync_call;
}
if (host != NULL && strlen(host) == 0) {
host = NULL;
}
if (host) {
target = host;
} else if (call_options & cib_scope_local) {
target = "local host";
} else {
target = "master";
}
if (from_peer) {
crm_trace("Processing peer %s operation from %s/%s on %s intended for %s (reply=%s)",
op, client_name, call_id, originator, target, reply_to);
} else {
crm_xml_add(request, F_ORIG, cib_our_uname);
crm_trace("Processing local %s operation from %s/%s intended for %s", op, client_name, call_id, target);
}
rc = cib_get_operation_id(op, &call_type);
if (rc != pcmk_ok) {
/* TODO: construct error reply? */
crm_err("Pre-processing of command failed: %s", pcmk_strerror(rc));
return;
}
if (from_peer == FALSE) {
parse_local_options(cib_client, call_type, call_options, host, op,
&local_notify, &needs_reply, &process, &needs_forward);
} else if (parse_peer_options(call_type, request, &local_notify,
&needs_reply, &process, &needs_forward) == FALSE) {
return;
}
is_update = cib_op_modifies(call_type);
if (call_options & cib_discard_reply) {
needs_reply = is_update;
local_notify = FALSE;
}
if (needs_forward) {
const char *host = crm_element_value(request, F_CIB_HOST);
const char *section = crm_element_value(request, F_CIB_SECTION);
int log_level = LOG_INFO;
if (safe_str_eq(op, CRM_OP_NOOP)) {
log_level = LOG_DEBUG;
}
do_crm_log(log_level,
"Forwarding %s operation for section %s to %s (origin=%s/%s/%s)",
op,
section ? section : "'all'",
host ? host : cib_legacy_mode() ? "master" : "all",
originator ? originator : "local",
client_name, call_id);
forward_request(request, cib_client, call_options);
return;
}
if (cib_status != pcmk_ok) {
const char *call = crm_element_value(request, F_CIB_CALLID);
rc = cib_status;
crm_err("Operation ignored, cluster configuration is invalid."
" Please repair and restart: %s", pcmk_strerror(cib_status));
op_reply = create_xml_node(NULL, "cib-reply");
crm_xml_add(op_reply, F_TYPE, T_CIB);
crm_xml_add(op_reply, F_CIB_OPERATION, op);
crm_xml_add(op_reply, F_CIB_CALLID, call);
crm_xml_add(op_reply, F_CIB_CLIENTID, client_id);
crm_xml_add_int(op_reply, F_CIB_CALLOPTS, call_options);
crm_xml_add_int(op_reply, F_CIB_RC, rc);
crm_trace("Attaching reply output");
add_message_xml(op_reply, F_CIB_CALLDATA, the_cib);
crm_log_xml_explicit(op_reply, "cib:reply");
} else if (process) {
time_t finished = 0;
time_t now = time(NULL);
int level = LOG_INFO;
const char *section = crm_element_value(request, F_CIB_SECTION);
rc = cib_process_command(request, &op_reply, &result_diff, privileged);
if (is_update == FALSE) {
level = LOG_TRACE;
} else if (global_update) {
switch (rc) {
case pcmk_ok:
level = LOG_INFO;
break;
case -pcmk_err_old_data:
case -pcmk_err_diff_resync:
case -pcmk_err_diff_failed:
level = LOG_TRACE;
break;
default:
level = LOG_ERR;
}
} else if (rc != pcmk_ok && is_update) {
level = LOG_WARNING;
}
do_crm_log(level,
"Completed %s operation for section %s: %s (rc=%d, origin=%s/%s/%s, version=%s.%s.%s)",
op, section ? section : "'all'", pcmk_strerror(rc), rc,
originator ? originator : "local", client_name, call_id,
the_cib ? crm_element_value(the_cib, XML_ATTR_GENERATION_ADMIN) : "0",
the_cib ? crm_element_value(the_cib, XML_ATTR_GENERATION) : "0",
the_cib ? crm_element_value(the_cib, XML_ATTR_NUMUPDATES) : "0");
finished = time(NULL);
if ((finished - now) > 3) {
crm_trace("%s operation took %lds to complete", op, (long)(finished - now));
crm_write_blackbox(0, NULL);
}
if (op_reply == NULL && (needs_reply || local_notify)) {
crm_err("Unexpected NULL reply to message");
crm_log_xml_err(request, "null reply");
needs_reply = FALSE;
local_notify = FALSE;
}
}
/* from now on we are the server */
if(is_update && cib_legacy_mode() == FALSE) {
crm_trace("Completed pre-sync update from %s/%s/%s%s",
originator ? originator : "local", client_name, call_id,
local_notify?" with local notification":"");
} else if (needs_reply == FALSE || stand_alone) {
/* nothing more to do...
* this was a non-originating slave update
*/
crm_trace("Completed slave update");
} else if (cib_legacy_mode() &&
rc == pcmk_ok && result_diff != NULL && !(call_options & cib_inhibit_bcast)) {
gboolean broadcast = FALSE;
cib_local_bcast_num++;
crm_xml_add_int(request, F_CIB_LOCAL_NOTIFY_ID, cib_local_bcast_num);
broadcast = send_peer_reply(request, result_diff, originator, TRUE);
if (broadcast && client_id && local_notify && op_reply) {
/* If we have been asked to sync the reply,
* and a bcast msg has gone out, we queue the local notify
* until we know the bcast message has been received */
local_notify = FALSE;
crm_trace("Queuing local %ssync notification for %s",
(call_options & cib_sync_call) ? "" : "a-", client_id);
queue_local_notify(op_reply, client_id, (call_options & cib_sync_call), from_peer);
op_reply = NULL; /* the reply is queued, so don't free here */
}
} else if (call_options & cib_discard_reply) {
crm_trace("Caller isn't interested in reply");
} else if (from_peer) {
if (is_update == FALSE || result_diff == NULL) {
crm_trace("Request not broadcast: R/O call");
} else if (call_options & cib_inhibit_bcast) {
crm_trace("Request not broadcast: inhibited");
} else if (rc != pcmk_ok) {
crm_trace("Request not broadcast: call failed: %s", pcmk_strerror(rc));
} else {
crm_trace("Directing reply to %s", originator);
}
send_peer_reply(op_reply, result_diff, originator, FALSE);
}
if (local_notify && client_id) {
crm_trace("Performing local %ssync notification for %s",
(call_options & cib_sync_call) ? "" : "a-", client_id);
if (process == FALSE) {
do_local_notify(request, client_id, call_options & cib_sync_call, from_peer);
} else {
do_local_notify(op_reply, client_id, call_options & cib_sync_call, from_peer);
}
}
free_xml(op_reply);
free_xml(result_diff);
return;
}
static int
cib_process_command(xmlNode * request, xmlNode ** reply, xmlNode ** cib_diff, gboolean privileged)
{
xmlNode *input = NULL;
xmlNode *output = NULL;
xmlNode *result_cib = NULL;
xmlNode *current_cib = NULL;
int call_type = 0;
int call_options = 0;
const char *op = NULL;
const char *section = NULL;
const char *call_id = crm_element_value(request, F_CIB_CALLID);
int rc = pcmk_ok;
int rc2 = pcmk_ok;
gboolean send_r_notify = FALSE;
gboolean global_update = FALSE;
gboolean config_changed = FALSE;
gboolean manage_counters = TRUE;
static mainloop_timer_t *digest_timer = NULL;
CRM_ASSERT(cib_status == pcmk_ok);
if(digest_timer == NULL) {
digest_timer = mainloop_timer_add("digester", 5000, FALSE, cib_digester_cb, NULL);
}
*reply = NULL;
*cib_diff = NULL;
current_cib = the_cib;
/* Start processing the request... */
op = crm_element_value(request, F_CIB_OPERATION);
crm_element_value_int(request, F_CIB_CALLOPTS, &call_options);
rc = cib_get_operation_id(op, &call_type);
if (rc == pcmk_ok && privileged == FALSE) {
rc = cib_op_can_run(call_type, call_options, privileged, global_update);
}
rc2 = cib_op_prepare(call_type, request, &input, &section);
if (rc == pcmk_ok) {
rc = rc2;
}
if (rc != pcmk_ok) {
crm_trace("Call setup failed: %s", pcmk_strerror(rc));
goto done;
} else if (cib_op_modifies(call_type) == FALSE) {
rc = cib_perform_op(op, call_options, cib_op_func(call_type), TRUE,
section, request, input, FALSE, &config_changed,
current_cib, &result_cib, NULL, &output);
CRM_CHECK(result_cib == NULL, free_xml(result_cib));
goto done;
}
/* Handle a valid write action */
global_update = crm_is_true(crm_element_value(request, F_CIB_GLOBAL_UPDATE));
if (global_update) {
/* legacy code */
manage_counters = FALSE;
call_options |= cib_force_diff;
crm_trace("Global update detected");
CRM_CHECK(call_type == 3 || call_type == 4, crm_err("Call type: %d", call_type);
crm_log_xml_err(request, "bad op"));
}
if (rc == pcmk_ok) {
ping_modified_since = TRUE;
if (call_options & cib_inhibit_bcast) {
/* skip */
crm_trace("Skipping update: inhibit broadcast");
manage_counters = FALSE;
}
if (is_not_set(call_options, cib_dryrun) && safe_str_eq(section, XML_CIB_TAG_STATUS)) {
/* Copying large CIBs accounts for a huge percentage of our CIB usage */
call_options |= cib_zero_copy;
} else {
clear_bit(call_options, cib_zero_copy);
}
/* result_cib must not be modified after cib_perform_op() returns */
rc = cib_perform_op(op, call_options, cib_op_func(call_type), FALSE,
section, request, input, manage_counters, &config_changed,
current_cib, &result_cib, cib_diff, &output);
if (manage_counters == FALSE) {
int format = 1;
/* Legacy code
* If the diff is NULL at this point, it's because nothing changed
*/
if (*cib_diff) {
crm_element_value_int(*cib_diff, "format", &format);
}
if (format == 1) {
config_changed = cib_config_changed(NULL, NULL, cib_diff);
}
}
/* Always write to disk for replace ops,
* this also negates the need to detect ordering changes
*/
if (crm_str_eq(CIB_OP_REPLACE, op, TRUE)) {
config_changed = TRUE;
}
}
if (rc == pcmk_ok && is_not_set(call_options, cib_dryrun)) {
crm_trace("Activating %s->%s%s%s",
crm_element_value(current_cib, XML_ATTR_NUMUPDATES),
crm_element_value(result_cib, XML_ATTR_NUMUPDATES),
(is_set(call_options, cib_zero_copy)? " zero-copy" : ""),
(config_changed? " changed" : ""));
if(is_not_set(call_options, cib_zero_copy)) {
rc = activateCibXml(result_cib, config_changed, op);
crm_trace("Activated %s (%d)",
crm_element_value(current_cib, XML_ATTR_NUMUPDATES), rc);
}
if (rc == pcmk_ok && cib_internal_config_changed(*cib_diff)) {
cib_read_config(config_hash, result_cib);
}
if (crm_str_eq(CIB_OP_REPLACE, op, TRUE)) {
if (section == NULL) {
send_r_notify = TRUE;
} else if (safe_str_eq(section, XML_TAG_CIB)) {
send_r_notify = TRUE;
} else if (safe_str_eq(section, XML_CIB_TAG_NODES)) {
send_r_notify = TRUE;
} else if (safe_str_eq(section, XML_CIB_TAG_STATUS)) {
send_r_notify = TRUE;
}
} else if (crm_str_eq(CIB_OP_ERASE, op, TRUE)) {
send_r_notify = TRUE;
}
mainloop_timer_stop(digest_timer);
mainloop_timer_start(digest_timer);
} else if (rc == -pcmk_err_schema_validation) {
CRM_ASSERT(is_not_set(call_options, cib_zero_copy));
if (output != NULL) {
crm_log_xml_info(output, "cib:output");
free_xml(output);
}
output = result_cib;
} else {
crm_trace("Not activating %d %d %s", rc, is_set(call_options, cib_dryrun), crm_element_value(result_cib, XML_ATTR_NUMUPDATES));
if(is_not_set(call_options, cib_zero_copy)) {
free_xml(result_cib);
}
}
if ((call_options & (cib_inhibit_notify|cib_dryrun)) == 0) {
const char *client = crm_element_value(request, F_CIB_CLIENTNAME);
crm_trace("Sending notifications %d", is_set(call_options, cib_dryrun));
cib_diff_notify(call_options, client, call_id, op, input, rc, *cib_diff);
}
if (send_r_notify) {
const char *origin = crm_element_value(request, F_ORIG);
cib_replace_notify(origin, the_cib, rc, *cib_diff);
}
xml_log_patchset(LOG_TRACE, "cib:diff", *cib_diff);
done:
if ((call_options & cib_discard_reply) == 0) {
const char *caller = crm_element_value(request, F_CIB_CLIENTID);
*reply = create_xml_node(NULL, "cib-reply");
crm_xml_add(*reply, F_TYPE, T_CIB);
crm_xml_add(*reply, F_CIB_OPERATION, op);
crm_xml_add(*reply, F_CIB_CALLID, call_id);
crm_xml_add(*reply, F_CIB_CLIENTID, caller);
crm_xml_add_int(*reply, F_CIB_CALLOPTS, call_options);
crm_xml_add_int(*reply, F_CIB_RC, rc);
if (output != NULL) {
crm_trace("Attaching reply output");
add_message_xml(*reply, F_CIB_CALLDATA, output);
}
crm_log_xml_explicit(*reply, "cib:reply");
}
crm_trace("cleanup");
if (cib_op_modifies(call_type) == FALSE && output != current_cib) {
free_xml(output);
output = NULL;
}
if (call_type >= 0) {
cib_op_cleanup(call_type, call_options, &input, &output);
}
crm_trace("done");
return rc;
}
void
cib_peer_callback(xmlNode * msg, void *private_data)
{
const char *reason = NULL;
const char *originator = crm_element_value(msg, F_ORIG);
if (cib_legacy_mode() && (originator == NULL || crm_str_eq(originator, cib_our_uname, TRUE))) {
/* message is from ourselves */
int bcast_id = 0;
if (!(crm_element_value_int(msg, F_CIB_LOCAL_NOTIFY_ID, &bcast_id))) {
check_local_notify(bcast_id);
}
return;
} else if (crm_peer_cache == NULL) {
reason = "membership not established";
goto bail;
}
if (crm_element_value(msg, F_CIB_CLIENTNAME) == NULL) {
crm_xml_add(msg, F_CIB_CLIENTNAME, originator);
}
/* crm_log_xml_trace("Peer[inbound]", msg); */
cib_process_request(msg, FALSE, TRUE, NULL);
return;
bail:
if (reason) {
const char *seq = crm_element_value(msg, F_SEQ);
const char *op = crm_element_value(msg, F_CIB_OPERATION);
crm_warn("Discarding %s message (%s) from %s: %s", op, seq, originator, reason);
}
}
static gboolean
cib_force_exit(gpointer data)
{
crm_notice("Forcing exit!");
terminate_cib(__FUNCTION__, CRM_EX_ERROR);
return FALSE;
}
static void
disconnect_remote_client(gpointer key, gpointer value, gpointer user_data)
{
pcmk__client_t *a_client = value;
crm_err("Disconnecting %s... Not implemented", crm_str(a_client->name));
}
void
cib_shutdown(int nsig)
{
struct qb_ipcs_stats srv_stats;
if (cib_shutdown_flag == FALSE) {
int disconnects = 0;
qb_ipcs_connection_t *c = NULL;
cib_shutdown_flag = TRUE;
c = qb_ipcs_connection_first_get(ipcs_rw);
while (c != NULL) {
qb_ipcs_connection_t *last = c;
c = qb_ipcs_connection_next_get(ipcs_rw, last);
crm_debug("Disconnecting r/w client %p...", last);
qb_ipcs_disconnect(last);
qb_ipcs_connection_unref(last);
disconnects++;
}
c = qb_ipcs_connection_first_get(ipcs_ro);
while (c != NULL) {
qb_ipcs_connection_t *last = c;
c = qb_ipcs_connection_next_get(ipcs_ro, last);
crm_debug("Disconnecting r/o client %p...", last);
qb_ipcs_disconnect(last);
qb_ipcs_connection_unref(last);
disconnects++;
}
c = qb_ipcs_connection_first_get(ipcs_shm);
while (c != NULL) {
qb_ipcs_connection_t *last = c;
c = qb_ipcs_connection_next_get(ipcs_shm, last);
crm_debug("Disconnecting non-blocking r/w client %p...", last);
qb_ipcs_disconnect(last);
qb_ipcs_connection_unref(last);
disconnects++;
}
disconnects += pcmk__ipc_client_count();
crm_debug("Disconnecting %d remote clients", pcmk__ipc_client_count());
pcmk__foreach_ipc_client(disconnect_remote_client, NULL);
crm_info("Disconnected %d clients", disconnects);
}
qb_ipcs_stats_get(ipcs_rw, &srv_stats, QB_FALSE);
if (pcmk__ipc_client_count() == 0) {
crm_info("All clients disconnected (%d)", srv_stats.active_connections);
initiate_exit();
} else {
crm_info("Waiting on %d clients to disconnect (%d)",
pcmk__ipc_client_count(), srv_stats.active_connections);
}
}
void
initiate_exit(void)
{
int active = 0;
xmlNode *leaving = NULL;
active = crm_active_peers();
if (active < 2) {
terminate_cib(__FUNCTION__, 0);
return;
}
crm_info("Sending disconnect notification to %d peers...", active);
leaving = create_xml_node(NULL, "exit-notification");
crm_xml_add(leaving, F_TYPE, "cib");
crm_xml_add(leaving, F_CIB_OPERATION, "cib_shutdown_req");
send_cluster_message(NULL, crm_msg_cib, leaving, TRUE);
free_xml(leaving);
g_timeout_add(EXIT_ESCALATION_MS, cib_force_exit, NULL);
}
extern int remote_fd;
extern int remote_tls_fd;
/*!
* \internal
* \brief Close remote sockets, free the global CIB and quit
*
* \param[in] caller Name of calling function (for log message)
* \param[in] fast If -1, skip disconnect; if positive, exit that
*/
void
terminate_cib(const char *caller, int fast)
{
crm_info("%s: Exiting%s...", caller,
(fast > 0)? " fast" : mainloop ? " from mainloop" : "");
if (remote_fd > 0) {
close(remote_fd);
remote_fd = 0;
}
if (remote_tls_fd > 0) {
close(remote_tls_fd);
remote_tls_fd = 0;
}
uninitializeCib();
if (fast > 0) {
/* Quit fast on error */
pcmk__stop_based_ipc(ipcs_ro, ipcs_rw, ipcs_shm);
crm_exit(fast);
} else if ((mainloop != NULL) && g_main_loop_is_running(mainloop)) {
/* Quit via returning from the main loop. If fast == -1, we skip the
* disconnect here, and it will be done when the main loop returns
* (this allows the peer status callback to avoid messing with the
* peer caches).
*/
if (fast == 0) {
crm_cluster_disconnect(&crm_cluster);
}
g_main_loop_quit(mainloop);
} else {
/* Quit via clean exit. Even the peer status callback can disconnect
* here, because we're not returning control to the caller. */
crm_cluster_disconnect(&crm_cluster);
pcmk__stop_based_ipc(ipcs_ro, ipcs_rw, ipcs_shm);
crm_exit(CRM_EX_OK);
}
}
diff --git a/daemons/based/cibmon.c b/daemons/based/cibmon.c
index c9b0784617..5bdc27e06c 100644
--- a/daemons/based/cibmon.c
+++ b/daemons/based/cibmon.c
@@ -1,251 +1,250 @@
/*
- * Copyright 2004-2019 the Pacemaker project contributors
+ * Copyright 2004-2020 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 <sys/param.h>
#include <crm/crm.h>
#include <stdio.h>
#include <sys/types.h>
#include <unistd.h>
#include <stdlib.h>
#include <errno.h>
#include <fcntl.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <crm/common/mainloop.h>
#include <crm/cib/internal.h>
#include <crm/common/ipc.h>
#include <crm/pengine/status.h>
-#include <../lib/pengine/unpack.h>
#include <crm/cib.h>
#ifdef HAVE_GETOPT_H
# include <getopt.h>
#endif
static int max_failures = 30;
static gboolean log_diffs = FALSE;
static gboolean log_updates = FALSE;
static GMainLoop *mainloop = NULL;
void usage(const char *cmd, crm_exit_t exit_status);
void cib_connection_destroy(gpointer user_data);
void cibmon_shutdown(int nsig);
void cibmon_diff(const char *event, xmlNode * msg);
static cib_t *cib = NULL;
static xmlNode *cib_copy = NULL;
#define OPTARGS "V?m:du"
int
main(int argc, char **argv)
{
int argerr = 0;
int flag;
int attempts = 0;
int rc = pcmk_ok;
#ifdef HAVE_GETOPT_H
int option_index = 0;
static struct option long_options[] = {
/* Top-level Options */
{"verbose", 0, 0, 'V'},
{"help", 0, 0, '?'},
{"log-diffs", 0, 0, 'd'},
{"log-updates", 0, 0, 'u'},
{"max-conn-fail", 1, 0, 'm'},
{0, 0, 0, 0}
};
#endif
crm_log_cli_init("cibmon");
crm_signal_handler(SIGTERM, cibmon_shutdown);
while (1) {
#ifdef HAVE_GETOPT_H
flag = getopt_long(argc, argv, OPTARGS, long_options, &option_index);
#else
flag = getopt(argc, argv, OPTARGS);
#endif
if (flag == -1)
break;
switch (flag) {
case 'V':
crm_bump_log_level(argc, argv);
break;
case '?':
usage(crm_system_name, CRM_EX_OK);
break;
case 'd':
log_diffs = TRUE;
break;
case 'u':
log_updates = TRUE;
break;
case 'm':
max_failures = crm_parse_int(optarg, "30");
break;
default:
printf("Argument code 0%o (%c)" " is not (?yet?) supported\n", flag, flag);
++argerr;
break;
}
}
if (optind < argc) {
printf("non-option ARGV-elements: ");
while (optind < argc)
printf("%s ", argv[optind++]);
printf("\n");
}
if (optind > argc) {
++argerr;
}
if (argerr) {
usage(crm_system_name, CRM_EX_USAGE);
}
cib = cib_new();
do {
sleep(1);
rc = cib->cmds->signon(cib, crm_system_name, cib_query);
} while (rc == -ENOTCONN && attempts++ < max_failures);
if (rc != pcmk_ok) {
crm_err("Signon to CIB failed: %s", pcmk_strerror(rc));
goto fail;
}
crm_debug("Setting dnotify");
rc = cib->cmds->set_connection_dnotify(cib, cib_connection_destroy);
if (rc != pcmk_ok) {
crm_err("Failed to set dnotify callback: %s", pcmk_strerror(rc));
goto fail;
}
crm_debug("Setting diff callback");
rc = cib->cmds->add_notify_callback(cib, T_CIB_DIFF_NOTIFY, cibmon_diff);
if (rc != pcmk_ok) {
crm_err("Failed to set diff callback: %s", pcmk_strerror(rc));
goto fail;
}
mainloop = g_main_loop_new(NULL, FALSE);
crm_info("Starting mainloop");
g_main_loop_run(mainloop);
crm_trace("%s exiting normally", crm_system_name);
fflush(stderr);
return CRM_EX_OK;
fail:
crm_err("Setup failed, could not monitor CIB actions");
return CRM_EX_ERROR;
}
void
usage(const char *cmd, crm_exit_t exit_status)
{
FILE *stream;
stream = (exit_status == CRM_EX_OK)? stdout : stderr;
fflush(stream);
crm_exit(exit_status);
}
void
cib_connection_destroy(gpointer user_data)
{
cib_t *conn = user_data;
crm_err("Connection to the CIB terminated... exiting");
conn->cmds->signoff(conn); /* Ensure IPC is cleaned up */
g_main_loop_quit(mainloop);
return;
}
void
cibmon_diff(const char *event, xmlNode * msg)
{
int rc = -1;
const char *op = NULL;
unsigned int log_level = LOG_INFO;
xmlNode *diff = NULL;
xmlNode *cib_last = NULL;
xmlNode *update = get_message_xml(msg, F_CIB_UPDATE);
if (msg == NULL) {
crm_err("NULL update");
return;
}
crm_element_value_int(msg, F_CIB_RC, &rc);
op = crm_element_value(msg, F_CIB_OPERATION);
diff = get_message_xml(msg, F_CIB_UPDATE_RESULT);
if (rc < pcmk_ok) {
log_level = LOG_WARNING;
do_crm_log(log_level, "[%s] %s ABORTED: %s", event, op, pcmk_strerror(rc));
return;
}
if (log_diffs) {
xml_log_patchset(log_level, op, diff);
}
if (log_updates && update != NULL) {
crm_log_xml_trace(update, "raw_update");
}
if (cib_copy != NULL) {
cib_last = cib_copy;
cib_copy = NULL;
rc = cib_process_diff(op, cib_force_diff, NULL, NULL, diff, cib_last, &cib_copy, NULL);
if (rc != pcmk_ok) {
crm_debug("Update didn't apply, requesting full copy: %s", pcmk_strerror(rc));
free_xml(cib_copy);
cib_copy = NULL;
}
}
if (cib_copy == NULL) {
rc = cib->cmds->query(cib, NULL, &cib_copy, cib_scope_local | cib_sync_call);
}
if(rc == -EACCES) {
crm_exit(CRM_EX_INSUFFICIENT_PRIV);
}
free_xml(cib_last);
}
void
cibmon_shutdown(int nsig)
{
crm_exit(CRM_EX_OK);
}
diff --git a/daemons/controld/controld_control.c b/daemons/controld/controld_control.c
index 24fbcdc825..1ddcada9e4 100644
--- a/daemons/controld/controld_control.c
+++ b/daemons/controld/controld_control.c
@@ -1,820 +1,818 @@
/*
* Copyright 2004-2020 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 <sys/param.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <crm/crm.h>
#include <crm/msg_xml.h>
#include <crm/pengine/rules.h>
#include <crm/cluster/internal.h>
#include <crm/cluster/election.h>
#include <crm/common/ipcs_internal.h>
#include <pacemaker-controld.h>
qb_ipcs_service_t *ipcs = NULL;
#if SUPPORT_COROSYNC
extern gboolean crm_connect_corosync(crm_cluster_t * cluster);
#endif
void crm_shutdown(int nsig);
gboolean crm_read_options(gpointer user_data);
gboolean fsa_has_quorum = FALSE;
crm_trigger_t *fsa_source = NULL;
crm_trigger_t *config_read = NULL;
bool no_quorum_suicide_escalation = FALSE;
bool controld_shutdown_lock_enabled = false;
/* A_HA_CONNECT */
void
do_ha_control(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state,
enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
gboolean registered = FALSE;
static crm_cluster_t *cluster = NULL;
if (cluster == NULL) {
cluster = calloc(1, sizeof(crm_cluster_t));
}
if (action & A_HA_DISCONNECT) {
crm_cluster_disconnect(cluster);
crm_info("Disconnected from the cluster");
set_bit(fsa_input_register, R_HA_DISCONNECTED);
}
if (action & A_HA_CONNECT) {
crm_set_status_callback(&peer_update_callback);
crm_set_autoreap(FALSE);
if (is_corosync_cluster()) {
#if SUPPORT_COROSYNC
registered = crm_connect_corosync(cluster);
#endif
}
if (registered == TRUE) {
controld_election_init(cluster->uname);
fsa_our_uname = cluster->uname;
fsa_our_uuid = cluster->uuid;
if(cluster->uuid == NULL) {
crm_err("Could not obtain local uuid");
registered = FALSE;
}
}
if (registered == FALSE) {
set_bit(fsa_input_register, R_HA_DISCONNECTED);
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
return;
}
populate_cib_nodes(node_update_none, __FUNCTION__);
clear_bit(fsa_input_register, R_HA_DISCONNECTED);
crm_info("Connected to the cluster");
}
if (action & ~(A_HA_CONNECT | A_HA_DISCONNECT)) {
crm_err("Unexpected action %s in %s", fsa_action2string(action), __FUNCTION__);
}
}
/* A_SHUTDOWN */
void
do_shutdown(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
/* just in case */
set_bit(fsa_input_register, R_SHUTDOWN);
controld_disconnect_fencer(FALSE);
}
/* A_SHUTDOWN_REQ */
void
do_shutdown_req(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state,
enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
xmlNode *msg = NULL;
set_bit(fsa_input_register, R_SHUTDOWN);
crm_info("Sending shutdown request to all peers (DC is %s)",
(fsa_our_dc? fsa_our_dc : "not set"));
msg = create_request(CRM_OP_SHUTDOWN_REQ, NULL, NULL, CRM_SYSTEM_CRMD, CRM_SYSTEM_CRMD, NULL);
/* set_bit(fsa_input_register, R_STAYDOWN); */
if (send_cluster_message(NULL, crm_msg_crmd, msg, TRUE) == FALSE) {
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
}
free_xml(msg);
}
extern char *max_generation_from;
extern xmlNode *max_generation_xml;
extern GHashTable *resource_history;
extern GHashTable *voted;
void
crmd_fast_exit(crm_exit_t exit_code)
{
if (is_set(fsa_input_register, R_STAYDOWN)) {
crm_warn("Inhibiting respawn "CRM_XS" remapping exit code %d to %d",
exit_code, CRM_EX_FATAL);
exit_code = CRM_EX_FATAL;
} else if ((exit_code == CRM_EX_OK)
&& is_set(fsa_input_register, R_IN_RECOVERY)) {
crm_err("Could not recover from internal error");
exit_code = CRM_EX_ERROR;
}
crm_exit(exit_code);
}
crm_exit_t
crmd_exit(crm_exit_t exit_code)
{
GListPtr gIter = NULL;
GMainLoop *mloop = crmd_mainloop;
static bool in_progress = FALSE;
if (in_progress && (exit_code == CRM_EX_OK)) {
crm_debug("Exit is already in progress");
return exit_code;
} else if(in_progress) {
crm_notice("Error during shutdown process, exiting now with status %d (%s)",
exit_code, crm_exit_str(exit_code));
crm_write_blackbox(SIGTRAP, NULL);
crmd_fast_exit(exit_code);
}
in_progress = TRUE;
crm_trace("Preparing to exit with status %d (%s)",
exit_code, crm_exit_str(exit_code));
/* Suppress secondary errors resulting from us disconnecting everything */
set_bit(fsa_input_register, R_HA_DISCONNECTED);
/* Close all IPC servers and clients to ensure any and all shared memory files are cleaned up */
if(ipcs) {
crm_trace("Closing IPC server");
mainloop_del_ipc_server(ipcs);
ipcs = NULL;
}
controld_close_attrd_ipc();
pe_subsystem_free();
controld_disconnect_fencer(TRUE);
if ((exit_code == CRM_EX_OK) && (crmd_mainloop == NULL)) {
crm_debug("No mainloop detected");
exit_code = CRM_EX_ERROR;
}
/* On an error, just get out.
*
* Otherwise, make the effort to have mainloop exit gracefully so
* that it (mostly) cleans up after itself and valgrind has less
* to report on - allowing real errors stand out
*/
if (exit_code != CRM_EX_OK) {
crm_notice("Forcing immediate exit with status %d (%s)",
exit_code, crm_exit_str(exit_code));
crm_write_blackbox(SIGTRAP, NULL);
crmd_fast_exit(exit_code);
}
/* Clean up as much memory as possible for valgrind */
for (gIter = fsa_message_queue; gIter != NULL; gIter = gIter->next) {
fsa_data_t *fsa_data = gIter->data;
crm_info("Dropping %s: [ state=%s cause=%s origin=%s ]",
fsa_input2string(fsa_data->fsa_input),
fsa_state2string(fsa_state),
fsa_cause2string(fsa_data->fsa_cause), fsa_data->origin);
delete_fsa_input(fsa_data);
}
clear_bit(fsa_input_register, R_MEMBERSHIP);
g_list_free(fsa_message_queue); fsa_message_queue = NULL;
metadata_cache_fini();
controld_election_fini();
/* Tear down the CIB manager connection, but don't free it yet -- it could
* be used when we drain the mainloop later.
*/
cib_free_callbacks(fsa_cib_conn);
fsa_cib_conn->cmds->signoff(fsa_cib_conn);
verify_stopped(fsa_state, LOG_WARNING);
clear_bit(fsa_input_register, R_LRM_CONNECTED);
lrm_state_destroy_all();
/* This basically will not work, since mainloop has a reference to it */
mainloop_destroy_trigger(fsa_source); fsa_source = NULL;
mainloop_destroy_trigger(config_read); config_read = NULL;
mainloop_destroy_trigger(transition_trigger); transition_trigger = NULL;
pcmk__client_cleanup();
crm_peer_destroy();
controld_free_fsa_timers();
te_cleanup_stonith_history_sync(NULL, TRUE);
controld_free_sched_timer();
free(fsa_our_dc_version); fsa_our_dc_version = NULL;
free(fsa_our_uname); fsa_our_uname = NULL;
free(fsa_our_uuid); fsa_our_uuid = NULL;
free(fsa_our_dc); fsa_our_dc = NULL;
free(fsa_cluster_name); fsa_cluster_name = NULL;
free(te_uuid); te_uuid = NULL;
free(failed_stop_offset); failed_stop_offset = NULL;
free(failed_start_offset); failed_start_offset = NULL;
free(max_generation_from); max_generation_from = NULL;
free_xml(max_generation_xml); max_generation_xml = NULL;
mainloop_destroy_signal(SIGPIPE);
mainloop_destroy_signal(SIGUSR1);
mainloop_destroy_signal(SIGTERM);
mainloop_destroy_signal(SIGTRAP);
/* leave SIGCHLD engaged as we might still want to drain some service-actions */
if (mloop) {
GMainContext *ctx = g_main_loop_get_context(crmd_mainloop);
/* Don't re-enter this block */
crmd_mainloop = NULL;
/* no signals on final draining anymore */
mainloop_destroy_signal(SIGCHLD);
crm_trace("Draining mainloop %d %d", g_main_loop_is_running(mloop), g_main_context_pending(ctx));
{
int lpc = 0;
while((g_main_context_pending(ctx) && lpc < 10)) {
lpc++;
crm_trace("Iteration %d", lpc);
g_main_context_dispatch(ctx);
}
}
crm_trace("Closing mainloop %d %d", g_main_loop_is_running(mloop), g_main_context_pending(ctx));
g_main_loop_quit(mloop);
/* Won't do anything yet, since we're inside it now */
g_main_loop_unref(mloop);
} else {
mainloop_destroy_signal(SIGCHLD);
}
cib_delete(fsa_cib_conn);
fsa_cib_conn = NULL;
throttle_fini();
/* Graceful */
crm_trace("Done preparing for exit with status %d (%s)",
exit_code, crm_exit_str(exit_code));
return exit_code;
}
/* A_EXIT_0, A_EXIT_1 */
void
do_exit(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
crm_exit_t exit_code = CRM_EX_OK;
int log_level = LOG_INFO;
const char *exit_type = "gracefully";
if (action & A_EXIT_1) {
log_level = LOG_ERR;
exit_type = "forcefully";
exit_code = CRM_EX_ERROR;
}
verify_stopped(cur_state, LOG_ERR);
do_crm_log(log_level, "Performing %s - %s exiting the controller",
fsa_action2string(action), exit_type);
crm_info("[%s] stopped (%d)", crm_system_name, exit_code);
crmd_exit(exit_code);
}
static void sigpipe_ignore(int nsig) { return; }
/* A_STARTUP */
void
do_startup(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
crm_debug("Registering Signal Handlers");
mainloop_add_signal(SIGTERM, crm_shutdown);
mainloop_add_signal(SIGPIPE, sigpipe_ignore);
fsa_source = mainloop_add_trigger(G_PRIORITY_HIGH, crm_fsa_trigger, NULL);
config_read = mainloop_add_trigger(G_PRIORITY_HIGH, crm_read_options, NULL);
transition_trigger = mainloop_add_trigger(G_PRIORITY_LOW, te_graph_trigger, NULL);
crm_debug("Creating CIB manager and executor objects");
fsa_cib_conn = cib_new();
lrm_state_init_local();
if (controld_init_fsa_timers() == FALSE) {
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
}
}
// \return libqb error code (0 on success, -errno on error)
static int32_t
accept_controller_client(qb_ipcs_connection_t *c, uid_t uid, gid_t gid)
{
crm_trace("Accepting new IPC client connection");
if (pcmk__new_client(c, uid, gid) == NULL) {
return -EIO;
}
return 0;
}
// \return libqb error code (0 on success, -errno on error)
static int32_t
dispatch_controller_ipc(qb_ipcs_connection_t * c, void *data, size_t size)
{
uint32_t id = 0;
uint32_t flags = 0;
pcmk__client_t *client = pcmk__find_client(c);
xmlNode *msg = pcmk__client_data2xml(client, data, &id, &flags);
pcmk__ipc_send_ack(client, id, flags, "ack");
if (msg == NULL) {
return 0;
}
#if ENABLE_ACL
CRM_ASSERT(client->user != NULL);
pcmk__update_acl_user(msg, F_CRM_USER, client->user);
#endif
crm_xml_add(msg, F_CRM_SYS_FROM, client->id);
if (controld_authorize_ipc_message(msg, client, NULL)) {
crm_trace("Processing IPC message from %s", pcmk__client_name(client));
route_message(C_IPC_MESSAGE, msg);
}
trigger_fsa(fsa_source);
free_xml(msg);
return 0;
}
static int32_t
crmd_ipc_closed(qb_ipcs_connection_t * c)
{
pcmk__client_t *client = pcmk__find_client(c);
if (client) {
crm_trace("Disconnecting %sregistered client %s (%p/%p)",
(client->userdata? "" : "un"), pcmk__client_name(client),
c, client);
free(client->userdata);
pcmk__free_client(client);
trigger_fsa(fsa_source);
}
return 0;
}
static void
crmd_ipc_destroy(qb_ipcs_connection_t * c)
{
crm_trace("Connection %p", c);
crmd_ipc_closed(c);
}
/* A_STOP */
void
do_stop(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
crm_trace("Closing IPC server");
mainloop_del_ipc_server(ipcs); ipcs = NULL;
register_fsa_input(C_FSA_INTERNAL, I_TERMINATE, NULL);
}
/* A_STARTED */
void
do_started(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
static struct qb_ipcs_service_handlers crmd_callbacks = {
.connection_accept = accept_controller_client,
.connection_created = NULL,
.msg_process = dispatch_controller_ipc,
.connection_closed = crmd_ipc_closed,
.connection_destroyed = crmd_ipc_destroy
};
if (cur_state != S_STARTING) {
crm_err("Start cancelled... %s", fsa_state2string(cur_state));
return;
} else if (is_set(fsa_input_register, R_MEMBERSHIP) == FALSE) {
crm_info("Delaying start, no membership data (%.16llx)", R_MEMBERSHIP);
crmd_fsa_stall(TRUE);
return;
} else if (is_set(fsa_input_register, R_LRM_CONNECTED) == FALSE) {
crm_info("Delaying start, not connected to executor (%.16llx)", R_LRM_CONNECTED);
crmd_fsa_stall(TRUE);
return;
} else if (is_set(fsa_input_register, R_CIB_CONNECTED) == FALSE) {
crm_info("Delaying start, CIB not connected (%.16llx)", R_CIB_CONNECTED);
crmd_fsa_stall(TRUE);
return;
} else if (is_set(fsa_input_register, R_READ_CONFIG) == FALSE) {
crm_info("Delaying start, Config not read (%.16llx)", R_READ_CONFIG);
crmd_fsa_stall(TRUE);
return;
} else if (is_set(fsa_input_register, R_PEER_DATA) == FALSE) {
crm_info("Delaying start, No peer data (%.16llx)", R_PEER_DATA);
crmd_fsa_stall(TRUE);
return;
}
crm_debug("Init server comms");
ipcs = pcmk__serve_controld_ipc(&crmd_callbacks);
if (ipcs == NULL) {
crm_err("Failed to create IPC server: shutting down and inhibiting respawn");
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
} else {
crm_notice("Pacemaker controller successfully started and accepting connections");
}
controld_trigger_fencer_connect();
clear_bit(fsa_input_register, R_STARTING);
register_fsa_input(msg_data->fsa_cause, I_PENDING, NULL);
}
/* A_RECOVER */
void
do_recover(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
set_bit(fsa_input_register, R_IN_RECOVERY);
crm_warn("Fast-tracking shutdown in response to errors");
register_fsa_input(C_FSA_INTERNAL, I_TERMINATE, NULL);
}
static pcmk__cluster_option_t crmd_opts[] = {
/* name, old name, type, allowed values,
* default value, validator,
* short description,
* long description
*/
{
"dc-version", NULL, "string", NULL, "none", NULL,
"Pacemaker version on cluster node elected Designated Controller (DC)",
"Includes a hash which identifies the exact changeset the code was "
"built from. Used for diagnostic purposes."
},
{
"cluster-infrastructure", NULL, "string", NULL, "corosync", NULL,
"The messaging stack on which Pacemaker is currently running",
"Used for informational and diagnostic purposes."
},
{
"cluster-name", NULL, "string", NULL, NULL, NULL,
"An arbitrary name for the cluster",
"This optional value is mostly for users' convenience as desired "
"in administration, but may also be used in Pacemaker "
"configuration rules via the #cluster-name node attribute, and "
"by higher-level tools and resource agents."
},
{
XML_CONFIG_ATTR_DC_DEADTIME, NULL, "time",
NULL, "20s", pcmk__valid_interval_spec,
"How long to wait for a response from other nodes during start-up",
"The optimal value will depend on the speed and load of your network "
"and the type of switches used."
},
{
XML_CONFIG_ATTR_RECHECK, NULL, "time",
"Zero disables polling, while positive values are an interval in seconds"
"(unless other units are specified, for example \"5min\")",
"15min", pcmk__valid_interval_spec,
"Polling interval to recheck cluster state and evalute rules "
"with date specifications",
"Pacemaker is primarily event-driven, and looks ahead to know when to "
"recheck cluster state for failure timeouts and most time-based "
"rules. However, it will also recheck the cluster after this "
"amount of inactivity, to evaluate rules with date specifications "
"and serve as a fail-safe for certain types of scheduler bugs."
},
{
"load-threshold", NULL, "percentage", NULL,
"80%", pcmk__valid_utilization,
"Maximum amount of system load that should be used by cluster nodes",
"The cluster will slow down its recovery process when the amount of "
"system resources used (currently CPU) approaches this limit",
},
{
"node-action-limit", NULL, "integer", NULL,
"0", pcmk__valid_number,
"Maximum number of jobs that can be scheduled per node "
"(defaults to 2x cores)"
},
{ XML_CONFIG_ATTR_FENCE_REACTION, NULL, "string", NULL, "stop", NULL,
"How a cluster node should react if notified of its own fencing",
"A cluster node may receive notification of its own fencing if fencing "
"is misconfigured, or if fabric fencing is in use that doesn't cut "
"cluster communication. Allowed values are \"stop\" to attempt to "
"immediately stop pacemaker and stay stopped, or \"panic\" to attempt "
"to immediately reboot the local node, falling back to stop on failure."
},
{
XML_CONFIG_ATTR_ELECTION_FAIL, NULL, "time", NULL,
"2min", pcmk__valid_interval_spec,
"*** Advanced Use Only ***",
"Declare an election failed if it is not decided within this much "
"time. If you need to adjust this value, it probably indicates "
"the presence of a bug."
},
{
XML_CONFIG_ATTR_FORCE_QUIT, NULL, "time", NULL,
"20min", pcmk__valid_interval_spec,
"*** Advanced Use Only ***",
"Exit immediately if shutdown does not complete within this much "
"time. If you need to adjust this value, it probably indicates "
"the presence of a bug."
},
{
"join-integration-timeout", "crmd-integration-timeout", "time", NULL,
"3min", pcmk__valid_interval_spec,
"*** Advanced Use Only ***",
"If you need to adjust this value, it probably indicates "
"the presence of a bug."
},
{
"join-finalization-timeout", "crmd-finalization-timeout", "time", NULL,
"30min", pcmk__valid_interval_spec,
"*** Advanced Use Only ***",
"If you need to adjust this value, it probably indicates "
"the presence of a bug."
},
{
"transition-delay", "crmd-transition-delay", "time", NULL,
"0s", pcmk__valid_interval_spec,
"*** Advanced Use Only *** Enabling this option will slow down "
"cluster recovery under all conditions",
"Delay cluster recovery for this much time to allow for additional "
"events to occur. Useful if your configuration is sensitive to "
"the order in which ping updates arrive."
},
{
"stonith-watchdog-timeout", NULL, "time", NULL,
NULL, pcmk__valid_sbd_timeout,
"How long to wait before we can assume nodes are safely down "
"when sbd is in use",
NULL
},
{
"stonith-max-attempts", NULL, "integer", NULL,
"10", pcmk__valid_positive_number,
"How many times fencing can fail before it will no longer be "
"immediately re-attempted on a target"
},
// Already documented in libpe_status (other values must be kept identical)
{
"no-quorum-policy", NULL, "enum", "stop, freeze, ignore, suicide",
"stop", pcmk__valid_quorum, NULL, NULL
},
{
XML_CONFIG_ATTR_SHUTDOWN_LOCK, NULL, "boolean", NULL,
"false", pcmk__valid_boolean, NULL, NULL
},
};
void
crmd_metadata(void)
{
pcmk__print_option_metadata("pacemaker-controld", "1.0",
"Pacemaker controller options",
"Cluster options used by Pacemaker's "
"controller (formerly called crmd)",
crmd_opts, DIMOF(crmd_opts));
}
static void
verify_crmd_options(GHashTable * options)
{
pcmk__validate_cluster_options(options, crmd_opts, DIMOF(crmd_opts));
}
static const char *
crmd_pref(GHashTable * options, const char *name)
{
return pcmk__cluster_option(options, crmd_opts, DIMOF(crmd_opts), name);
}
static void
config_query_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data)
{
const char *value = NULL;
GHashTable *config_hash = NULL;
crm_time_t *now = crm_time_new(NULL);
xmlNode *crmconfig = NULL;
xmlNode *alerts = NULL;
if (rc != pcmk_ok) {
fsa_data_t *msg_data = NULL;
crm_err("Local CIB query resulted in an error: %s", pcmk_strerror(rc));
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
if (rc == -EACCES || rc == -pcmk_err_schema_validation) {
crm_err("The cluster is mis-configured - shutting down and staying down");
set_bit(fsa_input_register, R_STAYDOWN);
}
goto bail;
}
crmconfig = output;
if ((crmconfig) &&
(crm_element_name(crmconfig)) &&
(strcmp(crm_element_name(crmconfig), XML_CIB_TAG_CRMCONFIG) != 0)) {
crmconfig = first_named_child(crmconfig, XML_CIB_TAG_CRMCONFIG);
}
if (!crmconfig) {
fsa_data_t *msg_data = NULL;
crm_err("Local CIB query for " XML_CIB_TAG_CRMCONFIG " section failed");
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
goto bail;
}
crm_debug("Call %d : Parsing CIB options", call_id);
config_hash = crm_str_table_new();
pe_unpack_nvpairs(crmconfig, crmconfig, XML_CIB_TAG_PROPSET, NULL,
config_hash, CIB_OPTIONS_FIRST, FALSE, now, NULL);
verify_crmd_options(config_hash);
value = crmd_pref(config_hash, XML_CONFIG_ATTR_DC_DEADTIME);
election_trigger->period_ms = crm_parse_interval_spec(value);
value = crmd_pref(config_hash, "node-action-limit"); /* Also checks migration-limit */
throttle_update_job_max(value);
value = crmd_pref(config_hash, "load-threshold");
if(value) {
throttle_set_load_target(strtof(value, NULL) / 100.0);
}
value = crmd_pref(config_hash, "no-quorum-policy");
if (safe_str_eq(value, "suicide") && pcmk_locate_sbd()) {
no_quorum_suicide_escalation = TRUE;
}
set_fence_reaction(crmd_pref(config_hash, XML_CONFIG_ATTR_FENCE_REACTION));
value = crmd_pref(config_hash,"stonith-max-attempts");
update_stonith_max_attempts(value);
value = crmd_pref(config_hash, XML_CONFIG_ATTR_FORCE_QUIT);
shutdown_escalation_timer->period_ms = crm_parse_interval_spec(value);
crm_debug("Shutdown escalation occurs if DC has not responded to request in %ums",
shutdown_escalation_timer->period_ms);
value = crmd_pref(config_hash, XML_CONFIG_ATTR_ELECTION_FAIL);
controld_set_election_period(value);
value = crmd_pref(config_hash, XML_CONFIG_ATTR_RECHECK);
recheck_interval_ms = crm_parse_interval_spec(value);
crm_debug("Re-run scheduler after %dms of inactivity", recheck_interval_ms);
value = crmd_pref(config_hash, "transition-delay");
transition_timer->period_ms = crm_parse_interval_spec(value);
value = crmd_pref(config_hash, "join-integration-timeout");
integration_timer->period_ms = crm_parse_interval_spec(value);
value = crmd_pref(config_hash, "join-finalization-timeout");
finalization_timer->period_ms = crm_parse_interval_spec(value);
value = crmd_pref(config_hash, XML_CONFIG_ATTR_SHUTDOWN_LOCK);
controld_shutdown_lock_enabled = crm_is_true(value);
free(fsa_cluster_name);
fsa_cluster_name = NULL;
value = g_hash_table_lookup(config_hash, "cluster-name");
if (value) {
fsa_cluster_name = strdup(value);
}
alerts = first_named_child(output, XML_CIB_TAG_ALERTS);
crmd_unpack_alerts(alerts);
set_bit(fsa_input_register, R_READ_CONFIG);
crm_trace("Triggering FSA: %s", __FUNCTION__);
mainloop_set_trigger(fsa_source);
g_hash_table_destroy(config_hash);
bail:
crm_time_free(now);
}
gboolean
crm_read_options(gpointer user_data)
{
int call_id =
fsa_cib_conn->cmds->query(fsa_cib_conn,
"//" XML_CIB_TAG_CRMCONFIG " | //" XML_CIB_TAG_ALERTS,
NULL, cib_xpath | cib_scope_local);
fsa_register_cib_callback(call_id, FALSE, NULL, config_query_callback);
crm_trace("Querying the CIB... call %d", call_id);
return TRUE;
}
/* A_READCONFIG */
void
do_read_config(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state,
enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
throttle_init();
mainloop_set_trigger(config_read);
}
void
crm_shutdown(int nsig)
{
- if (crmd_mainloop != NULL && g_main_loop_is_running(crmd_mainloop)) {
- if (is_set(fsa_input_register, R_SHUTDOWN)) {
- crm_err("Escalating the shutdown");
- register_fsa_input_before(C_SHUTDOWN, I_ERROR, NULL);
-
- } else {
- set_bit(fsa_input_register, R_SHUTDOWN);
- register_fsa_input(C_SHUTDOWN, I_SHUTDOWN, NULL);
+ if ((crmd_mainloop == NULL) || !g_main_loop_is_running(crmd_mainloop)) {
+ crmd_exit(CRM_EX_OK);
+ return;
+ }
- if (shutdown_escalation_timer->period_ms == 0) {
- const char *value = crmd_pref(NULL, XML_CONFIG_ATTR_FORCE_QUIT);
+ if (is_set(fsa_input_register, R_SHUTDOWN)) {
+ crm_err("Escalating shutdown");
+ register_fsa_input_before(C_SHUTDOWN, I_ERROR, NULL);
+ return;
+ }
- shutdown_escalation_timer->period_ms = crm_parse_interval_spec(value);
- }
+ set_bit(fsa_input_register, R_SHUTDOWN);
+ register_fsa_input(C_SHUTDOWN, I_SHUTDOWN, NULL);
- /* can't rely on this... */
- crm_notice("Shutting down cluster resource manager " CRM_XS
- " limit=%ums", shutdown_escalation_timer->period_ms);
- controld_start_timer(shutdown_escalation_timer);
- }
+ if (shutdown_escalation_timer->period_ms == 0) {
+ const char *value = crmd_pref(NULL, XML_CONFIG_ATTR_FORCE_QUIT);
- } else {
- crm_info("exit from shutdown");
- crmd_exit(CRM_EX_OK);
+ shutdown_escalation_timer->period_ms = crm_parse_interval_spec(value);
}
+
+ crm_notice("Initiating controller shutdown sequence " CRM_XS
+ " limit=%ums", shutdown_escalation_timer->period_ms);
+ controld_start_timer(shutdown_escalation_timer);
}
diff --git a/daemons/controld/controld_messages.c b/daemons/controld/controld_messages.c
index 6835149142..62719ad77c 100644
--- a/daemons/controld/controld_messages.c
+++ b/daemons/controld/controld_messages.c
@@ -1,1209 +1,1244 @@
/*
* Copyright 2004-2020 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 <sys/param.h>
#include <string.h>
#include <time.h>
#include <crm/crm.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <crm/cluster/internal.h>
#include <crm/cib.h>
#include <crm/common/ipcs_internal.h>
#include <pacemaker-controld.h>
GListPtr fsa_message_queue = NULL;
extern void crm_shutdown(int nsig);
static enum crmd_fsa_input handle_message(xmlNode *msg,
enum crmd_fsa_cause cause);
static void handle_response(xmlNode *stored_msg);
static enum crmd_fsa_input handle_request(xmlNode *stored_msg,
enum crmd_fsa_cause cause);
static enum crmd_fsa_input handle_shutdown_request(xmlNode *stored_msg);
static void send_msg_via_ipc(xmlNode * msg, const char *sys);
/* debug only, can wrap all it likes */
int last_data_id = 0;
void
register_fsa_error_adv(enum crmd_fsa_cause cause, enum crmd_fsa_input input,
fsa_data_t * cur_data, void *new_data, const char *raised_from)
{
/* save the current actions if any */
if (fsa_actions != A_NOTHING) {
register_fsa_input_adv(cur_data ? cur_data->fsa_cause : C_FSA_INTERNAL,
I_NULL, cur_data ? cur_data->data : NULL,
fsa_actions, TRUE, __FUNCTION__);
}
/* reset the action list */
crm_info("Resetting the current action list");
fsa_dump_actions(fsa_actions, "Drop");
fsa_actions = A_NOTHING;
/* register the error */
register_fsa_input_adv(cause, input, new_data, A_NOTHING, TRUE, raised_from);
}
int
register_fsa_input_adv(enum crmd_fsa_cause cause, enum crmd_fsa_input input,
void *data, long long with_actions,
gboolean prepend, const char *raised_from)
{
unsigned old_len = g_list_length(fsa_message_queue);
fsa_data_t *fsa_data = NULL;
if (raised_from == NULL) {
raised_from = "<unknown>";
}
if (input == I_NULL && with_actions == A_NOTHING /* && data == NULL */ ) {
/* no point doing anything */
crm_err("Cannot add entry to queue: no input and no action");
return 0;
}
if (input == I_WAIT_FOR_EVENT) {
do_fsa_stall = TRUE;
crm_debug("Stalling the FSA pending further input: source=%s cause=%s data=%p queue=%d",
raised_from, fsa_cause2string(cause), data, old_len);
if (old_len > 0) {
fsa_dump_queue(LOG_TRACE);
prepend = FALSE;
}
if (data == NULL) {
fsa_actions |= with_actions;
fsa_dump_actions(with_actions, "Restored");
return 0;
}
/* Store everything in the new event and reset fsa_actions */
with_actions |= fsa_actions;
fsa_actions = A_NOTHING;
}
last_data_id++;
crm_trace("%s %s FSA input %d (%s) due to %s, %s data",
raised_from, (prepend? "prepended" : "appended"), last_data_id,
fsa_input2string(input), fsa_cause2string(cause),
(data? "with" : "without"));
fsa_data = calloc(1, sizeof(fsa_data_t));
fsa_data->id = last_data_id;
fsa_data->fsa_input = input;
fsa_data->fsa_cause = cause;
fsa_data->origin = raised_from;
fsa_data->data = NULL;
fsa_data->data_type = fsa_dt_none;
fsa_data->actions = with_actions;
if (with_actions != A_NOTHING) {
crm_trace("Adding actions %.16llx to input", with_actions);
}
if (data != NULL) {
switch (cause) {
case C_FSA_INTERNAL:
case C_CRMD_STATUS_CALLBACK:
case C_IPC_MESSAGE:
case C_HA_MESSAGE:
CRM_CHECK(((ha_msg_input_t *) data)->msg != NULL,
crm_err("Bogus data from %s", raised_from));
crm_trace("Copying %s data from %s as cluster message data",
fsa_cause2string(cause), raised_from);
fsa_data->data = copy_ha_msg_input(data);
fsa_data->data_type = fsa_dt_ha_msg;
break;
case C_LRM_OP_CALLBACK:
crm_trace("Copying %s data from %s as lrmd_event_data_t",
fsa_cause2string(cause), raised_from);
fsa_data->data = lrmd_copy_event((lrmd_event_data_t *) data);
fsa_data->data_type = fsa_dt_lrm;
break;
case C_TIMER_POPPED:
case C_SHUTDOWN:
case C_UNKNOWN:
case C_STARTUP:
crm_crit("Copying %s data (from %s) is not yet implemented",
fsa_cause2string(cause), raised_from);
crmd_exit(CRM_EX_SOFTWARE);
break;
}
}
/* make sure to free it properly later */
if (prepend) {
fsa_message_queue = g_list_prepend(fsa_message_queue, fsa_data);
} else {
fsa_message_queue = g_list_append(fsa_message_queue, fsa_data);
}
crm_trace("FSA message queue length is %d",
g_list_length(fsa_message_queue));
/* fsa_dump_queue(LOG_TRACE); */
if (old_len == g_list_length(fsa_message_queue)) {
crm_err("Couldn't add message to the queue");
}
if (fsa_source && input != I_WAIT_FOR_EVENT) {
crm_trace("Triggering FSA");
mainloop_set_trigger(fsa_source);
}
return last_data_id;
}
void
fsa_dump_queue(int log_level)
{
int offset = 0;
GListPtr lpc = NULL;
for (lpc = fsa_message_queue; lpc != NULL; lpc = lpc->next) {
fsa_data_t *data = (fsa_data_t *) lpc->data;
do_crm_log_unlikely(log_level,
"queue[%d.%d]: input %s raised by %s(%p.%d)\t(cause=%s)",
offset++, data->id, fsa_input2string(data->fsa_input),
data->origin, data->data, data->data_type,
fsa_cause2string(data->fsa_cause));
}
}
ha_msg_input_t *
copy_ha_msg_input(ha_msg_input_t * orig)
{
ha_msg_input_t *copy = calloc(1, sizeof(ha_msg_input_t));
CRM_ASSERT(copy != NULL);
copy->msg = (orig && orig->msg)? copy_xml(orig->msg) : NULL;
copy->xml = get_message_xml(copy->msg, F_CRM_DATA);
return copy;
}
void
delete_fsa_input(fsa_data_t * fsa_data)
{
lrmd_event_data_t *op = NULL;
xmlNode *foo = NULL;
if (fsa_data == NULL) {
return;
}
crm_trace("About to free %s data", fsa_cause2string(fsa_data->fsa_cause));
if (fsa_data->data != NULL) {
switch (fsa_data->data_type) {
case fsa_dt_ha_msg:
delete_ha_msg_input(fsa_data->data);
break;
case fsa_dt_xml:
foo = fsa_data->data;
free_xml(foo);
break;
case fsa_dt_lrm:
op = (lrmd_event_data_t *) fsa_data->data;
lrmd_free_event(op);
break;
case fsa_dt_none:
if (fsa_data->data != NULL) {
crm_err("Don't know how to free %s data from %s",
fsa_cause2string(fsa_data->fsa_cause), fsa_data->origin);
crmd_exit(CRM_EX_SOFTWARE);
}
break;
}
crm_trace("%s data freed", fsa_cause2string(fsa_data->fsa_cause));
}
free(fsa_data);
}
/* returns the next message */
fsa_data_t *
get_message(void)
{
fsa_data_t *message = g_list_nth_data(fsa_message_queue, 0);
fsa_message_queue = g_list_remove(fsa_message_queue, message);
crm_trace("Processing input %d", message->id);
return message;
}
void *
fsa_typed_data_adv(fsa_data_t * fsa_data, enum fsa_data_type a_type, const char *caller)
{
void *ret_val = NULL;
if (fsa_data == NULL) {
crm_err("%s: No FSA data available", caller);
} else if (fsa_data->data == NULL) {
crm_err("%s: No message data available. Origin: %s", caller, fsa_data->origin);
} else if (fsa_data->data_type != a_type) {
crm_crit("%s: Message data was the wrong type! %d vs. requested=%d. Origin: %s",
caller, fsa_data->data_type, a_type, fsa_data->origin);
CRM_ASSERT(fsa_data->data_type == a_type);
} else {
ret_val = fsa_data->data;
}
return ret_val;
}
/* A_MSG_ROUTE */
void
do_msg_route(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state,
enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
ha_msg_input_t *input = fsa_typed_data(fsa_dt_ha_msg);
route_message(msg_data->fsa_cause, input->msg);
}
void
route_message(enum crmd_fsa_cause cause, xmlNode * input)
{
ha_msg_input_t fsa_input;
enum crmd_fsa_input result = I_NULL;
fsa_input.msg = input;
CRM_CHECK(cause == C_IPC_MESSAGE || cause == C_HA_MESSAGE, return);
/* try passing the buck first */
if (relay_message(input, cause == C_IPC_MESSAGE)) {
return;
}
/* handle locally */
result = handle_message(input, cause);
/* done or process later? */
switch (result) {
case I_NULL:
case I_CIB_OP:
case I_ROUTER:
case I_NODE_JOIN:
case I_JOIN_REQUEST:
case I_JOIN_RESULT:
break;
default:
/* Defering local processing of message */
register_fsa_input_later(cause, result, &fsa_input);
return;
}
if (result != I_NULL) {
/* add to the front of the queue */
register_fsa_input(cause, result, &fsa_input);
}
}
gboolean
relay_message(xmlNode * msg, gboolean originated_locally)
{
int dest = 1;
int is_for_dc = 0;
int is_for_dcib = 0;
int is_for_te = 0;
int is_for_crm = 0;
int is_for_cib = 0;
int is_local = 0;
const char *host_to = crm_element_value(msg, F_CRM_HOST_TO);
const char *sys_to = crm_element_value(msg, F_CRM_SYS_TO);
const char *sys_from = crm_element_value(msg, F_CRM_SYS_FROM);
const char *type = crm_element_value(msg, F_TYPE);
const char *task = crm_element_value(msg, F_CRM_TASK);
const char *ref = crm_element_value(msg, XML_ATTR_REFERENCE);
if (ref == NULL) {
ref = "without reference ID";
}
if (msg == NULL) {
crm_warn("Cannot route empty message");
return TRUE;
} else if (safe_str_eq(task, CRM_OP_HELLO)) {
/* quietly ignore */
crm_trace("No routing needed for hello message %s", ref);
return TRUE;
} else if (safe_str_neq(type, T_CRM)) {
crm_warn("Cannot route message %s: Type is '%s' not '" T_CRM "'",
ref, (type? type : "missing"));
crm_log_xml_warn(msg, "[bad message type]");
return TRUE;
} else if (sys_to == NULL) {
crm_warn("Cannot route message %s: No subsystem specified", ref);
crm_log_xml_warn(msg, "[no subsystem]");
return TRUE;
}
is_for_dc = (strcasecmp(CRM_SYSTEM_DC, sys_to) == 0);
is_for_dcib = (strcasecmp(CRM_SYSTEM_DCIB, sys_to) == 0);
is_for_te = (strcasecmp(CRM_SYSTEM_TENGINE, sys_to) == 0);
is_for_cib = (strcasecmp(CRM_SYSTEM_CIB, sys_to) == 0);
is_for_crm = (strcasecmp(CRM_SYSTEM_CRMD, sys_to) == 0);
is_local = 0;
if (host_to == NULL || strlen(host_to) == 0) {
if (is_for_dc || is_for_te) {
is_local = 0;
} else if (is_for_crm) {
if (safe_str_eq(task, CRM_OP_NODE_INFO)) {
/* Node info requests do not specify a host, which is normally
* treated as "all hosts", because the whole point is that the
* client doesn't know the local node name. Always handle these
* requests locally.
*/
is_local = 1;
} else {
is_local = !originated_locally;
}
} else {
is_local = 1;
}
} else if (safe_str_eq(fsa_our_uname, host_to)) {
is_local = 1;
} else if (is_for_crm && safe_str_eq(task, CRM_OP_LRM_DELETE)) {
xmlNode *msg_data = get_message_xml(msg, F_CRM_DATA);
const char *mode = crm_element_value(msg_data, PCMK__XA_MODE);
if (safe_str_eq(mode, XML_TAG_CIB)) {
// Local delete of an offline node's resource history
is_local = 1;
}
}
if (is_for_dc || is_for_dcib || is_for_te) {
if (AM_I_DC && is_for_te) {
crm_trace("Route message %s locally as transition request", ref);
send_msg_via_ipc(msg, sys_to);
} else if (AM_I_DC) {
crm_trace("Route message %s locally as DC request", ref);
return FALSE; // More to be done by caller
} else if (originated_locally && safe_str_neq(sys_from, CRM_SYSTEM_PENGINE)
&& safe_str_neq(sys_from, CRM_SYSTEM_TENGINE)) {
#if SUPPORT_COROSYNC
if (is_corosync_cluster()) {
dest = text2msg_type(sys_to);
}
#endif
crm_trace("Relay message %s to DC", ref);
send_cluster_message(host_to ? crm_get_peer(0, host_to) : NULL, dest, msg, TRUE);
} else {
/* Neither the TE nor the scheduler should be sending messages
* to DCs on other nodes. By definition, if we are no longer the DC,
* then the scheduler's or TE's data should be discarded.
*/
crm_trace("Discard message %s because we are not DC", ref);
}
} else if (is_local && (is_for_crm || is_for_cib)) {
crm_trace("Route message %s locally as controller request", ref);
return FALSE; // More to be done by caller
} else if (is_local) {
crm_trace("Relay message %s locally to %s",
ref, (sys_to? sys_to : "unknown client"));
crm_log_xml_trace(msg, "[IPC relay]");
send_msg_via_ipc(msg, sys_to);
} else {
crm_node_t *node_to = NULL;
#if SUPPORT_COROSYNC
if (is_corosync_cluster()) {
dest = text2msg_type(sys_to);
if (dest == crm_msg_none || dest > crm_msg_stonith_ng) {
dest = crm_msg_crmd;
}
}
#endif
if (host_to) {
node_to = crm_find_peer(0, host_to);
if (node_to == NULL) {
crm_warn("Cannot route message %s: Unknown node %s",
ref, host_to);
return TRUE;
}
crm_trace("Relay message %s to %s",
ref, (node_to->uname? node_to->uname : "peer"));
} else {
crm_trace("Broadcast message %s to all peers", ref);
}
send_cluster_message(host_to ? node_to : NULL, dest, msg, TRUE);
}
return TRUE; // No further processing of message is needed
}
// Return true if field contains a positive integer
static bool
authorize_version(xmlNode *message_data, const char *field,
const char *client_name, const char *ref, const char *uuid)
{
const char *version = crm_element_value(message_data, field);
if (pcmk__str_empty(version)) {
crm_warn("IPC hello from %s rejected: No protocol %s",
CRM_XS " ref=%s uuid=%s",
client_name, field, (ref? ref : "none"), uuid);
return false;
} else {
int version_num = crm_parse_int(version, NULL);
if (version_num < 0) {
crm_warn("IPC hello from %s rejected: Protocol %s '%s' "
"not recognized", CRM_XS " ref=%s uuid=%s",
client_name, field, version, (ref? ref : "none"), uuid);
return false;
}
}
return true;
}
/*!
* \internal
* \brief Check whether a client IPC message is acceptable
*
* If a given client IPC message is a hello, "authorize" it by ensuring it has
* valid information such as a protocol version, and return false indicating
* that nothing further needs to be done with the message. If the message is not
* a hello, just return true to indicate it needs further processing.
*
* \param[in] client_msg XML of IPC message
* \param[in] curr_client If IPC is not proxied, client that sent message
* \param[in] proxy_session If IPC is proxied, the session ID
*
* \return true if message needs further processing, false if it doesn't
*/
bool
controld_authorize_ipc_message(xmlNode *client_msg, pcmk__client_t *curr_client,
const char *proxy_session)
{
xmlNode *message_data = NULL;
const char *client_name = NULL;
const char *op = crm_element_value(client_msg, F_CRM_TASK);
const char *ref = crm_element_value(client_msg, XML_ATTR_REFERENCE);
const char *uuid = (curr_client? curr_client->id : proxy_session);
if (uuid == NULL) {
crm_warn("IPC message from client rejected: No client identifier "
CRM_XS " ref=%s", (ref? ref : "none"));
goto rejected;
}
if (safe_str_neq(CRM_OP_HELLO, op)) {
// Only hello messages need to be authorized
return true;
}
message_data = get_message_xml(client_msg, F_CRM_DATA);
client_name = crm_element_value(message_data, "client_name");
if (pcmk__str_empty(client_name)) {
crm_warn("IPC hello from client rejected: No client name",
CRM_XS " ref=%s uuid=%s", (ref? ref : "none"), uuid);
goto rejected;
}
if (!authorize_version(message_data, "major_version", client_name, ref,
uuid)) {
goto rejected;
}
if (!authorize_version(message_data, "minor_version", client_name, ref,
uuid)) {
goto rejected;
}
crm_trace("Validated IPC hello from client %s", client_name);
if (curr_client) {
curr_client->userdata = strdup(client_name);
}
mainloop_set_trigger(fsa_source);
return false;
rejected:
if (curr_client) {
qb_ipcs_disconnect(curr_client->ipcs);
}
return false;
}
static enum crmd_fsa_input
handle_message(xmlNode *msg, enum crmd_fsa_cause cause)
{
const char *type = NULL;
CRM_CHECK(msg != NULL, return I_NULL);
type = crm_element_value(msg, F_CRM_MSG_TYPE);
if (crm_str_eq(type, XML_ATTR_REQUEST, TRUE)) {
return handle_request(msg, cause);
} else if (crm_str_eq(type, XML_ATTR_RESPONSE, TRUE)) {
handle_response(msg);
return I_NULL;
}
crm_err("Unknown message type: %s", type);
return I_NULL;
}
static enum crmd_fsa_input
handle_failcount_op(xmlNode * stored_msg)
{
const char *rsc = NULL;
const char *uname = NULL;
const char *op = NULL;
char *interval_spec = NULL;
guint interval_ms = 0;
gboolean is_remote_node = FALSE;
xmlNode *xml_op = get_message_xml(stored_msg, F_CRM_DATA);
if (xml_op) {
xmlNode *xml_rsc = first_named_child(xml_op, XML_CIB_TAG_RESOURCE);
xmlNode *xml_attrs = first_named_child(xml_op, XML_TAG_ATTRS);
if (xml_rsc) {
rsc = ID(xml_rsc);
}
if (xml_attrs) {
op = crm_element_value(xml_attrs,
CRM_META "_" XML_RSC_ATTR_CLEAR_OP);
crm_element_value_ms(xml_attrs,
CRM_META "_" XML_RSC_ATTR_CLEAR_INTERVAL,
&interval_ms);
}
}
uname = crm_element_value(xml_op, XML_LRM_ATTR_TARGET);
if ((rsc == NULL) || (uname == NULL)) {
crm_log_xml_warn(stored_msg, "invalid failcount op");
return I_NULL;
}
if (crm_element_value(xml_op, XML_LRM_ATTR_ROUTER_NODE)) {
is_remote_node = TRUE;
}
if (interval_ms) {
interval_spec = crm_strdup_printf("%ums", interval_ms);
}
update_attrd_clear_failures(uname, rsc, op, interval_spec, is_remote_node);
free(interval_spec);
lrm_clear_last_failure(rsc, uname, op, interval_ms);
return I_NULL;
}
static enum crmd_fsa_input
handle_lrm_delete(xmlNode *stored_msg)
{
const char *mode = NULL;
xmlNode *msg_data = get_message_xml(stored_msg, F_CRM_DATA);
CRM_CHECK(msg_data != NULL, return I_NULL);
/* CRM_OP_LRM_DELETE has two distinct modes. The default behavior is to
* relay the operation to the affected node, which will unregister the
* resource from the local executor, clear the resource's history from the
* CIB, and do some bookkeeping in the controller.
*
* However, if the affected node is offline, the client will specify
* mode="cib" which means the controller receiving the operation should
* clear the resource's history from the CIB and nothing else. This is used
* to clear shutdown locks.
*/
mode = crm_element_value(msg_data, PCMK__XA_MODE);
if ((mode == NULL) || strcmp(mode, XML_TAG_CIB)) {
// Relay to affected node
crm_xml_add(stored_msg, F_CRM_SYS_TO, CRM_SYSTEM_LRMD);
return I_ROUTER;
} else {
// Delete CIB history locally (compare with do_lrm_delete())
const char *from_sys = NULL;
const char *user_name = NULL;
const char *rsc_id = NULL;
const char *node = NULL;
xmlNode *rsc_xml = NULL;
int rc = pcmk_rc_ok;
rsc_xml = first_named_child(msg_data, XML_CIB_TAG_RESOURCE);
CRM_CHECK(rsc_xml != NULL, return I_NULL);
rsc_id = ID(rsc_xml);
from_sys = crm_element_value(stored_msg, F_CRM_SYS_FROM);
node = crm_element_value(msg_data, XML_LRM_ATTR_TARGET);
#if ENABLE_ACL
user_name = pcmk__update_acl_user(stored_msg, F_CRM_USER, NULL);
#endif
crm_debug("Handling " CRM_OP_LRM_DELETE " for %s on %s locally%s%s "
"(clearing CIB resource history only)", rsc_id, node,
(user_name? " for user " : ""), (user_name? user_name : ""));
#if ENABLE_ACL
rc = controld_delete_resource_history(rsc_id, node, user_name,
cib_dryrun|cib_sync_call);
#endif
if (rc == pcmk_rc_ok) {
rc = controld_delete_resource_history(rsc_id, node, user_name,
crmd_cib_smart_opt());
}
// Notify client if not from graph (compare with notify_deleted())
if (from_sys && strcmp(from_sys, CRM_SYSTEM_TENGINE)) {
lrmd_event_data_t *op = NULL;
const char *from_host = crm_element_value(stored_msg,
F_CRM_HOST_FROM);
const char *transition = crm_element_value(msg_data,
XML_ATTR_TRANSITION_KEY);
crm_info("Notifying %s on %s that %s was%s deleted",
from_sys, (from_host? from_host : "local node"), rsc_id,
((rc == pcmk_rc_ok)? "" : " not"));
op = lrmd_new_event(rsc_id, CRMD_ACTION_DELETE, 0);
op->type = lrmd_event_exec_complete;
op->user_data = strdup(transition? transition : FAKE_TE_ID);
op->params = crm_str_table_new();
g_hash_table_insert(op->params, strdup(XML_ATTR_CRM_VERSION),
strdup(CRM_FEATURE_SET));
controld_rc2event(op, rc);
controld_ack_event_directly(from_host, from_sys, NULL, op, rsc_id);
lrmd_free_event(op);
controld_trigger_delete_refresh(from_sys, rsc_id);
}
return I_NULL;
}
}
/*!
* \brief Handle a CRM_OP_REMOTE_STATE message by updating remote peer cache
*
* \param[in] msg Message XML
*
* \return Next FSA input
*/
static enum crmd_fsa_input
handle_remote_state(xmlNode *msg)
{
const char *remote_uname = ID(msg);
const char *remote_is_up = crm_element_value(msg, XML_NODE_IN_CLUSTER);
crm_node_t *remote_peer;
CRM_CHECK(remote_uname && remote_is_up, return I_NULL);
remote_peer = crm_remote_peer_get(remote_uname);
CRM_CHECK(remote_peer, return I_NULL);
crm_update_peer_state(__FUNCTION__, remote_peer,
crm_is_true(remote_is_up)?
CRM_NODE_MEMBER : CRM_NODE_LOST, 0);
return I_NULL;
}
/*!
* \brief Handle a CRM_OP_PING message
*
* \param[in] msg Message XML
*
* \return Next FSA input
*/
static enum crmd_fsa_input
handle_ping(xmlNode *msg)
{
const char *value = NULL;
xmlNode *ping = NULL;
// Build reply
ping = create_xml_node(NULL, XML_CRM_TAG_PING);
value = crm_element_value(msg, F_CRM_SYS_TO);
crm_xml_add(ping, XML_PING_ATTR_SYSFROM, value);
// Add controller state
value = fsa_state2string(fsa_state);
crm_xml_add(ping, XML_PING_ATTR_CRMDSTATE, value);
crm_notice("Current ping state: %s", value); // CTS needs this
// Add controller health
// @TODO maybe do some checks to determine meaningful status
crm_xml_add(ping, XML_PING_ATTR_STATUS, "ok");
// Send reply
msg = create_reply(msg, ping);
free_xml(ping);
if (msg) {
(void) relay_message(msg, TRUE);
free_xml(msg);
}
// Nothing further to do
return I_NULL;
}
/*!
* \brief Handle a CRM_OP_NODE_INFO request
*
* \param[in] msg Message XML
*
* \return Next FSA input
*/
static enum crmd_fsa_input
handle_node_info_request(xmlNode *msg)
{
const char *value = NULL;
crm_node_t *node = NULL;
int node_id = 0;
xmlNode *reply = NULL;
// Build reply
reply = create_xml_node(NULL, XML_CIB_TAG_NODE);
crm_xml_add(reply, XML_PING_ATTR_SYSFROM, CRM_SYSTEM_CRMD);
// Add whether current partition has quorum
crm_xml_add_boolean(reply, XML_ATTR_HAVE_QUORUM, fsa_has_quorum);
// Check whether client requested node info by ID and/or name
crm_element_value_int(msg, XML_ATTR_ID, &node_id);
if (node_id < 0) {
node_id = 0;
}
value = crm_element_value(msg, XML_ATTR_UNAME);
// Default to local node if none given
if ((node_id == 0) && (value == NULL)) {
value = fsa_our_uname;
}
node = crm_find_peer_full(node_id, value, CRM_GET_PEER_ANY);
if (node) {
crm_xml_add_int(reply, XML_ATTR_ID, node->id);
crm_xml_add(reply, XML_ATTR_UUID, node->uuid);
crm_xml_add(reply, XML_ATTR_UNAME, node->uname);
crm_xml_add(reply, XML_NODE_IS_PEER, node->state);
crm_xml_add_boolean(reply, XML_NODE_IS_REMOTE,
node->flags & crm_remote_node);
}
// Send reply
msg = create_reply(msg, reply);
free_xml(reply);
if (msg) {
(void) relay_message(msg, TRUE);
free_xml(msg);
}
// Nothing further to do
return I_NULL;
}
static void
verify_feature_set(xmlNode *msg)
{
const char *dc_version = crm_element_value(msg, XML_ATTR_CRM_VERSION);
if (dc_version == NULL) {
/* All we really know is that the DC feature set is older than 3.1.0,
* but that's also all that really matters.
*/
dc_version = "3.0.14";
}
if (feature_set_compatible(dc_version, CRM_FEATURE_SET)) {
crm_trace("Local feature set (%s) is compatible with DC's (%s)",
CRM_FEATURE_SET, dc_version);
} else {
crm_err("Local feature set (%s) is incompatible with DC's (%s)",
CRM_FEATURE_SET, dc_version);
// Nothing is likely to improve without administrator involvement
set_bit(fsa_input_register, R_STAYDOWN);
crmd_exit(CRM_EX_FATAL);
}
}
+// DC gets own shutdown all-clear
+static enum crmd_fsa_input
+handle_shutdown_self_ack(xmlNode *stored_msg)
+{
+ const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM);
+
+ if (is_set(fsa_input_register, R_SHUTDOWN)) {
+ // The expected case -- we initiated own shutdown sequence
+ crm_info("Shutting down controller");
+ return I_STOP;
+ }
+
+ if (safe_str_eq(host_from, fsa_our_dc)) {
+ // Must be logic error -- DC confirming its own unrequested shutdown
+ crm_err("Shutting down controller immediately due to "
+ "unexpected shutdown confirmation");
+ return I_TERMINATE;
+ }
+
+ if (fsa_state != S_STOPPING) {
+ // Shouldn't happen -- non-DC confirming unrequested shutdown
+ crm_err("Starting new DC election because %s is "
+ "confirming shutdown we did not request",
+ (host_from? host_from : "another node"));
+ return I_ELECTION;
+ }
+
+ // Shouldn't happen, but we are already stopping anyway
+ crm_debug("Ignoring unexpected shutdown confirmation from %s",
+ (host_from? host_from : "another node"));
+ return I_NULL;
+}
+
+// Non-DC gets shutdown all-clear from DC
+static enum crmd_fsa_input
+handle_shutdown_ack(xmlNode *stored_msg)
+{
+ const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM);
+
+ if (host_from == NULL) {
+ crm_warn("Ignoring shutdown request without origin specified");
+ return I_NULL;
+ }
+
+ if ((fsa_our_dc == NULL) || (strcmp(host_from, fsa_our_dc) == 0)) {
+
+ if (is_set(fsa_input_register, R_SHUTDOWN)) {
+ crm_info("Shutting down controller after confirmation from %s",
+ host_from);
+ } else {
+ crm_err("Shutting down controller after unexpected "
+ "shutdown request from %s", host_from);
+ set_bit(fsa_input_register, R_STAYDOWN);
+ }
+ return I_STOP;
+ }
+
+ crm_warn("Ignoring shutdown request from %s because DC is %s",
+ host_from, fsa_our_dc);
+ return I_NULL;
+}
+
static enum crmd_fsa_input
handle_request(xmlNode *stored_msg, enum crmd_fsa_cause cause)
{
xmlNode *msg = NULL;
const char *op = crm_element_value(stored_msg, F_CRM_TASK);
/* Optimize this for the DC - it has the most to do */
if (op == NULL) {
crm_log_xml_warn(stored_msg, "[request without " F_CRM_TASK "]");
return I_NULL;
}
if (strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0) {
const char *from = crm_element_value(stored_msg, F_CRM_HOST_FROM);
crm_node_t *node = crm_find_peer(0, from);
crm_update_peer_expected(__FUNCTION__, node, CRMD_JOINSTATE_DOWN);
if(AM_I_DC == FALSE) {
return I_NULL; /* Done */
}
}
/*========== DC-Only Actions ==========*/
if (AM_I_DC) {
if (strcmp(op, CRM_OP_JOIN_ANNOUNCE) == 0) {
return I_NODE_JOIN;
} else if (strcmp(op, CRM_OP_JOIN_REQUEST) == 0) {
return I_JOIN_REQUEST;
} else if (strcmp(op, CRM_OP_JOIN_CONFIRM) == 0) {
return I_JOIN_RESULT;
} else if (strcmp(op, CRM_OP_SHUTDOWN) == 0) {
- const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM);
- gboolean dc_match = safe_str_eq(host_from, fsa_our_dc);
-
- if (is_set(fsa_input_register, R_SHUTDOWN)) {
- crm_info("Shutting ourselves down (DC)");
- return I_STOP;
-
- } else if (dc_match) {
- crm_err("We didn't ask to be shut down, yet our"
- " TE is telling us to. Better get out now!");
- return I_TERMINATE;
-
- } else if (fsa_state != S_STOPPING) {
- crm_err("Another node is asking us to shutdown" " but we think we're ok.");
- return I_ELECTION;
- }
+ return handle_shutdown_self_ack(stored_msg);
} else if (strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0) {
/* a slave wants to shut down */
/* create cib fragment and add to message */
return handle_shutdown_request(stored_msg);
} else if (strcmp(op, CRM_OP_REMOTE_STATE) == 0) {
/* a remote connection host is letting us know the node state */
return handle_remote_state(stored_msg);
}
}
/*========== common actions ==========*/
if (strcmp(op, CRM_OP_NOVOTE) == 0) {
ha_msg_input_t fsa_input;
fsa_input.msg = stored_msg;
register_fsa_input_adv(C_HA_MESSAGE, I_NULL, &fsa_input,
A_ELECTION_COUNT | A_ELECTION_CHECK, FALSE, __FUNCTION__);
} else if (strcmp(op, CRM_OP_THROTTLE) == 0) {
throttle_update(stored_msg);
if (AM_I_DC && transition_graph != NULL) {
if (transition_graph->complete == FALSE) {
crm_debug("The throttle changed. Trigger a graph.");
trigger_graph();
}
}
return I_NULL;
} else if (strcmp(op, CRM_OP_CLEAR_FAILCOUNT) == 0) {
return handle_failcount_op(stored_msg);
} else if (strcmp(op, CRM_OP_VOTE) == 0) {
/* count the vote and decide what to do after that */
ha_msg_input_t fsa_input;
fsa_input.msg = stored_msg;
register_fsa_input_adv(C_HA_MESSAGE, I_NULL, &fsa_input,
A_ELECTION_COUNT | A_ELECTION_CHECK, FALSE, __FUNCTION__);
/* Sometimes we _must_ go into S_ELECTION */
if (fsa_state == S_HALT) {
crm_debug("Forcing an election from S_HALT");
return I_ELECTION;
#if 0
} else if (AM_I_DC) {
/* This is the old way of doing things but what is gained? */
return I_ELECTION;
#endif
}
} else if (strcmp(op, CRM_OP_JOIN_OFFER) == 0) {
verify_feature_set(stored_msg);
crm_debug("Raising I_JOIN_OFFER: join-%s", crm_element_value(stored_msg, F_CRM_JOIN_ID));
return I_JOIN_OFFER;
} else if (strcmp(op, CRM_OP_JOIN_ACKNAK) == 0) {
crm_debug("Raising I_JOIN_RESULT: join-%s", crm_element_value(stored_msg, F_CRM_JOIN_ID));
return I_JOIN_RESULT;
} else if (strcmp(op, CRM_OP_LRM_DELETE) == 0) {
return handle_lrm_delete(stored_msg);
} else if ((strcmp(op, CRM_OP_LRM_FAIL) == 0)
|| (strcmp(op, CRM_OP_LRM_REFRESH) == 0)
|| (strcmp(op, CRM_OP_REPROBE) == 0)) {
crm_xml_add(stored_msg, F_CRM_SYS_TO, CRM_SYSTEM_LRMD);
return I_ROUTER;
} else if (strcmp(op, CRM_OP_NOOP) == 0) {
return I_NULL;
} else if (strcmp(op, CRM_OP_LOCAL_SHUTDOWN) == 0) {
crm_shutdown(SIGTERM);
/*return I_SHUTDOWN; */
return I_NULL;
} else if (strcmp(op, CRM_OP_PING) == 0) {
return handle_ping(stored_msg);
} else if (strcmp(op, CRM_OP_NODE_INFO) == 0) {
return handle_node_info_request(stored_msg);
} else if (strcmp(op, CRM_OP_RM_NODE_CACHE) == 0) {
int id = 0;
const char *name = NULL;
crm_element_value_int(stored_msg, XML_ATTR_ID, &id);
name = crm_element_value(stored_msg, XML_ATTR_UNAME);
if(cause == C_IPC_MESSAGE) {
msg = create_request(CRM_OP_RM_NODE_CACHE, NULL, NULL, CRM_SYSTEM_CRMD, CRM_SYSTEM_CRMD, NULL);
if (send_cluster_message(NULL, crm_msg_crmd, msg, TRUE) == FALSE) {
crm_err("Could not instruct peers to remove references to node %s/%u", name, id);
} else {
crm_notice("Instructing peers to remove references to node %s/%u", name, id);
}
free_xml(msg);
} else {
reap_crm_member(id, name);
/* If we're forgetting this node, also forget any failures to fence
* it, so we don't carry that over to any node added later with the
* same name.
*/
st_fail_count_reset(name);
}
} else if (strcmp(op, CRM_OP_MAINTENANCE_NODES) == 0) {
xmlNode *xml = get_message_xml(stored_msg, F_CRM_DATA);
remote_ra_process_maintenance_nodes(xml);
/*========== (NOT_DC)-Only Actions ==========*/
- } else if (AM_I_DC == FALSE && strcmp(op, CRM_OP_SHUTDOWN) == 0) {
+ } else if (!AM_I_DC) {
- const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM);
- gboolean dc_match = safe_str_eq(host_from, fsa_our_dc);
-
- if (dc_match || fsa_our_dc == NULL) {
- if (is_set(fsa_input_register, R_SHUTDOWN) == FALSE) {
- crm_err("We didn't ask to be shut down, yet our DC is telling us to.");
- set_bit(fsa_input_register, R_STAYDOWN);
- return I_STOP;
- }
- crm_info("Shutting down");
- return I_STOP;
-
- } else {
- crm_warn("Discarding %s op from %s", op, host_from);
+ if (strcmp(op, CRM_OP_SHUTDOWN) == 0) {
+ return handle_shutdown_ack(stored_msg);
}
} else {
crm_err("Unexpected request (%s) sent to %s", op, AM_I_DC ? "the DC" : "non-DC node");
crm_log_xml_err(stored_msg, "Unexpected");
}
return I_NULL;
}
static void
handle_response(xmlNode *stored_msg)
{
const char *op = crm_element_value(stored_msg, F_CRM_TASK);
if (op == NULL) {
crm_log_xml_err(stored_msg, "Bad message");
} else if (AM_I_DC && strcmp(op, CRM_OP_PECALC) == 0) {
// Check whether scheduler answer been superseded by subsequent request
const char *msg_ref = crm_element_value(stored_msg, XML_ATTR_REFERENCE);
if (msg_ref == NULL) {
crm_err("%s - Ignoring calculation with no reference", op);
} else if (safe_str_eq(msg_ref, fsa_pe_ref)) {
ha_msg_input_t fsa_input;
controld_stop_sched_timer();
fsa_input.msg = stored_msg;
register_fsa_input_later(C_IPC_MESSAGE, I_PE_SUCCESS, &fsa_input);
} else {
crm_info("%s calculation %s is obsolete", op, msg_ref);
}
} else if (strcmp(op, CRM_OP_VOTE) == 0
|| strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0 || strcmp(op, CRM_OP_SHUTDOWN) == 0) {
} else {
const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM);
crm_err("Unexpected response (op=%s, src=%s) sent to the %s",
op, host_from, AM_I_DC ? "DC" : "controller");
}
}
static enum crmd_fsa_input
handle_shutdown_request(xmlNode * stored_msg)
{
/* handle here to avoid potential version issues
* where the shutdown message/procedure may have
* been changed in later versions.
*
* This way the DC is always in control of the shutdown
*/
char *now_s = NULL;
time_t now = time(NULL);
const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM);
if (host_from == NULL) {
/* we're shutting down and the DC */
host_from = fsa_our_uname;
}
crm_info("Creating shutdown request for %s (state=%s)", host_from, fsa_state2string(fsa_state));
crm_log_xml_trace(stored_msg, "message");
now_s = crm_itoa(now);
update_attrd(host_from, XML_CIB_ATTR_SHUTDOWN, now_s, NULL, FALSE);
free(now_s);
/* will be picked up by the TE as long as its running */
return I_NULL;
}
/* msg is deleted by the time this returns */
extern gboolean process_te_message(xmlNode * msg, xmlNode * xml_data);
static void
send_msg_via_ipc(xmlNode * msg, const char *sys)
{
pcmk__client_t *client_channel = pcmk__find_client_by_id(sys);
if (crm_element_value(msg, F_CRM_HOST_FROM) == NULL) {
crm_xml_add(msg, F_CRM_HOST_FROM, fsa_our_uname);
}
if (client_channel != NULL) {
/* Transient clients such as crmadmin */
pcmk__ipc_send_xml(client_channel, 0, msg, crm_ipc_server_event);
} else if (sys != NULL && strcmp(sys, CRM_SYSTEM_TENGINE) == 0) {
xmlNode *data = get_message_xml(msg, F_CRM_DATA);
process_te_message(msg, data);
} else if (sys != NULL && strcmp(sys, CRM_SYSTEM_LRMD) == 0) {
fsa_data_t fsa_data;
ha_msg_input_t fsa_input;
fsa_input.msg = msg;
fsa_input.xml = get_message_xml(msg, F_CRM_DATA);
fsa_data.id = 0;
fsa_data.actions = 0;
fsa_data.data = &fsa_input;
fsa_data.fsa_input = I_MESSAGE;
fsa_data.fsa_cause = C_IPC_MESSAGE;
fsa_data.origin = __FUNCTION__;
fsa_data.data_type = fsa_dt_ha_msg;
do_lrm_invoke(A_LRM_INVOKE, C_IPC_MESSAGE, fsa_state, I_MESSAGE, &fsa_data);
} else if (sys != NULL && crmd_is_proxy_session(sys)) {
crmd_proxy_send(sys, msg);
} else {
crm_debug("Unknown Sub-system (%s)... discarding message.", crm_str(sys));
}
}
void
delete_ha_msg_input(ha_msg_input_t * orig)
{
if (orig == NULL) {
return;
}
free_xml(orig->msg);
free(orig);
}
/*!
* \internal
* \brief Notify the DC of a remote node state change
*
* \param[in] node_name Node's name
* \param[in] node_up TRUE if node is up, FALSE if down
*/
void
send_remote_state_message(const char *node_name, gboolean node_up)
{
/* If we don't have a DC, or the message fails, we have a failsafe:
* the DC will eventually pick up the change via the CIB node state.
* The message allows it to happen sooner if possible.
*/
if (fsa_our_dc) {
xmlNode *msg = create_request(CRM_OP_REMOTE_STATE, NULL, fsa_our_dc,
CRM_SYSTEM_DC, CRM_SYSTEM_CRMD, NULL);
crm_info("Notifying DC %s of pacemaker_remote node %s %s",
fsa_our_dc, node_name, (node_up? "coming up" : "going down"));
crm_xml_add(msg, XML_ATTR_ID, node_name);
crm_xml_add_boolean(msg, XML_NODE_IN_CLUSTER, node_up);
send_cluster_message(crm_get_peer(0, fsa_our_dc), crm_msg_crmd, msg,
TRUE);
free_xml(msg);
} else {
crm_debug("No DC to notify of pacemaker_remote node %s %s",
node_name, (node_up? "coming up" : "going down"));
}
}
diff --git a/include/crm/common/iso8601.h b/include/crm/common/iso8601.h
index 13b5453d38..47c15d6cc0 100644
--- a/include/crm/common/iso8601.h
+++ b/include/crm/common/iso8601.h
@@ -1,121 +1,122 @@
/*
- * Copyright 2005-2019 the Pacemaker project contributors
+ * Copyright 2005-2020 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.
*/
#ifndef CRM_COMMON_ISO8601
# define CRM_COMMON_ISO8601
#ifdef __cplusplus
extern "C" {
#endif
/**
* \file
* \brief ISO_8601 Date handling
* \ingroup date
*/
/*
* See https://en.wikipedia.org/wiki/ISO_8601
*/
# include <time.h>
# include <ctype.h>
# include <stdint.h> // uint32_t
# include <stdbool.h> // bool
typedef struct crm_time_s crm_time_t;
typedef struct crm_time_period_s {
crm_time_t *start;
crm_time_t *end;
crm_time_t *diff;
} crm_time_period_t;
/* Creates a new date/time object conforming to ISO 8601, for example:
* Ordinal: 2010-01 12:00:00 +10:00
* Gregorian: 2010-01-01 12:00:00 +10:00
* ISO Week: 2010-W53-6 12:00:00 +10:00
*
* Notes:
* Only one of date, time is required
* If date or timezone is unspecified, they default to the current one
* Supplying NULL results in the current date/time
* Dashes may be omitted from dates
* Colons may be omitted from times and timezones
* A timezone of 'Z' denotes UTC time
*/
crm_time_t *crm_time_new(const char *string);
crm_time_t *crm_time_new_undefined(void);
void crm_time_free(crm_time_t * dt);
bool crm_time_is_defined(const crm_time_t *t);
char *crm_time_as_string(crm_time_t * dt, int flags);
# define crm_time_log(level, prefix, dt, flags) crm_time_log_alias(level, __FILE__, __FUNCTION__, __LINE__, prefix, dt, flags)
void crm_time_log_alias(int log_level, const char *file, const char *function, int line,
const char *prefix, crm_time_t * date_time, int flags);
# define crm_time_log_date 0x001
# define crm_time_log_timeofday 0x002
# define crm_time_log_with_timezone 0x004
# define crm_time_log_duration 0x008
# define crm_time_ordinal 0x010
# define crm_time_weeks 0x020
# define crm_time_seconds 0x100
# define crm_time_epoch 0x200
crm_time_t *crm_time_parse_duration(const char *duration_str);
crm_time_t *crm_time_calculate_duration(crm_time_t * dt, crm_time_t * value);
crm_time_period_t *crm_time_parse_period(const char *period_str);
void crm_time_free_period(crm_time_period_t *period);
int crm_time_compare(crm_time_t * dt, crm_time_t * rhs);
int crm_time_get_timeofday(crm_time_t * dt, uint32_t * h, uint32_t * m, uint32_t * s);
int crm_time_get_timezone(crm_time_t * dt, uint32_t * h, uint32_t * m);
int crm_time_get_gregorian(crm_time_t * dt, uint32_t * y, uint32_t * m, uint32_t * d);
int crm_time_get_ordinal(crm_time_t * dt, uint32_t * y, uint32_t * d);
int crm_time_get_isoweek(crm_time_t * dt, uint32_t * y, uint32_t * w, uint32_t * d);
/* Time in seconds since 0000-01-01 00:00:00Z */
long long int crm_time_get_seconds(crm_time_t * dt);
/* Time in seconds since 1970-01-01 00:00:00Z */
long long int crm_time_get_seconds_since_epoch(crm_time_t * dt);
void crm_time_set(crm_time_t * target, crm_time_t * source);
void crm_time_set_timet(crm_time_t * target, time_t * source);
/* Returns a new time object */
+crm_time_t *pcmk_copy_time(crm_time_t *source);
crm_time_t *crm_time_add(crm_time_t * dt, crm_time_t * value);
crm_time_t *crm_time_subtract(crm_time_t * dt, crm_time_t * value);
/* All crm_time_add_... functions support negative values */
void crm_time_add_seconds(crm_time_t * dt, int value);
void crm_time_add_minutes(crm_time_t * dt, int value);
void crm_time_add_hours(crm_time_t * dt, int value);
void crm_time_add_days(crm_time_t * dt, int value);
void crm_time_add_weeks(crm_time_t * dt, int value);
void crm_time_add_months(crm_time_t * dt, int value);
void crm_time_add_years(crm_time_t * dt, int value);
/* Useful helper functions */
int crm_time_january1_weekday(int year);
int crm_time_weeks_in_year(int year);
int crm_time_days_in_month(int month, int year);
bool crm_time_leapyear(int year);
bool crm_time_check(crm_time_t * dt);
#ifdef __cplusplus
}
#endif
#endif
diff --git a/include/crm/common/util.h b/include/crm/common/util.h
index de6808c210..93fb9767ee 100644
--- a/include/crm/common/util.h
+++ b/include/crm/common/util.h
@@ -1,228 +1,229 @@
/*
* Copyright 2004-2020 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.
*/
#ifndef CRM_COMMON_UTIL__H
# define CRM_COMMON_UTIL__H
#ifdef __cplusplus
extern "C" {
#endif
/**
* \file
* \brief Utility functions
* \ingroup core
*/
# include <sys/types.h> // gid_t, mode_t, size_t, time_t, uid_t
# include <stdlib.h>
# include <stdbool.h>
# include <stdint.h> // uint32_t
# include <limits.h>
# include <signal.h>
# include <glib.h>
# include <libxml/tree.h>
# include <crm/lrmd.h>
# include <crm/common/acl.h>
# include <crm/common/results.h>
# define ONLINESTATUS "online" // Status of an online client
# define OFFLINESTATUS "offline" // Status of an offline client
// public name/value pair functions (from nvpair.c)
int pcmk_scan_nvpair(const char *input, char **name, char **value);
char *pcmk_format_nvpair(const char *name, const char *value, const char *units);
char *pcmk_format_named_time(const char *name, time_t epoch_time);
/* public Pacemaker Remote functions (from remote.c) */
int crm_default_remote_port(void);
/* public string functions (from strings.c) */
char *crm_itoa_stack(int an_int, char *buf, size_t len);
gboolean crm_is_true(const char *s);
int crm_str_to_boolean(const char *s, int *ret);
long long crm_parse_ll(const char *text, const char *default_text);
int crm_parse_int(const char *text, const char *default_text);
long long crm_get_msec(const char *input);
char * crm_strip_trailing_newline(char *str);
gboolean crm_str_eq(const char *a, const char *b, gboolean use_case);
gboolean safe_str_neq(const char *a, const char *b);
gboolean crm_strcase_equal(gconstpointer a, gconstpointer b);
guint crm_strcase_hash(gconstpointer v);
guint g_str_hash_traditional(gconstpointer v);
char *crm_strdup_printf(char const *format, ...) __attribute__ ((__format__ (__printf__, 1, 2)));
int pcmk__parse_ll_range(const char *srcstring, long long *start, long long *end);
# define safe_str_eq(a, b) crm_str_eq(a, b, FALSE)
# define crm_str_hash g_str_hash_traditional
static inline char *
crm_itoa(int an_int)
{
return crm_strdup_printf("%d", an_int);
}
static inline char *
crm_ftoa(double a_float)
{
return crm_strdup_printf("%f", a_float);
}
static inline char *
crm_ttoa(time_t epoch_time)
{
return crm_strdup_printf("%lld", (long long) epoch_time);
}
/*!
* \brief Create hash table with dynamically allocated string keys/values
*
* \return Newly allocated hash table
* \note It is the caller's responsibility to free the result, using
* g_hash_table_destroy().
*/
static inline GHashTable *
crm_str_table_new()
{
return g_hash_table_new_full(crm_str_hash, g_str_equal, free, free);
}
/*!
* \brief Create hash table with case-insensitive dynamically allocated string keys/values
*
* \return Newly allocated hash table
* \note It is the caller's responsibility to free the result, using
* g_hash_table_destroy().
*/
static inline GHashTable *
crm_strcase_table_new()
{
return g_hash_table_new_full(crm_strcase_hash, crm_strcase_equal, free, free);
}
GHashTable *crm_str_table_dup(GHashTable *old_table);
# define crm_atoi(text, default_text) crm_parse_int(text, default_text)
/* public I/O functions (from io.c) */
void crm_build_path(const char *path_c, mode_t mode);
guint crm_parse_interval_spec(const char *input);
int char2score(const char *score);
char *score2char(int score);
char *score2char_stack(int score, char *buf, size_t len);
/* public operation functions (from operations.c) */
gboolean parse_op_key(const char *key, char **rsc_id, char **op_type,
guint *interval_ms);
gboolean decode_transition_key(const char *key, char **uuid, int *transition_id,
int *action_id, int *target_rc);
gboolean decode_transition_magic(const char *magic, char **uuid,
int *transition_id, int *action_id,
int *op_status, int *op_rc, int *target_rc);
int rsc_op_expected_rc(lrmd_event_data_t *event);
gboolean did_rsc_op_fail(lrmd_event_data_t *event, int target_rc);
bool crm_op_needs_metadata(const char *rsc_class, const char *op);
xmlNode *crm_create_op_xml(xmlNode *parent, const char *prefix,
const char *task, const char *interval_spec,
const char *timeout);
#define CRM_DEFAULT_OP_TIMEOUT_S "20s"
// Public resource agent functions (from agents.c)
// Capabilities supported by a resource agent standard
enum pcmk_ra_caps {
- pcmk_ra_cap_none = 0x000,
- pcmk_ra_cap_provider = 0x001, // Requires provider
- pcmk_ra_cap_status = 0x002, // Supports status instead of monitor
- pcmk_ra_cap_params = 0x004, // Supports parameters
- pcmk_ra_cap_unique = 0x008, // Supports unique clones
- pcmk_ra_cap_promotable = 0x010, // Supports promotable clones
+ pcmk_ra_cap_none = 0,
+ pcmk_ra_cap_provider = (1 << 0), // Requires provider
+ pcmk_ra_cap_status = (1 << 1), // Supports status instead of monitor
+ pcmk_ra_cap_params = (1 << 2), // Supports parameters
+ pcmk_ra_cap_unique = (1 << 3), // Supports unique clones
+ pcmk_ra_cap_promotable = (1 << 4), // Supports promotable clones
+ pcmk_ra_cap_stdin = (1 << 5), // Reads from standard input
};
uint32_t pcmk_get_ra_caps(const char *standard);
char *crm_generate_ra_key(const char *standard, const char *provider,
const char *type);
int crm_parse_agent_spec(const char *spec, char **standard, char **provider,
char **type);
int compare_version(const char *version1, const char *version2);
/* coverity[+kill] */
void crm_abort(const char *file, const char *function, int line,
const char *condition, gboolean do_core, gboolean do_fork);
static inline gboolean
is_not_set(long long word, long long bit)
{
return ((word & bit) == 0);
}
static inline gboolean
is_set(long long word, long long bit)
{
return ((word & bit) == bit);
}
static inline gboolean
is_set_any(long long word, long long bit)
{
return ((word & bit) != 0);
}
static inline guint
crm_hash_table_size(GHashTable * hashtable)
{
if (hashtable == NULL) {
return 0;
}
return g_hash_table_size(hashtable);
}
char *crm_meta_name(const char *field);
const char *crm_meta_value(GHashTable * hash, const char *field);
char *crm_md5sum(const char *buffer);
char *crm_generate_uuid(void);
bool crm_is_daemon_name(const char *name);
int crm_user_lookup(const char *name, uid_t * uid, gid_t * gid);
int pcmk_daemon_user(uid_t *uid, gid_t *gid);
#ifdef HAVE_GNUTLS_GNUTLS_H
void crm_gnutls_global_init(void);
#endif
char *pcmk_hostname(void);
bool pcmk_str_is_infinity(const char *s);
bool pcmk_str_is_minus_infinity(const char *s);
#ifndef PCMK__NO_COMPAT
/* Everything here is deprecated and kept only for public API backward
* compatibility. It will be moved to compatibility.h when 2.1.0 is released.
*/
//! \deprecated Use crm_parse_interval_spec() instead
#define crm_get_interval crm_parse_interval_spec
//! \deprecated Use pcmk_get_ra_caps() instead
bool crm_provider_required(const char *standard);
#endif // PCMK__NO_COMPAT
#ifdef __cplusplus
}
#endif
#endif
diff --git a/include/crm/pengine/internal.h b/include/crm/pengine/internal.h
index cc0f168fd7..4e1a42e8a7 100644
--- a/include/crm/pengine/internal.h
+++ b/include/crm/pengine/internal.h
@@ -1,447 +1,473 @@
/*
* Copyright 2004-2020 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.
*/
#ifndef PE_INTERNAL__H
# define PE_INTERNAL__H
# include <string.h>
# include <crm/pengine/status.h>
# include <crm/pengine/remote_internal.h>
# include <crm/common/output.h>
# define pe_rsc_info(rsc, fmt, args...) crm_log_tag(LOG_INFO, rsc ? rsc->id : "<NULL>", fmt, ##args)
# define pe_rsc_debug(rsc, fmt, args...) crm_log_tag(LOG_DEBUG, rsc ? rsc->id : "<NULL>", fmt, ##args)
# define pe_rsc_trace(rsc, fmt, args...) crm_log_tag(LOG_TRACE, rsc ? rsc->id : "<NULL>", fmt, ##args)
# define pe_err(fmt...) { was_processing_error = TRUE; crm_config_error = TRUE; crm_err(fmt); }
# define pe_warn(fmt...) { was_processing_warning = TRUE; crm_config_warning = TRUE; crm_warn(fmt); }
# define pe_proc_err(fmt...) { was_processing_error = TRUE; crm_err(fmt); }
# define pe_proc_warn(fmt...) { was_processing_warning = TRUE; crm_warn(fmt); }
# define pe_set_action_bit(action, bit) action->flags = crm_set_bit(__FUNCTION__, __LINE__, action->uuid, action->flags, bit)
# define pe_clear_action_bit(action, bit) action->flags = crm_clear_bit(__FUNCTION__, __LINE__, action->uuid, action->flags, bit)
+// Some warnings we don't want to print every transition
+
+enum pe_warn_once_e {
+ pe_wo_blind = 0x0001,
+ pe_wo_restart_type = 0x0002,
+ pe_wo_role_after = 0x0004,
+ pe_wo_poweroff = 0x0008,
+ pe_wo_require_all = 0x0010,
+ pe_wo_order_score = 0x0020,
+ pe_wo_neg_threshold = 0x0040,
+};
+
+extern uint32_t pe_wo;
+
+#define pe_warn_once(pe_wo_bit, fmt...) do { \
+ if (is_not_set(pe_wo, pe_wo_bit)) { \
+ if (pe_wo_bit == pe_wo_blind) { \
+ crm_warn(fmt); \
+ } else { \
+ pe_warn(fmt); \
+ } \
+ set_bit(pe_wo, pe_wo_bit); \
+ } \
+ } while (0);
+
+
typedef struct pe__location_constraint_s {
char *id; // Constraint XML ID
pe_resource_t *rsc_lh; // Resource being located
enum rsc_role_e role_filter; // Role to locate
enum pe_discover_e discover_mode; // Resource discovery
GListPtr node_list_rh; // List of pe_node_t*
} pe__location_t;
typedef struct pe__order_constraint_s {
int id;
enum pe_ordering type;
void *lh_opaque;
pe_resource_t *lh_rsc;
pe_action_t *lh_action;
char *lh_action_task;
void *rh_opaque;
pe_resource_t *rh_rsc;
pe_action_t *rh_action;
char *rh_action_task;
} pe__ordering_t;
typedef struct notify_data_s {
GSList *keys; // Environment variable name/value pairs
const char *action;
pe_action_t *pre;
pe_action_t *post;
pe_action_t *pre_done;
pe_action_t *post_done;
GListPtr active; /* notify_entry_t* */
GListPtr inactive; /* notify_entry_t* */
GListPtr start; /* notify_entry_t* */
GListPtr stop; /* notify_entry_t* */
GListPtr demote; /* notify_entry_t* */
GListPtr promote; /* notify_entry_t* */
GListPtr master; /* notify_entry_t* */
GListPtr slave; /* notify_entry_t* */
GHashTable *allowed_nodes;
} notify_data_t;
bool pe_can_fence(pe_working_set_t *data_set, pe_node_t *node);
int merge_weights(int w1, int w2);
void add_hash_param(GHashTable * hash, const char *name, const char *value);
char *native_parameter(pe_resource_t * rsc, pe_node_t * node, gboolean create, const char *name,
pe_working_set_t * data_set);
pe_node_t *native_location(const pe_resource_t *rsc, GList **list, int current);
void pe_metadata(void);
void verify_pe_options(GHashTable * options);
void common_update_score(pe_resource_t * rsc, const char *id, int score);
void native_add_running(pe_resource_t * rsc, pe_node_t * node, pe_working_set_t * data_set);
gboolean native_unpack(pe_resource_t * rsc, pe_working_set_t * data_set);
gboolean group_unpack(pe_resource_t * rsc, pe_working_set_t * data_set);
gboolean clone_unpack(pe_resource_t * rsc, pe_working_set_t * data_set);
gboolean pe__unpack_bundle(pe_resource_t *rsc, pe_working_set_t *data_set);
pe_resource_t *native_find_rsc(pe_resource_t *rsc, const char *id, const pe_node_t *node,
int flags);
gboolean native_active(pe_resource_t * rsc, gboolean all);
gboolean group_active(pe_resource_t * rsc, gboolean all);
gboolean clone_active(pe_resource_t * rsc, gboolean all);
gboolean pe__bundle_active(pe_resource_t *rsc, gboolean all);
void native_print(pe_resource_t * rsc, const char *pre_text, long options, void *print_data);
void group_print(pe_resource_t * rsc, const char *pre_text, long options, void *print_data);
void clone_print(pe_resource_t * rsc, const char *pre_text, long options, void *print_data);
void pe__print_bundle(pe_resource_t *rsc, const char *pre_text, long options,
void *print_data);
int pe__name_and_nvpairs_xml(pcmk__output_t *out, bool is_list, const char *tag_name
, size_t pairs_count, ...);
char *pe__node_display_name(pe_node_t *node, bool print_detail);
int pe__ban_html(pcmk__output_t *out, va_list args);
int pe__ban_text(pcmk__output_t *out, va_list args);
int pe__ban_xml(pcmk__output_t *out, va_list args);
int pe__clone_xml(pcmk__output_t *out, va_list args);
int pe__clone_html(pcmk__output_t *out, va_list args);
int pe__clone_text(pcmk__output_t *out, va_list args);
int pe__cluster_counts_html(pcmk__output_t *out, va_list args);
int pe__cluster_counts_text(pcmk__output_t *out, va_list args);
int pe__cluster_counts_xml(pcmk__output_t *out, va_list args);
int pe__cluster_dc_html(pcmk__output_t *out, va_list args);
int pe__cluster_dc_text(pcmk__output_t *out, va_list args);
int pe__cluster_dc_xml(pcmk__output_t *out, va_list args);
int pe__cluster_maint_mode_html(pcmk__output_t *out, va_list args);
int pe__cluster_maint_mode_text(pcmk__output_t *out, va_list args);
int pe__cluster_options_html(pcmk__output_t *out, va_list args);
int pe__cluster_options_log(pcmk__output_t *out, va_list args);
int pe__cluster_options_text(pcmk__output_t *out, va_list args);
int pe__cluster_options_xml(pcmk__output_t *out, va_list args);
int pe__cluster_stack_html(pcmk__output_t *out, va_list args);
int pe__cluster_stack_text(pcmk__output_t *out, va_list args);
int pe__cluster_stack_xml(pcmk__output_t *out, va_list args);
int pe__cluster_summary(pcmk__output_t *out, va_list args);
int pe__cluster_summary_html(pcmk__output_t *out, va_list args);
int pe__cluster_times_html(pcmk__output_t *out, va_list args);
int pe__cluster_times_xml(pcmk__output_t *out, va_list args);
int pe__cluster_times_text(pcmk__output_t *out, va_list args);
int pe__failed_action_text(pcmk__output_t *out, va_list args);
int pe__failed_action_xml(pcmk__output_t *out, va_list args);
int pe__group_xml(pcmk__output_t *out, va_list args);
int pe__group_html(pcmk__output_t *out, va_list args);
int pe__group_text(pcmk__output_t *out, va_list args);
int pe__bundle_xml(pcmk__output_t *out, va_list args);
int pe__bundle_html(pcmk__output_t *out, va_list args);
int pe__bundle_text(pcmk__output_t *out, va_list args);
int pe__node_html(pcmk__output_t *out, va_list args);
int pe__node_text(pcmk__output_t *out, va_list args);
int pe__node_xml(pcmk__output_t *out, va_list args);
int pe__node_attribute_html(pcmk__output_t *out, va_list args);
int pe__node_attribute_text(pcmk__output_t *out, va_list args);
int pe__node_attribute_xml(pcmk__output_t *out, va_list args);
int pe__op_history_text(pcmk__output_t *out, va_list args);
int pe__op_history_xml(pcmk__output_t *out, va_list args);
int pe__resource_history_text(pcmk__output_t *out, va_list args);
int pe__resource_history_xml(pcmk__output_t *out, va_list args);
int pe__resource_xml(pcmk__output_t *out, va_list args);
int pe__resource_html(pcmk__output_t *out, va_list args);
int pe__resource_text(pcmk__output_t *out, va_list args);
int pe__ticket_html(pcmk__output_t *out, va_list args);
int pe__ticket_text(pcmk__output_t *out, va_list args);
int pe__ticket_xml(pcmk__output_t *out, va_list args);
void native_free(pe_resource_t * rsc);
void group_free(pe_resource_t * rsc);
void clone_free(pe_resource_t * rsc);
void pe__free_bundle(pe_resource_t *rsc);
enum rsc_role_e native_resource_state(const pe_resource_t * rsc, gboolean current);
enum rsc_role_e group_resource_state(const pe_resource_t * rsc, gboolean current);
enum rsc_role_e clone_resource_state(const pe_resource_t * rsc, gboolean current);
enum rsc_role_e pe__bundle_resource_state(const pe_resource_t *rsc,
gboolean current);
void pe__count_common(pe_resource_t *rsc);
void pe__count_bundle(pe_resource_t *rsc);
gboolean common_unpack(xmlNode * xml_obj, pe_resource_t ** rsc, pe_resource_t * parent,
pe_working_set_t * data_set);
void common_free(pe_resource_t * rsc);
extern pe_node_t *node_copy(const pe_node_t *this_node);
extern time_t get_effective_time(pe_working_set_t * data_set);
/* Failure handling utilities (from failcounts.c) */
// bit flags for fail count handling options
enum pe_fc_flags_e {
pe_fc_default = 0x00,
pe_fc_effective = 0x01, // don't count expired failures
pe_fc_fillers = 0x02, // if container, include filler failures in count
};
int pe_get_failcount(pe_node_t *node, pe_resource_t *rsc, time_t *last_failure,
uint32_t flags, xmlNode *xml_op,
pe_working_set_t *data_set);
pe_action_t *pe__clear_failcount(pe_resource_t *rsc, pe_node_t *node,
const char *reason,
pe_working_set_t *data_set);
/* Functions for finding/counting a resource's active nodes */
pe_node_t *pe__find_active_on(const pe_resource_t *rsc,
unsigned int *count_all,
unsigned int *count_clean);
pe_node_t *pe__find_active_requires(const pe_resource_t *rsc,
unsigned int *count);
static inline pe_node_t *
pe__current_node(const pe_resource_t *rsc)
{
return pe__find_active_on(rsc, NULL, NULL);
}
/* Binary like operators for lists of nodes */
extern void node_list_exclude(GHashTable * list, GListPtr list2, gboolean merge_scores);
extern GListPtr node_list_dup(GListPtr list, gboolean reset, gboolean filter);
extern GHashTable *node_hash_from_list(GListPtr list);
static inline gpointer
pe_hash_table_lookup(GHashTable * hash, gconstpointer key)
{
if (hash) {
return g_hash_table_lookup(hash, key);
}
return NULL;
}
extern pe_action_t *get_pseudo_op(const char *name, pe_working_set_t * data_set);
extern gboolean order_actions(pe_action_t * lh_action, pe_action_t * rh_action, enum pe_ordering order);
GHashTable *node_hash_dup(GHashTable * hash);
/* Printing functions for debug */
extern void print_node(const char *pre_text, pe_node_t * node, gboolean details);
extern void print_str_str(gpointer key, gpointer value, gpointer user_data);
extern void pe__output_node(pe_node_t * node, gboolean details, pcmk__output_t *out);
extern void dump_node_capacity(int level, const char *comment, pe_node_t * node);
extern void dump_rsc_utilization(int level, const char *comment, pe_resource_t * rsc, pe_node_t * node);
void pe__show_node_weights_as(const char *file, const char *function,
int line, bool to_log, pe_resource_t *rsc,
const char *comment, GHashTable *nodes);
#define pe__show_node_weights(level, rsc, text, nodes) \
pe__show_node_weights_as(__FILE__, __FUNCTION__, __LINE__, \
(level), (rsc), (text), (nodes))
/* Sorting functions */
extern gint sort_rsc_priority(gconstpointer a, gconstpointer b);
extern gint sort_rsc_index(gconstpointer a, gconstpointer b);
extern xmlNode *find_rsc_op_entry(pe_resource_t * rsc, const char *key);
extern pe_action_t *custom_action(pe_resource_t * rsc, char *key, const char *task, pe_node_t * on_node,
gboolean optional, gboolean foo, pe_working_set_t * data_set);
# define delete_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_DELETE, 0)
# define delete_action(rsc, node, optional) custom_action( \
rsc, delete_key(rsc), CRMD_ACTION_DELETE, node, \
optional, TRUE, data_set);
# define stopped_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_STOPPED, 0)
# define stopped_action(rsc, node, optional) custom_action( \
rsc, stopped_key(rsc), CRMD_ACTION_STOPPED, node, \
optional, TRUE, data_set);
# define stop_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_STOP, 0)
# define stop_action(rsc, node, optional) custom_action( \
rsc, stop_key(rsc), CRMD_ACTION_STOP, node, \
optional, TRUE, data_set);
# define reload_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_RELOAD, 0)
# define start_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_START, 0)
# define start_action(rsc, node, optional) custom_action( \
rsc, start_key(rsc), CRMD_ACTION_START, node, \
optional, TRUE, data_set)
# define started_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_STARTED, 0)
# define started_action(rsc, node, optional) custom_action( \
rsc, started_key(rsc), CRMD_ACTION_STARTED, node, \
optional, TRUE, data_set)
# define promote_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_PROMOTE, 0)
# define promote_action(rsc, node, optional) custom_action( \
rsc, promote_key(rsc), CRMD_ACTION_PROMOTE, node, \
optional, TRUE, data_set)
# define promoted_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_PROMOTED, 0)
# define promoted_action(rsc, node, optional) custom_action( \
rsc, promoted_key(rsc), CRMD_ACTION_PROMOTED, node, \
optional, TRUE, data_set)
# define demote_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_DEMOTE, 0)
# define demote_action(rsc, node, optional) custom_action( \
rsc, demote_key(rsc), CRMD_ACTION_DEMOTE, node, \
optional, TRUE, data_set)
# define demoted_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_DEMOTED, 0)
# define demoted_action(rsc, node, optional) custom_action( \
rsc, demoted_key(rsc), CRMD_ACTION_DEMOTED, node, \
optional, TRUE, data_set)
extern int pe_get_configured_timeout(pe_resource_t *rsc, const char *action,
pe_working_set_t *data_set);
extern pe_action_t *find_first_action(GListPtr input, const char *uuid, const char *task,
pe_node_t * on_node);
extern enum action_tasks get_complex_task(pe_resource_t * rsc, const char *name,
gboolean allow_non_atomic);
extern GListPtr find_actions(GListPtr input, const char *key, const pe_node_t *on_node);
GList *find_actions_exact(GList *input, const char *key,
const pe_node_t *on_node);
extern GListPtr find_recurring_actions(GListPtr input, pe_node_t * not_on_node);
GList *pe__resource_actions(const pe_resource_t *rsc, const pe_node_t *node,
const char *task, bool require_node);
extern void pe_free_action(pe_action_t * action);
extern void resource_location(pe_resource_t * rsc, pe_node_t * node, int score, const char *tag,
pe_working_set_t * data_set);
extern gint sort_op_by_callid(gconstpointer a, gconstpointer b);
extern gboolean get_target_role(pe_resource_t * rsc, enum rsc_role_e *role);
extern pe_resource_t *find_clone_instance(pe_resource_t * rsc, const char *sub_id,
pe_working_set_t * data_set);
extern void destroy_ticket(gpointer data);
extern pe_ticket_t *ticket_new(const char *ticket_id, pe_working_set_t * data_set);
// Resources for manipulating resource names
const char *pe_base_name_end(const char *id);
char *clone_strip(const char *last_rsc_id);
char *clone_zero(const char *last_rsc_id);
static inline bool
pe_base_name_eq(pe_resource_t *rsc, const char *id)
{
if (id && rsc && rsc->id) {
// Number of characters in rsc->id before any clone suffix
size_t base_len = pe_base_name_end(rsc->id) - rsc->id + 1;
return (strlen(id) == base_len) && !strncmp(id, rsc->id, base_len);
}
return FALSE;
}
int pe__target_rc_from_xml(xmlNode *xml_op);
gint sort_node_uname(gconstpointer a, gconstpointer b);
bool is_set_recursive(pe_resource_t * rsc, long long flag, bool any);
enum rsc_digest_cmp_val {
/*! Digests are the same */
RSC_DIGEST_MATCH = 0,
/*! Params that require a restart changed */
RSC_DIGEST_RESTART,
/*! Some parameter changed. */
RSC_DIGEST_ALL,
/*! rsc op didn't have a digest associated with it, so
* it is unknown if parameters changed or not. */
RSC_DIGEST_UNKNOWN,
};
typedef struct op_digest_cache_s {
enum rsc_digest_cmp_val rc;
xmlNode *params_all;
xmlNode *params_secure;
xmlNode *params_restart;
char *digest_all_calc;
char *digest_secure_calc;
char *digest_restart_calc;
} op_digest_cache_t;
op_digest_cache_t *rsc_action_digest_cmp(pe_resource_t * rsc, xmlNode * xml_op, pe_node_t * node,
pe_working_set_t * data_set);
pe_action_t *pe_fence_op(pe_node_t * node, const char *op, bool optional, const char *reason, bool priority_delay, pe_working_set_t * data_set);
void trigger_unfencing(
pe_resource_t * rsc, pe_node_t *node, const char *reason, pe_action_t *dependency, pe_working_set_t * data_set);
void pe_action_set_reason(pe_action_t *action, const char *reason, bool overwrite);
void pe_action_set_flag_reason(const char *function, long line, pe_action_t *action, pe_action_t *reason, const char *text, enum pe_action_flags flags, bool overwrite);
#define pe_action_required(action, reason, text) pe_action_set_flag_reason(__FUNCTION__, __LINE__, action, reason, text, pe_action_optional, FALSE)
#define pe_action_implies(action, reason, flag) pe_action_set_flag_reason(__FUNCTION__, __LINE__, action, reason, NULL, flag, FALSE)
void set_bit_recursive(pe_resource_t * rsc, unsigned long long flag);
void clear_bit_recursive(pe_resource_t * rsc, unsigned long long flag);
gboolean add_tag_ref(GHashTable * tags, const char * tag_name, const char * obj_ref);
void print_rscs_brief(GListPtr rsc_list, const char * pre_text, long options,
void * print_data, gboolean print_all);
int pe__rscs_brief_output(pcmk__output_t *out, GListPtr rsc_list, long options, gboolean print_all);
void pe_fence_node(pe_working_set_t * data_set, pe_node_t * node, const char *reason, bool priority_delay);
pe_node_t *pe_create_node(const char *id, const char *uname, const char *type,
const char *score, pe_working_set_t * data_set);
void common_print(pe_resource_t * rsc, const char *pre_text, const char *name, pe_node_t *node, long options, void *print_data);
int pe__common_output_text(pcmk__output_t *out, pe_resource_t * rsc, const char *name, pe_node_t *node, long options);
int pe__common_output_html(pcmk__output_t *out, pe_resource_t * rsc, const char *name, pe_node_t *node, long options);
pe_resource_t *pe__find_bundle_replica(const pe_resource_t *bundle,
const pe_node_t *node);
bool pe__bundle_needs_remote_name(pe_resource_t *rsc);
const char *pe__add_bundle_remote_name(pe_resource_t *rsc, xmlNode *xml,
const char *field);
const char *pe_node_attribute_calculated(const pe_node_t *node,
const char *name,
const pe_resource_t *rsc);
const char *pe_node_attribute_raw(pe_node_t *node, const char *name);
bool pe__is_universal_clone(pe_resource_t *rsc,
pe_working_set_t *data_set);
void pe__add_param_check(xmlNode *rsc_op, pe_resource_t *rsc, pe_node_t *node,
enum pe_check_parameters, pe_working_set_t *data_set);
void pe__foreach_param_check(pe_working_set_t *data_set,
void (*cb)(pe_resource_t*, pe_node_t*, xmlNode*,
enum pe_check_parameters,
pe_working_set_t*));
void pe__free_param_checks(pe_working_set_t *data_set);
bool pe__shutdown_requested(pe_node_t *node);
void pe__update_recheck_time(time_t recheck, pe_working_set_t *data_set);
#define BOOL2STR(x) ((x) ? "true" : "false")
/*!
* \internal
* \brief Register xml formatting message functions.
*/
void pe__register_messages(pcmk__output_t *out);
void pe__unpack_dataset_nvpairs(xmlNode *xml_obj, const char *set_name,
GHashTable *node_hash, GHashTable *hash,
const char *always_first, gboolean overwrite,
pe_working_set_t *data_set);
bool pe__resource_is_disabled(pe_resource_t *rsc);
pe_action_t *pe__clear_resource_history(pe_resource_t *rsc, pe_node_t *node,
pe_working_set_t *data_set);
gchar **pe__unames_with_tag(pe_working_set_t *data_set, const char *tag_name);
bool pe__uname_has_tag(pe_working_set_t *data_set, const char *node, const char *tag);
#endif
diff --git a/include/pcmki/pcmki_scheduler.h b/include/pcmki/pcmki_scheduler.h
index d34faa329c..342d57a2bf 100644
--- a/include/pcmki/pcmki_scheduler.h
+++ b/include/pcmki/pcmki_scheduler.h
@@ -1,111 +1,110 @@
/*
- * Copyright 2014-2019 the Pacemaker project contributors
+ * Copyright 2014-2020 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.
*/
#ifndef PENGINE__H
# define PENGINE__H
typedef struct rsc_colocation_s rsc_colocation_t;
typedef struct rsc_ticket_s rsc_ticket_t;
typedef struct lrm_agent_s lrm_agent_t;
# include <glib.h>
# include <crm/crm.h>
# include <crm/common/iso8601.h>
# include <crm/pengine/rules.h>
# include <crm/pengine/common.h>
# include <crm/pengine/status.h>
# include <crm/pengine/complex.h>
enum pe_stop_fail {
pesf_block,
pesf_stonith,
pesf_ignore
};
enum pe_weights {
pe_weights_none = 0x0,
pe_weights_init = 0x1,
pe_weights_forward = 0x4,
pe_weights_positive = 0x8,
pe_weights_rollback = 0x10,
};
struct rsc_colocation_s {
const char *id;
const char *node_attribute;
pe_resource_t *rsc_lh;
pe_resource_t *rsc_rh;
int role_lh;
int role_rh;
int score;
};
enum loss_ticket_policy_e {
loss_ticket_stop,
loss_ticket_demote,
loss_ticket_fence,
loss_ticket_freeze
};
struct rsc_ticket_s {
const char *id;
pe_resource_t *rsc_lh;
pe_ticket_t *ticket;
enum loss_ticket_policy_e loss_policy;
int role_lh;
};
extern gboolean stage0(pe_working_set_t * data_set);
extern gboolean probe_resources(pe_working_set_t * data_set);
extern gboolean stage2(pe_working_set_t * data_set);
extern gboolean stage3(pe_working_set_t * data_set);
extern gboolean stage4(pe_working_set_t * data_set);
extern gboolean stage5(pe_working_set_t * data_set);
extern gboolean stage6(pe_working_set_t * data_set);
extern gboolean stage7(pe_working_set_t * data_set);
extern gboolean stage8(pe_working_set_t * data_set);
extern gboolean summary(GListPtr resources);
extern gboolean unpack_constraints(xmlNode * xml_constraints, pe_working_set_t * data_set);
extern gboolean shutdown_constraints(pe_node_t * node, pe_action_t * shutdown_op,
pe_working_set_t * data_set);
void pcmk__order_vs_fence(pe_action_t *stonith_op, pe_working_set_t *data_set);
extern int custom_action_order(pe_resource_t * lh_rsc, char *lh_task, pe_action_t * lh_action,
pe_resource_t * rh_rsc, char *rh_task, pe_action_t * rh_action,
enum pe_ordering type, pe_working_set_t * data_set);
extern int new_rsc_order(pe_resource_t * lh_rsc, const char *lh_task,
pe_resource_t * rh_rsc, const char *rh_task,
enum pe_ordering type, pe_working_set_t * data_set);
# define order_start_start(rsc1,rsc2, type) \
new_rsc_order(rsc1, CRMD_ACTION_START, rsc2, CRMD_ACTION_START, type, data_set)
# define order_stop_stop(rsc1, rsc2, type) \
new_rsc_order(rsc1, CRMD_ACTION_STOP, rsc2, CRMD_ACTION_STOP, type, data_set)
extern void graph_element_from_action(pe_action_t * action, pe_working_set_t * data_set);
extern void add_maintenance_update(pe_working_set_t *data_set);
xmlNode *pcmk__schedule_actions(pe_working_set_t *data_set, xmlNode *xml_input,
crm_time_t *now);
bool pcmk__ordering_is_invalid(pe_action_t *action, pe_action_wrapper_t *input);
extern gboolean show_scores;
extern gboolean show_utilization;
-extern int utilization_log_level;
extern const char *transition_idle_timeout;
#endif
diff --git a/lib/common/agents.c b/lib/common/agents.c
index a4add2a152..113007c52a 100644
--- a/lib/common/agents.c
+++ b/lib/common/agents.c
@@ -1,151 +1,151 @@
/*
* Copyright 2004-2020 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>
#ifndef _GNU_SOURCE
# define _GNU_SOURCE
#endif
#include <stdio.h>
#include <string.h>
#include <strings.h>
#include <crm/crm.h>
#include <crm/common/util.h>
/*!
* \brief Get capabilities of a resource agent standard
*
* \param[in] standard Standard name
*
* \return Bitmask of enum pcmk_ra_caps values
*/
uint32_t
pcmk_get_ra_caps(const char *standard)
{
/* @COMPAT This should probably be case-sensitive, but isn't,
* for backward compatibility.
*/
if (standard == NULL) {
return pcmk_ra_cap_none;
} else if (!strcasecmp(standard, PCMK_RESOURCE_CLASS_OCF)) {
return pcmk_ra_cap_provider | pcmk_ra_cap_params
| pcmk_ra_cap_unique | pcmk_ra_cap_promotable;
} else if (!strcasecmp(standard, PCMK_RESOURCE_CLASS_STONITH)) {
/* @COMPAT Stonith resources can't really be unique clones, but we've
* allowed it in the past and have it in some scheduler regression tests
* (which were likely never used as real configurations).
*
* @TODO Remove pcmk_ra_cap_unique at the next major schema version
* bump, with a transform to remove globally-unique from the config.
*/
- return pcmk_ra_cap_params | pcmk_ra_cap_unique;
+ return pcmk_ra_cap_params | pcmk_ra_cap_unique | pcmk_ra_cap_stdin;
} else if (!strcasecmp(standard, PCMK_RESOURCE_CLASS_SYSTEMD)
|| !strcasecmp(standard, PCMK_RESOURCE_CLASS_SERVICE)
|| !strcasecmp(standard, PCMK_RESOURCE_CLASS_LSB)
|| !strcasecmp(standard, PCMK_RESOURCE_CLASS_UPSTART)) {
/* Since service can map to LSB, systemd, or upstart, these should
* have identical capabilities
*/
return pcmk_ra_cap_status;
} else if (!strcasecmp(standard, PCMK_RESOURCE_CLASS_NAGIOS)) {
return pcmk_ra_cap_params;
}
return pcmk_ra_cap_none;
}
char *
crm_generate_ra_key(const char *standard, const char *provider,
const char *type)
{
if (!standard && !provider && !type) {
return NULL;
}
return crm_strdup_printf("%s%s%s:%s",
(standard? standard : ""),
(provider? ":" : ""), (provider? provider : ""),
(type? type : ""));
}
/*!
* \brief Parse a "standard[:provider]:type" agent specification
*
* \param[in] spec Agent specification
* \param[out] standard Newly allocated memory containing agent standard (or NULL)
* \param[out] provider Newly allocated memory containing agent provider (or NULL)
* \param[put] type Newly allocated memory containing agent type (or NULL)
*
* \return pcmk_ok if the string could be parsed, -EINVAL otherwise
*
* \note It is acceptable for the type to contain a ':' if the standard supports
* that. For example, systemd supports the form "systemd:UNIT@A:B".
* \note It is the caller's responsibility to free the returned values.
*/
int
crm_parse_agent_spec(const char *spec, char **standard, char **provider,
char **type)
{
char *colon;
CRM_CHECK(spec && standard && provider && type, return -EINVAL);
*standard = NULL;
*provider = NULL;
*type = NULL;
colon = strchr(spec, ':');
if ((colon == NULL) || (colon == spec)) {
return -EINVAL;
}
*standard = strndup(spec, colon - spec);
spec = colon + 1;
if (is_set(pcmk_get_ra_caps(*standard), pcmk_ra_cap_provider)) {
colon = strchr(spec, ':');
if ((colon == NULL) || (colon == spec)) {
free(*standard);
return -EINVAL;
}
*provider = strndup(spec, colon - spec);
spec = colon + 1;
}
if (*spec == '\0') {
free(*standard);
free(*provider);
return -EINVAL;
}
*type = strdup(spec);
return pcmk_ok;
}
// Deprecated functions kept only for backward API compatibility
bool crm_provider_required(const char *standard);
/*!
* \deprecated
* \brief Check whether a resource standard requires a provider to be specified
*
* \param[in] standard Standard name
*
* \return TRUE if standard requires a provider, FALSE otherwise
*/
bool
crm_provider_required(const char *standard)
{
return is_set(pcmk_get_ra_caps(standard), pcmk_ra_cap_provider);
}
diff --git a/lib/common/iso8601.c b/lib/common/iso8601.c
index 1e9ce6066d..8657d5835c 100644
--- a/lib/common/iso8601.c
+++ b/lib/common/iso8601.c
@@ -1,1729 +1,1736 @@
/*
* Copyright 2005-2020 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.
*/
/*
* References:
* https://en.wikipedia.org/wiki/ISO_8601
* http://www.staff.science.uu.nl/~gent0113/calendar/isocalendar.htm
*/
#include <crm_internal.h>
#include <crm/crm.h>
#include <time.h>
#include <ctype.h>
#include <string.h>
#include <stdbool.h>
#include <crm/common/iso8601.h>
#include <crm/common/iso8601_internal.h>
/*
* Andrew's code was originally written for OSes whose "struct tm" contains:
* long tm_gmtoff; :: Seconds east of UTC
* const char *tm_zone; :: Timezone abbreviation
* Some OSes lack these, instead having:
* time_t (or long) timezone;
:: "difference between UTC and local standard time"
* char *tzname[2] = { "...", "..." };
* I (David Lee) confess to not understanding the details. So my attempted
* generalisations for where their use is necessary may be flawed.
*
* 1. Does "difference between ..." subtract the same or opposite way?
* 2. Should it use "altzone" instead of "timezone"?
* 3. Should it use tzname[0] or tzname[1]? Interaction with timezone/altzone?
*/
#if defined(HAVE_STRUCT_TM_TM_GMTOFF)
# define GMTOFF(tm) ((tm)->tm_gmtoff)
#else
/* Note: extern variable; macro argument not actually used. */
# define GMTOFF(tm) (-timezone+daylight)
#endif
#define HOUR_SECONDS (60 * 60)
#define DAY_SECONDS (HOUR_SECONDS * 24)
// A date/time or duration
struct crm_time_s {
int years; // Calendar year (date/time) or number of years (duration)
int months; // Number of months (duration only)
int days; // Ordinal day of year (date/time) or number of days (duration)
int seconds; // Seconds of day (date/time) or number of seconds (duration)
int offset; // Seconds offset from UTC (date/time only)
bool duration; // True if duration
};
static crm_time_t *parse_date(const char *date_str);
static crm_time_t *
crm_get_utc_time(crm_time_t *dt)
{
crm_time_t *utc = NULL;
if (dt == NULL) {
errno = EINVAL;
return NULL;
}
utc = crm_time_new_undefined();
utc->years = dt->years;
utc->days = dt->days;
utc->seconds = dt->seconds;
utc->offset = 0;
if (dt->offset) {
crm_time_add_seconds(utc, -dt->offset);
} else {
/* Durations (which are the only things that can include months, never have a timezone */
utc->months = dt->months;
}
crm_time_log(LOG_TRACE, "utc-source", dt,
crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone);
crm_time_log(LOG_TRACE, "utc-target", utc,
crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone);
return utc;
}
crm_time_t *
crm_time_new(const char *date_time)
{
time_t tm_now;
crm_time_t *dt = NULL;
tzset();
if (date_time == NULL) {
tm_now = time(NULL);
dt = crm_time_new_undefined();
crm_time_set_timet(dt, &tm_now);
} else {
dt = parse_date(date_time);
}
return dt;
}
/*!
* \brief Allocate memory for an uninitialized time object
*
* \return Newly allocated time object
* \note The caller is responsible for freeing the return value using
* crm_time_free().
*/
crm_time_t *
crm_time_new_undefined()
{
crm_time_t *result = calloc(1, sizeof(crm_time_t));
CRM_ASSERT(result != NULL);
return result;
}
/*!
* \brief Check whether a time object has been initialized yet
*
* \param[in] t Time object to check
*
* \return TRUE if time object has been initialized, FALSE otherwise
*/
bool
crm_time_is_defined(const crm_time_t *t)
{
// Any nonzero member indicates something has been done to t
return (t != NULL) && (t->years || t->months || t->days || t->seconds
|| t->offset || t->duration);
}
void
crm_time_free(crm_time_t * dt)
{
if (dt == NULL) {
return;
}
free(dt);
}
static int
year_days(int year)
{
int d = 365;
if (crm_time_leapyear(year)) {
d++;
}
return d;
}
/* From http://myweb.ecu.edu/mccartyr/ISOwdALG.txt :
*
* 5. Find the Jan1Weekday for Y (Monday=1, Sunday=7)
* YY = (Y-1) % 100
* C = (Y-1) - YY
* G = YY + YY/4
* Jan1Weekday = 1 + (((((C / 100) % 4) x 5) + G) % 7)
*/
int
crm_time_january1_weekday(int year)
{
int YY = (year - 1) % 100;
int C = (year - 1) - YY;
int G = YY + YY / 4;
int jan1 = 1 + (((((C / 100) % 4) * 5) + G) % 7);
crm_trace("YY=%d, C=%d, G=%d", YY, C, G);
crm_trace("January 1 %.4d: %d", year, jan1);
return jan1;
}
int
crm_time_weeks_in_year(int year)
{
int weeks = 52;
int jan1 = crm_time_january1_weekday(year);
/* if jan1 == thursday */
if (jan1 == 4) {
weeks++;
} else {
jan1 = crm_time_january1_weekday(year + 1);
/* if dec31 == thursday aka. jan1 of next year is a friday */
if (jan1 == 5) {
weeks++;
}
}
return weeks;
}
// Jan-Dec plus Feb of leap years
static int month_days[13] = {
31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31, 29
};
/*!
* \brief Return number of days in given month of given year
*
* \param[in] Ordinal month (1-12)
* \param[in] Gregorian year
*
* \return Number of days in given month (0 if given month is invalid)
*/
int
crm_time_days_in_month(int month, int year)
{
if ((month < 1) || (month > 12)) {
return 0;
}
if ((month == 2) && crm_time_leapyear(year)) {
month = 13;
}
return month_days[month - 1];
}
bool
crm_time_leapyear(int year)
{
gboolean is_leap = FALSE;
if (year % 4 == 0) {
is_leap = TRUE;
}
if (year % 100 == 0 && year % 400 != 0) {
is_leap = FALSE;
}
return is_leap;
}
static uint32_t
get_ordinal_days(uint32_t y, uint32_t m, uint32_t d)
{
int lpc;
for (lpc = 1; lpc < m; lpc++) {
d += crm_time_days_in_month(lpc, y);
}
return d;
}
void
crm_time_log_alias(int log_level, const char *file, const char *function, int line,
const char *prefix, crm_time_t * date_time, int flags)
{
char *date_s = crm_time_as_string(date_time, flags);
if (log_level == LOG_STDOUT) {
printf("%s%s%s\n",
(prefix? prefix : ""), (prefix? ": " : ""), date_s);
} else {
do_crm_log_alias(log_level, file, function, line, "%s%s%s",
(prefix? prefix : ""), (prefix? ": " : ""), date_s);
}
free(date_s);
}
static void
crm_time_get_sec(int sec, uint * h, uint * m, uint * s)
{
uint hours, minutes, seconds;
if (sec < 0) {
seconds = 0 - sec;
} else {
seconds = sec;
}
hours = seconds / HOUR_SECONDS;
seconds -= HOUR_SECONDS * hours;
minutes = seconds / 60;
seconds -= 60 * minutes;
crm_trace("%d == %.2d:%.2d:%.2d", sec, hours, minutes, seconds);
*h = hours;
*m = minutes;
*s = seconds;
}
int
crm_time_get_timeofday(crm_time_t * dt, uint * h, uint * m, uint * s)
{
crm_time_get_sec(dt->seconds, h, m, s);
return TRUE;
}
int
crm_time_get_timezone(crm_time_t * dt, uint * h, uint * m)
{
uint s;
crm_time_get_sec(dt->seconds, h, m, &s);
return TRUE;
}
long long
crm_time_get_seconds(crm_time_t * dt)
{
int lpc;
crm_time_t *utc = NULL;
long long in_seconds = 0;
if (dt == NULL) {
return 0;
}
utc = crm_get_utc_time(dt);
if (utc == NULL) {
return 0;
}
for (lpc = 1; lpc < utc->years; lpc++) {
int dmax = year_days(lpc);
in_seconds += DAY_SECONDS * dmax;
}
/* utc->months is an offset that can only be set for a duration.
* By definition, the value is variable depending on the date to
* which it is applied.
*
* Force 30-day months so that something vaguely sane happens
* for anyone that tries to use a month in this way.
*/
if (utc->months > 0) {
in_seconds += DAY_SECONDS * 30 * utc->months;
}
if (utc->days > 0) {
in_seconds += DAY_SECONDS * (utc->days - 1);
}
in_seconds += utc->seconds;
crm_time_free(utc);
return in_seconds;
}
#define EPOCH_SECONDS 62135596800ULL /* Calculated using crm_time_get_seconds() */
long long
crm_time_get_seconds_since_epoch(crm_time_t * dt)
{
return (dt == NULL)? 0 : (crm_time_get_seconds(dt) - EPOCH_SECONDS);
}
int
crm_time_get_gregorian(crm_time_t * dt, uint * y, uint * m, uint * d)
{
int months = 0;
int days = dt->days;
if(dt->years != 0) {
for (months = 1; months <= 12 && days > 0; months++) {
int mdays = crm_time_days_in_month(months, dt->years);
if (mdays >= days) {
break;
} else {
days -= mdays;
}
}
} else if (dt->months) {
/* This is a duration including months, don't convert the days field */
months = dt->months;
} else {
/* This is a duration not including months, still don't convert the days field */
}
*y = dt->years;
*m = months;
*d = days;
crm_trace("%.4d-%.3d -> %.4d-%.2d-%.2d", dt->years, dt->days, dt->years, months, days);
return TRUE;
}
int
crm_time_get_ordinal(crm_time_t * dt, uint * y, uint * d)
{
*y = dt->years;
*d = dt->days;
return TRUE;
}
int
crm_time_get_isoweek(crm_time_t * dt, uint * y, uint * w, uint * d)
{
/*
* Monday 29 December 2008 is written "2009-W01-1"
* Sunday 3 January 2010 is written "2009-W53-7"
*/
int year_num = 0;
int jan1 = crm_time_january1_weekday(dt->years);
int h = -1;
CRM_CHECK(dt->days > 0, return FALSE);
/* 6. Find the Weekday for Y M D */
h = dt->days + jan1 - 1;
*d = 1 + ((h - 1) % 7);
/* 7. Find if Y M D falls in YearNumber Y-1, WeekNumber 52 or 53 */
if (dt->days <= (8 - jan1) && jan1 > 4) {
crm_trace("year--, jan1=%d", jan1);
year_num = dt->years - 1;
*w = crm_time_weeks_in_year(year_num);
} else {
year_num = dt->years;
}
/* 8. Find if Y M D falls in YearNumber Y+1, WeekNumber 1 */
if (year_num == dt->years) {
int dmax = year_days(year_num);
int correction = 4 - *d;
if ((dmax - dt->days) < correction) {
crm_trace("year++, jan1=%d, i=%d vs. %d", jan1, dmax - dt->days, correction);
year_num = dt->years + 1;
*w = 1;
}
}
/* 9. Find if Y M D falls in YearNumber Y, WeekNumber 1 through 53 */
if (year_num == dt->years) {
int j = dt->days + (7 - *d) + (jan1 - 1);
*w = j / 7;
if (jan1 > 4) {
*w -= 1;
}
}
*y = year_num;
crm_trace("Converted %.4d-%.3d to %.4d-W%.2d-%d", dt->years, dt->days, *y, *w, *d);
return TRUE;
}
#define DATE_MAX 128
static void
crm_duration_as_string(crm_time_t *dt, char *result)
{
size_t offset = 0;
if (dt->years) {
offset += snprintf(result + offset, DATE_MAX - offset, "%4d year%s ",
dt->years, pcmk__plural_s(dt->years));
}
if (dt->months) {
offset += snprintf(result + offset, DATE_MAX - offset, "%2d month%s ",
dt->months, pcmk__plural_s(dt->months));
}
if (dt->days) {
offset += snprintf(result + offset, DATE_MAX - offset, "%2d day%s ",
dt->days, pcmk__plural_s(dt->days));
}
if (((offset == 0) || (dt->seconds != 0))
&& (dt->seconds > -60) && (dt->seconds < 60)) {
offset += snprintf(result + offset, DATE_MAX - offset, "%d second%s",
dt->seconds, pcmk__plural_s(dt->seconds));
} else if (dt->seconds) {
uint h = 0, m = 0, s = 0;
offset += snprintf(result + offset, DATE_MAX - offset, "%d seconds (",
dt->seconds);
crm_time_get_sec(dt->seconds, &h, &m, &s);
if (h) {
offset += snprintf(result + offset, DATE_MAX - offset, "%u hour%s%s",
h, pcmk__plural_s(h), ((m || s)? " " : ""));
}
if (m) {
offset += snprintf(result + offset, DATE_MAX - offset, "%u minute%s%s",
m, pcmk__plural_s(m), (s? " " : ""));
}
if (s) {
offset += snprintf(result + offset, DATE_MAX - offset, "%u second%s",
s, pcmk__plural_s(s));
}
offset += snprintf(result + offset, DATE_MAX - offset, ")");
}
}
char *
crm_time_as_string(crm_time_t * date_time, int flags)
{
crm_time_t *dt = NULL;
crm_time_t *utc = NULL;
char result[DATE_MAX] = { '\0', };
char *result_copy = NULL;
size_t offset = 0;
// Convert to UTC if local timezone was not requested
if (date_time && date_time->offset
&& is_not_set(flags, crm_time_log_with_timezone)) {
crm_trace("UTC conversion");
utc = crm_get_utc_time(date_time);
dt = utc;
} else {
dt = date_time;
}
if (!crm_time_is_defined(dt)) {
strcpy(result, "<undefined time>");
goto done;
}
// Simple cases: as duration, seconds, or seconds since epoch
if (flags & crm_time_log_duration) {
crm_duration_as_string(date_time, result);
goto done;
}
if (flags & crm_time_seconds) {
snprintf(result, DATE_MAX, "%lld", crm_time_get_seconds(date_time));
goto done;
}
if (flags & crm_time_epoch) {
snprintf(result, DATE_MAX, "%lld",
crm_time_get_seconds_since_epoch(date_time));
goto done;
}
// As readable string
if (flags & crm_time_log_date) {
if (flags & crm_time_weeks) { // YYYY-WW-D
uint y, w, d;
if (crm_time_get_isoweek(dt, &y, &w, &d)) {
offset += snprintf(result + offset, DATE_MAX - offset,
"%u-W%.2u-%u", y, w, d);
}
} else if (flags & crm_time_ordinal) { // YYYY-DDD
uint y, d;
if (crm_time_get_ordinal(dt, &y, &d)) {
offset += snprintf(result + offset, DATE_MAX - offset,
"%u-%.3u", y, d);
}
} else { // YYYY-MM-DD
uint y, m, d;
if (crm_time_get_gregorian(dt, &y, &m, &d)) {
offset += snprintf(result + offset, DATE_MAX - offset,
"%.4u-%.2u-%.2u", y, m, d);
}
}
}
if (flags & crm_time_log_timeofday) {
uint h = 0, m = 0, s = 0;
if (offset > 0) {
offset += snprintf(result + offset, DATE_MAX - offset, " ");
}
if (crm_time_get_timeofday(dt, &h, &m, &s)) {
offset += snprintf(result + offset, DATE_MAX - offset,
"%.2u:%.2u:%.2u", h, m, s);
}
if ((flags & crm_time_log_with_timezone) && (dt->offset != 0)) {
crm_time_get_sec(dt->offset, &h, &m, &s);
offset += snprintf(result + offset, DATE_MAX - offset,
" %c%.2u:%.2u",
((dt->offset < 0)? '-' : '+'), h, m);
} else {
offset += snprintf(result + offset, DATE_MAX - offset, "Z");
}
}
done:
crm_time_free(utc);
result_copy = strdup(result);
CRM_ASSERT(result_copy != NULL);
return result_copy;
}
/*!
* \internal
* \brief Determine number of seconds from an hour:minute:second string
*
* \param[in] time_str Time specification string
* \param[out] result Number of seconds equivalent to time_str
*
* \return TRUE if specification was valid, FALSE (and set errno) otherwise
* \note This may return the number of seconds in a day (which is out of bounds
* for a time object) if given 24:00:00.
*/
static bool
crm_time_parse_sec(const char *time_str, int *result)
{
int rc;
uint hour = 0;
uint minute = 0;
uint second = 0;
*result = 0;
// Must have at least hour, but minutes and seconds are optional
rc = sscanf(time_str, "%d:%d:%d", &hour, &minute, &second);
if (rc == 1) {
rc = sscanf(time_str, "%2d%2d%2d", &hour, &minute, &second);
}
if (rc == 0) {
crm_err("%s is not a valid ISO 8601 time specification", time_str);
errno = EINVAL;
return FALSE;
}
crm_trace("Got valid time: %.2d:%.2d:%.2d", hour, minute, second);
if ((hour == 24) && (minute == 0) && (second == 0)) {
// Equivalent to 00:00:00 of next day, return number of seconds in day
} else if (hour >= 24) {
crm_err("%s is not a valid ISO 8601 time specification "
"because %d is not a valid hour", time_str, hour);
errno = EINVAL;
return FALSE;
}
if (minute >= 60) {
crm_err("%s is not a valid ISO 8601 time specification "
"because %d is not a valid minute", time_str, minute);
errno = EINVAL;
return FALSE;
}
if (second >= 60) {
crm_err("%s is not a valid ISO 8601 time specification "
"because %d is not a valid second", time_str, second);
errno = EINVAL;
return FALSE;
}
*result = (hour * HOUR_SECONDS) + (minute * 60) + second;
return TRUE;
}
static bool
crm_time_parse_offset(const char *offset_str, int *offset)
{
tzset();
if (offset_str == NULL) {
// Use local offset
#if defined(HAVE_STRUCT_TM_TM_GMTOFF)
time_t now = time(NULL);
struct tm *now_tm = localtime(&now);
#endif
int h_offset = GMTOFF(now_tm) / HOUR_SECONDS;
int m_offset = (GMTOFF(now_tm) - (HOUR_SECONDS * h_offset)) / 60;
if (h_offset < 0 && m_offset < 0) {
m_offset = 0 - m_offset;
}
*offset = (HOUR_SECONDS * h_offset) + (60 * m_offset);
return TRUE;
}
if (offset_str[0] == 'Z') { // @TODO invalid if anything after?
*offset = 0;
return TRUE;
}
*offset = 0;
if ((offset_str[0] == '+') || (offset_str[0] == '-')
|| isdigit((int)offset_str[0])) {
gboolean negate = FALSE;
if (offset_str[0] == '+') {
offset_str++;
} else if (offset_str[0] == '-') {
negate = TRUE;
offset_str++;
}
if (crm_time_parse_sec(offset_str, offset) == FALSE) {
return FALSE;
}
if (negate) {
*offset = 0 - *offset;
}
} // @TODO else invalid?
return TRUE;
}
/*!
* \internal
* \brief Parse the time portion of an ISO 8601 date/time string
*
* \param[in] time_str Time portion of specification (after any 'T')
* \param[in,out] a_time Time object to parse into
*
* \return TRUE if valid time was parsed, FALSE (and set errno) otherwise
* \note This may add a day to a_time (if the time is 24:00:00).
*/
static bool
crm_time_parse(const char *time_str, crm_time_t *a_time)
{
uint h, m, s;
char *offset_s = NULL;
tzset();
if (time_str) {
if (crm_time_parse_sec(time_str, &(a_time->seconds)) == FALSE) {
return FALSE;
}
offset_s = strstr(time_str, "Z");
if (offset_s == NULL) {
offset_s = strstr(time_str, " ");
if (offset_s) {
while (isspace(offset_s[0])) {
offset_s++;
}
}
}
}
if (crm_time_parse_offset(offset_s, &(a_time->offset)) == FALSE) {
return FALSE;
}
crm_time_get_sec(a_time->offset, &h, &m, &s);
crm_trace("Got tz: %c%2.d:%.2d", ((a_time->offset < 0)? '-' : '+'), h, m);
if (a_time->seconds == DAY_SECONDS) {
// 24:00:00 == 00:00:00 of next day
a_time->seconds = 0;
crm_time_add_days(a_time, 1);
}
return TRUE;
}
/*
* \internal
* \brief Parse a time object from an ISO 8601 date/time specification
*
* \param[in] date_str ISO 8601 date/time specification (or "epoch")
*
* \return New time object on success, NULL (and set errno) otherwise
*/
static crm_time_t *
parse_date(const char *date_str)
{
const char *time_s = NULL;
crm_time_t *dt = NULL;
int year = 0;
int month = 0;
int week = 0;
int day = 0;
int rc = 0;
if (pcmk__str_empty(date_str)) {
crm_err("No ISO 8601 date/time specification given");
goto invalid;
}
if ((date_str[0] == 'T') || (date_str[2] == ':')) {
/* Just a time supplied - Infer current date */
dt = crm_time_new(NULL);
if (date_str[0] == 'T') {
time_s = date_str + 1;
} else {
time_s = date_str;
}
goto parse_time;
}
dt = crm_time_new_undefined();
if (!strncasecmp("epoch", date_str, 5)
&& ((date_str[5] == '\0') || (date_str[5] == '/') || isspace(date_str[5]))) {
dt->days = 1;
dt->years = 1970;
crm_time_log(LOG_TRACE, "Unpacked", dt, crm_time_log_date | crm_time_log_timeofday);
return dt;
}
/* YYYY-MM-DD */
rc = sscanf(date_str, "%d-%d-%d", &year, &month, &day);
if (rc == 1) {
/* YYYYMMDD */
rc = sscanf(date_str, "%4d%2d%2d", &year, &month, &day);
}
if (rc == 3) {
if (month > 12) {
crm_err("'%s' is not a valid ISO 8601 date/time specification "
"because '%d' is not a valid month", date_str, month);
goto invalid;
} else if (day > crm_time_days_in_month(month, year)) {
crm_err("'%s' is not a valid ISO 8601 date/time specification "
"because '%d' is not a valid day of the month",
date_str, day);
goto invalid;
} else {
dt->years = year;
dt->days = get_ordinal_days(year, month, day);
crm_trace("Parsed Gregorian date '%.4d-%.3d' from date string '%s'",
year, dt->days, date_str);
}
goto parse_time;
}
/* YYYY-DDD */
rc = sscanf(date_str, "%d-%d", &year, &day);
if (rc == 2) {
if (day > year_days(year)) {
crm_err("'%s' is not a valid ISO 8601 date/time specification "
"because '%d' is not a valid day of the year (max %d)",
date_str, day, year_days(year));
goto invalid;
}
crm_trace("Parsed ordinal year %d and days %d from date string '%s'",
year, day, date_str);
dt->days = day;
dt->years = year;
goto parse_time;
}
/* YYYY-Www-D */
rc = sscanf(date_str, "%d-W%d-%d", &year, &week, &day);
if (rc == 3) {
if (week > crm_time_weeks_in_year(year)) {
crm_err("'%s' is not a valid ISO 8601 date/time specification "
"because '%d' is not a valid week of the year (max %d)",
date_str, week, crm_time_weeks_in_year(year));
goto invalid;
} else if (day < 1 || day > 7) {
crm_err("'%s' is not a valid ISO 8601 date/time specification "
"because '%d' is not a valid day of the week",
date_str, day);
goto invalid;
} else {
/*
* See https://en.wikipedia.org/wiki/ISO_week_date
*
* Monday 29 December 2008 is written "2009-W01-1"
* Sunday 3 January 2010 is written "2009-W53-7"
* Saturday 27 September 2008 is written "2008-W37-6"
*
* If 1 January is on a Monday, Tuesday, Wednesday or Thursday, it is in week 01.
* If 1 January is on a Friday, Saturday or Sunday, it is in week 52 or 53 of the previous year.
*/
int jan1 = crm_time_january1_weekday(year);
crm_trace("Got year %d (Jan 1 = %d), week %d, and day %d from date string '%s'",
year, jan1, week, day, date_str);
dt->years = year;
crm_time_add_days(dt, (week - 1) * 7);
if (jan1 <= 4) {
crm_time_add_days(dt, 1 - jan1);
} else {
crm_time_add_days(dt, 8 - jan1);
}
crm_time_add_days(dt, day);
}
goto parse_time;
}
crm_err("'%s' is not a valid ISO 8601 date/time specification", date_str);
goto invalid;
parse_time:
if (time_s == NULL) {
time_s = date_str + strspn(date_str, "0123456789-W");
if ((time_s[0] == ' ') || (time_s[0] == 'T')) {
++time_s;
} else {
time_s = NULL;
}
}
if ((time_s != NULL) && (crm_time_parse(time_s, dt) == FALSE)) {
goto invalid;
}
crm_time_log(LOG_TRACE, "Unpacked", dt, crm_time_log_date | crm_time_log_timeofday);
if (crm_time_check(dt) == FALSE) {
crm_err("'%s' is not a valid ISO 8601 date/time specification",
date_str);
goto invalid;
}
return dt;
invalid:
crm_time_free(dt);
errno = EINVAL;
return NULL;
}
// Parse an ISO 8601 numeric value and return number of characters consumed
// @TODO This cannot handle >INT_MAX int values
// @TODO Fractions appear to be not working
// @TODO Error out on invalid specifications
static int
parse_int(const char *str, int field_width, int upper_bound, int *result)
{
int lpc = 0;
int offset = 0;
int intermediate = 0;
gboolean fraction = FALSE;
gboolean negate = FALSE;
*result = 0;
if (*str == '\0') {
return 0;
}
if (str[offset] == 'T') {
offset++;
}
if (str[offset] == '.' || str[offset] == ',') {
fraction = TRUE;
field_width = -1;
offset++;
} else if (str[offset] == '-') {
negate = TRUE;
offset++;
} else if (str[offset] == '+' || str[offset] == ':') {
offset++;
}
for (; (fraction || lpc < field_width) && isdigit((int)str[offset]); lpc++) {
if (fraction) {
intermediate = (str[offset] - '0') / (10 ^ lpc);
} else {
*result *= 10;
intermediate = str[offset] - '0';
}
*result += intermediate;
offset++;
}
if (fraction) {
*result = (int)(*result * upper_bound);
} else if (upper_bound > 0 && *result > upper_bound) {
*result = upper_bound;
}
if (negate) {
*result = 0 - *result;
}
if (lpc > 0) {
crm_trace("Found int: %d. Stopped at str[%d]='%c'", *result, lpc, str[lpc]);
return offset;
}
return 0;
}
/*!
* \brief Parse a time duration from an ISO 8601 duration specification
*
* \param[in] period_s ISO 8601 duration specification (optionally followed by
* whitespace, after which the rest of the string will be
* ignored)
*
* \return New time object on success, NULL (and set errno) otherwise
* \note It is the caller's responsibility to return the result using
* crm_time_free().
*/
crm_time_t *
crm_time_parse_duration(const char *period_s)
{
gboolean is_time = FALSE;
crm_time_t *diff = NULL;
if (pcmk__str_empty(period_s)) {
crm_err("No ISO 8601 time duration given");
goto invalid;
}
if (period_s[0] != 'P') {
crm_err("'%s' is not a valid ISO 8601 time duration "
"because it does not start with a 'P'", period_s);
goto invalid;
}
if ((period_s[1] == '\0') || isspace(period_s[1])) {
crm_err("'%s' is not a valid ISO 8601 time duration "
"because nothing follows 'P'", period_s);
goto invalid;
}
diff = crm_time_new_undefined();
diff->duration = TRUE;
for (const char *current = period_s + 1;
current[0] && (current[0] != '/') && !isspace(current[0]);
++current) {
int an_int = 0, rc;
if (current[0] == 'T') {
/* A 'T' separates year/month/day from hour/minute/seconds. We don't
* require it strictly, but just use it to differentiate month from
* minutes.
*/
is_time = TRUE;
continue;
}
// An integer must be next
rc = parse_int(current, 10, 0, &an_int);
if (rc == 0) {
crm_err("'%s' is not a valid ISO 8601 time duration "
"because no integer at '%s'", period_s, current);
goto invalid;
}
current += rc;
// A time unit must be next (we're not strict about the order)
switch (current[0]) {
case 'Y':
diff->years = an_int;
break;
case 'M':
if (is_time) {
/* Minutes */
diff->seconds += an_int * 60;
} else {
diff->months = an_int;
}
break;
case 'W':
diff->days += an_int * 7;
break;
case 'D':
diff->days += an_int;
break;
case 'H':
diff->seconds += an_int * HOUR_SECONDS;
break;
case 'S':
diff->seconds += an_int;
break;
case '\0':
crm_err("'%s' is not a valid ISO 8601 time duration "
"because no units after %d", period_s, an_int);
goto invalid;
default:
crm_err("'%s' is not a valid ISO 8601 time duration "
"because '%c' is not a valid time unit",
period_s, current[0]);
goto invalid;
}
}
if (!crm_time_is_defined(diff)) {
crm_err("'%s' is not a valid ISO 8601 time duration "
"because no amounts and units given", period_s);
goto invalid;
}
return diff;
invalid:
crm_time_free(diff);
errno = EINVAL;
return NULL;
}
/*!
* \brief Parse a time period from an ISO 8601 interval specification
*
* \param[in] period_str ISO 8601 interval specification (start/end,
* start/duration, or duration/end)
*
* \return New time period object on success, NULL (and set errno) otherwise
* \note The caller is responsible for freeing the result using
* crm_time_free_period().
*/
crm_time_period_t *
crm_time_parse_period(const char *period_str)
{
const char *original = period_str;
crm_time_period_t *period = NULL;
if (pcmk__str_empty(period_str)) {
crm_err("No ISO 8601 time period given");
goto invalid;
}
tzset();
period = calloc(1, sizeof(crm_time_period_t));
CRM_ASSERT(period != NULL);
if (period_str[0] == 'P') {
period->diff = crm_time_parse_duration(period_str);
if (period->diff == NULL) {
goto error;
}
} else {
period->start = parse_date(period_str);
if (period->start == NULL) {
goto error;
}
}
period_str = strstr(original, "/");
if (period_str) {
++period_str;
if (period_str[0] == 'P') {
if (period->diff != NULL) {
crm_err("'%s' is not a valid ISO 8601 time period "
"because it has two durations",
original);
goto invalid;
}
period->diff = crm_time_parse_duration(period_str);
if (period->diff == NULL) {
goto error;
}
} else {
period->end = parse_date(period_str);
if (period->end == NULL) {
goto error;
}
}
} else if (period->diff != NULL) {
// Only duration given, assume start is now
period->start = crm_time_new(NULL);
} else {
// Only start given
crm_err("'%s' is not a valid ISO 8601 time period "
"because it has no duration or ending time",
original);
goto invalid;
}
if (period->start == NULL) {
period->start = crm_time_subtract(period->end, period->diff);
} else if (period->end == NULL) {
period->end = crm_time_add(period->start, period->diff);
}
if (crm_time_check(period->start) == FALSE) {
crm_err("'%s' is not a valid ISO 8601 time period "
"because the start is invalid", period_str);
goto invalid;
}
if (crm_time_check(period->end) == FALSE) {
crm_err("'%s' is not a valid ISO 8601 time period "
"because the end is invalid", period_str);
goto invalid;
}
return period;
invalid:
errno = EINVAL;
error:
crm_time_free_period(period);
return NULL;
}
/*!
* \brief Free a dynamically allocated time period object
*
* \param[in] period Time period to free
*/
void
crm_time_free_period(crm_time_period_t *period)
{
if (period) {
crm_time_free(period->start);
crm_time_free(period->end);
crm_time_free(period->diff);
free(period);
}
}
void
crm_time_set(crm_time_t * target, crm_time_t * source)
{
crm_trace("target=%p, source=%p", target, source);
CRM_CHECK(target != NULL && source != NULL, return);
target->years = source->years;
target->days = source->days;
target->months = source->months; /* Only for durations */
target->seconds = source->seconds;
target->offset = source->offset;
crm_time_log(LOG_TRACE, "source", source,
crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone);
crm_time_log(LOG_TRACE, "target", target,
crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone);
}
static void
ha_set_tm_time(crm_time_t * target, struct tm *source)
{
int h_offset = 0;
int m_offset = 0;
/* Ensure target is fully initialized */
target->years = 0;
target->months = 0;
target->days = 0;
target->seconds = 0;
target->offset = 0;
target->duration = FALSE;
if (source->tm_year > 0) {
/* years since 1900 */
target->years = 1900 + source->tm_year;
}
if (source->tm_yday >= 0) {
/* days since January 1 [0-365] */
target->days = 1 + source->tm_yday;
}
if (source->tm_hour >= 0) {
target->seconds += HOUR_SECONDS * source->tm_hour;
}
if (source->tm_min >= 0) {
target->seconds += 60 * source->tm_min;
}
if (source->tm_sec >= 0) {
target->seconds += source->tm_sec;
}
/* tm_gmtoff == offset from UTC in seconds */
h_offset = GMTOFF(source) / HOUR_SECONDS;
m_offset = (GMTOFF(source) - (HOUR_SECONDS * h_offset)) / 60;
crm_trace("Offset (s): %ld, offset (hh:mm): %.2d:%.2d", GMTOFF(source), h_offset, m_offset);
target->offset += HOUR_SECONDS * h_offset;
target->offset += 60 * m_offset;
}
void
crm_time_set_timet(crm_time_t * target, time_t * source)
{
ha_set_tm_time(target, localtime(source));
}
+crm_time_t *
+pcmk_copy_time(crm_time_t *source)
+{
+ crm_time_t *target = crm_time_new_undefined();
+
+ crm_time_set(target, source);
+ return target;
+}
+
crm_time_t *
crm_time_add(crm_time_t * dt, crm_time_t * value)
{
crm_time_t *utc = NULL;
crm_time_t *answer = NULL;
if ((dt == NULL) || (value == NULL)) {
errno = EINVAL;
return NULL;
}
- answer = crm_time_new_undefined();
- crm_time_set(answer, dt);
+ answer = pcmk_copy_time(dt);
utc = crm_get_utc_time(value);
if (utc == NULL) {
crm_time_free(answer);
return NULL;
}
answer->years += utc->years;
crm_time_add_months(answer, utc->months);
crm_time_add_days(answer, utc->days);
crm_time_add_seconds(answer, utc->seconds);
crm_time_free(utc);
return answer;
}
crm_time_t *
crm_time_calculate_duration(crm_time_t * dt, crm_time_t * value)
{
crm_time_t *utc = NULL;
crm_time_t *answer = NULL;
if ((dt == NULL) || (value == NULL)) {
errno = EINVAL;
return NULL;
}
utc = crm_get_utc_time(value);
if (utc == NULL) {
return NULL;
}
answer = crm_get_utc_time(dt);
if (answer == NULL) {
crm_time_free(utc);
return NULL;
}
answer->duration = TRUE;
answer->years -= utc->years;
if(utc->months != 0) {
crm_time_add_months(answer, -utc->months);
}
crm_time_add_days(answer, -utc->days);
crm_time_add_seconds(answer, -utc->seconds);
crm_time_free(utc);
return answer;
}
crm_time_t *
crm_time_subtract(crm_time_t * dt, crm_time_t * value)
{
crm_time_t *utc = NULL;
crm_time_t *answer = NULL;
if ((dt == NULL) || (value == NULL)) {
errno = EINVAL;
return NULL;
}
utc = crm_get_utc_time(value);
if (utc == NULL) {
return NULL;
}
- answer = crm_time_new_undefined();
- crm_time_set(answer, dt);
+ answer = pcmk_copy_time(dt);
answer->years -= utc->years;
if(utc->months != 0) {
crm_time_add_months(answer, -utc->months);
}
crm_time_add_days(answer, -utc->days);
crm_time_add_seconds(answer, -utc->seconds);
return answer;
}
/*!
* \brief Check whether a time object represents a sensible date/time
*
* \param[in] dt Date/time object to check
*
* \return TRUE if years, days, and seconds are sensible, FALSE otherwise
*/
bool
crm_time_check(crm_time_t * dt)
{
return (dt != NULL)
&& (dt->days > 0) && (dt->days <= year_days(dt->years))
&& (dt->seconds >= 0) && (dt->seconds < DAY_SECONDS);
}
#define do_cmp_field(l, r, field) \
if(rc == 0) { \
if(l->field > r->field) { \
crm_trace("%s: %d > %d", \
#field, l->field, r->field); \
rc = 1; \
} else if(l->field < r->field) { \
crm_trace("%s: %d < %d", \
#field, l->field, r->field); \
rc = -1; \
} \
}
int
crm_time_compare(crm_time_t *a, crm_time_t *b)
{
int rc = 0;
crm_time_t *t1 = crm_get_utc_time(a);
crm_time_t *t2 = crm_get_utc_time(b);
if ((t1 == NULL) && (t2 == NULL)) {
rc = 0;
} else if (t1 == NULL) {
rc = -1;
} else if (t2 == NULL) {
rc = 1;
} else {
do_cmp_field(t1, t2, years);
do_cmp_field(t1, t2, days);
do_cmp_field(t1, t2, seconds);
}
crm_time_free(t1);
crm_time_free(t2);
return rc;
}
/*!
* \brief Add a given number of seconds to a date/time or duration
*
* \param[in] a_time Date/time or duration to add seconds to
* \param[in] extra Number of seconds to add
*/
void
crm_time_add_seconds(crm_time_t *a_time, int extra)
{
int days = 0;
crm_trace("Adding %d seconds to %d (max=%d)",
extra, a_time->seconds, DAY_SECONDS);
a_time->seconds += extra;
days = a_time->seconds / DAY_SECONDS;
a_time->seconds %= DAY_SECONDS;
// Don't have negative seconds
if (a_time->seconds < 0) {
a_time->seconds += DAY_SECONDS;
--days;
}
crm_time_add_days(a_time, days);
}
void
crm_time_add_days(crm_time_t * a_time, int extra)
{
int lower_bound = 1;
int ydays = crm_time_leapyear(a_time->years) ? 366 : 365;
crm_trace("Adding %d days to %.4d-%.3d", extra, a_time->years, a_time->days);
a_time->days += extra;
while (a_time->days > ydays) {
a_time->years++;
a_time->days -= ydays;
ydays = crm_time_leapyear(a_time->years) ? 366 : 365;
}
if(a_time->duration) {
lower_bound = 0;
}
while (a_time->days < lower_bound) {
a_time->years--;
a_time->days += crm_time_leapyear(a_time->years) ? 366 : 365;
}
}
void
crm_time_add_months(crm_time_t * a_time, int extra)
{
int lpc;
uint32_t y, m, d, dmax;
crm_time_get_gregorian(a_time, &y, &m, &d);
crm_trace("Adding %d months to %.4d-%.2d-%.2d", extra, y, m, d);
if (extra > 0) {
for (lpc = extra; lpc > 0; lpc--) {
m++;
if (m == 13) {
m = 1;
y++;
}
}
} else {
for (lpc = -extra; lpc > 0; lpc--) {
m--;
if (m == 0) {
m = 12;
y--;
}
}
}
dmax = crm_time_days_in_month(m, y);
if (dmax < d) {
/* Preserve day-of-month unless the month doesn't have enough days */
d = dmax;
}
crm_trace("Calculated %.4d-%.2d-%.2d", y, m, d);
a_time->years = y;
a_time->days = get_ordinal_days(y, m, d);
crm_time_get_gregorian(a_time, &y, &m, &d);
crm_trace("Got %.4d-%.2d-%.2d", y, m, d);
}
void
crm_time_add_minutes(crm_time_t * a_time, int extra)
{
crm_time_add_seconds(a_time, extra * 60);
}
void
crm_time_add_hours(crm_time_t * a_time, int extra)
{
crm_time_add_seconds(a_time, extra * HOUR_SECONDS);
}
void
crm_time_add_weeks(crm_time_t * a_time, int extra)
{
crm_time_add_days(a_time, extra * 7);
}
void
crm_time_add_years(crm_time_t * a_time, int extra)
{
a_time->years += extra;
}
static void
ha_get_tm_time( struct tm *target, crm_time_t *source)
{
*target = (struct tm) {
.tm_year = source->years - 1900,
.tm_mday = source->days,
.tm_sec = source->seconds % 60,
.tm_min = ( source->seconds / 60 ) % 60,
.tm_hour = source->seconds / HOUR_SECONDS,
.tm_isdst = -1, /* don't adjust */
#if defined(HAVE_STRUCT_TM_TM_GMTOFF)
.tm_gmtoff = source->offset
#endif
};
mktime(target);
}
/* The high-resolution variant of time object was added to meet an immediate
* need, and is kept internal API.
*
* @TODO The long-term goal is to come up with a clean, unified design for a
* time type (or types) that meets all the various needs, to replace
* crm_time_t, pcmk__time_hr_t, and struct timespec (in lrmd_cmd_t).
* Using glib's GDateTime is a possibility (if we are willing to require
* glib >= 2.26).
*/
pcmk__time_hr_t *
pcmk__time_hr_convert(pcmk__time_hr_t *target, crm_time_t *dt)
{
pcmk__time_hr_t *hr_dt = NULL;
if (dt) {
hr_dt = target?target:calloc(1, sizeof(pcmk__time_hr_t));
CRM_ASSERT(hr_dt != NULL);
*hr_dt = (pcmk__time_hr_t) {
.years = dt->years,
.months = dt->months,
.days = dt->days,
.seconds = dt->seconds,
.offset = dt->offset,
.duration = dt->duration
};
}
return hr_dt;
}
void
pcmk__time_set_hr_dt(crm_time_t *target, pcmk__time_hr_t *hr_dt)
{
CRM_ASSERT((hr_dt) && (target));
*target = (crm_time_t) {
.years = hr_dt->years,
.months = hr_dt->months,
.days = hr_dt->days,
.seconds = hr_dt->seconds,
.offset = hr_dt->offset,
.duration = hr_dt->duration
};
}
pcmk__time_hr_t *
pcmk__time_timeval_hr_convert(pcmk__time_hr_t *target, struct timeval *tv)
{
crm_time_t dt;
pcmk__time_hr_t *ret;
crm_time_set_timet(&dt, &tv->tv_sec);
ret = pcmk__time_hr_convert(target, &dt);
if (ret) {
ret->useconds = tv->tv_usec;
}
return ret;
}
pcmk__time_hr_t *
pcmk__time_hr_new(const char *date_time)
{
pcmk__time_hr_t *hr_dt = NULL;
struct timeval tv_now;
if (!date_time) {
if (gettimeofday(&tv_now, NULL) == 0) {
hr_dt = pcmk__time_timeval_hr_convert(NULL, &tv_now);
}
} else {
crm_time_t *dt;
dt = parse_date(date_time);
hr_dt = pcmk__time_hr_convert(NULL, dt);
crm_time_free(dt);
}
return hr_dt;
}
void
pcmk__time_hr_free(pcmk__time_hr_t * hr_dt)
{
free(hr_dt);
}
char *
pcmk__time_format_hr(const char *format, pcmk__time_hr_t * hr_dt)
{
const char *mark_s;
int max = 128, scanned_pos = 0, printed_pos = 0, fmt_pos = 0,
date_len = 0, nano_digits = 0;
char nano_s[10], date_s[max+1], nanofmt_s[5] = "%", *tmp_fmt_s;
struct tm tm;
crm_time_t dt;
if (!format) {
return NULL;
}
pcmk__time_set_hr_dt(&dt, hr_dt);
ha_get_tm_time(&tm, &dt);
sprintf(nano_s, "%06d000", hr_dt->useconds);
while ((format[scanned_pos]) != '\0') {
mark_s = strchr(&format[scanned_pos], '%');
if (mark_s) {
int fmt_len = 1;
fmt_pos = mark_s - format;
while ((format[fmt_pos+fmt_len] != '\0') &&
(format[fmt_pos+fmt_len] >= '0') &&
(format[fmt_pos+fmt_len] <= '9')) {
fmt_len++;
}
scanned_pos = fmt_pos + fmt_len + 1;
if (format[fmt_pos+fmt_len] == 'N') {
nano_digits = atoi(&format[fmt_pos+1]);
nano_digits = (nano_digits > 6)?6:nano_digits;
nano_digits = (nano_digits < 0)?0:nano_digits;
sprintf(&nanofmt_s[1], ".%ds", nano_digits);
} else {
if (format[scanned_pos] != '\0') {
continue;
}
fmt_pos = scanned_pos; /* print till end */
}
} else {
scanned_pos = strlen(format);
fmt_pos = scanned_pos; /* print till end */
}
tmp_fmt_s = strndup(&format[printed_pos], fmt_pos - printed_pos);
#ifdef GCC_FORMAT_NONLITERAL_CHECKING_ENABLED
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wformat-nonliteral"
#endif
date_len += strftime(&date_s[date_len], max-date_len, tmp_fmt_s, &tm);
#ifdef GCC_FORMAT_NONLITERAL_CHECKING_ENABLED
#pragma GCC diagnostic pop
#endif
printed_pos = scanned_pos;
free(tmp_fmt_s);
if (nano_digits) {
#ifdef GCC_FORMAT_NONLITERAL_CHECKING_ENABLED
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wformat-nonliteral"
#endif
date_len += snprintf(&date_s[date_len], max-date_len,
nanofmt_s, nano_s);
#ifdef GCC_FORMAT_NONLITERAL_CHECKING_ENABLED
#pragma GCC diagnostic pop
#endif
nano_digits = 0;
}
}
return (date_len == 0)?NULL:strdup(date_s);
}
/*!
* \internal
* \brief Return human-friendly string corresponding to a time
*
* \param[in] when Pointer to epoch time value (or NULL for current time)
*
* \return Current time as string (as by ctime() but without newline) on
* success, NULL otherwise
* \note The return value points to a statically allocated string which might be
* overwritten by subsequent calls to any of the C library date and time
* functions.
*/
const char *
pcmk__epoch2str(time_t *when)
{
char *since_epoch = NULL;
if (when == NULL) {
time_t a_time = time(NULL);
if (a_time == (time_t) -1) {
return NULL;
} else {
since_epoch = ctime(&a_time);
}
} else {
since_epoch = ctime(when);
}
if (since_epoch == NULL) {
return NULL;
} else {
return crm_strip_trailing_newline(since_epoch);
}
}
diff --git a/lib/pacemaker/pcmk_sched_allocate.c b/lib/pacemaker/pcmk_sched_allocate.c
index 2c227672c8..2bb388c610 100644
--- a/lib/pacemaker/pcmk_sched_allocate.c
+++ b/lib/pacemaker/pcmk_sched_allocate.c
@@ -1,3068 +1,3058 @@
/*
* Copyright 2004-2020 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 <sys/param.h>
#include <crm/crm.h>
#include <crm/cib.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <glib.h>
#include <crm/pengine/status.h>
#include <pacemaker-internal.h>
CRM_TRACE_INIT_DATA(pacemaker);
void set_alloc_actions(pe_working_set_t * data_set);
extern void ReloadRsc(pe_resource_t * rsc, pe_node_t *node, pe_working_set_t * data_set);
extern gboolean DeleteRsc(pe_resource_t * rsc, pe_node_t * node, gboolean optional, pe_working_set_t * data_set);
static void apply_remote_node_ordering(pe_working_set_t *data_set);
static enum remote_connection_state get_remote_node_state(pe_node_t *node);
enum remote_connection_state {
remote_state_unknown = 0,
remote_state_alive = 1,
remote_state_resting = 2,
remote_state_failed = 3,
remote_state_stopped = 4
};
static const char *
state2text(enum remote_connection_state state)
{
switch (state) {
case remote_state_unknown:
return "unknown";
case remote_state_alive:
return "alive";
case remote_state_resting:
return "resting";
case remote_state_failed:
return "failed";
case remote_state_stopped:
return "stopped";
}
return "impossible";
}
resource_alloc_functions_t resource_class_alloc_functions[] = {
{
pcmk__native_merge_weights,
pcmk__native_allocate,
native_create_actions,
native_create_probe,
native_internal_constraints,
native_rsc_colocation_lh,
native_rsc_colocation_rh,
native_rsc_location,
native_action_flags,
native_update_actions,
native_expand,
native_append_meta,
},
{
pcmk__group_merge_weights,
pcmk__group_allocate,
group_create_actions,
native_create_probe,
group_internal_constraints,
group_rsc_colocation_lh,
group_rsc_colocation_rh,
group_rsc_location,
group_action_flags,
group_update_actions,
group_expand,
group_append_meta,
},
{
pcmk__native_merge_weights,
pcmk__clone_allocate,
clone_create_actions,
clone_create_probe,
clone_internal_constraints,
clone_rsc_colocation_lh,
clone_rsc_colocation_rh,
clone_rsc_location,
clone_action_flags,
pcmk__multi_update_actions,
clone_expand,
clone_append_meta,
},
{
pcmk__native_merge_weights,
pcmk__bundle_allocate,
pcmk__bundle_create_actions,
pcmk__bundle_create_probe,
pcmk__bundle_internal_constraints,
pcmk__bundle_rsc_colocation_lh,
pcmk__bundle_rsc_colocation_rh,
pcmk__bundle_rsc_location,
pcmk__bundle_action_flags,
pcmk__multi_update_actions,
pcmk__bundle_expand,
pcmk__bundle_append_meta,
}
};
gboolean
update_action_flags(pe_action_t * action, enum pe_action_flags flags, const char *source, int line)
{
static unsigned long calls = 0;
gboolean changed = FALSE;
gboolean clear = is_set(flags, pe_action_clear);
enum pe_action_flags last = action->flags;
if (clear) {
action->flags = crm_clear_bit(source, line, action->uuid, action->flags, flags);
} else {
action->flags = crm_set_bit(source, line, action->uuid, action->flags, flags);
}
if (last != action->flags) {
calls++;
changed = TRUE;
/* Useful for tracking down _who_ changed a specific flag */
/* CRM_ASSERT(calls != 534); */
clear_bit(flags, pe_action_clear);
crm_trace("%s on %s: %sset flags 0x%.6x (was 0x%.6x, now 0x%.6x, %lu, %s)",
action->uuid, action->node ? action->node->details->uname : "[none]",
clear ? "un-" : "", flags, last, action->flags, calls, source);
}
return changed;
}
static gboolean
check_rsc_parameters(pe_resource_t * rsc, pe_node_t * node, xmlNode * rsc_entry,
gboolean active_here, pe_working_set_t * data_set)
{
int attr_lpc = 0;
gboolean force_restart = FALSE;
gboolean delete_resource = FALSE;
gboolean changed = FALSE;
const char *value = NULL;
const char *old_value = NULL;
const char *attr_list[] = {
XML_ATTR_TYPE,
XML_AGENT_ATTR_CLASS,
XML_AGENT_ATTR_PROVIDER
};
for (; attr_lpc < DIMOF(attr_list); attr_lpc++) {
value = crm_element_value(rsc->xml, attr_list[attr_lpc]);
old_value = crm_element_value(rsc_entry, attr_list[attr_lpc]);
if (value == old_value /* i.e. NULL */
|| crm_str_eq(value, old_value, TRUE)) {
continue;
}
changed = TRUE;
trigger_unfencing(rsc, node, "Device definition changed", NULL, data_set);
if (active_here) {
force_restart = TRUE;
crm_notice("Forcing restart of %s on %s, %s changed: %s -> %s",
rsc->id, node->details->uname, attr_list[attr_lpc],
crm_str(old_value), crm_str(value));
}
}
if (force_restart) {
/* make sure the restart happens */
stop_action(rsc, node, FALSE);
set_bit(rsc->flags, pe_rsc_start_pending);
delete_resource = TRUE;
} else if (changed) {
delete_resource = TRUE;
}
return delete_resource;
}
static void
CancelXmlOp(pe_resource_t * rsc, xmlNode * xml_op, pe_node_t * active_node,
const char *reason, pe_working_set_t * data_set)
{
guint interval_ms = 0;
pe_action_t *cancel = NULL;
const char *task = NULL;
const char *call_id = NULL;
CRM_CHECK(xml_op != NULL, return);
CRM_CHECK(active_node != NULL, return);
task = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
call_id = crm_element_value(xml_op, XML_LRM_ATTR_CALLID);
crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
crm_info("Action " PCMK__OP_FMT " on %s will be stopped: %s",
rsc->id, task, interval_ms,
active_node->details->uname, (reason? reason : "unknown"));
cancel = pe_cancel_op(rsc, task, interval_ms, active_node, data_set);
add_hash_param(cancel->meta, XML_LRM_ATTR_CALLID, call_id);
custom_action_order(rsc, stop_key(rsc), NULL, rsc, NULL, cancel, pe_order_optional, data_set);
}
static gboolean
check_action_definition(pe_resource_t * rsc, pe_node_t * active_node, xmlNode * xml_op,
pe_working_set_t * data_set)
{
char *key = NULL;
guint interval_ms = 0;
const op_digest_cache_t *digest_data = NULL;
gboolean did_change = FALSE;
const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
const char *digest_secure = NULL;
CRM_CHECK(active_node != NULL, return FALSE);
crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
if (interval_ms > 0) {
xmlNode *op_match = NULL;
/* we need to reconstruct the key because of the way we used to construct resource IDs */
key = pcmk__op_key(rsc->id, task, interval_ms);
pe_rsc_trace(rsc, "Checking parameters for %s", key);
op_match = find_rsc_op_entry(rsc, key);
if (op_match == NULL && is_set(data_set->flags, pe_flag_stop_action_orphans)) {
CancelXmlOp(rsc, xml_op, active_node, "orphan", data_set);
free(key);
return TRUE;
} else if (op_match == NULL) {
pe_rsc_debug(rsc, "Orphan action detected: %s on %s", key, active_node->details->uname);
free(key);
return TRUE;
}
free(key);
key = NULL;
}
crm_trace("Testing " PCMK__OP_FMT " on %s",
rsc->id, task, interval_ms, active_node->details->uname);
if ((interval_ms == 0) && safe_str_eq(task, RSC_STATUS)) {
/* Reload based on the start action not a probe */
task = RSC_START;
} else if ((interval_ms == 0) && safe_str_eq(task, RSC_MIGRATED)) {
/* Reload based on the start action not a migrate */
task = RSC_START;
} else if ((interval_ms == 0) && safe_str_eq(task, RSC_PROMOTE)) {
/* Reload based on the start action not a promote */
task = RSC_START;
}
digest_data = rsc_action_digest_cmp(rsc, xml_op, active_node, data_set);
if(is_set(data_set->flags, pe_flag_sanitized)) {
digest_secure = crm_element_value(xml_op, XML_LRM_ATTR_SECURE_DIGEST);
}
if(digest_data->rc != RSC_DIGEST_MATCH
&& digest_secure
&& digest_data->digest_secure_calc
&& strcmp(digest_data->digest_secure_calc, digest_secure) == 0) {
if (is_set(data_set->flags, pe_flag_stdout)) {
printf("Only 'private' parameters to " PCMK__OP_FMT
" on %s changed: %s\n",
rsc->id, task, interval_ms, active_node->details->uname,
crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC));
}
} else if (digest_data->rc == RSC_DIGEST_RESTART) {
/* Changes that force a restart */
pe_action_t *required = NULL;
did_change = TRUE;
key = pcmk__op_key(rsc->id, task, interval_ms);
crm_log_xml_info(digest_data->params_restart, "params:restart");
required = custom_action(rsc, key, task, NULL, TRUE, TRUE, data_set);
pe_action_set_flag_reason(__FUNCTION__, __LINE__, required, NULL,
"resource definition change", pe_action_optional, TRUE);
trigger_unfencing(rsc, active_node, "Device parameters changed", NULL, data_set);
} else if ((digest_data->rc == RSC_DIGEST_ALL) || (digest_data->rc == RSC_DIGEST_UNKNOWN)) {
/* Changes that can potentially be handled by a reload */
const char *digest_restart = crm_element_value(xml_op, XML_LRM_ATTR_RESTART_DIGEST);
did_change = TRUE;
trigger_unfencing(rsc, active_node, "Device parameters changed (reload)", NULL, data_set);
crm_log_xml_info(digest_data->params_all, "params:reload");
key = pcmk__op_key(rsc->id, task, interval_ms);
if (interval_ms > 0) {
pe_action_t *op = NULL;
#if 0
/* Always reload/restart the entire resource */
ReloadRsc(rsc, active_node, data_set);
#else
/* Re-sending the recurring op is sufficient - the old one will be cancelled automatically */
op = custom_action(rsc, key, task, active_node, TRUE, TRUE, data_set);
set_bit(op->flags, pe_action_reschedule);
#endif
} else if (digest_restart) {
pe_rsc_trace(rsc, "Reloading '%s' action for resource %s", task, rsc->id);
/* Reload this resource */
ReloadRsc(rsc, active_node, data_set);
free(key);
} else {
pe_action_t *required = NULL;
pe_rsc_trace(rsc, "Resource %s doesn't know how to reload", rsc->id);
/* Re-send the start/demote/promote op
* Recurring ops will be detected independently
*/
required = custom_action(rsc, key, task, NULL, TRUE, TRUE, data_set);
pe_action_set_flag_reason(__FUNCTION__, __LINE__, required, NULL,
"resource definition change", pe_action_optional, TRUE);
}
}
return did_change;
}
/*!
* \internal
* \brief Do deferred action checks after allocation
*
* \param[in] data_set Working set for cluster
*/
static void
check_params(pe_resource_t *rsc, pe_node_t *node, xmlNode *rsc_op,
enum pe_check_parameters check, pe_working_set_t *data_set)
{
const char *reason = NULL;
op_digest_cache_t *digest_data = NULL;
switch (check) {
case pe_check_active:
if (check_action_definition(rsc, node, rsc_op, data_set)
&& pe_get_failcount(node, rsc, NULL, pe_fc_effective, NULL,
data_set)) {
reason = "action definition changed";
}
break;
case pe_check_last_failure:
digest_data = rsc_action_digest_cmp(rsc, rsc_op, node, data_set);
switch (digest_data->rc) {
case RSC_DIGEST_UNKNOWN:
crm_trace("Resource %s history entry %s on %s has no digest to compare",
rsc->id, ID(rsc_op), node->details->id);
break;
case RSC_DIGEST_MATCH:
break;
default:
reason = "resource parameters have changed";
break;
}
break;
}
if (reason) {
pe__clear_failcount(rsc, node, reason, data_set);
}
}
static void
check_actions_for(xmlNode * rsc_entry, pe_resource_t * rsc, pe_node_t * node, pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
int offset = -1;
int stop_index = 0;
int start_index = 0;
const char *task = NULL;
xmlNode *rsc_op = NULL;
GListPtr op_list = NULL;
GListPtr sorted_op_list = NULL;
CRM_CHECK(node != NULL, return);
if (is_set(rsc->flags, pe_rsc_orphan)) {
pe_resource_t *parent = uber_parent(rsc);
if(parent == NULL
|| pe_rsc_is_clone(parent) == FALSE
|| is_set(parent->flags, pe_rsc_unique)) {
pe_rsc_trace(rsc, "Skipping param check for %s and deleting: orphan", rsc->id);
DeleteRsc(rsc, node, FALSE, data_set);
} else {
pe_rsc_trace(rsc, "Skipping param check for %s (orphan clone)", rsc->id);
}
return;
} else if (pe_find_node_id(rsc->running_on, node->details->id) == NULL) {
if (check_rsc_parameters(rsc, node, rsc_entry, FALSE, data_set)) {
DeleteRsc(rsc, node, FALSE, data_set);
}
pe_rsc_trace(rsc, "Skipping param check for %s: no longer active on %s",
rsc->id, node->details->uname);
return;
}
pe_rsc_trace(rsc, "Processing %s on %s", rsc->id, node->details->uname);
if (check_rsc_parameters(rsc, node, rsc_entry, TRUE, data_set)) {
DeleteRsc(rsc, node, FALSE, data_set);
}
for (rsc_op = __xml_first_child_element(rsc_entry); rsc_op != NULL;
rsc_op = __xml_next_element(rsc_op)) {
if (crm_str_eq((const char *)rsc_op->name, XML_LRM_TAG_RSC_OP, TRUE)) {
op_list = g_list_prepend(op_list, rsc_op);
}
}
sorted_op_list = g_list_sort(op_list, sort_op_by_callid);
calculate_active_ops(sorted_op_list, &start_index, &stop_index);
for (gIter = sorted_op_list; gIter != NULL; gIter = gIter->next) {
xmlNode *rsc_op = (xmlNode *) gIter->data;
guint interval_ms = 0;
offset++;
if (start_index < stop_index) {
/* stopped */
continue;
} else if (offset < start_index) {
/* action occurred prior to a start */
continue;
}
task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK);
crm_element_value_ms(rsc_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
if ((interval_ms > 0) &&
(is_set(rsc->flags, pe_rsc_maintenance) || node->details->maintenance)) {
// Maintenance mode cancels recurring operations
CancelXmlOp(rsc, rsc_op, node, "maintenance mode", data_set);
} else if ((interval_ms > 0)
|| safe_str_eq(task, RSC_STATUS)
|| safe_str_eq(task, RSC_START)
|| safe_str_eq(task, RSC_PROMOTE)
|| safe_str_eq(task, RSC_MIGRATED)) {
/* If a resource operation failed, and the operation's definition
* has changed, clear any fail count so they can be retried fresh.
*/
if (pe__bundle_needs_remote_name(rsc)) {
/* We haven't allocated resources to nodes yet, so if the
* REMOTE_CONTAINER_HACK is used, we may calculate the digest
* based on the literal "#uname" value rather than the properly
* substituted value. That would mistakenly make the action
* definition appear to have been changed. Defer the check until
* later in this case.
*/
pe__add_param_check(rsc_op, rsc, node, pe_check_active,
data_set);
} else if (check_action_definition(rsc, node, rsc_op, data_set)
&& pe_get_failcount(node, rsc, NULL, pe_fc_effective, NULL,
data_set)) {
pe__clear_failcount(rsc, node, "action definition changed",
data_set);
}
}
}
g_list_free(sorted_op_list);
}
static GListPtr
find_rsc_list(GListPtr result, pe_resource_t * rsc, const char *id, gboolean renamed_clones,
gboolean partial, pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
gboolean match = FALSE;
if (id == NULL) {
return NULL;
}
if (rsc == NULL) {
if (data_set == NULL) {
return NULL;
}
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child = (pe_resource_t *) gIter->data;
result = find_rsc_list(result, child, id, renamed_clones, partial,
NULL);
}
return result;
}
if (partial) {
if (strstr(rsc->id, id)) {
match = TRUE;
} else if (renamed_clones && rsc->clone_name && strstr(rsc->clone_name, id)) {
match = TRUE;
}
} else {
if (strcmp(rsc->id, id) == 0) {
match = TRUE;
} else if (renamed_clones && rsc->clone_name && strcmp(rsc->clone_name, id) == 0) {
match = TRUE;
}
}
if (match) {
result = g_list_prepend(result, rsc);
}
if (rsc->children) {
gIter = rsc->children;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child = (pe_resource_t *) gIter->data;
result = find_rsc_list(result, child, id, renamed_clones, partial, NULL);
}
}
return result;
}
static void
check_actions(pe_working_set_t * data_set)
{
const char *id = NULL;
pe_node_t *node = NULL;
xmlNode *lrm_rscs = NULL;
xmlNode *status = get_object_root(XML_CIB_TAG_STATUS, data_set->input);
xmlNode *node_state = NULL;
for (node_state = __xml_first_child_element(status); node_state != NULL;
node_state = __xml_next_element(node_state)) {
if (crm_str_eq((const char *)node_state->name, XML_CIB_TAG_STATE, TRUE)) {
id = crm_element_value(node_state, XML_ATTR_ID);
lrm_rscs = find_xml_node(node_state, XML_CIB_TAG_LRM, FALSE);
lrm_rscs = find_xml_node(lrm_rscs, XML_LRM_TAG_RESOURCES, FALSE);
node = pe_find_node_id(data_set->nodes, id);
if (node == NULL) {
continue;
/* Still need to check actions for a maintenance node to cancel existing monitor operations */
} else if (can_run_resources(node) == FALSE && node->details->maintenance == FALSE) {
crm_trace("Skipping param check for %s: can't run resources",
node->details->uname);
continue;
}
crm_trace("Processing node %s", node->details->uname);
if (node->details->online || is_set(data_set->flags, pe_flag_stonith_enabled)) {
xmlNode *rsc_entry = NULL;
for (rsc_entry = __xml_first_child_element(lrm_rscs);
rsc_entry != NULL;
rsc_entry = __xml_next_element(rsc_entry)) {
if (crm_str_eq((const char *)rsc_entry->name, XML_LRM_TAG_RESOURCE, TRUE)) {
if (xml_has_children(rsc_entry)) {
GListPtr gIter = NULL;
GListPtr result = NULL;
const char *rsc_id = ID(rsc_entry);
CRM_CHECK(rsc_id != NULL, return);
result = find_rsc_list(NULL, NULL, rsc_id, TRUE, FALSE, data_set);
for (gIter = result; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
if (rsc->variant != pe_native) {
continue;
}
check_actions_for(rsc_entry, rsc, node, data_set);
}
g_list_free(result);
}
}
}
}
}
}
}
static void
apply_placement_constraints(pe_working_set_t * data_set)
{
for (GList *gIter = data_set->placement_constraints;
gIter != NULL; gIter = gIter->next) {
pe__location_t *cons = gIter->data;
cons->rsc_lh->cmds->rsc_location(cons->rsc_lh, cons);
}
}
static gboolean
failcount_clear_action_exists(pe_node_t * node, pe_resource_t * rsc)
{
gboolean rc = FALSE;
GList *list = pe__resource_actions(rsc, node, CRM_OP_CLEAR_FAILCOUNT, TRUE);
if (list) {
rc = TRUE;
}
g_list_free(list);
return rc;
}
/*!
* \internal
* \brief Force resource away if failures hit migration threshold
*
* \param[in,out] rsc Resource to check for failures
* \param[in,out] node Node to check for failures
* \param[in,out] data_set Cluster working set to update
*/
static void
check_migration_threshold(pe_resource_t *rsc, pe_node_t *node,
pe_working_set_t *data_set)
{
int fail_count, countdown;
pe_resource_t *failed;
/* Migration threshold of 0 means never force away */
if (rsc->migration_threshold == 0) {
return;
}
// If we're ignoring failures, also ignore the migration threshold
if (is_set(rsc->flags, pe_rsc_failure_ignored)) {
return;
}
/* If there are no failures, there's no need to force away */
fail_count = pe_get_failcount(node, rsc, NULL,
pe_fc_effective|pe_fc_fillers, NULL,
data_set);
if (fail_count <= 0) {
return;
}
/* How many more times recovery will be tried on this node */
countdown = QB_MAX(rsc->migration_threshold - fail_count, 0);
/* If failed resource has a parent, we'll force the parent away */
failed = rsc;
if (is_not_set(rsc->flags, pe_rsc_unique)) {
failed = uber_parent(rsc);
}
if (countdown == 0) {
resource_location(failed, node, -INFINITY, "__fail_limit__", data_set);
crm_warn("Forcing %s away from %s after %d failures (max=%d)",
failed->id, node->details->uname, fail_count,
rsc->migration_threshold);
} else {
crm_info("%s can fail %d more times on %s before being forced off",
failed->id, countdown, node->details->uname);
}
}
static void
common_apply_stickiness(pe_resource_t * rsc, pe_node_t * node, pe_working_set_t * data_set)
{
if (rsc->children) {
GListPtr gIter = rsc->children;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
common_apply_stickiness(child_rsc, node, data_set);
}
return;
}
if (is_set(rsc->flags, pe_rsc_managed)
&& rsc->stickiness != 0 && pcmk__list_of_1(rsc->running_on)) {
pe_node_t *current = pe_find_node_id(rsc->running_on, node->details->id);
pe_node_t *match = pe_hash_table_lookup(rsc->allowed_nodes, node->details->id);
if (current == NULL) {
} else if (match != NULL || is_set(data_set->flags, pe_flag_symmetric_cluster)) {
pe_resource_t *sticky_rsc = rsc;
resource_location(sticky_rsc, node, rsc->stickiness, "stickiness", data_set);
pe_rsc_debug(sticky_rsc, "Resource %s: preferring current location"
" (node=%s, weight=%d)", sticky_rsc->id,
node->details->uname, rsc->stickiness);
} else {
GHashTableIter iter;
pe_node_t *nIter = NULL;
pe_rsc_debug(rsc, "Ignoring stickiness for %s: the cluster is asymmetric"
" and node %s is not explicitly allowed", rsc->id, node->details->uname);
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&nIter)) {
crm_err("%s[%s] = %d", rsc->id, nIter->details->uname, nIter->weight);
}
}
}
/* Check the migration threshold only if a failcount clear action
* has not already been placed for this resource on the node.
* There is no sense in potentially forcing the resource from this
* node if the failcount is being reset anyway.
*
* @TODO A clear_failcount operation can be scheduled in stage4() via
* check_actions_for(), or in stage5() via check_params(). This runs in
* stage2(), so it cannot detect those, meaning we might check the migration
* threshold when we shouldn't -- worst case, we stop or move the resource,
* then move it back next transition.
*/
if (failcount_clear_action_exists(node, rsc) == FALSE) {
check_migration_threshold(rsc, node, data_set);
}
}
void
complex_set_cmds(pe_resource_t * rsc)
{
GListPtr gIter = rsc->children;
rsc->cmds = &resource_class_alloc_functions[rsc->variant];
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
complex_set_cmds(child_rsc);
}
}
void
set_alloc_actions(pe_working_set_t * data_set)
{
GListPtr gIter = data_set->resources;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
complex_set_cmds(rsc);
}
}
static void
calculate_system_health(gpointer gKey, gpointer gValue, gpointer user_data)
{
const char *key = (const char *)gKey;
const char *value = (const char *)gValue;
int *system_health = (int *)user_data;
if (!gKey || !gValue || !user_data) {
return;
}
if (pcmk__starts_with(key, "#health")) {
int score;
/* Convert the value into an integer */
score = char2score(value);
/* Add it to the running total */
*system_health = merge_weights(score, *system_health);
}
}
static gboolean
apply_system_health(pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
const char *health_strategy = pe_pref(data_set->config_hash, "node-health-strategy");
int base_health = 0;
if (health_strategy == NULL || safe_str_eq(health_strategy, "none")) {
/* Prevent any accidental health -> score translation */
pcmk__score_red = 0;
pcmk__score_yellow = 0;
pcmk__score_green = 0;
return TRUE;
} else if (safe_str_eq(health_strategy, "migrate-on-red")) {
/* Resources on nodes which have health values of red are
* weighted away from that node.
*/
pcmk__score_red = -INFINITY;
pcmk__score_yellow = 0;
pcmk__score_green = 0;
} else if (safe_str_eq(health_strategy, "only-green")) {
/* Resources on nodes which have health values of red or yellow
* are forced away from that node.
*/
pcmk__score_red = -INFINITY;
pcmk__score_yellow = -INFINITY;
pcmk__score_green = 0;
} else if (safe_str_eq(health_strategy, "progressive")) {
/* Same as the above, but use the r/y/g scores provided by the user
* Defaults are provided by the pe_prefs table
* Also, custom health "base score" can be used
*/
base_health = crm_parse_int(pe_pref(data_set->config_hash, "node-health-base"), "0");
} else if (safe_str_eq(health_strategy, "custom")) {
/* Requires the admin to configure the rsc_location constaints for
* processing the stored health scores
*/
/* TODO: Check for the existence of appropriate node health constraints */
return TRUE;
} else {
crm_err("Unknown node health strategy: %s", health_strategy);
return FALSE;
}
crm_info("Applying automated node health strategy: %s", health_strategy);
for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
int system_health = base_health;
pe_node_t *node = (pe_node_t *) gIter->data;
/* Search through the node hash table for system health entries. */
g_hash_table_foreach(node->details->attrs, calculate_system_health, &system_health);
crm_info(" Node %s has an combined system health of %d",
node->details->uname, system_health);
/* If the health is non-zero, then create a new rsc2node so that the
* weight will be added later on.
*/
if (system_health != 0) {
GListPtr gIter2 = data_set->resources;
for (; gIter2 != NULL; gIter2 = gIter2->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter2->data;
rsc2node_new(health_strategy, rsc, system_health, NULL, node, data_set);
}
}
}
return TRUE;
}
gboolean
stage0(pe_working_set_t * data_set)
{
xmlNode *cib_constraints = get_object_root(XML_CIB_TAG_CONSTRAINTS, data_set->input);
if (data_set->input == NULL) {
return FALSE;
}
if (is_set(data_set->flags, pe_flag_have_status) == FALSE) {
crm_trace("Calculating status");
cluster_status(data_set);
}
set_alloc_actions(data_set);
apply_system_health(data_set);
unpack_constraints(cib_constraints, data_set);
return TRUE;
}
/*
* Check nodes for resources started outside of the LRM
*/
gboolean
probe_resources(pe_working_set_t * data_set)
{
pe_action_t *probe_node_complete = NULL;
for (GListPtr gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
const char *probed = pe_node_attribute_raw(node, CRM_OP_PROBED);
if (node->details->online == FALSE) {
if (pe__is_remote_node(node) && node->details->remote_rsc
&& (get_remote_node_state(node) == remote_state_failed)) {
pe_fence_node(data_set, node, "the connection is unrecoverable", FALSE);
}
continue;
} else if (node->details->unclean) {
continue;
} else if (node->details->rsc_discovery_enabled == FALSE) {
/* resource discovery is disabled for this node */
continue;
}
if (probed != NULL && crm_is_true(probed) == FALSE) {
pe_action_t *probe_op = custom_action(NULL, crm_strdup_printf("%s-%s", CRM_OP_REPROBE, node->details->uname),
CRM_OP_REPROBE, node, FALSE, TRUE, data_set);
add_hash_param(probe_op->meta, XML_ATTR_TE_NOWAIT, XML_BOOLEAN_TRUE);
continue;
}
for (GListPtr gIter2 = data_set->resources; gIter2 != NULL; gIter2 = gIter2->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter2->data;
rsc->cmds->create_probe(rsc, node, probe_node_complete, FALSE, data_set);
}
}
return TRUE;
}
static void
rsc_discover_filter(pe_resource_t *rsc, pe_node_t *node)
{
GListPtr gIter = rsc->children;
pe_resource_t *top = uber_parent(rsc);
pe_node_t *match;
if (rsc->exclusive_discover == FALSE && top->exclusive_discover == FALSE) {
return;
}
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
rsc_discover_filter(child_rsc, node);
}
match = g_hash_table_lookup(rsc->allowed_nodes, node->details->id);
if (match && match->rsc_discover_mode != pe_discover_exclusive) {
match->weight = -INFINITY;
}
}
static time_t
shutdown_time(pe_node_t *node, pe_working_set_t *data_set)
{
const char *shutdown = pe_node_attribute_raw(node, XML_CIB_ATTR_SHUTDOWN);
time_t result = 0;
if (shutdown) {
errno = 0;
result = (time_t) crm_parse_ll(shutdown, NULL);
if (errno != 0) {
result = 0;
}
}
return result? result : get_effective_time(data_set);
}
static void
apply_shutdown_lock(pe_resource_t *rsc, pe_working_set_t *data_set)
{
const char *class;
// Only primitives and (uncloned) groups may be locked
if (rsc->variant == pe_group) {
for (GList *item = rsc->children; item != NULL;
item = item->next) {
apply_shutdown_lock((pe_resource_t *) item->data, data_set);
}
} else if (rsc->variant != pe_native) {
return;
}
// Fence devices and remote connections can't be locked
class = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS);
if ((class == NULL) || !strcmp(class, PCMK_RESOURCE_CLASS_STONITH)
|| pe__resource_is_remote_conn(rsc, data_set)) {
return;
}
if (rsc->lock_node != NULL) {
// The lock was obtained from resource history
if (rsc->running_on != NULL) {
/* The resource was started elsewhere even though it is now
* considered locked. This shouldn't be possible, but as a
* failsafe, we don't want to disturb the resource now.
*/
pe_rsc_info(rsc,
"Cancelling shutdown lock because %s is already active",
rsc->id);
pe__clear_resource_history(rsc, rsc->lock_node, data_set);
rsc->lock_node = NULL;
rsc->lock_time = 0;
}
// Only a resource active on exactly one node can be locked
} else if (pcmk__list_of_1(rsc->running_on)) {
pe_node_t *node = rsc->running_on->data;
if (node->details->shutdown) {
if (node->details->unclean) {
pe_rsc_debug(rsc, "Not locking %s to unclean %s for shutdown",
rsc->id, node->details->uname);
} else {
rsc->lock_node = node;
rsc->lock_time = shutdown_time(node, data_set);
}
}
}
if (rsc->lock_node == NULL) {
// No lock needed
return;
}
if (data_set->shutdown_lock > 0) {
time_t lock_expiration = rsc->lock_time + data_set->shutdown_lock;
pe_rsc_info(rsc, "Locking %s to %s due to shutdown (expires @%lld)",
rsc->id, rsc->lock_node->details->uname,
(long long) lock_expiration);
pe__update_recheck_time(++lock_expiration, data_set);
} else {
pe_rsc_info(rsc, "Locking %s to %s due to shutdown",
rsc->id, rsc->lock_node->details->uname);
}
// If resource is locked to one node, ban it from all other nodes
for (GList *item = data_set->nodes; item != NULL; item = item->next) {
pe_node_t *node = item->data;
if (strcmp(node->details->uname, rsc->lock_node->details->uname)) {
resource_location(rsc, node, -CRM_SCORE_INFINITY,
XML_CONFIG_ATTR_SHUTDOWN_LOCK, data_set);
}
}
}
/*
* \internal
* \brief Stage 2 of cluster status: apply node-specific criteria
*
* Count known nodes, and apply location constraints, stickiness, and exclusive
* resource discovery.
*/
gboolean
stage2(pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
if (is_set(data_set->flags, pe_flag_shutdown_lock)) {
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
apply_shutdown_lock((pe_resource_t *) gIter->data, data_set);
}
}
if (is_not_set(data_set->flags, pe_flag_no_compat)) {
// @COMPAT API backward compatibility
for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
if (node && (node->weight >= 0) && node->details->online
&& (node->details->type != node_ping)) {
data_set->max_valid_nodes++;
}
}
}
crm_trace("Applying placement constraints");
apply_placement_constraints(data_set);
gIter = data_set->nodes;
for (; gIter != NULL; gIter = gIter->next) {
GListPtr gIter2 = NULL;
pe_node_t *node = (pe_node_t *) gIter->data;
gIter2 = data_set->resources;
for (; gIter2 != NULL; gIter2 = gIter2->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter2->data;
common_apply_stickiness(rsc, node, data_set);
rsc_discover_filter(rsc, node);
}
}
return TRUE;
}
/*
* Create internal resource constraints before allocation
*/
gboolean
stage3(pe_working_set_t * data_set)
{
GListPtr gIter = data_set->resources;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
rsc->cmds->internal_constraints(rsc, data_set);
}
return TRUE;
}
/*
* Check for orphaned or redefined actions
*/
gboolean
stage4(pe_working_set_t * data_set)
{
check_actions(data_set);
return TRUE;
}
static void *
convert_const_pointer(const void *ptr)
{
/* Worst function ever */
return (void *)ptr;
}
static gint
sort_rsc_process_order(gconstpointer a, gconstpointer b, gpointer data)
{
int rc = 0;
int r1_weight = -INFINITY;
int r2_weight = -INFINITY;
const char *reason = "existence";
const GListPtr nodes = (GListPtr) data;
const pe_resource_t *resource1 = a;
const pe_resource_t *resource2 = b;
pe_node_t *r1_node = NULL;
pe_node_t *r2_node = NULL;
GListPtr gIter = NULL;
GHashTable *r1_nodes = NULL;
GHashTable *r2_nodes = NULL;
- if (a == NULL && b == NULL) {
- goto done;
- }
- if (a == NULL) {
- return 1;
- }
- if (b == NULL) {
- return -1;
- }
-
reason = "priority";
r1_weight = resource1->priority;
r2_weight = resource2->priority;
if (r1_weight > r2_weight) {
rc = -1;
goto done;
}
if (r1_weight < r2_weight) {
rc = 1;
goto done;
}
reason = "no node list";
if (nodes == NULL) {
goto done;
}
r1_nodes = pcmk__native_merge_weights(convert_const_pointer(resource1),
resource1->id, NULL, NULL, 1,
pe_weights_forward | pe_weights_init);
pe__show_node_weights(true, NULL, resource1->id, r1_nodes);
r2_nodes = pcmk__native_merge_weights(convert_const_pointer(resource2),
resource2->id, NULL, NULL, 1,
pe_weights_forward | pe_weights_init);
pe__show_node_weights(true, NULL, resource2->id, r2_nodes);
/* Current location score */
reason = "current location";
r1_weight = -INFINITY;
r2_weight = -INFINITY;
if (resource1->running_on) {
r1_node = pe__current_node(resource1);
r1_node = g_hash_table_lookup(r1_nodes, r1_node->details->id);
if (r1_node != NULL) {
r1_weight = r1_node->weight;
}
}
if (resource2->running_on) {
r2_node = pe__current_node(resource2);
r2_node = g_hash_table_lookup(r2_nodes, r2_node->details->id);
if (r2_node != NULL) {
r2_weight = r2_node->weight;
}
}
if (r1_weight > r2_weight) {
rc = -1;
goto done;
}
if (r1_weight < r2_weight) {
rc = 1;
goto done;
}
reason = "score";
for (gIter = nodes; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
r1_node = NULL;
r2_node = NULL;
r1_weight = -INFINITY;
if (r1_nodes) {
r1_node = g_hash_table_lookup(r1_nodes, node->details->id);
}
if (r1_node) {
r1_weight = r1_node->weight;
}
r2_weight = -INFINITY;
if (r2_nodes) {
r2_node = g_hash_table_lookup(r2_nodes, node->details->id);
}
if (r2_node) {
r2_weight = r2_node->weight;
}
if (r1_weight > r2_weight) {
rc = -1;
goto done;
}
if (r1_weight < r2_weight) {
rc = 1;
goto done;
}
}
done:
crm_trace("%s (%d) on %s %c %s (%d) on %s: %s",
resource1->id, r1_weight, r1_node ? r1_node->details->id : "n/a",
rc < 0 ? '>' : rc > 0 ? '<' : '=',
resource2->id, r2_weight, r2_node ? r2_node->details->id : "n/a", reason);
if (r1_nodes) {
g_hash_table_destroy(r1_nodes);
}
if (r2_nodes) {
g_hash_table_destroy(r2_nodes);
}
return rc;
}
static void
allocate_resources(pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
if (is_set(data_set->flags, pe_flag_have_remote_nodes)) {
/* Allocate remote connection resources first (which will also allocate
* any colocation dependencies). If the connection is migrating, always
* prefer the partial migration target.
*/
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
if (rsc->is_remote_node == FALSE) {
continue;
}
pe_rsc_trace(rsc, "Allocating remote connection resource '%s'",
rsc->id);
rsc->cmds->allocate(rsc, rsc->partial_migration_target, data_set);
}
}
/* now do the rest of the resources */
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
if (rsc->is_remote_node == TRUE) {
continue;
}
pe_rsc_trace(rsc, "Allocating resource '%s'", rsc->id);
rsc->cmds->allocate(rsc, NULL, data_set);
}
}
/* We always use pe_order_preserve with these convenience functions to exempt
* internally generated constraints from the prohibition of user constraints
* involving remote connection resources.
*
* The start ordering additionally uses pe_order_runnable_left so that the
* specified action is not runnable if the start is not runnable.
*/
static inline void
order_start_then_action(pe_resource_t *lh_rsc, pe_action_t *rh_action,
enum pe_ordering extra, pe_working_set_t *data_set)
{
if (lh_rsc && rh_action && data_set) {
custom_action_order(lh_rsc, start_key(lh_rsc), NULL,
rh_action->rsc, NULL, rh_action,
pe_order_preserve | pe_order_runnable_left | extra,
data_set);
}
}
static inline void
order_action_then_stop(pe_action_t *lh_action, pe_resource_t *rh_rsc,
enum pe_ordering extra, pe_working_set_t *data_set)
{
if (lh_action && rh_rsc && data_set) {
custom_action_order(lh_action->rsc, NULL, lh_action,
rh_rsc, stop_key(rh_rsc), NULL,
pe_order_preserve | extra, data_set);
}
}
// Clear fail counts for orphaned rsc on all online nodes
static void
cleanup_orphans(pe_resource_t * rsc, pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
if (node->details->online
&& pe_get_failcount(node, rsc, NULL, pe_fc_effective, NULL,
data_set)) {
pe_action_t *clear_op = NULL;
clear_op = pe__clear_failcount(rsc, node, "it is orphaned",
data_set);
/* We can't use order_action_then_stop() here because its
* pe_order_preserve breaks things
*/
custom_action_order(clear_op->rsc, NULL, clear_op,
rsc, stop_key(rsc), NULL,
pe_order_optional, data_set);
}
}
}
gboolean
stage5(pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
- int log_prio = show_utilization? LOG_STDOUT : utilization_log_level;
+ int log_prio = show_utilization? LOG_STDOUT : LOG_TRACE;
if (safe_str_neq(data_set->placement_strategy, "default")) {
GListPtr nodes = g_list_copy(data_set->nodes);
nodes = sort_nodes_by_weight(nodes, NULL, data_set);
data_set->resources =
g_list_sort_with_data(data_set->resources, sort_rsc_process_order, nodes);
g_list_free(nodes);
}
gIter = data_set->nodes;
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
dump_node_capacity(log_prio, "Original", node);
}
crm_trace("Allocating services");
/* Take (next) highest resource, assign it and create its actions */
allocate_resources(data_set);
gIter = data_set->nodes;
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
dump_node_capacity(log_prio, "Remaining", node);
}
// Process deferred action checks
pe__foreach_param_check(data_set, check_params);
pe__free_param_checks(data_set);
if (is_set(data_set->flags, pe_flag_startup_probes)) {
crm_trace("Calculating needed probes");
/* This code probably needs optimization
* ptest -x with 100 nodes, 100 clones and clone-max=100:
With probes:
ptest[14781]: 2010/09/27_17:56:46 notice: TRACE: do_calculations: pengine.c:258 Calculate cluster status
ptest[14781]: 2010/09/27_17:56:46 notice: TRACE: do_calculations: pengine.c:278 Applying placement constraints
ptest[14781]: 2010/09/27_17:56:47 notice: TRACE: do_calculations: pengine.c:285 Create internal constraints
ptest[14781]: 2010/09/27_17:56:47 notice: TRACE: do_calculations: pengine.c:292 Check actions
ptest[14781]: 2010/09/27_17:56:48 notice: TRACE: do_calculations: pengine.c:299 Allocate resources
ptest[14781]: 2010/09/27_17:56:48 notice: TRACE: stage5: allocate.c:881 Allocating services
ptest[14781]: 2010/09/27_17:56:49 notice: TRACE: stage5: allocate.c:894 Calculating needed probes
ptest[14781]: 2010/09/27_17:56:51 notice: TRACE: stage5: allocate.c:899 Creating actions
ptest[14781]: 2010/09/27_17:56:52 notice: TRACE: stage5: allocate.c:905 Creating done
ptest[14781]: 2010/09/27_17:56:52 notice: TRACE: do_calculations: pengine.c:306 Processing fencing and shutdown cases
ptest[14781]: 2010/09/27_17:56:52 notice: TRACE: do_calculations: pengine.c:313 Applying ordering constraints
36s
ptest[14781]: 2010/09/27_17:57:28 notice: TRACE: do_calculations: pengine.c:320 Create transition graph
Without probes:
ptest[14637]: 2010/09/27_17:56:21 notice: TRACE: do_calculations: pengine.c:258 Calculate cluster status
ptest[14637]: 2010/09/27_17:56:22 notice: TRACE: do_calculations: pengine.c:278 Applying placement constraints
ptest[14637]: 2010/09/27_17:56:22 notice: TRACE: do_calculations: pengine.c:285 Create internal constraints
ptest[14637]: 2010/09/27_17:56:22 notice: TRACE: do_calculations: pengine.c:292 Check actions
ptest[14637]: 2010/09/27_17:56:23 notice: TRACE: do_calculations: pengine.c:299 Allocate resources
ptest[14637]: 2010/09/27_17:56:23 notice: TRACE: stage5: allocate.c:881 Allocating services
ptest[14637]: 2010/09/27_17:56:24 notice: TRACE: stage5: allocate.c:899 Creating actions
ptest[14637]: 2010/09/27_17:56:25 notice: TRACE: stage5: allocate.c:905 Creating done
ptest[14637]: 2010/09/27_17:56:25 notice: TRACE: do_calculations: pengine.c:306 Processing fencing and shutdown cases
ptest[14637]: 2010/09/27_17:56:25 notice: TRACE: do_calculations: pengine.c:313 Applying ordering constraints
ptest[14637]: 2010/09/27_17:56:25 notice: TRACE: do_calculations: pengine.c:320 Create transition graph
*/
probe_resources(data_set);
}
crm_trace("Handle orphans");
if (is_set(data_set->flags, pe_flag_stop_rsc_orphans)) {
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
/* There's no need to recurse into rsc->children because those
* should just be unallocated clone instances.
*/
if (is_set(rsc->flags, pe_rsc_orphan)) {
cleanup_orphans(rsc, data_set);
}
}
}
crm_trace("Creating actions");
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
rsc->cmds->create_actions(rsc, data_set);
}
crm_trace("Creating done");
return TRUE;
}
static gboolean
is_managed(const pe_resource_t * rsc)
{
GListPtr gIter = rsc->children;
if (is_set(rsc->flags, pe_rsc_managed)) {
return TRUE;
}
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
if (is_managed(child_rsc)) {
return TRUE;
}
}
return FALSE;
}
static gboolean
any_managed_resources(pe_working_set_t * data_set)
{
GListPtr gIter = data_set->resources;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
if (is_managed(rsc)) {
return TRUE;
}
}
return FALSE;
}
/*!
* \internal
* \brief Create pseudo-op for guest node fence, and order relative to it
*
* \param[in] node Guest node to fence
* \param[in] data_set Working set of CIB state
*/
static void
fence_guest(pe_node_t *node, pe_working_set_t *data_set)
{
pe_resource_t *container = node->details->remote_rsc->container;
pe_action_t *stop = NULL;
pe_action_t *stonith_op = NULL;
/* The fence action is just a label; we don't do anything differently for
* off vs. reboot. We specify it explicitly, rather than let it default to
* cluster's default action, because we are not _initiating_ fencing -- we
* are creating a pseudo-event to describe fencing that is already occurring
* by other means (container recovery).
*/
const char *fence_action = "off";
/* Check whether guest's container resource has any explicit stop or
* start (the stop may be implied by fencing of the guest's host).
*/
if (container) {
stop = find_first_action(container->actions, NULL, CRMD_ACTION_STOP, NULL);
if (find_first_action(container->actions, NULL, CRMD_ACTION_START, NULL)) {
fence_action = "reboot";
}
}
/* Create a fence pseudo-event, so we have an event to order actions
* against, and the controller can always detect it.
*/
stonith_op = pe_fence_op(node, fence_action, FALSE, "guest is unclean", FALSE, data_set);
update_action_flags(stonith_op, pe_action_pseudo | pe_action_runnable,
__FUNCTION__, __LINE__);
/* We want to imply stops/demotes after the guest is stopped, not wait until
* it is restarted, so we always order pseudo-fencing after stop, not start
* (even though start might be closer to what is done for a real reboot).
*/
if(stop && is_set(stop->flags, pe_action_pseudo)) {
pe_action_t *parent_stonith_op = pe_fence_op(stop->node, NULL, FALSE, NULL, FALSE, data_set);
crm_info("Implying guest node %s is down (action %d) after %s fencing",
node->details->uname, stonith_op->id, stop->node->details->uname);
order_actions(parent_stonith_op, stonith_op,
pe_order_runnable_left|pe_order_implies_then);
} else if (stop) {
order_actions(stop, stonith_op,
pe_order_runnable_left|pe_order_implies_then);
crm_info("Implying guest node %s is down (action %d) "
"after container %s is stopped (action %d)",
node->details->uname, stonith_op->id,
container->id, stop->id);
} else {
/* If we're fencing the guest node but there's no stop for the guest
* resource, we must think the guest is already stopped. However, we may
* think so because its resource history was just cleaned. To avoid
* unnecessarily considering the guest node down if it's really up,
* order the pseudo-fencing after any stop of the connection resource,
* which will be ordered after any container (re-)probe.
*/
stop = find_first_action(node->details->remote_rsc->actions, NULL,
RSC_STOP, NULL);
if (stop) {
order_actions(stop, stonith_op, pe_order_optional);
crm_info("Implying guest node %s is down (action %d) "
"after connection is stopped (action %d)",
node->details->uname, stonith_op->id, stop->id);
} else {
/* Not sure why we're fencing, but everything must already be
* cleanly stopped.
*/
crm_info("Implying guest node %s is down (action %d) ",
node->details->uname, stonith_op->id);
}
}
/* Order/imply other actions relative to pseudo-fence as with real fence */
pcmk__order_vs_fence(stonith_op, data_set);
}
/*
* Create dependencies for stonith and shutdown operations
*/
gboolean
stage6(pe_working_set_t * data_set)
{
pe_action_t *dc_down = NULL;
pe_action_t *stonith_op = NULL;
gboolean integrity_lost = FALSE;
gboolean need_stonith = TRUE;
GListPtr gIter;
GListPtr stonith_ops = NULL;
GList *shutdown_ops = NULL;
/* Remote ordering constraints need to happen prior to calculating fencing
* because it is one more place we will mark the node as dirty.
*
* A nice side effect of doing them early is that apply_*_ordering() can be
* simpler because pe_fence_node() has already done some of the work.
*/
crm_trace("Creating remote ordering constraints");
apply_remote_node_ordering(data_set);
crm_trace("Processing fencing and shutdown cases");
if (any_managed_resources(data_set) == FALSE) {
crm_notice("Delaying fencing operations until there are resources to manage");
need_stonith = FALSE;
}
/* Check each node for stonith/shutdown */
for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
/* Guest nodes are "fenced" by recovering their container resource,
* so handle them separately.
*/
if (pe__is_guest_node(node)) {
if (node->details->remote_requires_reset && need_stonith
&& pe_can_fence(data_set, node)) {
fence_guest(node, data_set);
}
continue;
}
stonith_op = NULL;
if (node->details->unclean
&& need_stonith && pe_can_fence(data_set, node)) {
stonith_op = pe_fence_op(node, NULL, FALSE, "node is unclean", FALSE, data_set);
pe_warn("Scheduling Node %s for STONITH", node->details->uname);
pcmk__order_vs_fence(stonith_op, data_set);
if (node->details->is_dc) {
// Remember if the DC is being fenced
dc_down = stonith_op;
} else {
if (is_not_set(data_set->flags, pe_flag_concurrent_fencing)
&& (stonith_ops != NULL)) {
/* Concurrent fencing is disabled, so order each non-DC
* fencing in a chain. If there is any DC fencing or
* shutdown, it will be ordered after the last action in the
* chain later.
*/
order_actions((pe_action_t *) stonith_ops->data,
stonith_op, pe_order_optional);
}
// Remember all non-DC fencing actions in a separate list
stonith_ops = g_list_prepend(stonith_ops, stonith_op);
}
} else if (node->details->online && node->details->shutdown &&
/* TODO define what a shutdown op means for a remote node.
* For now we do not send shutdown operations for remote nodes, but
* if we can come up with a good use for this in the future, we will. */
pe__is_guest_or_remote_node(node) == FALSE) {
pe_action_t *down_op = sched_shutdown_op(node, data_set);
if (node->details->is_dc) {
// Remember if the DC is being shut down
dc_down = down_op;
} else {
// Remember non-DC shutdowns for later ordering
shutdown_ops = g_list_prepend(shutdown_ops, down_op);
}
}
if (node->details->unclean && stonith_op == NULL) {
integrity_lost = TRUE;
pe_warn("Node %s is unclean!", node->details->uname);
}
}
if (integrity_lost) {
if (is_set(data_set->flags, pe_flag_stonith_enabled) == FALSE) {
pe_warn("YOUR RESOURCES ARE NOW LIKELY COMPROMISED");
pe_err("ENABLE STONITH TO KEEP YOUR RESOURCES SAFE");
} else if (is_set(data_set->flags, pe_flag_have_quorum) == FALSE) {
crm_notice("Cannot fence unclean nodes until quorum is"
" attained (or no-quorum-policy is set to ignore)");
}
}
if (dc_down != NULL) {
/* Order any non-DC shutdowns before any DC shutdown, to avoid repeated
* DC elections. However, we don't want to order non-DC shutdowns before
* a DC *fencing*, because even though we don't want a node that's
* shutting down to become DC, the DC fencing could be ordered before a
* clone stop that's also ordered before the shutdowns, thus leading to
* a graph loop.
*/
if (safe_str_eq(dc_down->task, CRM_OP_SHUTDOWN)) {
for (gIter = shutdown_ops; gIter != NULL; gIter = gIter->next) {
pe_action_t *node_stop = (pe_action_t *) gIter->data;
crm_debug("Ordering shutdown on %s before %s on DC %s",
node_stop->node->details->uname,
dc_down->task, dc_down->node->details->uname);
order_actions(node_stop, dc_down, pe_order_optional);
}
}
// Order any non-DC fencing before any DC fencing or shutdown
if (is_set(data_set->flags, pe_flag_concurrent_fencing)) {
/* With concurrent fencing, order each non-DC fencing action
* separately before any DC fencing or shutdown.
*/
for (gIter = stonith_ops; gIter != NULL; gIter = gIter->next) {
order_actions((pe_action_t *) gIter->data, dc_down,
pe_order_optional);
}
} else if (stonith_ops) {
/* Without concurrent fencing, the non-DC fencing actions are
* already ordered relative to each other, so we just need to order
* the DC fencing after the last action in the chain (which is the
* first item in the list).
*/
order_actions((pe_action_t *) stonith_ops->data, dc_down,
pe_order_optional);
}
}
g_list_free(stonith_ops);
g_list_free(shutdown_ops);
return TRUE;
}
/*
* Determine the sets of independent actions and the correct order for the
* actions in each set.
*
* Mark dependencies of un-runnable actions un-runnable
*
*/
static GListPtr
find_actions_by_task(GListPtr actions, pe_resource_t * rsc, const char *original_key)
{
GListPtr list = NULL;
list = find_actions(actions, original_key, NULL);
if (list == NULL) {
/* we're potentially searching a child of the original resource */
char *key = NULL;
char *task = NULL;
guint interval_ms = 0;
if (parse_op_key(original_key, NULL, &task, &interval_ms)) {
key = pcmk__op_key(rsc->id, task, interval_ms);
list = find_actions(actions, key, NULL);
} else {
crm_err("search key: %s", original_key);
}
free(key);
free(task);
}
return list;
}
static void
rsc_order_then(pe_action_t *lh_action, pe_resource_t *rsc,
pe__ordering_t *order)
{
GListPtr gIter = NULL;
GListPtr rh_actions = NULL;
pe_action_t *rh_action = NULL;
enum pe_ordering type;
CRM_CHECK(rsc != NULL, return);
CRM_CHECK(order != NULL, return);
type = order->type;
rh_action = order->rh_action;
crm_trace("Processing RH of ordering constraint %d", order->id);
if (rh_action != NULL) {
rh_actions = g_list_prepend(NULL, rh_action);
} else if (rsc != NULL) {
rh_actions = find_actions_by_task(rsc->actions, rsc, order->rh_action_task);
}
if (rh_actions == NULL) {
pe_rsc_trace(rsc, "No RH-Side (%s/%s) found for constraint..."
" ignoring", rsc->id, order->rh_action_task);
if (lh_action) {
pe_rsc_trace(rsc, "LH-Side was: %s", lh_action->uuid);
}
return;
}
if (lh_action && lh_action->rsc == rsc && is_set(lh_action->flags, pe_action_dangle)) {
pe_rsc_trace(rsc, "Detected dangling operation %s -> %s", lh_action->uuid,
order->rh_action_task);
clear_bit(type, pe_order_implies_then);
}
gIter = rh_actions;
for (; gIter != NULL; gIter = gIter->next) {
pe_action_t *rh_action_iter = (pe_action_t *) gIter->data;
if (lh_action) {
order_actions(lh_action, rh_action_iter, type);
} else if (type & pe_order_implies_then) {
update_action_flags(rh_action_iter, pe_action_runnable | pe_action_clear, __FUNCTION__, __LINE__);
crm_warn("Unrunnable %s 0x%.6x", rh_action_iter->uuid, type);
} else {
crm_warn("neither %s 0x%.6x", rh_action_iter->uuid, type);
}
}
g_list_free(rh_actions);
}
static void
rsc_order_first(pe_resource_t *lh_rsc, pe__ordering_t *order,
pe_working_set_t *data_set)
{
GListPtr gIter = NULL;
GListPtr lh_actions = NULL;
pe_action_t *lh_action = order->lh_action;
pe_resource_t *rh_rsc = order->rh_rsc;
crm_trace("Processing LH of ordering constraint %d", order->id);
CRM_ASSERT(lh_rsc != NULL);
if (lh_action != NULL) {
lh_actions = g_list_prepend(NULL, lh_action);
} else {
lh_actions = find_actions_by_task(lh_rsc->actions, lh_rsc, order->lh_action_task);
}
if (lh_actions == NULL && lh_rsc != rh_rsc) {
char *key = NULL;
char *op_type = NULL;
guint interval_ms = 0;
parse_op_key(order->lh_action_task, NULL, &op_type, &interval_ms);
key = pcmk__op_key(lh_rsc->id, op_type, interval_ms);
if (lh_rsc->fns->state(lh_rsc, TRUE) == RSC_ROLE_STOPPED && safe_str_eq(op_type, RSC_STOP)) {
free(key);
pe_rsc_trace(lh_rsc, "No LH-Side (%s/%s) found for constraint %d with %s - ignoring",
lh_rsc->id, order->lh_action_task, order->id, order->rh_action_task);
} else if (lh_rsc->fns->state(lh_rsc, TRUE) == RSC_ROLE_SLAVE && safe_str_eq(op_type, RSC_DEMOTE)) {
free(key);
pe_rsc_trace(lh_rsc, "No LH-Side (%s/%s) found for constraint %d with %s - ignoring",
lh_rsc->id, order->lh_action_task, order->id, order->rh_action_task);
} else {
pe_rsc_trace(lh_rsc, "No LH-Side (%s/%s) found for constraint %d with %s - creating",
lh_rsc->id, order->lh_action_task, order->id, order->rh_action_task);
lh_action = custom_action(lh_rsc, key, op_type, NULL, TRUE, TRUE, data_set);
lh_actions = g_list_prepend(NULL, lh_action);
}
free(op_type);
}
gIter = lh_actions;
for (; gIter != NULL; gIter = gIter->next) {
pe_action_t *lh_action_iter = (pe_action_t *) gIter->data;
if (rh_rsc == NULL && order->rh_action) {
rh_rsc = order->rh_action->rsc;
}
if (rh_rsc) {
rsc_order_then(lh_action_iter, rh_rsc, order);
} else if (order->rh_action) {
order_actions(lh_action_iter, order->rh_action, order->type);
}
}
g_list_free(lh_actions);
}
extern void update_colo_start_chain(pe_action_t *action,
pe_working_set_t *data_set);
static int
is_recurring_action(pe_action_t *action)
{
guint interval_ms;
if (pcmk__guint_from_hash(action->meta,
XML_LRM_ATTR_INTERVAL_MS, 0,
&interval_ms) != pcmk_rc_ok) {
return 0;
}
return (interval_ms > 0);
}
static void
apply_container_ordering(pe_action_t *action, pe_working_set_t *data_set)
{
/* VMs are also classified as containers for these purposes... in
* that they both involve a 'thing' running on a real or remote
* cluster node.
*
* This allows us to be smarter about the type and extent of
* recovery actions required in various scenarios
*/
pe_resource_t *remote_rsc = NULL;
pe_resource_t *container = NULL;
enum action_tasks task = text2task(action->task);
CRM_ASSERT(action->rsc);
CRM_ASSERT(action->node);
CRM_ASSERT(pe__is_guest_or_remote_node(action->node));
remote_rsc = action->node->details->remote_rsc;
CRM_ASSERT(remote_rsc);
container = remote_rsc->container;
CRM_ASSERT(container);
if(is_set(container->flags, pe_rsc_failed)) {
pe_fence_node(data_set, action->node, "container failed", FALSE);
}
crm_trace("Order %s action %s relative to %s%s for %s%s",
action->task, action->uuid,
is_set(remote_rsc->flags, pe_rsc_failed)? "failed " : "",
remote_rsc->id,
is_set(container->flags, pe_rsc_failed)? "failed " : "",
container->id);
if (safe_str_eq(action->task, CRMD_ACTION_MIGRATE)
|| safe_str_eq(action->task, CRMD_ACTION_MIGRATED)) {
/* Migration ops map to "no_action", but we need to apply the same
* ordering as for stop or demote (see get_router_node()).
*/
task = stop_rsc;
}
switch (task) {
case start_rsc:
case action_promote:
/* Force resource recovery if the container is recovered */
order_start_then_action(container, action, pe_order_implies_then,
data_set);
/* Wait for the connection resource to be up too */
order_start_then_action(remote_rsc, action, pe_order_none,
data_set);
break;
case stop_rsc:
case action_demote:
if (is_set(container->flags, pe_rsc_failed)) {
/* When the container representing a guest node fails, any stop
* or demote actions for resources running on the guest node
* are implied by the container stopping. This is similar to
* how fencing operations work for cluster nodes and remote
* nodes.
*/
} else {
/* Ensure the operation happens before the connection is brought
* down.
*
* If we really wanted to, we could order these after the
* connection start, IFF the container's current role was
* stopped (otherwise we re-introduce an ordering loop when the
* connection is restarting).
*/
order_action_then_stop(action, remote_rsc, pe_order_none,
data_set);
}
break;
default:
/* Wait for the connection resource to be up */
if (is_recurring_action(action)) {
/* In case we ever get the recovery logic wrong, force
* recurring monitors to be restarted, even if just
* the connection was re-established
*/
if(task != no_action) {
order_start_then_action(remote_rsc, action,
pe_order_implies_then, data_set);
}
} else {
order_start_then_action(remote_rsc, action, pe_order_none,
data_set);
}
break;
}
}
static enum remote_connection_state
get_remote_node_state(pe_node_t *node)
{
pe_resource_t *remote_rsc = NULL;
pe_node_t *cluster_node = NULL;
CRM_ASSERT(node);
remote_rsc = node->details->remote_rsc;
CRM_ASSERT(remote_rsc);
cluster_node = pe__current_node(remote_rsc);
/* If the cluster node the remote connection resource resides on
* is unclean or went offline, we can't process any operations
* on that remote node until after it starts elsewhere.
*/
if(remote_rsc->next_role == RSC_ROLE_STOPPED || remote_rsc->allocated_to == NULL) {
/* The connection resource is not going to run anywhere */
if (cluster_node && cluster_node->details->unclean) {
/* The remote connection is failed because its resource is on a
* failed node and can't be recovered elsewhere, so we must fence.
*/
return remote_state_failed;
}
if (is_not_set(remote_rsc->flags, pe_rsc_failed)) {
/* Connection resource is cleanly stopped */
return remote_state_stopped;
}
/* Connection resource is failed */
if ((remote_rsc->next_role == RSC_ROLE_STOPPED)
&& remote_rsc->remote_reconnect_ms
&& node->details->remote_was_fenced
&& !pe__shutdown_requested(node)) {
/* We won't know whether the connection is recoverable until the
* reconnect interval expires and we reattempt connection.
*/
return remote_state_unknown;
}
/* The remote connection is in a failed state. If there are any
* resources known to be active on it (stop) or in an unknown state
* (probe), we must assume the worst and fence it.
*/
return remote_state_failed;
} else if (cluster_node == NULL) {
/* Connection is recoverable but not currently running anywhere, see if we can recover it first */
return remote_state_unknown;
} else if(cluster_node->details->unclean == TRUE
|| cluster_node->details->online == FALSE) {
/* Connection is running on a dead node, see if we can recover it first */
return remote_state_resting;
} else if (pcmk__list_of_multiple(remote_rsc->running_on)
&& remote_rsc->partial_migration_source
&& remote_rsc->partial_migration_target) {
/* We're in the middle of migrating a connection resource,
* wait until after the resource migrates before performing
* any actions.
*/
return remote_state_resting;
}
return remote_state_alive;
}
/*!
* \internal
* \brief Order actions on remote node relative to actions for the connection
*/
static void
apply_remote_ordering(pe_action_t *action, pe_working_set_t *data_set)
{
pe_resource_t *remote_rsc = NULL;
enum action_tasks task = text2task(action->task);
enum remote_connection_state state = get_remote_node_state(action->node);
enum pe_ordering order_opts = pe_order_none;
if (action->rsc == NULL) {
return;
}
CRM_ASSERT(action->node);
CRM_ASSERT(pe__is_guest_or_remote_node(action->node));
remote_rsc = action->node->details->remote_rsc;
CRM_ASSERT(remote_rsc);
crm_trace("Order %s action %s relative to %s%s (state: %s)",
action->task, action->uuid,
is_set(remote_rsc->flags, pe_rsc_failed)? "failed " : "",
remote_rsc->id, state2text(state));
if (safe_str_eq(action->task, CRMD_ACTION_MIGRATE)
|| safe_str_eq(action->task, CRMD_ACTION_MIGRATED)) {
/* Migration ops map to "no_action", but we need to apply the same
* ordering as for stop or demote (see get_router_node()).
*/
task = stop_rsc;
}
switch (task) {
case start_rsc:
case action_promote:
order_opts = pe_order_none;
if (state == remote_state_failed) {
/* Force recovery, by making this action required */
order_opts |= pe_order_implies_then;
}
/* Ensure connection is up before running this action */
order_start_then_action(remote_rsc, action, order_opts, data_set);
break;
case stop_rsc:
if(state == remote_state_alive) {
order_action_then_stop(action, remote_rsc,
pe_order_implies_first, data_set);
} else if(state == remote_state_failed) {
/* The resource is active on the node, but since we don't have a
* valid connection, the only way to stop the resource is by
* fencing the node. There is no need to order the stop relative
* to the remote connection, since the stop will become implied
* by the fencing.
*/
pe_fence_node(data_set, action->node, "resources are active and the connection is unrecoverable", FALSE);
} else if(remote_rsc->next_role == RSC_ROLE_STOPPED) {
/* State must be remote_state_unknown or remote_state_stopped.
* Since the connection is not coming back up in this
* transition, stop this resource first.
*/
order_action_then_stop(action, remote_rsc,
pe_order_implies_first, data_set);
} else {
/* The connection is going to be started somewhere else, so
* stop this resource after that completes.
*/
order_start_then_action(remote_rsc, action, pe_order_none, data_set);
}
break;
case action_demote:
/* Only order this demote relative to the connection start if the
* connection isn't being torn down. Otherwise, the demote would be
* blocked because the connection start would not be allowed.
*/
if(state == remote_state_resting || state == remote_state_unknown) {
order_start_then_action(remote_rsc, action, pe_order_none,
data_set);
} /* Otherwise we can rely on the stop ordering */
break;
default:
/* Wait for the connection resource to be up */
if (is_recurring_action(action)) {
/* In case we ever get the recovery logic wrong, force
* recurring monitors to be restarted, even if just
* the connection was re-established
*/
order_start_then_action(remote_rsc, action,
pe_order_implies_then, data_set);
} else {
pe_node_t *cluster_node = pe__current_node(remote_rsc);
if(task == monitor_rsc && state == remote_state_failed) {
/* We would only be here if we do not know the
* state of the resource on the remote node.
* Since we have no way to find out, it is
* necessary to fence the node.
*/
pe_fence_node(data_set, action->node, "resources are in an unknown state and the connection is unrecoverable", FALSE);
}
if(cluster_node && state == remote_state_stopped) {
/* The connection is currently up, but is going
* down permanently.
*
* Make sure we check services are actually
* stopped _before_ we let the connection get
* closed
*/
order_action_then_stop(action, remote_rsc,
pe_order_runnable_left, data_set);
} else {
order_start_then_action(remote_rsc, action, pe_order_none,
data_set);
}
}
break;
}
}
static void
apply_remote_node_ordering(pe_working_set_t *data_set)
{
if (is_set(data_set->flags, pe_flag_have_remote_nodes) == FALSE) {
return;
}
for (GListPtr gIter = data_set->actions; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
pe_resource_t *remote = NULL;
// We are only interested in resource actions
if (action->rsc == NULL) {
continue;
}
/* Special case: If we are clearing the failcount of an actual
* remote connection resource, then make sure this happens before
* any start of the resource in this transition.
*/
if (action->rsc->is_remote_node &&
safe_str_eq(action->task, CRM_OP_CLEAR_FAILCOUNT)) {
custom_action_order(action->rsc,
NULL,
action,
action->rsc,
pcmk__op_key(action->rsc->id, RSC_START, 0),
NULL,
pe_order_optional,
data_set);
continue;
}
// We are only interested in actions allocated to a node
if (action->node == NULL) {
continue;
}
if (!pe__is_guest_or_remote_node(action->node)) {
continue;
}
/* We are only interested in real actions.
*
* @TODO This is probably wrong; pseudo-actions might be converted to
* real actions and vice versa later in update_actions() at the end of
* stage7().
*/
if (is_set(action->flags, pe_action_pseudo)) {
continue;
}
remote = action->node->details->remote_rsc;
if (remote == NULL) {
// Orphaned
continue;
}
/* Another special case: if a resource is moving to a Pacemaker Remote
* node, order the stop on the original node after any start of the
* remote connection. This ensures that if the connection fails to
* start, we leave the resource running on the original node.
*/
if (safe_str_eq(action->task, RSC_START)) {
for (GList *item = action->rsc->actions; item != NULL;
item = item->next) {
pe_action_t *rsc_action = item->data;
if ((rsc_action->node->details != action->node->details)
&& safe_str_eq(rsc_action->task, RSC_STOP)) {
custom_action_order(remote, start_key(remote), NULL,
action->rsc, NULL, rsc_action,
pe_order_optional, data_set);
}
}
}
/* The action occurs across a remote connection, so create
* ordering constraints that guarantee the action occurs while the node
* is active (after start, before stop ... things like that).
*
* This is somewhat brittle in that we need to make sure the results of
* this ordering are compatible with the result of get_router_node().
* It would probably be better to add XML_LRM_ATTR_ROUTER_NODE as part
* of this logic rather than action2xml().
*/
if (remote->container) {
crm_trace("Container ordering for %s", action->uuid);
apply_container_ordering(action, data_set);
} else {
crm_trace("Remote ordering for %s", action->uuid);
apply_remote_ordering(action, data_set);
}
}
}
static gboolean
order_first_probe_unneeded(pe_action_t * probe, pe_action_t * rh_action)
{
/* No need to probe the resource on the node that is being
* unfenced. Otherwise it might introduce transition loop
* since probe will be performed after the node is
* unfenced.
*/
if (safe_str_eq(rh_action->task, CRM_OP_FENCE)
&& probe->node && rh_action->node
&& probe->node->details == rh_action->node->details) {
const char *op = g_hash_table_lookup(rh_action->meta, "stonith_action");
if (safe_str_eq(op, "on")) {
return TRUE;
}
}
// Shutdown waits for probe to complete only if it's on the same node
if ((safe_str_eq(rh_action->task, CRM_OP_SHUTDOWN))
&& probe->node && rh_action->node
&& probe->node->details != rh_action->node->details) {
return TRUE;
}
return FALSE;
}
static void
order_first_probes_imply_stops(pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
for (gIter = data_set->ordering_constraints; gIter != NULL; gIter = gIter->next) {
pe__ordering_t *order = gIter->data;
enum pe_ordering order_type = pe_order_optional;
pe_resource_t *lh_rsc = order->lh_rsc;
pe_resource_t *rh_rsc = order->rh_rsc;
pe_action_t *lh_action = order->lh_action;
pe_action_t *rh_action = order->rh_action;
const char *lh_action_task = order->lh_action_task;
const char *rh_action_task = order->rh_action_task;
GListPtr probes = NULL;
GListPtr rh_actions = NULL;
GListPtr pIter = NULL;
if (lh_rsc == NULL) {
continue;
} else if (rh_rsc && lh_rsc == rh_rsc) {
continue;
}
if (lh_action == NULL && lh_action_task == NULL) {
continue;
}
if (rh_action == NULL && rh_action_task == NULL) {
continue;
}
/* Technically probe is expected to return "not running", which could be
* the alternative of stop action if the status of the resource is
* unknown yet.
*/
if (lh_action && safe_str_neq(lh_action->task, RSC_STOP)) {
continue;
} else if (lh_action == NULL
&& lh_action_task
&& !pcmk__ends_with(lh_action_task, "_" RSC_STOP "_0")) {
continue;
}
/* Do not probe the resource inside of a stopping container. Otherwise
* it might introduce transition loop since probe will be performed
* after the container starts again.
*/
if (rh_rsc && lh_rsc->container == rh_rsc) {
if (rh_action && safe_str_eq(rh_action->task, RSC_STOP)) {
continue;
} else if (rh_action == NULL && rh_action_task
&& pcmk__ends_with(rh_action_task,"_" RSC_STOP "_0")) {
continue;
}
}
if (order->type == pe_order_none) {
continue;
}
// Preserve the order options for future filtering
if (is_set(order->type, pe_order_apply_first_non_migratable)) {
set_bit(order_type, pe_order_apply_first_non_migratable);
}
if (is_set(order->type, pe_order_same_node)) {
set_bit(order_type, pe_order_same_node);
}
// Keep the order types for future filtering
if (order->type == pe_order_anti_colocation
|| order->type == pe_order_load) {
order_type = order->type;
}
probes = pe__resource_actions(lh_rsc, NULL, RSC_STATUS, FALSE);
if (probes == NULL) {
continue;
}
if (rh_action) {
rh_actions = g_list_prepend(rh_actions, rh_action);
} else if (rh_rsc && rh_action_task) {
rh_actions = find_actions(rh_rsc->actions, rh_action_task, NULL);
}
if (rh_actions == NULL) {
g_list_free(probes);
continue;
}
crm_trace("Processing for LH probe based on ordering constraint %s -> %s"
" (id=%d, type=%.6x)",
lh_action ? lh_action->uuid : lh_action_task,
rh_action ? rh_action->uuid : rh_action_task,
order->id, order->type);
for (pIter = probes; pIter != NULL; pIter = pIter->next) {
pe_action_t *probe = (pe_action_t *) pIter->data;
GListPtr rIter = NULL;
for (rIter = rh_actions; rIter != NULL; rIter = rIter->next) {
pe_action_t *rh_action_iter = (pe_action_t *) rIter->data;
if (order_first_probe_unneeded(probe, rh_action_iter)) {
continue;
}
order_actions(probe, rh_action_iter, order_type);
}
}
g_list_free(rh_actions);
g_list_free(probes);
}
}
static void
order_first_probe_then_restart_repromote(pe_action_t * probe,
pe_action_t * after,
pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
bool interleave = FALSE;
pe_resource_t *compatible_rsc = NULL;
if (probe == NULL
|| probe->rsc == NULL
|| probe->rsc->variant != pe_native) {
return;
}
if (after == NULL
// Avoid running into any possible loop
|| is_set(after->flags, pe_action_tracking)) {
return;
}
if (safe_str_neq(probe->task, RSC_STATUS)) {
return;
}
pe_set_action_bit(after, pe_action_tracking);
crm_trace("Processing based on %s %s -> %s %s",
probe->uuid,
probe->node ? probe->node->details->uname: "",
after->uuid,
after->node ? after->node->details->uname : "");
if (after->rsc
/* Better not build a dependency directly with a clone/group.
* We are going to proceed through the ordering chain and build
* dependencies with its children.
*/
&& after->rsc->variant == pe_native
&& probe->rsc != after->rsc) {
GListPtr then_actions = NULL;
enum pe_ordering probe_order_type = pe_order_optional;
if (safe_str_eq(after->task, RSC_START)) {
then_actions = pe__resource_actions(after->rsc, NULL, RSC_STOP, FALSE);
} else if (safe_str_eq(after->task, RSC_PROMOTE)) {
then_actions = pe__resource_actions(after->rsc, NULL, RSC_DEMOTE, FALSE);
}
for (gIter = then_actions; gIter != NULL; gIter = gIter->next) {
pe_action_t *then = (pe_action_t *) gIter->data;
// Skip any pseudo action which for example is implied by fencing
if (is_set(then->flags, pe_action_pseudo)) {
continue;
}
order_actions(probe, then, probe_order_type);
}
g_list_free(then_actions);
}
if (after->rsc
&& after->rsc->variant > pe_group) {
const char *interleave_s = g_hash_table_lookup(after->rsc->meta,
XML_RSC_ATTR_INTERLEAVE);
interleave = crm_is_true(interleave_s);
if (interleave) {
/* For an interleaved clone, we should build a dependency only
* with the relevant clone child.
*/
compatible_rsc = find_compatible_child(probe->rsc,
after->rsc,
RSC_ROLE_UNKNOWN,
FALSE, data_set);
}
}
for (gIter = after->actions_after; gIter != NULL; gIter = gIter->next) {
pe_action_wrapper_t *after_wrapper = (pe_action_wrapper_t *) gIter->data;
/* pe_order_implies_then is the reason why a required A.start
* implies/enforces B.start to be required too, which is the cause of
* B.restart/re-promote.
*
* Not sure about pe_order_implies_then_on_node though. It's now only
* used for unfencing case, which tends to introduce transition
* loops...
*/
if (is_not_set(after_wrapper->type, pe_order_implies_then)) {
/* The order type between a group/clone and its child such as
* B.start-> B_child.start is:
* pe_order_implies_first_printed | pe_order_runnable_left
*
* Proceed through the ordering chain and build dependencies with
* its children.
*/
if (after->rsc == NULL
|| after->rsc->variant < pe_group
|| probe->rsc->parent == after->rsc
|| after_wrapper->action->rsc == NULL
|| after_wrapper->action->rsc->variant > pe_group
|| after->rsc != after_wrapper->action->rsc->parent) {
continue;
}
/* Proceed to the children of a group or a non-interleaved clone.
* For an interleaved clone, proceed only to the relevant child.
*/
if (after->rsc->variant > pe_group
&& interleave == TRUE
&& (compatible_rsc == NULL
|| compatible_rsc != after_wrapper->action->rsc)) {
continue;
}
}
crm_trace("Proceeding through %s %s -> %s %s (type=0x%.6x)",
after->uuid,
after->node ? after->node->details->uname: "",
after_wrapper->action->uuid,
after_wrapper->action->node ? after_wrapper->action->node->details->uname : "",
after_wrapper->type);
order_first_probe_then_restart_repromote(probe, after_wrapper->action, data_set);
}
}
static void clear_actions_tracking_flag(pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
if (is_set(action->flags, pe_action_tracking)) {
pe_clear_action_bit(action, pe_action_tracking);
}
}
}
static void
order_first_rsc_probes(pe_resource_t * rsc, pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
GListPtr probes = NULL;
for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) {
pe_resource_t * child = (pe_resource_t *) gIter->data;
order_first_rsc_probes(child, data_set);
}
if (rsc->variant != pe_native) {
return;
}
probes = pe__resource_actions(rsc, NULL, RSC_STATUS, FALSE);
for (gIter = probes; gIter != NULL; gIter= gIter->next) {
pe_action_t *probe = (pe_action_t *) gIter->data;
GListPtr aIter = NULL;
for (aIter = probe->actions_after; aIter != NULL; aIter = aIter->next) {
pe_action_wrapper_t *after_wrapper = (pe_action_wrapper_t *) aIter->data;
order_first_probe_then_restart_repromote(probe, after_wrapper->action, data_set);
clear_actions_tracking_flag(data_set);
}
}
g_list_free(probes);
}
static void
order_first_probes(pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
order_first_rsc_probes(rsc, data_set);
}
order_first_probes_imply_stops(data_set);
}
static void
order_then_probes(pe_working_set_t * data_set)
{
#if 0
GListPtr gIter = NULL;
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
/* Given "A then B", we would prefer to wait for A to be
* started before probing B.
*
* If A was a filesystem on which the binaries and data for B
* lived, it would have been useful if the author of B's agent
* could assume that A is running before B.monitor will be
* called.
*
* However we can't _only_ probe once A is running, otherwise
* we'd not detect the state of B if A could not be started
* for some reason.
*
* In practice however, we cannot even do an opportunistic
* version of this because B may be moving:
*
* B.probe -> B.start
* B.probe -> B.stop
* B.stop -> B.start
* A.stop -> A.start
* A.start -> B.probe
*
* So far so good, but if we add the result of this code:
*
* B.stop -> A.stop
*
* Then we get a loop:
*
* B.probe -> B.stop -> A.stop -> A.start -> B.probe
*
* We could kill the 'B.probe -> B.stop' dependency, but that
* could mean stopping B "too" soon, because B.start must wait
* for the probes to complete.
*
* Another option is to allow it only if A is a non-unique
* clone with clone-max == node-max (since we'll never be
* moving it). However, we could still be stopping one
* instance at the same time as starting another.
* The complexity of checking for allowed conditions combined
* with the ever narrowing usecase suggests that this code
* should remain disabled until someone gets smarter.
*/
pe_action_t *start = NULL;
GListPtr actions = NULL;
GListPtr probes = NULL;
actions = pe__resource_actions(rsc, NULL, RSC_START, FALSE);
if (actions) {
start = actions->data;
g_list_free(actions);
}
if(start == NULL) {
crm_err("No start action for %s", rsc->id);
continue;
}
probes = pe__resource_actions(rsc, NULL, RSC_STATUS, FALSE);
for (actions = start->actions_before; actions != NULL; actions = actions->next) {
pe_action_wrapper_t *before = (pe_action_wrapper_t *) actions->data;
GListPtr pIter = NULL;
pe_action_t *first = before->action;
pe_resource_t *first_rsc = first->rsc;
if(first->required_runnable_before) {
GListPtr clone_actions = NULL;
for (clone_actions = first->actions_before; clone_actions != NULL; clone_actions = clone_actions->next) {
before = (pe_action_wrapper_t *) clone_actions->data;
crm_trace("Testing %s -> %s (%p) for %s", first->uuid, before->action->uuid, before->action->rsc, start->uuid);
CRM_ASSERT(before->action->rsc);
first_rsc = before->action->rsc;
break;
}
} else if(safe_str_neq(first->task, RSC_START)) {
crm_trace("Not a start op %s for %s", first->uuid, start->uuid);
}
if(first_rsc == NULL) {
continue;
} else if(uber_parent(first_rsc) == uber_parent(start->rsc)) {
crm_trace("Same parent %s for %s", first_rsc->id, start->uuid);
continue;
} else if(FALSE && pe_rsc_is_clone(uber_parent(first_rsc)) == FALSE) {
crm_trace("Not a clone %s for %s", first_rsc->id, start->uuid);
continue;
}
crm_err("Applying %s before %s %d", first->uuid, start->uuid, uber_parent(first_rsc)->variant);
for (pIter = probes; pIter != NULL; pIter = pIter->next) {
pe_action_t *probe = (pe_action_t *) pIter->data;
crm_err("Ordering %s before %s", first->uuid, probe->uuid);
order_actions(first, probe, pe_order_optional);
}
}
}
#endif
}
static void
order_probes(pe_working_set_t * data_set)
{
order_first_probes(data_set);
order_then_probes(data_set);
}
gboolean
stage7(pe_working_set_t * data_set)
{
GList *gIter = NULL;
crm_trace("Applying ordering constraints");
/* Don't ask me why, but apparently they need to be processed in
* the order they were created in... go figure
*
* Also g_list_append() has horrendous performance characteristics
* So we need to use g_list_prepend() and then reverse the list here
*/
data_set->ordering_constraints = g_list_reverse(data_set->ordering_constraints);
for (gIter = data_set->ordering_constraints; gIter != NULL; gIter = gIter->next) {
pe__ordering_t *order = gIter->data;
pe_resource_t *rsc = order->lh_rsc;
crm_trace("Applying ordering constraint: %d", order->id);
if (rsc != NULL) {
crm_trace("rsc_action-to-*");
rsc_order_first(rsc, order, data_set);
continue;
}
rsc = order->rh_rsc;
if (rsc != NULL) {
crm_trace("action-to-rsc_action");
rsc_order_then(order->lh_action, rsc, order);
} else {
crm_trace("action-to-action");
order_actions(order->lh_action, order->rh_action, order->type);
}
}
for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
update_colo_start_chain(action, data_set);
}
crm_trace("Ordering probes");
order_probes(data_set);
crm_trace("Updating %d actions", g_list_length(data_set->actions));
for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
update_action(action, data_set);
}
// Check for invalid orderings
for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
pe_action_wrapper_t *input = NULL;
for (GList *input_iter = action->actions_before;
input_iter != NULL; input_iter = input_iter->next) {
input = (pe_action_wrapper_t *) input_iter->data;
if (pcmk__ordering_is_invalid(action, input)) {
input->type = pe_order_none;
}
}
}
LogNodeActions(data_set, FALSE);
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
LogActions(rsc, data_set, FALSE);
}
return TRUE;
}
static int transition_id = -1;
/*!
* \internal
* \brief Log a message after calculating a transition
*
* \param[in] filename Where transition input is stored
*/
void
pcmk__log_transition_summary(const char *filename)
{
if (was_processing_error) {
crm_err("Calculated transition %d (with errors), saving inputs in %s",
transition_id, filename);
} else if (was_processing_warning) {
crm_warn("Calculated transition %d (with warnings), saving inputs in %s",
transition_id, filename);
} else {
crm_notice("Calculated transition %d, saving inputs in %s",
transition_id, filename);
}
if (crm_config_error) {
crm_notice("Configuration errors found during scheduler processing,"
" please run \"crm_verify -L\" to identify issues");
}
}
/*
* Create a dependency graph to send to the transitioner (via the controller)
*/
gboolean
stage8(pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
const char *value = NULL;
transition_id++;
crm_trace("Creating transition graph %d.", transition_id);
data_set->graph = create_xml_node(NULL, XML_TAG_GRAPH);
value = pe_pref(data_set->config_hash, "cluster-delay");
crm_xml_add(data_set->graph, "cluster-delay", value);
value = pe_pref(data_set->config_hash, "stonith-timeout");
crm_xml_add(data_set->graph, "stonith-timeout", value);
crm_xml_add(data_set->graph, "failed-stop-offset", "INFINITY");
if (is_set(data_set->flags, pe_flag_start_failure_fatal)) {
crm_xml_add(data_set->graph, "failed-start-offset", "INFINITY");
} else {
crm_xml_add(data_set->graph, "failed-start-offset", "1");
}
value = pe_pref(data_set->config_hash, "batch-limit");
crm_xml_add(data_set->graph, "batch-limit", value);
crm_xml_add_int(data_set->graph, "transition_id", transition_id);
value = pe_pref(data_set->config_hash, "migration-limit");
if (crm_parse_ll(value, NULL) > 0) {
crm_xml_add(data_set->graph, "migration-limit", value);
}
if (data_set->recheck_by > 0) {
char *recheck_epoch = NULL;
recheck_epoch = crm_strdup_printf("%llu",
(long long) data_set->recheck_by);
crm_xml_add(data_set->graph, "recheck-by", recheck_epoch);
free(recheck_epoch);
}
/* errors...
slist_iter(action, pe_action_t, action_list, lpc,
if(action->optional == FALSE && action->runnable == FALSE) {
print_action("Ignoring", action, TRUE);
}
);
*/
/* The following code will de-duplicate action inputs, so nothing past this
* should rely on the action input type flags retaining their original
* values.
*/
gIter = data_set->resources;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
pe_rsc_trace(rsc, "processing actions for rsc=%s", rsc->id);
rsc->cmds->expand(rsc, data_set);
}
crm_log_xml_trace(data_set->graph, "created resource-driven action list");
/* pseudo action to distribute list of nodes with maintenance state update */
add_maintenance_update(data_set);
/* catch any non-resource specific actions */
crm_trace("processing non-resource actions");
gIter = data_set->actions;
for (; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
if (action->rsc
&& action->node
&& action->node->details->shutdown
&& is_not_set(action->rsc->flags, pe_rsc_maintenance)
&& is_not_set(action->flags, pe_action_optional)
&& is_not_set(action->flags, pe_action_runnable)
&& crm_str_eq(action->task, RSC_STOP, TRUE)
) {
/* Eventually we should just ignore the 'fence' case
* But for now it's the best way to detect (in CTS) when
* CIB resource updates are being lost
*/
if (is_set(data_set->flags, pe_flag_have_quorum)
|| data_set->no_quorum_policy == no_quorum_ignore) {
crm_crit("Cannot %s node '%s' because of %s:%s%s (%s)",
action->node->details->unclean ? "fence" : "shut down",
action->node->details->uname, action->rsc->id,
is_not_set(action->rsc->flags, pe_rsc_managed) ? " unmanaged" : " blocked",
is_set(action->rsc->flags, pe_rsc_failed) ? " failed" : "",
action->uuid);
}
}
graph_element_from_action(action, data_set);
}
crm_log_xml_trace(data_set->graph, "created generic action list");
crm_trace("Created transition graph %d.", transition_id);
return TRUE;
}
void
LogNodeActions(pe_working_set_t * data_set, gboolean terminal)
{
GListPtr gIter = NULL;
for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) {
char *node_name = NULL;
char *task = NULL;
pe_action_t *action = (pe_action_t *) gIter->data;
if (action->rsc != NULL) {
continue;
} else if (is_set(action->flags, pe_action_optional)) {
continue;
}
if (pe__is_guest_node(action->node)) {
node_name = crm_strdup_printf("%s (resource: %s)", action->node->details->uname, action->node->details->remote_rsc->container->id);
} else if(action->node) {
node_name = crm_strdup_printf("%s", action->node->details->uname);
}
if (safe_str_eq(action->task, CRM_OP_SHUTDOWN)) {
task = strdup("Shutdown");
} else if (safe_str_eq(action->task, CRM_OP_FENCE)) {
const char *op = g_hash_table_lookup(action->meta, "stonith_action");
task = crm_strdup_printf("Fence (%s)", op);
}
if(task == NULL) {
/* Nothing to report */
} else if(terminal && action->reason) {
printf(" * %s %s '%s'\n", task, node_name, action->reason);
} else if(terminal) {
printf(" * %s %s\n", task, node_name);
} else if(action->reason) {
crm_notice(" * %s %s '%s'\n", task, node_name, action->reason);
} else {
crm_notice(" * %s %s\n", task, node_name);
}
free(node_name);
free(task);
}
}
diff --git a/lib/pacemaker/pcmk_sched_constraints.c b/lib/pacemaker/pcmk_sched_constraints.c
index fdc4f3a6e2..08dac7d8cb 100644
--- a/lib/pacemaker/pcmk_sched_constraints.c
+++ b/lib/pacemaker/pcmk_sched_constraints.c
@@ -1,3064 +1,3061 @@
/*
* Copyright 2004-2020 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 <sys/param.h>
#include <sys/types.h>
#include <regex.h>
+#include <glib.h>
#include <crm/crm.h>
#include <crm/cib.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <crm/common/iso8601.h>
-
-#include <glib.h>
-
#include <crm/pengine/status.h>
-#include <pacemaker-internal.h>
+#include <crm/pengine/internal.h>
#include <crm/pengine/rules.h>
-
-#include <../lib/pengine/unpack.h>
+#include <pacemaker-internal.h>
enum pe_order_kind {
pe_order_kind_optional,
pe_order_kind_mandatory,
pe_order_kind_serialize,
};
#define EXPAND_CONSTRAINT_IDREF(__set, __rsc, __name) do { \
__rsc = pe_find_constraint_resource(data_set->resources, __name); \
if(__rsc == NULL) { \
pcmk__config_err("%s: No resource found for %s", __set, __name); \
return FALSE; \
} \
} while(0)
enum pe_ordering get_flags(const char *id, enum pe_order_kind kind,
const char *action_first, const char *action_then, gboolean invert);
enum pe_ordering get_asymmetrical_flags(enum pe_order_kind kind);
static pe__location_t *generate_location_rule(pe_resource_t *rsc,
xmlNode *rule_xml,
const char *discovery,
crm_time_t *next_change,
pe_working_set_t *data_set,
pe_match_data_t *match_data);
static bool
evaluate_lifetime(xmlNode *lifetime, pe_working_set_t *data_set)
{
bool result = FALSE;
crm_time_t *next_change = crm_time_new_undefined();
result = pe_evaluate_rules(lifetime, NULL, data_set->now, next_change);
if (crm_time_is_defined(next_change)) {
time_t recheck = (time_t) crm_time_get_seconds_since_epoch(next_change);
pe__update_recheck_time(recheck, data_set);
}
crm_time_free(next_change);
return result;
}
gboolean
unpack_constraints(xmlNode * xml_constraints, pe_working_set_t * data_set)
{
xmlNode *xml_obj = NULL;
xmlNode *lifetime = NULL;
for (xml_obj = __xml_first_child_element(xml_constraints); xml_obj != NULL;
xml_obj = __xml_next_element(xml_obj)) {
const char *id = crm_element_value(xml_obj, XML_ATTR_ID);
const char *tag = crm_element_name(xml_obj);
if (id == NULL) {
pcmk__config_err("Ignoring <%s> constraint without "
XML_ATTR_ID, tag);
continue;
}
crm_trace("Processing constraint %s %s", tag, id);
lifetime = first_named_child(xml_obj, "lifetime");
if (lifetime) {
pcmk__config_warn("Support for 'lifetime' attribute (in %s) is "
"deprecated (the rules it contains should "
"instead be direct descendents of the "
"constraint object)", id);
}
if (lifetime && !evaluate_lifetime(lifetime, data_set)) {
crm_info("Constraint %s %s is not active", tag, id);
} else if (safe_str_eq(XML_CONS_TAG_RSC_ORDER, tag)) {
unpack_rsc_order(xml_obj, data_set);
} else if (safe_str_eq(XML_CONS_TAG_RSC_DEPEND, tag)) {
unpack_rsc_colocation(xml_obj, data_set);
} else if (safe_str_eq(XML_CONS_TAG_RSC_LOCATION, tag)) {
unpack_location(xml_obj, data_set);
} else if (safe_str_eq(XML_CONS_TAG_RSC_TICKET, tag)) {
unpack_rsc_ticket(xml_obj, data_set);
} else {
pe_err("Unsupported constraint type: %s", tag);
}
}
return TRUE;
}
static const char *
invert_action(const char *action)
{
if (safe_str_eq(action, RSC_START)) {
return RSC_STOP;
} else if (safe_str_eq(action, RSC_STOP)) {
return RSC_START;
} else if (safe_str_eq(action, RSC_PROMOTE)) {
return RSC_DEMOTE;
} else if (safe_str_eq(action, RSC_DEMOTE)) {
return RSC_PROMOTE;
} else if (safe_str_eq(action, RSC_PROMOTED)) {
return RSC_DEMOTED;
} else if (safe_str_eq(action, RSC_DEMOTED)) {
return RSC_PROMOTED;
} else if (safe_str_eq(action, RSC_STARTED)) {
return RSC_STOPPED;
} else if (safe_str_eq(action, RSC_STOPPED)) {
return RSC_STARTED;
}
crm_warn("Unknown action '%s' specified in order constraint", action);
return NULL;
}
static enum pe_order_kind
get_ordering_type(xmlNode * xml_obj)
{
enum pe_order_kind kind_e = pe_order_kind_mandatory;
const char *kind = crm_element_value(xml_obj, XML_ORDER_ATTR_KIND);
if (kind == NULL) {
const char *score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE);
kind_e = pe_order_kind_mandatory;
if (score) {
// @COMPAT deprecated informally since 1.0.7, formally since 2.0.1
int score_i = char2score(score);
if (score_i == 0) {
kind_e = pe_order_kind_optional;
}
pe_warn_once(pe_wo_order_score,
"Support for 'score' in rsc_order is deprecated "
"and will be removed in a future release (use 'kind' instead)");
}
} else if (safe_str_eq(kind, "Mandatory")) {
kind_e = pe_order_kind_mandatory;
} else if (safe_str_eq(kind, "Optional")) {
kind_e = pe_order_kind_optional;
} else if (safe_str_eq(kind, "Serialize")) {
kind_e = pe_order_kind_serialize;
} else {
pcmk__config_err("Resetting '" XML_ORDER_ATTR_KIND "' for constraint "
"'%s' to Mandatory because '%s' is not valid",
crm_str(ID(xml_obj)), kind);
}
return kind_e;
}
static pe_resource_t *
pe_find_constraint_resource(GListPtr rsc_list, const char *id)
{
GListPtr rIter = NULL;
for (rIter = rsc_list; id && rIter; rIter = rIter->next) {
pe_resource_t *parent = rIter->data;
pe_resource_t *match = parent->fns->find_rsc(parent, id, NULL,
pe_find_renamed);
if (match != NULL) {
if(safe_str_neq(match->id, id)) {
/* We found an instance of a clone instead */
match = uber_parent(match);
crm_debug("Found %s for %s", match->id, id);
}
return match;
}
}
crm_trace("No match for %s", id);
return NULL;
}
static gboolean
pe_find_constraint_tag(pe_working_set_t * data_set, const char * id, pe_tag_t ** tag)
{
gboolean rc = FALSE;
*tag = NULL;
rc = g_hash_table_lookup_extended(data_set->template_rsc_sets, id,
NULL, (gpointer*) tag);
if (rc == FALSE) {
rc = g_hash_table_lookup_extended(data_set->tags, id,
NULL, (gpointer*) tag);
if (rc == FALSE) {
crm_warn("No template or tag named '%s'", id);
return FALSE;
} else if (*tag == NULL) {
crm_warn("No resource is tagged with '%s'", id);
return FALSE;
}
} else if (*tag == NULL) {
crm_warn("No resource is derived from template '%s'", id);
return FALSE;
}
return rc;
}
static gboolean
valid_resource_or_tag(pe_working_set_t * data_set, const char * id,
pe_resource_t ** rsc, pe_tag_t ** tag)
{
gboolean rc = FALSE;
if (rsc) {
*rsc = NULL;
*rsc = pe_find_constraint_resource(data_set->resources, id);
if (*rsc) {
return TRUE;
}
}
if (tag) {
*tag = NULL;
rc = pe_find_constraint_tag(data_set, id, tag);
}
return rc;
}
static gboolean
order_is_symmetrical(xmlNode * xml_obj,
enum pe_order_kind parent_kind, const char * parent_symmetrical_s)
{
const char *id = crm_element_value(xml_obj, XML_ATTR_ID);
const char *kind_s = crm_element_value(xml_obj, XML_ORDER_ATTR_KIND);
const char *score_s = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE);
const char *symmetrical_s = crm_element_value(xml_obj, XML_CONS_ATTR_SYMMETRICAL);
enum pe_order_kind kind = parent_kind;
if (kind_s || score_s) {
kind = get_ordering_type(xml_obj);
}
if (symmetrical_s == NULL) {
symmetrical_s = parent_symmetrical_s;
}
if (symmetrical_s) {
gboolean symmetrical = crm_is_true(symmetrical_s);
if (symmetrical && kind == pe_order_kind_serialize) {
pcmk__config_warn("Ignoring " XML_CONS_ATTR_SYMMETRICAL
" for '%s' because not valid with "
XML_ORDER_ATTR_KIND " of 'Serialize'", id);
return FALSE;
}
return symmetrical;
} else {
if (kind == pe_order_kind_serialize) {
return FALSE;
} else {
return TRUE;
}
}
}
static gboolean
unpack_simple_rsc_order(xmlNode * xml_obj, pe_working_set_t * data_set)
{
int order_id = 0;
pe_resource_t *rsc_then = NULL;
pe_resource_t *rsc_first = NULL;
gboolean invert_bool = TRUE;
int min_required_before = 0;
enum pe_order_kind kind = pe_order_kind_mandatory;
enum pe_ordering cons_weight = pe_order_optional;
const char *id_first = NULL;
const char *id_then = NULL;
const char *action_then = NULL;
const char *action_first = NULL;
const char *instance_then = NULL;
const char *instance_first = NULL;
const char *id = NULL;
CRM_CHECK(xml_obj != NULL, return FALSE);
id = crm_element_value(xml_obj, XML_ATTR_ID);
if (id == NULL) {
pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID,
crm_element_name(xml_obj));
return FALSE;
}
invert_bool = order_is_symmetrical(xml_obj, kind, NULL);
id_then = crm_element_value(xml_obj, XML_ORDER_ATTR_THEN);
id_first = crm_element_value(xml_obj, XML_ORDER_ATTR_FIRST);
action_then = crm_element_value(xml_obj, XML_ORDER_ATTR_THEN_ACTION);
action_first = crm_element_value(xml_obj, XML_ORDER_ATTR_FIRST_ACTION);
instance_then = crm_element_value(xml_obj, XML_ORDER_ATTR_THEN_INSTANCE);
instance_first = crm_element_value(xml_obj, XML_ORDER_ATTR_FIRST_INSTANCE);
if (action_first == NULL) {
action_first = RSC_START;
}
if (action_then == NULL) {
action_then = action_first;
}
if (id_first == NULL) {
pcmk__config_err("Ignoring constraint '%s' without "
XML_ORDER_ATTR_FIRST, id);
return FALSE;
}
if (id_then == NULL) {
pcmk__config_err("Ignoring constraint '%s' without "
XML_ORDER_ATTR_THEN, id);
return FALSE;
}
rsc_then = pe_find_constraint_resource(data_set->resources, id_then);
rsc_first = pe_find_constraint_resource(data_set->resources, id_first);
if (rsc_then == NULL) {
pcmk__config_err("Ignoring constraint '%s' because resource '%s' "
"does not exist", id, id_then);
return FALSE;
} else if (rsc_first == NULL) {
pcmk__config_err("Ignoring constraint '%s' because resource '%s' "
"does not exist", id, id_first);
return FALSE;
} else if (instance_then && pe_rsc_is_clone(rsc_then) == FALSE) {
pcmk__config_err("Ignoring constraint '%s' because resource '%s' "
"is not a clone but instance '%s' was requested",
id, id_then, instance_then);
return FALSE;
} else if (instance_first && pe_rsc_is_clone(rsc_first) == FALSE) {
pcmk__config_err("Ignoring constraint '%s' because resource '%s' "
"is not a clone but instance '%s' was requested",
id, id_first, instance_first);
return FALSE;
}
if (instance_then) {
rsc_then = find_clone_instance(rsc_then, instance_then, data_set);
if (rsc_then == NULL) {
pcmk__config_warn("Ignoring constraint '%s' because resource '%s' "
"does not have an instance '%s'",
id, id_then, instance_then);
return FALSE;
}
}
if (instance_first) {
rsc_first = find_clone_instance(rsc_first, instance_first, data_set);
if (rsc_first == NULL) {
pcmk__config_warn("Ignoring constraint '%s' because resource '%s' "
"does not have an instance '%s'",
"'%s'", id, id_first, instance_first);
return FALSE;
}
}
cons_weight = pe_order_optional;
kind = get_ordering_type(xml_obj);
if (kind == pe_order_kind_optional && rsc_then->restart_type == pe_restart_restart) {
crm_trace("Upgrade : recovery - implies right");
cons_weight |= pe_order_implies_then;
}
if (invert_bool == FALSE) {
cons_weight |= get_asymmetrical_flags(kind);
} else {
cons_weight |= get_flags(id, kind, action_first, action_then, FALSE);
}
if (pe_rsc_is_clone(rsc_first)) {
/* If clone-min is set, require at least that number of instances to be
* runnable before allowing dependencies to be runnable.
*/
const char *min_clones_s = g_hash_table_lookup(rsc_first->meta,
XML_RSC_ATTR_INCARNATION_MIN);
// @COMPAT 1.1.13: deprecated
const char *require_all_s = crm_element_value(xml_obj, "require-all");
if (min_clones_s) {
min_required_before = crm_parse_int(min_clones_s, "0");
} else if (require_all_s) {
pe_warn_once(pe_wo_require_all,
"Support for require-all in ordering constraints "
"is deprecated and will be removed in a future release"
" (use clone-min clone meta-attribute instead)");
if (crm_is_true(require_all_s) == FALSE) {
// require-all=false is deprecated equivalent of clone-min=1
min_required_before = 1;
}
}
}
/* If there is a minimum number of instances that must be runnable before
* the 'then' action is runnable, we use a pseudo action as an intermediate step
* start min number of clones -> pseudo action is runnable -> dependency runnable. */
if (min_required_before) {
GListPtr rIter = NULL;
char *task = crm_strdup_printf(CRM_OP_RELAXED_CLONE ":%s", id);
pe_action_t *unordered_action = get_pseudo_op(task, data_set);
free(task);
/* require the pseudo action to have "min_required_before" number of
* actions to be considered runnable before allowing the pseudo action
* to be runnable. */
unordered_action->required_runnable_before = min_required_before;
update_action_flags(unordered_action, pe_action_requires_any, __FUNCTION__, __LINE__);
for (rIter = rsc_first->children; id && rIter; rIter = rIter->next) {
pe_resource_t *child = rIter->data;
/* order each clone instance before the pseudo action */
custom_action_order(child, pcmk__op_key(child->id, action_first, 0),
NULL, NULL, NULL, unordered_action,
pe_order_one_or_more|pe_order_implies_then_printed,
data_set);
}
/* order the "then" dependency to occur after the pseudo action only if
* the pseudo action is runnable */
order_id = custom_action_order(NULL, NULL, unordered_action, rsc_then,
pcmk__op_key(rsc_then->id, action_then, 0),
NULL, cons_weight|pe_order_runnable_left,
data_set);
} else {
order_id = new_rsc_order(rsc_first, action_first, rsc_then, action_then, cons_weight, data_set);
}
pe_rsc_trace(rsc_first, "order-%d (%s): %s_%s before %s_%s flags=0x%.6x",
order_id, id, rsc_first->id, action_first, rsc_then->id, action_then, cons_weight);
if (invert_bool == FALSE) {
return TRUE;
}
action_then = invert_action(action_then);
action_first = invert_action(action_first);
if (action_then == NULL || action_first == NULL) {
pcmk__config_err("Cannot invert constraint '%s' "
"(please specify inverse manually)", id);
return TRUE;
}
cons_weight = pe_order_optional;
if (kind == pe_order_kind_optional && rsc_then->restart_type == pe_restart_restart) {
crm_trace("Upgrade : recovery - implies left");
cons_weight |= pe_order_implies_first;
}
cons_weight |= get_flags(id, kind, action_first, action_then, TRUE);
order_id = new_rsc_order(rsc_then, action_then, rsc_first, action_first, cons_weight, data_set);
pe_rsc_trace(rsc_then, "order-%d (%s): %s_%s before %s_%s flags=0x%.6x",
order_id, id, rsc_then->id, action_then, rsc_first->id, action_first, cons_weight);
return TRUE;
}
static gboolean
expand_tags_in_sets(xmlNode * xml_obj, xmlNode ** expanded_xml, pe_working_set_t * data_set)
{
xmlNode *new_xml = NULL;
xmlNode *set = NULL;
gboolean any_refs = FALSE;
const char *cons_id = NULL;
*expanded_xml = NULL;
CRM_CHECK(xml_obj != NULL, return FALSE);
new_xml = copy_xml(xml_obj);
cons_id = ID(new_xml);
for (set = __xml_first_child_element(new_xml); set != NULL;
set = __xml_next_element(set)) {
xmlNode *xml_rsc = NULL;
GListPtr tag_refs = NULL;
GListPtr gIter = NULL;
if (safe_str_neq((const char *)set->name, XML_CONS_TAG_RSC_SET)) {
continue;
}
for (xml_rsc = __xml_first_child_element(set); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
pe_resource_t *rsc = NULL;
pe_tag_t *tag = NULL;
const char *id = ID(xml_rsc);
if (safe_str_neq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF)) {
continue;
}
if (valid_resource_or_tag(data_set, id, &rsc, &tag) == FALSE) {
pcmk__config_err("Ignoring resource sets for constraint '%s' "
"because '%s' is not a valid resource or tag",
cons_id, id);
free_xml(new_xml);
return FALSE;
} else if (rsc) {
continue;
} else if (tag) {
/* The resource_ref under the resource_set references a template/tag */
xmlNode *last_ref = xml_rsc;
/* A sample:
Original XML:
<resource_set id="tag1-colocation-0" sequential="true">
<resource_ref id="rsc1"/>
<resource_ref id="tag1"/>
<resource_ref id="rsc4"/>
</resource_set>
Now we are appending rsc2 and rsc3 which are tagged with tag1 right after it:
<resource_set id="tag1-colocation-0" sequential="true">
<resource_ref id="rsc1"/>
<resource_ref id="tag1"/>
<resource_ref id="rsc2"/>
<resource_ref id="rsc3"/>
<resource_ref id="rsc4"/>
</resource_set>
*/
for (gIter = tag->refs; gIter != NULL; gIter = gIter->next) {
const char *obj_ref = (const char *) gIter->data;
xmlNode *new_rsc_ref = NULL;
new_rsc_ref = xmlNewDocRawNode(getDocPtr(set), NULL,
(pcmkXmlStr) XML_TAG_RESOURCE_REF, NULL);
crm_xml_add(new_rsc_ref, XML_ATTR_ID, obj_ref);
xmlAddNextSibling(last_ref, new_rsc_ref);
last_ref = new_rsc_ref;
}
any_refs = TRUE;
/* Do not directly free '<resource_ref id="tag1"/>'.
That would break the further __xml_next_element(xml_rsc)) and cause "Invalid read" seen by valgrind.
So just record it into a hash table for freeing it later.
*/
tag_refs = g_list_append(tag_refs, xml_rsc);
}
}
/* Now free '<resource_ref id="tag1"/>', and finally get:
<resource_set id="tag1-colocation-0" sequential="true">
<resource_ref id="rsc1"/>
<resource_ref id="rsc2"/>
<resource_ref id="rsc3"/>
<resource_ref id="rsc4"/>
</resource_set>
*/
for (gIter = tag_refs; gIter != NULL; gIter = gIter->next) {
xmlNode *tag_ref = gIter->data;
free_xml(tag_ref);
}
g_list_free(tag_refs);
}
if (any_refs) {
*expanded_xml = new_xml;
} else {
free_xml(new_xml);
}
return TRUE;
}
static gboolean
tag_to_set(xmlNode * xml_obj, xmlNode ** rsc_set, const char * attr,
gboolean convert_rsc, pe_working_set_t * data_set)
{
const char *cons_id = NULL;
const char *id = NULL;
pe_resource_t *rsc = NULL;
pe_tag_t *tag = NULL;
*rsc_set = NULL;
CRM_CHECK((xml_obj != NULL) && (attr != NULL), return FALSE);
cons_id = ID(xml_obj);
if (cons_id == NULL) {
pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID,
crm_element_name(xml_obj));
return FALSE;
}
id = crm_element_value(xml_obj, attr);
if (id == NULL) {
return TRUE;
}
if (valid_resource_or_tag(data_set, id, &rsc, &tag) == FALSE) {
pcmk__config_err("Ignoring constraint '%s' because '%s' is not a "
"valid resource or tag", cons_id, id);
return FALSE;
} else if (tag) {
GListPtr gIter = NULL;
/* A template/tag is referenced by the "attr" attribute (first, then, rsc or with-rsc).
Add the template/tag's corresponding "resource_set" which contains the resources derived
from it or tagged with it under the constraint. */
*rsc_set = create_xml_node(xml_obj, XML_CONS_TAG_RSC_SET);
crm_xml_add(*rsc_set, XML_ATTR_ID, id);
for (gIter = tag->refs; gIter != NULL; gIter = gIter->next) {
const char *obj_ref = (const char *) gIter->data;
xmlNode *rsc_ref = NULL;
rsc_ref = create_xml_node(*rsc_set, XML_TAG_RESOURCE_REF);
crm_xml_add(rsc_ref, XML_ATTR_ID, obj_ref);
}
/* Set sequential="false" for the resource_set */
crm_xml_add(*rsc_set, "sequential", XML_BOOLEAN_FALSE);
} else if (rsc && convert_rsc) {
/* Even a regular resource is referenced by "attr", convert it into a resource_set.
Because the other side of the constraint could be a template/tag reference. */
xmlNode *rsc_ref = NULL;
*rsc_set = create_xml_node(xml_obj, XML_CONS_TAG_RSC_SET);
crm_xml_add(*rsc_set, XML_ATTR_ID, id);
rsc_ref = create_xml_node(*rsc_set, XML_TAG_RESOURCE_REF);
crm_xml_add(rsc_ref, XML_ATTR_ID, id);
} else {
return TRUE;
}
/* Remove the "attr" attribute referencing the template/tag */
if (*rsc_set) {
xml_remove_prop(xml_obj, attr);
}
return TRUE;
}
static gboolean unpack_rsc_location(xmlNode * xml_obj, pe_resource_t * rsc_lh, const char * role,
const char * score, pe_working_set_t * data_set, pe_match_data_t * match_data);
static gboolean
unpack_simple_location(xmlNode * xml_obj, pe_working_set_t * data_set)
{
const char *id = crm_element_value(xml_obj, XML_ATTR_ID);
const char *value = crm_element_value(xml_obj, XML_LOC_ATTR_SOURCE);
if(value) {
pe_resource_t *rsc_lh = pe_find_constraint_resource(data_set->resources, value);
return unpack_rsc_location(xml_obj, rsc_lh, NULL, NULL, data_set, NULL);
}
value = crm_element_value(xml_obj, XML_LOC_ATTR_SOURCE_PATTERN);
if(value) {
regex_t *r_patt = calloc(1, sizeof(regex_t));
bool invert = FALSE;
GListPtr rIter = NULL;
if(value[0] == '!') {
value++;
invert = TRUE;
}
if (regcomp(r_patt, value, REG_EXTENDED)) {
pcmk__config_err("Ignoring constraint '%s' because "
XML_LOC_ATTR_SOURCE_PATTERN
" has invalid value '%s'", id, value);
regfree(r_patt);
free(r_patt);
return FALSE;
}
for (rIter = data_set->resources; rIter; rIter = rIter->next) {
pe_resource_t *r = rIter->data;
int nregs = 0;
regmatch_t *pmatch = NULL;
int status;
if(r_patt->re_nsub > 0) {
nregs = r_patt->re_nsub + 1;
} else {
nregs = 1;
}
pmatch = calloc(nregs, sizeof(regmatch_t));
status = regexec(r_patt, r->id, nregs, pmatch, 0);
if(invert == FALSE && status == 0) {
pe_re_match_data_t re_match_data = {
.string = r->id,
.nregs = nregs,
.pmatch = pmatch
};
pe_match_data_t match_data = {
.re = &re_match_data,
.params = r->parameters,
.meta = r->meta,
};
crm_debug("'%s' matched '%s' for %s", r->id, value, id);
unpack_rsc_location(xml_obj, r, NULL, NULL, data_set, &match_data);
} else if (invert && (status != 0)) {
crm_debug("'%s' is an inverted match of '%s' for %s", r->id, value, id);
unpack_rsc_location(xml_obj, r, NULL, NULL, data_set, NULL);
} else {
crm_trace("'%s' does not match '%s' for %s", r->id, value, id);
}
free(pmatch);
}
regfree(r_patt);
free(r_patt);
}
return FALSE;
}
static gboolean
unpack_rsc_location(xmlNode * xml_obj, pe_resource_t * rsc_lh, const char * role,
const char * score, pe_working_set_t * data_set, pe_match_data_t * match_data)
{
pe__location_t *location = NULL;
const char *id_lh = crm_element_value(xml_obj, XML_LOC_ATTR_SOURCE);
const char *id = crm_element_value(xml_obj, XML_ATTR_ID);
const char *node = crm_element_value(xml_obj, XML_CIB_TAG_NODE);
const char *discovery = crm_element_value(xml_obj, XML_LOCATION_ATTR_DISCOVERY);
if (rsc_lh == NULL) {
pcmk__config_warn("Ignoring constraint '%s' because resource '%s' "
"does not exist", id, id_lh);
return FALSE;
}
if (score == NULL) {
score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE);
}
if (node != NULL && score != NULL) {
int score_i = char2score(score);
pe_node_t *match = pe_find_node(data_set->nodes, node);
if (!match) {
return FALSE;
}
location = rsc2node_new(id, rsc_lh, score_i, discovery, match, data_set);
} else {
bool empty = TRUE;
crm_time_t *next_change = crm_time_new_undefined();
/* This loop is logically parallel to pe_evaluate_rules(), except
* instead of checking whether any rule is active, we set up location
* constraints for each active rule.
*/
for (xmlNode *rule_xml = first_named_child(xml_obj, XML_TAG_RULE);
rule_xml != NULL; rule_xml = crm_next_same_xml(rule_xml)) {
empty = FALSE;
crm_trace("Unpacking %s/%s", id, ID(rule_xml));
generate_location_rule(rsc_lh, rule_xml, discovery, next_change,
data_set, match_data);
}
if (empty) {
pcmk__config_err("Ignoring constraint '%s' because it contains "
"no rules", id);
}
/* If there is a point in the future when the evaluation of a rule will
* change, make sure the scheduler is re-run by that time.
*/
if (crm_time_is_defined(next_change)) {
time_t t = (time_t) crm_time_get_seconds_since_epoch(next_change);
pe__update_recheck_time(t, data_set);
}
crm_time_free(next_change);
return TRUE;
}
if (role == NULL) {
role = crm_element_value(xml_obj, XML_RULE_ATTR_ROLE);
}
if (location && role) {
if (text2role(role) == RSC_ROLE_UNKNOWN) {
pe_err("Invalid constraint %s: Bad role %s", id, role);
return FALSE;
} else {
enum rsc_role_e r = text2role(role);
switch(r) {
case RSC_ROLE_UNKNOWN:
case RSC_ROLE_STARTED:
case RSC_ROLE_SLAVE:
/* Applies to all */
location->role_filter = RSC_ROLE_UNKNOWN;
break;
default:
location->role_filter = r;
break;
}
}
}
return TRUE;
}
static gboolean
unpack_location_tags(xmlNode * xml_obj, xmlNode ** expanded_xml, pe_working_set_t * data_set)
{
const char *id = NULL;
const char *id_lh = NULL;
const char *state_lh = NULL;
pe_resource_t *rsc_lh = NULL;
pe_tag_t *tag_lh = NULL;
xmlNode *new_xml = NULL;
xmlNode *rsc_set_lh = NULL;
*expanded_xml = NULL;
CRM_CHECK(xml_obj != NULL, return FALSE);
id = ID(xml_obj);
if (id == NULL) {
pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID,
crm_element_name(xml_obj));
return FALSE;
}
/* Attempt to expand any template/tag references in possible resource sets. */
expand_tags_in_sets(xml_obj, &new_xml, data_set);
if (new_xml) {
/* There are resource sets referencing templates. Return with the expanded XML. */
crm_log_xml_trace(new_xml, "Expanded rsc_location...");
*expanded_xml = new_xml;
return TRUE;
}
id_lh = crm_element_value(xml_obj, XML_LOC_ATTR_SOURCE);
if (id_lh == NULL) {
return TRUE;
}
if (valid_resource_or_tag(data_set, id_lh, &rsc_lh, &tag_lh) == FALSE) {
pcmk__config_err("Ignoring constraint '%s' because '%s' is not a "
"valid resource or tag", id, id_lh);
return FALSE;
} else if (rsc_lh) {
/* No template is referenced. */
return TRUE;
}
state_lh = crm_element_value(xml_obj, XML_RULE_ATTR_ROLE);
new_xml = copy_xml(xml_obj);
/* Convert the template/tag reference in "rsc" into a resource_set under the rsc_location constraint. */
if (tag_to_set(new_xml, &rsc_set_lh, XML_LOC_ATTR_SOURCE, FALSE, data_set) == FALSE) {
free_xml(new_xml);
return FALSE;
}
if (rsc_set_lh) {
if (state_lh) {
/* A "rsc-role" is specified.
Move it into the converted resource_set as a "role"" attribute. */
crm_xml_add(rsc_set_lh, "role", state_lh);
xml_remove_prop(new_xml, XML_RULE_ATTR_ROLE);
}
crm_log_xml_trace(new_xml, "Expanded rsc_location...");
*expanded_xml = new_xml;
} else {
/* No sets */
free_xml(new_xml);
}
return TRUE;
}
static gboolean
unpack_location_set(xmlNode * location, xmlNode * set, pe_working_set_t * data_set)
{
xmlNode *xml_rsc = NULL;
pe_resource_t *resource = NULL;
const char *set_id;
const char *role;
const char *local_score;
CRM_CHECK(set != NULL, return FALSE);
set_id = ID(set);
if (set_id == NULL) {
pcmk__config_err("Ignoring " XML_CONS_TAG_RSC_SET " without "
XML_ATTR_ID " in constraint '%s'",
crm_str(ID(location)));
return FALSE;
}
role = crm_element_value(set, "role");
local_score = crm_element_value(set, XML_RULE_ATTR_SCORE);
for (xml_rsc = __xml_first_child_element(set); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
EXPAND_CONSTRAINT_IDREF(set_id, resource, ID(xml_rsc));
unpack_rsc_location(location, resource, role, local_score, data_set, NULL);
}
}
return TRUE;
}
gboolean
unpack_location(xmlNode * xml_obj, pe_working_set_t * data_set)
{
xmlNode *set = NULL;
gboolean any_sets = FALSE;
xmlNode *orig_xml = NULL;
xmlNode *expanded_xml = NULL;
if (unpack_location_tags(xml_obj, &expanded_xml, data_set) == FALSE) {
return FALSE;
}
if (expanded_xml) {
orig_xml = xml_obj;
xml_obj = expanded_xml;
}
for (set = __xml_first_child_element(xml_obj); set != NULL;
set = __xml_next_element(set)) {
if (crm_str_eq((const char *)set->name, XML_CONS_TAG_RSC_SET, TRUE)) {
any_sets = TRUE;
set = expand_idref(set, data_set->input);
if (unpack_location_set(xml_obj, set, data_set) == FALSE) {
if (expanded_xml) {
free_xml(expanded_xml);
}
return FALSE;
}
}
}
if (expanded_xml) {
free_xml(expanded_xml);
xml_obj = orig_xml;
}
if (any_sets == FALSE) {
return unpack_simple_location(xml_obj, data_set);
}
return TRUE;
}
static int
get_node_score(const char *rule, const char *score, gboolean raw, pe_node_t * node, pe_resource_t *rsc)
{
int score_f = 0;
if (score == NULL) {
pe_err("Rule %s: no score specified. Assuming 0.", rule);
} else if (raw) {
score_f = char2score(score);
} else {
const char *attr_score = pe_node_attribute_calculated(node, score, rsc);
if (attr_score == NULL) {
crm_debug("Rule %s: node %s did not have a value for %s",
rule, node->details->uname, score);
score_f = -INFINITY;
} else {
crm_debug("Rule %s: node %s had value %s for %s",
rule, node->details->uname, attr_score, score);
score_f = char2score(attr_score);
}
}
return score_f;
}
static pe__location_t *
generate_location_rule(pe_resource_t *rsc, xmlNode *rule_xml,
const char *discovery, crm_time_t *next_change,
pe_working_set_t *data_set, pe_match_data_t *match_data)
{
const char *rule_id = NULL;
const char *score = NULL;
const char *boolean = NULL;
const char *role = NULL;
GListPtr gIter = NULL;
GListPtr match_L = NULL;
gboolean do_and = TRUE;
gboolean accept = TRUE;
gboolean raw_score = TRUE;
gboolean score_allocated = FALSE;
pe__location_t *location_rule = NULL;
rule_xml = expand_idref(rule_xml, data_set->input);
rule_id = crm_element_value(rule_xml, XML_ATTR_ID);
boolean = crm_element_value(rule_xml, XML_RULE_ATTR_BOOLEAN_OP);
role = crm_element_value(rule_xml, XML_RULE_ATTR_ROLE);
crm_trace("Processing rule: %s", rule_id);
if (role != NULL && text2role(role) == RSC_ROLE_UNKNOWN) {
pe_err("Bad role specified for %s: %s", rule_id, role);
return NULL;
}
score = crm_element_value(rule_xml, XML_RULE_ATTR_SCORE);
if (score == NULL) {
score = crm_element_value(rule_xml, XML_RULE_ATTR_SCORE_ATTRIBUTE);
if (score != NULL) {
raw_score = FALSE;
}
}
if (safe_str_eq(boolean, "or")) {
do_and = FALSE;
}
location_rule = rsc2node_new(rule_id, rsc, 0, discovery, NULL, data_set);
if (location_rule == NULL) {
return NULL;
}
if (match_data && match_data->re && match_data->re->nregs > 0 && match_data->re->pmatch[0].rm_so != -1) {
if (raw_score == FALSE) {
char *result = pe_expand_re_matches(score, match_data->re);
if (result) {
score = (const char *) result;
score_allocated = TRUE;
}
}
}
if (role != NULL) {
crm_trace("Setting role filter: %s", role);
location_rule->role_filter = text2role(role);
if (location_rule->role_filter == RSC_ROLE_SLAVE) {
/* Any promotable clone cannot be promoted without being a slave first
* Ergo, any constraint for the slave role applies to every role
*/
location_rule->role_filter = RSC_ROLE_UNKNOWN;
}
}
if (do_and) {
GListPtr gIter = NULL;
match_L = node_list_dup(data_set->nodes, TRUE, FALSE);
for (gIter = match_L; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
node->weight = get_node_score(rule_id, score, raw_score, node, rsc);
}
}
for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
int score_f = 0;
pe_node_t *node = (pe_node_t *) gIter->data;
accept = pe_test_rule(rule_xml, node->details->attrs, RSC_ROLE_UNKNOWN,
data_set->now, next_change, match_data);
crm_trace("Rule %s %s on %s", ID(rule_xml), accept ? "passed" : "failed",
node->details->uname);
score_f = get_node_score(rule_id, score, raw_score, node, rsc);
/* if(accept && score_f == -INFINITY) { */
/* accept = FALSE; */
/* } */
if (accept) {
pe_node_t *local = pe_find_node_id(match_L, node->details->id);
if (local == NULL && do_and) {
continue;
} else if (local == NULL) {
local = node_copy(node);
match_L = g_list_append(match_L, local);
}
if (do_and == FALSE) {
local->weight = merge_weights(local->weight, score_f);
}
crm_trace("node %s now has weight %d", node->details->uname, local->weight);
} else if (do_and && !accept) {
/* remove it */
pe_node_t *delete = pe_find_node_id(match_L, node->details->id);
if (delete != NULL) {
match_L = g_list_remove(match_L, delete);
crm_trace("node %s did not match", node->details->uname);
}
free(delete);
}
}
if (score_allocated == TRUE) {
free((char *)score);
}
location_rule->node_list_rh = match_L;
if (location_rule->node_list_rh == NULL) {
crm_trace("No matching nodes for rule %s", rule_id);
return NULL;
}
crm_trace("%s: %d nodes matched", rule_id, g_list_length(location_rule->node_list_rh));
return location_rule;
}
static gint
sort_cons_priority_lh(gconstpointer a, gconstpointer b)
{
const rsc_colocation_t *rsc_constraint1 = (const rsc_colocation_t *)a;
const rsc_colocation_t *rsc_constraint2 = (const rsc_colocation_t *)b;
if (a == NULL) {
return 1;
}
if (b == NULL) {
return -1;
}
CRM_ASSERT(rsc_constraint1->rsc_lh != NULL);
CRM_ASSERT(rsc_constraint1->rsc_rh != NULL);
if (rsc_constraint1->rsc_lh->priority > rsc_constraint2->rsc_lh->priority) {
return -1;
}
if (rsc_constraint1->rsc_lh->priority < rsc_constraint2->rsc_lh->priority) {
return 1;
}
/* Process clones before primitives and groups */
if (rsc_constraint1->rsc_lh->variant > rsc_constraint2->rsc_lh->variant) {
return -1;
} else if (rsc_constraint1->rsc_lh->variant < rsc_constraint2->rsc_lh->variant) {
return 1;
}
/* @COMPAT scheduler <2.0.0: Process promotable clones before nonpromotable
* clones (probably unnecessary, but avoids having to update regression
* tests)
*/
if (rsc_constraint1->rsc_lh->variant == pe_clone) {
if (is_set(rsc_constraint1->rsc_lh->flags, pe_rsc_promotable)
&& is_not_set(rsc_constraint2->rsc_lh->flags, pe_rsc_promotable)) {
return -1;
} else if (is_not_set(rsc_constraint1->rsc_lh->flags, pe_rsc_promotable)
&& is_set(rsc_constraint2->rsc_lh->flags, pe_rsc_promotable)) {
return 1;
}
}
return strcmp(rsc_constraint1->rsc_lh->id, rsc_constraint2->rsc_lh->id);
}
static gint
sort_cons_priority_rh(gconstpointer a, gconstpointer b)
{
const rsc_colocation_t *rsc_constraint1 = (const rsc_colocation_t *)a;
const rsc_colocation_t *rsc_constraint2 = (const rsc_colocation_t *)b;
if (a == NULL) {
return 1;
}
if (b == NULL) {
return -1;
}
CRM_ASSERT(rsc_constraint1->rsc_lh != NULL);
CRM_ASSERT(rsc_constraint1->rsc_rh != NULL);
if (rsc_constraint1->rsc_rh->priority > rsc_constraint2->rsc_rh->priority) {
return -1;
}
if (rsc_constraint1->rsc_rh->priority < rsc_constraint2->rsc_rh->priority) {
return 1;
}
/* Process clones before primitives and groups */
if (rsc_constraint1->rsc_rh->variant > rsc_constraint2->rsc_rh->variant) {
return -1;
} else if (rsc_constraint1->rsc_rh->variant < rsc_constraint2->rsc_rh->variant) {
return 1;
}
/* @COMPAT scheduler <2.0.0: Process promotable clones before nonpromotable
* clones (probably unnecessary, but avoids having to update regression
* tests)
*/
if (rsc_constraint1->rsc_rh->variant == pe_clone) {
if (is_set(rsc_constraint1->rsc_rh->flags, pe_rsc_promotable)
&& is_not_set(rsc_constraint2->rsc_rh->flags, pe_rsc_promotable)) {
return -1;
} else if (is_not_set(rsc_constraint1->rsc_rh->flags, pe_rsc_promotable)
&& is_set(rsc_constraint2->rsc_rh->flags, pe_rsc_promotable)) {
return 1;
}
}
return strcmp(rsc_constraint1->rsc_rh->id, rsc_constraint2->rsc_rh->id);
}
static void
anti_colocation_order(pe_resource_t * first_rsc, int first_role,
pe_resource_t * then_rsc, int then_role,
pe_working_set_t * data_set)
{
const char *first_tasks[] = { NULL, NULL };
const char *then_tasks[] = { NULL, NULL };
int first_lpc = 0;
int then_lpc = 0;
/* Actions to make first_rsc lose first_role */
if (first_role == RSC_ROLE_MASTER) {
first_tasks[0] = CRMD_ACTION_DEMOTE;
} else {
first_tasks[0] = CRMD_ACTION_STOP;
if (first_role == RSC_ROLE_SLAVE) {
first_tasks[1] = CRMD_ACTION_PROMOTE;
}
}
/* Actions to make then_rsc gain then_role */
if (then_role == RSC_ROLE_MASTER) {
then_tasks[0] = CRMD_ACTION_PROMOTE;
} else {
then_tasks[0] = CRMD_ACTION_START;
if (then_role == RSC_ROLE_SLAVE) {
then_tasks[1] = CRMD_ACTION_DEMOTE;
}
}
for (first_lpc = 0; first_lpc <= 1 && first_tasks[first_lpc] != NULL; first_lpc++) {
for (then_lpc = 0; then_lpc <= 1 && then_tasks[then_lpc] != NULL; then_lpc++) {
new_rsc_order(first_rsc, first_tasks[first_lpc], then_rsc, then_tasks[then_lpc],
pe_order_anti_colocation, data_set);
}
}
}
gboolean
rsc_colocation_new(const char *id, const char *node_attr, int score,
pe_resource_t * rsc_lh, pe_resource_t * rsc_rh,
const char *state_lh, const char *state_rh, pe_working_set_t * data_set)
{
rsc_colocation_t *new_con = NULL;
if ((rsc_lh == NULL) || (rsc_rh == NULL)) {
pcmk__config_err("Ignoring colocation '%s' because resource "
"does not exist", id);
return FALSE;
}
new_con = calloc(1, sizeof(rsc_colocation_t));
if (new_con == NULL) {
return FALSE;
}
if (state_lh == NULL || safe_str_eq(state_lh, RSC_ROLE_STARTED_S)) {
state_lh = RSC_ROLE_UNKNOWN_S;
}
if (state_rh == NULL || safe_str_eq(state_rh, RSC_ROLE_STARTED_S)) {
state_rh = RSC_ROLE_UNKNOWN_S;
}
new_con->id = id;
new_con->rsc_lh = rsc_lh;
new_con->rsc_rh = rsc_rh;
new_con->score = score;
new_con->role_lh = text2role(state_lh);
new_con->role_rh = text2role(state_rh);
new_con->node_attribute = node_attr;
if (node_attr == NULL) {
node_attr = CRM_ATTR_UNAME;
}
pe_rsc_trace(rsc_lh, "%s ==> %s (%s %d)", rsc_lh->id, rsc_rh->id, node_attr, score);
rsc_lh->rsc_cons = g_list_insert_sorted(rsc_lh->rsc_cons, new_con, sort_cons_priority_rh);
rsc_rh->rsc_cons_lhs =
g_list_insert_sorted(rsc_rh->rsc_cons_lhs, new_con, sort_cons_priority_lh);
data_set->colocation_constraints = g_list_append(data_set->colocation_constraints, new_con);
if (score <= -INFINITY) {
anti_colocation_order(rsc_lh, new_con->role_lh, rsc_rh, new_con->role_rh, data_set);
anti_colocation_order(rsc_rh, new_con->role_rh, rsc_lh, new_con->role_lh, data_set);
}
return TRUE;
}
/* LHS before RHS */
int
new_rsc_order(pe_resource_t * lh_rsc, const char *lh_task,
pe_resource_t * rh_rsc, const char *rh_task,
enum pe_ordering type, pe_working_set_t * data_set)
{
char *lh_key = NULL;
char *rh_key = NULL;
CRM_CHECK(lh_rsc != NULL, return -1);
CRM_CHECK(lh_task != NULL, return -1);
CRM_CHECK(rh_rsc != NULL, return -1);
CRM_CHECK(rh_task != NULL, return -1);
#if 0
/* We do not need to test if these reference stonith resources
* because the fencer has access to them even when they're not "running"
*/
if (validate_order_resources(lh_rsc, lh_task, rh_rsc, rh_task)) {
return -1;
}
#endif
lh_key = pcmk__op_key(lh_rsc->id, lh_task, 0);
rh_key = pcmk__op_key(rh_rsc->id, rh_task, 0);
return custom_action_order(lh_rsc, lh_key, NULL, rh_rsc, rh_key, NULL, type, data_set);
}
static char *
task_from_action_or_key(pe_action_t *action, const char *key)
{
char *res = NULL;
if (action) {
res = strdup(action->task);
} else if (key) {
parse_op_key(key, NULL, &res, NULL);
}
return res;
}
/* when order constraints are made between two resources start and stop actions
* those constraints have to be mirrored against the corresponding
* migration actions to ensure start/stop ordering is preserved during
* a migration */
static void
handle_migration_ordering(pe__ordering_t *order, pe_working_set_t *data_set)
{
char *lh_task = NULL;
char *rh_task = NULL;
gboolean rh_migratable;
gboolean lh_migratable;
if (order->lh_rsc == NULL || order->rh_rsc == NULL) {
return;
} else if (order->lh_rsc == order->rh_rsc) {
return;
/* don't mess with those constraints built between parent
* resources and the children */
} else if (is_parent(order->lh_rsc, order->rh_rsc)) {
return;
} else if (is_parent(order->rh_rsc, order->lh_rsc)) {
return;
}
lh_migratable = is_set(order->lh_rsc->flags, pe_rsc_allow_migrate);
rh_migratable = is_set(order->rh_rsc->flags, pe_rsc_allow_migrate);
/* one of them has to be migratable for
* the migrate ordering logic to be applied */
if (lh_migratable == FALSE && rh_migratable == FALSE) {
return;
}
/* at this point we have two resources which allow migrations that have an
* order dependency set between them. If those order dependencies involve
* start/stop actions, we need to mirror the corresponding migrate actions
* so order will be preserved. */
lh_task = task_from_action_or_key(order->lh_action, order->lh_action_task);
rh_task = task_from_action_or_key(order->rh_action, order->rh_action_task);
if (lh_task == NULL || rh_task == NULL) {
goto cleanup_order;
}
if (safe_str_eq(lh_task, RSC_START) && safe_str_eq(rh_task, RSC_START)) {
int flags = pe_order_optional;
if (lh_migratable && rh_migratable) {
/* A start then B start
* A migrate_from then B migrate_to */
custom_action_order(order->lh_rsc,
pcmk__op_key(order->lh_rsc->id, RSC_MIGRATED, 0),
NULL, order->rh_rsc,
pcmk__op_key(order->rh_rsc->id, RSC_MIGRATE, 0),
NULL, flags, data_set);
}
if (rh_migratable) {
if (lh_migratable) {
flags |= pe_order_apply_first_non_migratable;
}
/* A start then B start
* A start then B migrate_to... only if A start is not a part of a migration*/
custom_action_order(order->lh_rsc,
pcmk__op_key(order->lh_rsc->id, RSC_START, 0),
NULL, order->rh_rsc,
pcmk__op_key(order->rh_rsc->id, RSC_MIGRATE, 0),
NULL, flags, data_set);
}
} else if (rh_migratable == TRUE && safe_str_eq(lh_task, RSC_STOP) && safe_str_eq(rh_task, RSC_STOP)) {
int flags = pe_order_optional;
if (lh_migratable) {
flags |= pe_order_apply_first_non_migratable;
}
/* rh side is at the bottom of the stack during a stop. If we have a constraint
* stop B then stop A, if B is migrating via stop/start, and A is migrating using migration actions,
* we need to enforce that A's migrate_to action occurs after B's stop action. */
custom_action_order(order->lh_rsc,
pcmk__op_key(order->lh_rsc->id, RSC_STOP, 0), NULL,
order->rh_rsc,
pcmk__op_key(order->rh_rsc->id, RSC_MIGRATE, 0),
NULL, flags, data_set);
/* We need to build the stop constraint against migrate_from as well
* to account for partial migrations. */
if (order->rh_rsc->partial_migration_target) {
custom_action_order(order->lh_rsc,
pcmk__op_key(order->lh_rsc->id, RSC_STOP, 0),
NULL, order->rh_rsc,
pcmk__op_key(order->rh_rsc->id, RSC_MIGRATED, 0),
NULL, flags, data_set);
}
} else if (safe_str_eq(lh_task, RSC_PROMOTE) && safe_str_eq(rh_task, RSC_START)) {
int flags = pe_order_optional;
if (rh_migratable) {
/* A promote then B start
* A promote then B migrate_to */
custom_action_order(order->lh_rsc,
pcmk__op_key(order->lh_rsc->id, RSC_PROMOTE, 0),
NULL, order->rh_rsc,
pcmk__op_key(order->rh_rsc->id, RSC_MIGRATE, 0),
NULL, flags, data_set);
}
} else if (safe_str_eq(lh_task, RSC_DEMOTE) && safe_str_eq(rh_task, RSC_STOP)) {
int flags = pe_order_optional;
if (rh_migratable) {
/* A demote then B stop
* A demote then B migrate_to */
custom_action_order(order->lh_rsc, pcmk__op_key(order->lh_rsc->id, RSC_DEMOTE, 0), NULL,
order->rh_rsc, pcmk__op_key(order->rh_rsc->id, RSC_MIGRATE, 0), NULL,
flags, data_set);
/* We need to build the demote constraint against migrate_from as well
* to account for partial migrations. */
if (order->rh_rsc->partial_migration_target) {
custom_action_order(order->lh_rsc,
pcmk__op_key(order->lh_rsc->id, RSC_DEMOTE, 0),
NULL, order->rh_rsc,
pcmk__op_key(order->rh_rsc->id, RSC_MIGRATED, 0),
NULL, flags, data_set);
}
}
}
cleanup_order:
free(lh_task);
free(rh_task);
}
/* LHS before RHS */
int
custom_action_order(pe_resource_t * lh_rsc, char *lh_action_task, pe_action_t * lh_action,
pe_resource_t * rh_rsc, char *rh_action_task, pe_action_t * rh_action,
enum pe_ordering type, pe_working_set_t * data_set)
{
pe__ordering_t *order = NULL;
if (lh_rsc == NULL && lh_action) {
lh_rsc = lh_action->rsc;
}
if (rh_rsc == NULL && rh_action) {
rh_rsc = rh_action->rsc;
}
if ((lh_action == NULL && lh_rsc == NULL)
|| (rh_action == NULL && rh_rsc == NULL)) {
crm_err("Invalid ordering (bug?)");
free(lh_action_task);
free(rh_action_task);
return -1;
}
order = calloc(1, sizeof(pe__ordering_t));
crm_trace("Creating[%d] %s %s %s - %s %s %s", data_set->order_id,
lh_rsc?lh_rsc->id:"NA", lh_action_task, lh_action?lh_action->uuid:"NA",
rh_rsc?rh_rsc->id:"NA", rh_action_task, rh_action?rh_action->uuid:"NA");
/* CRM_ASSERT(data_set->order_id != 291); */
order->id = data_set->order_id++;
order->type = type;
order->lh_rsc = lh_rsc;
order->rh_rsc = rh_rsc;
order->lh_action = lh_action;
order->rh_action = rh_action;
order->lh_action_task = lh_action_task;
order->rh_action_task = rh_action_task;
if (order->lh_action_task == NULL && lh_action) {
order->lh_action_task = strdup(lh_action->uuid);
}
if (order->rh_action_task == NULL && rh_action) {
order->rh_action_task = strdup(rh_action->uuid);
}
if (order->lh_rsc == NULL && lh_action) {
order->lh_rsc = lh_action->rsc;
}
if (order->rh_rsc == NULL && rh_action) {
order->rh_rsc = rh_action->rsc;
}
data_set->ordering_constraints = g_list_prepend(data_set->ordering_constraints, order);
handle_migration_ordering(order, data_set);
return order->id;
}
enum pe_ordering
get_asymmetrical_flags(enum pe_order_kind kind)
{
enum pe_ordering flags = pe_order_optional;
if (kind == pe_order_kind_mandatory) {
flags |= pe_order_asymmetrical;
} else if (kind == pe_order_kind_serialize) {
flags |= pe_order_serialize_only;
}
return flags;
}
enum pe_ordering
get_flags(const char *id, enum pe_order_kind kind,
const char *action_first, const char *action_then, gboolean invert)
{
enum pe_ordering flags = pe_order_optional;
if (invert && kind == pe_order_kind_mandatory) {
crm_trace("Upgrade %s: implies left", id);
flags |= pe_order_implies_first;
} else if (kind == pe_order_kind_mandatory) {
crm_trace("Upgrade %s: implies right", id);
flags |= pe_order_implies_then;
if (safe_str_eq(action_first, RSC_START)
|| safe_str_eq(action_first, RSC_PROMOTE)) {
crm_trace("Upgrade %s: runnable", id);
flags |= pe_order_runnable_left;
}
} else if (kind == pe_order_kind_serialize) {
flags |= pe_order_serialize_only;
}
return flags;
}
static gboolean
unpack_order_set(xmlNode * set, enum pe_order_kind parent_kind, pe_resource_t ** rsc,
pe_action_t ** begin, pe_action_t ** end, pe_action_t ** inv_begin,
pe_action_t ** inv_end, const char *parent_symmetrical_s,
pe_working_set_t * data_set)
{
xmlNode *xml_rsc = NULL;
GListPtr set_iter = NULL;
GListPtr resources = NULL;
pe_resource_t *last = NULL;
pe_resource_t *resource = NULL;
int local_kind = parent_kind;
gboolean sequential = FALSE;
enum pe_ordering flags = pe_order_optional;
gboolean symmetrical = TRUE;
char *key = NULL;
const char *id = ID(set);
const char *action = crm_element_value(set, "action");
const char *sequential_s = crm_element_value(set, "sequential");
const char *kind_s = crm_element_value(set, XML_ORDER_ATTR_KIND);
/*
char *pseudo_id = NULL;
char *end_id = NULL;
char *begin_id = NULL;
*/
if (action == NULL) {
action = RSC_START;
}
if (kind_s) {
local_kind = get_ordering_type(set);
}
if (sequential_s == NULL) {
sequential_s = "1";
}
sequential = crm_is_true(sequential_s);
symmetrical = order_is_symmetrical(set, parent_kind, parent_symmetrical_s);
if (symmetrical) {
flags = get_flags(id, local_kind, action, action, FALSE);
} else {
flags = get_asymmetrical_flags(local_kind);
}
for (xml_rsc = __xml_first_child_element(set); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
EXPAND_CONSTRAINT_IDREF(id, resource, ID(xml_rsc));
resources = g_list_append(resources, resource);
}
}
if (pcmk__list_of_1(resources)) {
crm_trace("Single set: %s", id);
*rsc = resource;
*end = NULL;
*begin = NULL;
*inv_end = NULL;
*inv_begin = NULL;
goto done;
}
/*
pseudo_id = crm_strdup_printf("%s-%s", id, action);
end_id = crm_strdup_printf("%s-%s", pseudo_id, "end");
begin_id = crm_strdup_printf("%s-%s", pseudo_id, "begin");
*/
*rsc = NULL;
/*
*end = get_pseudo_op(end_id, data_set);
*begin = get_pseudo_op(begin_id, data_set);
free(pseudo_id);
free(begin_id);
free(end_id);
*/
set_iter = resources;
while (set_iter != NULL) {
resource = (pe_resource_t *) set_iter->data;
set_iter = set_iter->next;
key = pcmk__op_key(resource->id, action, 0);
/*
custom_action_order(NULL, NULL, *begin, resource, strdup(key), NULL,
flags|pe_order_implies_first_printed, data_set);
custom_action_order(resource, strdup(key), NULL, NULL, NULL, *end,
flags|pe_order_implies_then_printed, data_set);
*/
if (local_kind == pe_order_kind_serialize) {
/* Serialize before everything that comes after */
GListPtr gIter = NULL;
for (gIter = set_iter; gIter != NULL; gIter = gIter->next) {
pe_resource_t *then_rsc = (pe_resource_t *) gIter->data;
char *then_key = pcmk__op_key(then_rsc->id, action, 0);
custom_action_order(resource, strdup(key), NULL, then_rsc, then_key, NULL,
flags, data_set);
}
} else if (sequential) {
if (last != NULL) {
new_rsc_order(last, action, resource, action, flags, data_set);
}
last = resource;
}
free(key);
}
if (symmetrical == FALSE) {
goto done;
}
last = NULL;
action = invert_action(action);
/*
pseudo_id = crm_strdup_printf("%s-%s", id, action);
end_id = crm_strdup_printf("%s-%s", pseudo_id, "end");
begin_id = crm_strdup_printf("%s-%s", pseudo_id, "begin");
*inv_end = get_pseudo_op(end_id, data_set);
*inv_begin = get_pseudo_op(begin_id, data_set);
free(pseudo_id);
free(begin_id);
free(end_id);
*/
flags = get_flags(id, local_kind, action, action, TRUE);
set_iter = resources;
while (set_iter != NULL) {
resource = (pe_resource_t *) set_iter->data;
set_iter = set_iter->next;
/*
key = pcmk__op_key(resource->id, action, 0);
custom_action_order(NULL, NULL, *inv_begin, resource, strdup(key), NULL,
flags|pe_order_implies_first_printed, data_set);
custom_action_order(resource, key, NULL, NULL, NULL, *inv_end,
flags|pe_order_implies_then_printed, data_set);
*/
if (sequential) {
if (last != NULL) {
new_rsc_order(resource, action, last, action, flags, data_set);
}
last = resource;
}
}
done:
g_list_free(resources);
return TRUE;
}
static gboolean
order_rsc_sets(const char *id, xmlNode * set1, xmlNode * set2, enum pe_order_kind kind,
pe_working_set_t * data_set, gboolean invert, gboolean symmetrical)
{
xmlNode *xml_rsc = NULL;
xmlNode *xml_rsc_2 = NULL;
pe_resource_t *rsc_1 = NULL;
pe_resource_t *rsc_2 = NULL;
const char *action_1 = crm_element_value(set1, "action");
const char *action_2 = crm_element_value(set2, "action");
const char *sequential_1 = crm_element_value(set1, "sequential");
const char *sequential_2 = crm_element_value(set2, "sequential");
const char *require_all_s = crm_element_value(set1, "require-all");
gboolean require_all = require_all_s ? crm_is_true(require_all_s) : TRUE;
enum pe_ordering flags = pe_order_none;
if (action_1 == NULL) {
action_1 = RSC_START;
};
if (action_2 == NULL) {
action_2 = RSC_START;
};
if (invert) {
action_1 = invert_action(action_1);
action_2 = invert_action(action_2);
}
if(safe_str_eq(RSC_STOP, action_1) || safe_str_eq(RSC_DEMOTE, action_1)) {
/* Assuming: A -> ( B || C) -> D
* The one-or-more logic only applies during the start/promote phase
* During shutdown neither B nor can shutdown until D is down, so simply turn require_all back on.
*/
require_all = TRUE;
}
if (symmetrical == FALSE) {
flags = get_asymmetrical_flags(kind);
} else {
flags = get_flags(id, kind, action_2, action_1, invert);
}
/* If we have an un-ordered set1, whether it is sequential or not is irrelevant in regards to set2. */
if (!require_all) {
char *task = crm_strdup_printf(CRM_OP_RELAXED_SET ":%s", ID(set1));
pe_action_t *unordered_action = get_pseudo_op(task, data_set);
free(task);
update_action_flags(unordered_action, pe_action_requires_any, __FUNCTION__, __LINE__);
for (xml_rsc = __xml_first_child_element(set1); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (!crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
continue;
}
EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc));
/* Add an ordering constraint between every element in set1 and the pseudo action.
* If any action in set1 is runnable the pseudo action will be runnable. */
custom_action_order(rsc_1, pcmk__op_key(rsc_1->id, action_1, 0),
NULL, NULL, NULL, unordered_action,
pe_order_one_or_more|pe_order_implies_then_printed,
data_set);
}
for (xml_rsc_2 = __xml_first_child_element(set2); xml_rsc_2 != NULL;
xml_rsc_2 = __xml_next_element(xml_rsc_2)) {
if (!crm_str_eq((const char *)xml_rsc_2->name, XML_TAG_RESOURCE_REF, TRUE)) {
continue;
}
EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc_2));
/* Add an ordering constraint between the pseudo action and every element in set2.
* If the pseudo action is runnable, every action in set2 will be runnable */
custom_action_order(NULL, NULL, unordered_action,
rsc_2, pcmk__op_key(rsc_2->id, action_2, 0),
NULL, flags|pe_order_runnable_left, data_set);
}
return TRUE;
}
if (crm_is_true(sequential_1)) {
if (invert == FALSE) {
/* get the last one */
const char *rid = NULL;
for (xml_rsc = __xml_first_child_element(set1); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
rid = ID(xml_rsc);
}
}
EXPAND_CONSTRAINT_IDREF(id, rsc_1, rid);
} else {
/* get the first one */
for (xml_rsc = __xml_first_child_element(set1); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc));
break;
}
}
}
}
if (crm_is_true(sequential_2)) {
if (invert == FALSE) {
/* get the first one */
for (xml_rsc = __xml_first_child_element(set2); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc));
break;
}
}
} else {
/* get the last one */
const char *rid = NULL;
for (xml_rsc = __xml_first_child_element(set2); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
rid = ID(xml_rsc);
}
}
EXPAND_CONSTRAINT_IDREF(id, rsc_2, rid);
}
}
if (rsc_1 != NULL && rsc_2 != NULL) {
new_rsc_order(rsc_1, action_1, rsc_2, action_2, flags, data_set);
} else if (rsc_1 != NULL) {
for (xml_rsc = __xml_first_child_element(set2); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc));
new_rsc_order(rsc_1, action_1, rsc_2, action_2, flags, data_set);
}
}
} else if (rsc_2 != NULL) {
xmlNode *xml_rsc = NULL;
for (xml_rsc = __xml_first_child_element(set1); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc));
new_rsc_order(rsc_1, action_1, rsc_2, action_2, flags, data_set);
}
}
} else {
for (xml_rsc = __xml_first_child_element(set1); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
xmlNode *xml_rsc_2 = NULL;
EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc));
for (xml_rsc_2 = __xml_first_child_element(set2);
xml_rsc_2 != NULL;
xml_rsc_2 = __xml_next_element(xml_rsc_2)) {
if (crm_str_eq((const char *)xml_rsc_2->name, XML_TAG_RESOURCE_REF, TRUE)) {
EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc_2));
new_rsc_order(rsc_1, action_1, rsc_2, action_2, flags, data_set);
}
}
}
}
}
return TRUE;
}
static gboolean
unpack_order_tags(xmlNode * xml_obj, xmlNode ** expanded_xml, pe_working_set_t * data_set)
{
const char *id = NULL;
const char *id_first = NULL;
const char *id_then = NULL;
const char *action_first = NULL;
const char *action_then = NULL;
pe_resource_t *rsc_first = NULL;
pe_resource_t *rsc_then = NULL;
pe_tag_t *tag_first = NULL;
pe_tag_t *tag_then = NULL;
xmlNode *new_xml = NULL;
xmlNode *rsc_set_first = NULL;
xmlNode *rsc_set_then = NULL;
gboolean any_sets = FALSE;
*expanded_xml = NULL;
CRM_CHECK(xml_obj != NULL, return FALSE);
id = ID(xml_obj);
if (id == NULL) {
pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID,
crm_element_name(xml_obj));
return FALSE;
}
/* Attempt to expand any template/tag references in possible resource sets. */
expand_tags_in_sets(xml_obj, &new_xml, data_set);
if (new_xml) {
/* There are resource sets referencing templates/tags. Return with the expanded XML. */
crm_log_xml_trace(new_xml, "Expanded rsc_order...");
*expanded_xml = new_xml;
return TRUE;
}
id_first = crm_element_value(xml_obj, XML_ORDER_ATTR_FIRST);
id_then = crm_element_value(xml_obj, XML_ORDER_ATTR_THEN);
if (id_first == NULL || id_then == NULL) {
return TRUE;
}
if (valid_resource_or_tag(data_set, id_first, &rsc_first, &tag_first) == FALSE) {
pcmk__config_err("Ignoring constraint '%s' because '%s' is not a "
"valid resource or tag", id, id_first);
return FALSE;
}
if (valid_resource_or_tag(data_set, id_then, &rsc_then, &tag_then) == FALSE) {
pcmk__config_err("Ignoring constraint '%s' because '%s' is not a "
"valid resource or tag", id, id_then);
return FALSE;
}
if (rsc_first && rsc_then) {
/* Neither side references any template/tag. */
return TRUE;
}
action_first = crm_element_value(xml_obj, XML_ORDER_ATTR_FIRST_ACTION);
action_then = crm_element_value(xml_obj, XML_ORDER_ATTR_THEN_ACTION);
new_xml = copy_xml(xml_obj);
/* Convert the template/tag reference in "first" into a resource_set under the order constraint. */
if (tag_to_set(new_xml, &rsc_set_first, XML_ORDER_ATTR_FIRST, TRUE, data_set) == FALSE) {
free_xml(new_xml);
return FALSE;
}
if (rsc_set_first) {
if (action_first) {
/* A "first-action" is specified.
Move it into the converted resource_set as an "action" attribute. */
crm_xml_add(rsc_set_first, "action", action_first);
xml_remove_prop(new_xml, XML_ORDER_ATTR_FIRST_ACTION);
}
any_sets = TRUE;
}
/* Convert the template/tag reference in "then" into a resource_set under the order constraint. */
if (tag_to_set(new_xml, &rsc_set_then, XML_ORDER_ATTR_THEN, TRUE, data_set) == FALSE) {
free_xml(new_xml);
return FALSE;
}
if (rsc_set_then) {
if (action_then) {
/* A "then-action" is specified.
Move it into the converted resource_set as an "action" attribute. */
crm_xml_add(rsc_set_then, "action", action_then);
xml_remove_prop(new_xml, XML_ORDER_ATTR_THEN_ACTION);
}
any_sets = TRUE;
}
if (any_sets) {
crm_log_xml_trace(new_xml, "Expanded rsc_order...");
*expanded_xml = new_xml;
} else {
free_xml(new_xml);
}
return TRUE;
}
gboolean
unpack_rsc_order(xmlNode * xml_obj, pe_working_set_t * data_set)
{
gboolean any_sets = FALSE;
pe_resource_t *rsc = NULL;
/*
pe_resource_t *last_rsc = NULL;
*/
pe_action_t *set_end = NULL;
pe_action_t *set_begin = NULL;
pe_action_t *set_inv_end = NULL;
pe_action_t *set_inv_begin = NULL;
xmlNode *set = NULL;
xmlNode *last = NULL;
xmlNode *orig_xml = NULL;
xmlNode *expanded_xml = NULL;
/*
pe_action_t *last_end = NULL;
pe_action_t *last_begin = NULL;
pe_action_t *last_inv_end = NULL;
pe_action_t *last_inv_begin = NULL;
*/
const char *id = crm_element_value(xml_obj, XML_ATTR_ID);
const char *invert = crm_element_value(xml_obj, XML_CONS_ATTR_SYMMETRICAL);
enum pe_order_kind kind = get_ordering_type(xml_obj);
gboolean invert_bool = order_is_symmetrical(xml_obj, kind, NULL);
gboolean rc = TRUE;
rc = unpack_order_tags(xml_obj, &expanded_xml, data_set);
if (expanded_xml) {
orig_xml = xml_obj;
xml_obj = expanded_xml;
} else if (rc == FALSE) {
return FALSE;
}
for (set = __xml_first_child_element(xml_obj); set != NULL;
set = __xml_next_element(set)) {
if (crm_str_eq((const char *)set->name, XML_CONS_TAG_RSC_SET, TRUE)) {
any_sets = TRUE;
set = expand_idref(set, data_set->input);
if (unpack_order_set(set, kind, &rsc, &set_begin, &set_end,
&set_inv_begin, &set_inv_end, invert, data_set) == FALSE) {
return FALSE;
/* Expand orders in order_rsc_sets() instead of via pseudo actions. */
/*
} else if(last) {
const char *set_action = crm_element_value(set, "action");
const char *last_action = crm_element_value(last, "action");
enum pe_ordering flags = get_flags(id, kind, last_action, set_action, FALSE);
if(!set_action) { set_action = RSC_START; }
if(!last_action) { last_action = RSC_START; }
if(rsc == NULL && last_rsc == NULL) {
order_actions(last_end, set_begin, flags);
} else {
custom_action_order(
last_rsc, null_or_opkey(last_rsc, last_action), last_end,
rsc, null_or_opkey(rsc, set_action), set_begin,
flags, data_set);
}
if(crm_is_true(invert)) {
set_action = invert_action(set_action);
last_action = invert_action(last_action);
flags = get_flags(id, kind, last_action, set_action, TRUE);
if(rsc == NULL && last_rsc == NULL) {
order_actions(last_inv_begin, set_inv_end, flags);
} else {
custom_action_order(
last_rsc, null_or_opkey(last_rsc, last_action), last_inv_begin,
rsc, null_or_opkey(rsc, set_action), set_inv_end,
flags, data_set);
}
}
*/
} else if ( /* never called -- Now call it for supporting clones in resource sets */
last) {
if (order_rsc_sets(id, last, set, kind, data_set, FALSE, invert_bool) == FALSE) {
return FALSE;
}
if (invert_bool
&& order_rsc_sets(id, set, last, kind, data_set, TRUE, invert_bool) == FALSE) {
return FALSE;
}
}
last = set;
/*
last_rsc = rsc;
last_end = set_end;
last_begin = set_begin;
last_inv_end = set_inv_end;
last_inv_begin = set_inv_begin;
*/
}
}
if (expanded_xml) {
free_xml(expanded_xml);
xml_obj = orig_xml;
}
if (any_sets == FALSE) {
return unpack_simple_rsc_order(xml_obj, data_set);
}
return TRUE;
}
static gboolean
unpack_colocation_set(xmlNode * set, int score, pe_working_set_t * data_set)
{
xmlNode *xml_rsc = NULL;
pe_resource_t *with = NULL;
pe_resource_t *resource = NULL;
const char *set_id = ID(set);
const char *role = crm_element_value(set, "role");
const char *sequential = crm_element_value(set, "sequential");
const char *ordering = crm_element_value(set, "ordering");
int local_score = score;
const char *score_s = crm_element_value(set, XML_RULE_ATTR_SCORE);
if (score_s) {
local_score = char2score(score_s);
}
if(ordering == NULL) {
ordering = "group";
}
if (sequential != NULL && crm_is_true(sequential) == FALSE) {
return TRUE;
} else if (local_score >= 0 && safe_str_eq(ordering, "group")) {
for (xml_rsc = __xml_first_child_element(set); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
EXPAND_CONSTRAINT_IDREF(set_id, resource, ID(xml_rsc));
if (with != NULL) {
pe_rsc_trace(resource, "Colocating %s with %s", resource->id, with->id);
rsc_colocation_new(set_id, NULL, local_score, resource, with, role, role,
data_set);
}
with = resource;
}
}
} else if (local_score >= 0) {
pe_resource_t *last = NULL;
for (xml_rsc = __xml_first_child_element(set); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
EXPAND_CONSTRAINT_IDREF(set_id, resource, ID(xml_rsc));
if (last != NULL) {
pe_rsc_trace(resource, "Colocating %s with %s", last->id, resource->id);
rsc_colocation_new(set_id, NULL, local_score, last, resource, role, role,
data_set);
}
last = resource;
}
}
} else {
/* Anti-colocating with every prior resource is
* the only way to ensure the intuitive result
* (i.e. that no one in the set can run with anyone else in the set)
*/
for (xml_rsc = __xml_first_child_element(set); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
xmlNode *xml_rsc_with = NULL;
EXPAND_CONSTRAINT_IDREF(set_id, resource, ID(xml_rsc));
for (xml_rsc_with = __xml_first_child_element(set);
xml_rsc_with != NULL;
xml_rsc_with = __xml_next_element(xml_rsc_with)) {
if (crm_str_eq((const char *)xml_rsc_with->name, XML_TAG_RESOURCE_REF, TRUE)) {
if (safe_str_eq(resource->id, ID(xml_rsc_with))) {
break;
}
EXPAND_CONSTRAINT_IDREF(set_id, with, ID(xml_rsc_with));
pe_rsc_trace(resource, "Anti-Colocating %s with %s", resource->id,
with->id);
rsc_colocation_new(set_id, NULL, local_score, resource, with, role, role,
data_set);
}
}
}
}
}
return TRUE;
}
static gboolean
colocate_rsc_sets(const char *id, xmlNode * set1, xmlNode * set2, int score,
pe_working_set_t * data_set)
{
xmlNode *xml_rsc = NULL;
pe_resource_t *rsc_1 = NULL;
pe_resource_t *rsc_2 = NULL;
const char *role_1 = crm_element_value(set1, "role");
const char *role_2 = crm_element_value(set2, "role");
const char *sequential_1 = crm_element_value(set1, "sequential");
const char *sequential_2 = crm_element_value(set2, "sequential");
if (sequential_1 == NULL || crm_is_true(sequential_1)) {
/* get the first one */
for (xml_rsc = __xml_first_child_element(set1); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc));
break;
}
}
}
if (sequential_2 == NULL || crm_is_true(sequential_2)) {
/* get the last one */
const char *rid = NULL;
for (xml_rsc = __xml_first_child_element(set2); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
rid = ID(xml_rsc);
}
}
EXPAND_CONSTRAINT_IDREF(id, rsc_2, rid);
}
if (rsc_1 != NULL && rsc_2 != NULL) {
rsc_colocation_new(id, NULL, score, rsc_1, rsc_2, role_1, role_2, data_set);
} else if (rsc_1 != NULL) {
for (xml_rsc = __xml_first_child_element(set2); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc));
rsc_colocation_new(id, NULL, score, rsc_1, rsc_2, role_1, role_2, data_set);
}
}
} else if (rsc_2 != NULL) {
for (xml_rsc = __xml_first_child_element(set1); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc));
rsc_colocation_new(id, NULL, score, rsc_1, rsc_2, role_1, role_2, data_set);
}
}
} else {
for (xml_rsc = __xml_first_child_element(set1); xml_rsc != NULL;
xml_rsc = __xml_next_element(xml_rsc)) {
if (crm_str_eq((const char *)xml_rsc->name, XML_TAG_RESOURCE_REF, TRUE)) {
xmlNode *xml_rsc_2 = NULL;
EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc));
for (xml_rsc_2 = __xml_first_child_element(set2);
xml_rsc_2 != NULL;
xml_rsc_2 = __xml_next_element(xml_rsc_2)) {
if (crm_str_eq((const char *)xml_rsc_2->name, XML_TAG_RESOURCE_REF, TRUE)) {
EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc_2));
rsc_colocation_new(id, NULL, score, rsc_1, rsc_2, role_1, role_2, data_set);
}
}
}
}
}
return TRUE;
}
static gboolean
unpack_simple_colocation(xmlNode * xml_obj, pe_working_set_t * data_set)
{
int score_i = 0;
const char *id = crm_element_value(xml_obj, XML_ATTR_ID);
const char *score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE);
const char *id_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE);
const char *id_rh = crm_element_value(xml_obj, XML_COLOC_ATTR_TARGET);
const char *state_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE_ROLE);
const char *state_rh = crm_element_value(xml_obj, XML_COLOC_ATTR_TARGET_ROLE);
const char *attr = crm_element_value(xml_obj, XML_COLOC_ATTR_NODE_ATTR);
const char *symmetrical = crm_element_value(xml_obj, XML_CONS_ATTR_SYMMETRICAL);
// experimental syntax from pacemaker-next (unlikely to be adopted as-is)
const char *instance_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE_INSTANCE);
const char *instance_rh = crm_element_value(xml_obj, XML_COLOC_ATTR_TARGET_INSTANCE);
pe_resource_t *rsc_lh = pe_find_constraint_resource(data_set->resources, id_lh);
pe_resource_t *rsc_rh = pe_find_constraint_resource(data_set->resources, id_rh);
if (rsc_lh == NULL) {
pcmk__config_err("Ignoring constraint '%s' because resource '%s' "
"does not exist", id, id_lh);
return FALSE;
} else if (rsc_rh == NULL) {
pcmk__config_err("Ignoring constraint '%s' because resource '%s' "
"does not exist", id, id_rh);
return FALSE;
} else if (instance_lh && pe_rsc_is_clone(rsc_lh) == FALSE) {
pcmk__config_err("Ignoring constraint '%s' because resource '%s' "
"is not a clone but instance '%s' was requested",
id, id_lh, instance_lh);
return FALSE;
} else if (instance_rh && pe_rsc_is_clone(rsc_rh) == FALSE) {
pcmk__config_err("Ignoring constraint '%s' because resource '%s' "
"is not a clone but instance '%s' was requested",
id, id_rh, instance_rh);
return FALSE;
}
if (instance_lh) {
rsc_lh = find_clone_instance(rsc_lh, instance_lh, data_set);
if (rsc_lh == NULL) {
pcmk__config_warn("Ignoring constraint '%s' because resource '%s' "
"does not have an instance '%s'",
id, id_lh, instance_lh);
return FALSE;
}
}
if (instance_rh) {
rsc_rh = find_clone_instance(rsc_rh, instance_rh, data_set);
if (rsc_rh == NULL) {
pcmk__config_warn("Ignoring constraint '%s' because resource '%s' "
"does not have an instance '%s'",
"'%s'", id, id_rh, instance_rh);
return FALSE;
}
}
if (crm_is_true(symmetrical)) {
pcmk__config_warn("The colocation constraint '"
XML_CONS_ATTR_SYMMETRICAL
"' attribute has been removed");
}
if (score) {
score_i = char2score(score);
}
rsc_colocation_new(id, attr, score_i, rsc_lh, rsc_rh, state_lh, state_rh, data_set);
return TRUE;
}
static gboolean
unpack_colocation_tags(xmlNode * xml_obj, xmlNode ** expanded_xml, pe_working_set_t * data_set)
{
const char *id = NULL;
const char *id_lh = NULL;
const char *id_rh = NULL;
const char *state_lh = NULL;
const char *state_rh = NULL;
pe_resource_t *rsc_lh = NULL;
pe_resource_t *rsc_rh = NULL;
pe_tag_t *tag_lh = NULL;
pe_tag_t *tag_rh = NULL;
xmlNode *new_xml = NULL;
xmlNode *rsc_set_lh = NULL;
xmlNode *rsc_set_rh = NULL;
gboolean any_sets = FALSE;
*expanded_xml = NULL;
CRM_CHECK(xml_obj != NULL, return FALSE);
id = ID(xml_obj);
if (id == NULL) {
pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID,
crm_element_name(xml_obj));
return FALSE;
}
/* Attempt to expand any template/tag references in possible resource sets. */
expand_tags_in_sets(xml_obj, &new_xml, data_set);
if (new_xml) {
/* There are resource sets referencing templates/tags. Return with the expanded XML. */
crm_log_xml_trace(new_xml, "Expanded rsc_colocation...");
*expanded_xml = new_xml;
return TRUE;
}
id_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE);
id_rh = crm_element_value(xml_obj, XML_COLOC_ATTR_TARGET);
if (id_lh == NULL || id_rh == NULL) {
return TRUE;
}
if (valid_resource_or_tag(data_set, id_lh, &rsc_lh, &tag_lh) == FALSE) {
pcmk__config_err("Ignoring constraint '%s' because '%s' is not a "
"valid resource or tag", id, id_lh);
return FALSE;
}
if (valid_resource_or_tag(data_set, id_rh, &rsc_rh, &tag_rh) == FALSE) {
pcmk__config_err("Ignoring constraint '%s' because '%s' is not a "
"valid resource or tag", id, id_rh);
return FALSE;
}
if (rsc_lh && rsc_rh) {
/* Neither side references any template/tag. */
return TRUE;
}
if (tag_lh && tag_rh) {
/* A colocation constraint between two templates/tags makes no sense. */
pcmk__config_err("Ignoring constraint '%s' because two templates or "
"tags cannot be colocated", id);
return FALSE;
}
state_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE_ROLE);
state_rh = crm_element_value(xml_obj, XML_COLOC_ATTR_TARGET_ROLE);
new_xml = copy_xml(xml_obj);
/* Convert the template/tag reference in "rsc" into a resource_set under the colocation constraint. */
if (tag_to_set(new_xml, &rsc_set_lh, XML_COLOC_ATTR_SOURCE, TRUE, data_set) == FALSE) {
free_xml(new_xml);
return FALSE;
}
if (rsc_set_lh) {
if (state_lh) {
/* A "rsc-role" is specified.
Move it into the converted resource_set as a "role"" attribute. */
crm_xml_add(rsc_set_lh, "role", state_lh);
xml_remove_prop(new_xml, XML_COLOC_ATTR_SOURCE_ROLE);
}
any_sets = TRUE;
}
/* Convert the template/tag reference in "with-rsc" into a resource_set under the colocation constraint. */
if (tag_to_set(new_xml, &rsc_set_rh, XML_COLOC_ATTR_TARGET, TRUE, data_set) == FALSE) {
free_xml(new_xml);
return FALSE;
}
if (rsc_set_rh) {
if (state_rh) {
/* A "with-rsc-role" is specified.
Move it into the converted resource_set as a "role"" attribute. */
crm_xml_add(rsc_set_rh, "role", state_rh);
xml_remove_prop(new_xml, XML_COLOC_ATTR_TARGET_ROLE);
}
any_sets = TRUE;
}
if (any_sets) {
crm_log_xml_trace(new_xml, "Expanded rsc_colocation...");
*expanded_xml = new_xml;
} else {
free_xml(new_xml);
}
return TRUE;
}
gboolean
unpack_rsc_colocation(xmlNode * xml_obj, pe_working_set_t * data_set)
{
int score_i = 0;
xmlNode *set = NULL;
xmlNode *last = NULL;
gboolean any_sets = FALSE;
xmlNode *orig_xml = NULL;
xmlNode *expanded_xml = NULL;
const char *id = crm_element_value(xml_obj, XML_ATTR_ID);
const char *score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE);
gboolean rc = TRUE;
if (score) {
score_i = char2score(score);
}
rc = unpack_colocation_tags(xml_obj, &expanded_xml, data_set);
if (expanded_xml) {
orig_xml = xml_obj;
xml_obj = expanded_xml;
} else if (rc == FALSE) {
return FALSE;
}
for (set = __xml_first_child_element(xml_obj); set != NULL;
set = __xml_next_element(set)) {
if (crm_str_eq((const char *)set->name, XML_CONS_TAG_RSC_SET, TRUE)) {
any_sets = TRUE;
set = expand_idref(set, data_set->input);
if (unpack_colocation_set(set, score_i, data_set) == FALSE) {
return FALSE;
} else if (last && colocate_rsc_sets(id, last, set, score_i, data_set) == FALSE) {
return FALSE;
}
last = set;
}
}
if (expanded_xml) {
free_xml(expanded_xml);
xml_obj = orig_xml;
}
if (any_sets == FALSE) {
return unpack_simple_colocation(xml_obj, data_set);
}
return TRUE;
}
gboolean
rsc_ticket_new(const char *id, pe_resource_t * rsc_lh, pe_ticket_t * ticket,
const char *state_lh, const char *loss_policy, pe_working_set_t * data_set)
{
rsc_ticket_t *new_rsc_ticket = NULL;
if (rsc_lh == NULL) {
pcmk__config_err("Ignoring ticket '%s' because resource "
"does not exist", id);
return FALSE;
}
new_rsc_ticket = calloc(1, sizeof(rsc_ticket_t));
if (new_rsc_ticket == NULL) {
return FALSE;
}
if (state_lh == NULL || safe_str_eq(state_lh, RSC_ROLE_STARTED_S)) {
state_lh = RSC_ROLE_UNKNOWN_S;
}
new_rsc_ticket->id = id;
new_rsc_ticket->ticket = ticket;
new_rsc_ticket->rsc_lh = rsc_lh;
new_rsc_ticket->role_lh = text2role(state_lh);
if (safe_str_eq(loss_policy, "fence")) {
if (is_set(data_set->flags, pe_flag_stonith_enabled)) {
new_rsc_ticket->loss_policy = loss_ticket_fence;
} else {
pcmk__config_err("Resetting '" XML_TICKET_ATTR_LOSS_POLICY
"' for ticket '%s' to 'stop' "
"because fencing is not configured", ticket->id);
loss_policy = "stop";
}
}
if (new_rsc_ticket->loss_policy == loss_ticket_fence) {
crm_debug("On loss of ticket '%s': Fence the nodes running %s (%s)",
new_rsc_ticket->ticket->id, new_rsc_ticket->rsc_lh->id,
role2text(new_rsc_ticket->role_lh));
} else if (safe_str_eq(loss_policy, "freeze")) {
crm_debug("On loss of ticket '%s': Freeze %s (%s)",
new_rsc_ticket->ticket->id, new_rsc_ticket->rsc_lh->id,
role2text(new_rsc_ticket->role_lh));
new_rsc_ticket->loss_policy = loss_ticket_freeze;
} else if (safe_str_eq(loss_policy, "demote")) {
crm_debug("On loss of ticket '%s': Demote %s (%s)",
new_rsc_ticket->ticket->id, new_rsc_ticket->rsc_lh->id,
role2text(new_rsc_ticket->role_lh));
new_rsc_ticket->loss_policy = loss_ticket_demote;
} else if (safe_str_eq(loss_policy, "stop")) {
crm_debug("On loss of ticket '%s': Stop %s (%s)",
new_rsc_ticket->ticket->id, new_rsc_ticket->rsc_lh->id,
role2text(new_rsc_ticket->role_lh));
new_rsc_ticket->loss_policy = loss_ticket_stop;
} else {
if (new_rsc_ticket->role_lh == RSC_ROLE_MASTER) {
crm_debug("On loss of ticket '%s': Default to demote %s (%s)",
new_rsc_ticket->ticket->id, new_rsc_ticket->rsc_lh->id,
role2text(new_rsc_ticket->role_lh));
new_rsc_ticket->loss_policy = loss_ticket_demote;
} else {
crm_debug("On loss of ticket '%s': Default to stop %s (%s)",
new_rsc_ticket->ticket->id, new_rsc_ticket->rsc_lh->id,
role2text(new_rsc_ticket->role_lh));
new_rsc_ticket->loss_policy = loss_ticket_stop;
}
}
pe_rsc_trace(rsc_lh, "%s (%s) ==> %s", rsc_lh->id, role2text(new_rsc_ticket->role_lh),
ticket->id);
rsc_lh->rsc_tickets = g_list_append(rsc_lh->rsc_tickets, new_rsc_ticket);
data_set->ticket_constraints = g_list_append(data_set->ticket_constraints, new_rsc_ticket);
if (new_rsc_ticket->ticket->granted == FALSE || new_rsc_ticket->ticket->standby) {
rsc_ticket_constraint(rsc_lh, new_rsc_ticket, data_set);
}
return TRUE;
}
static gboolean
unpack_rsc_ticket_set(xmlNode * set, pe_ticket_t * ticket, const char *loss_policy,
pe_working_set_t * data_set)
{
xmlNode *xml_rsc = NULL;
pe_resource_t *resource = NULL;
const char *set_id = NULL;
const char *role = NULL;
CRM_CHECK(set != NULL, return FALSE);
CRM_CHECK(ticket != NULL, return FALSE);
set_id = ID(set);
if (set_id == NULL) {
pcmk__config_err("Ignoring <" XML_CONS_TAG_RSC_SET "> without "
XML_ATTR_ID);
return FALSE;
}
role = crm_element_value(set, "role");
for (xml_rsc = first_named_child(set, XML_TAG_RESOURCE_REF);
xml_rsc != NULL; xml_rsc = crm_next_same_xml(xml_rsc)) {
EXPAND_CONSTRAINT_IDREF(set_id, resource, ID(xml_rsc));
pe_rsc_trace(resource, "Resource '%s' depends on ticket '%s'",
resource->id, ticket->id);
rsc_ticket_new(set_id, resource, ticket, role, loss_policy, data_set);
}
return TRUE;
}
static gboolean
unpack_simple_rsc_ticket(xmlNode * xml_obj, pe_working_set_t * data_set)
{
const char *id = NULL;
const char *ticket_str = crm_element_value(xml_obj, XML_TICKET_ATTR_TICKET);
const char *loss_policy = crm_element_value(xml_obj, XML_TICKET_ATTR_LOSS_POLICY);
pe_ticket_t *ticket = NULL;
const char *id_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE);
const char *state_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE_ROLE);
// experimental syntax from pacemaker-next (unlikely to be adopted as-is)
const char *instance_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE_INSTANCE);
pe_resource_t *rsc_lh = NULL;
CRM_CHECK(xml_obj != NULL, return FALSE);
id = ID(xml_obj);
if (id == NULL) {
pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID,
crm_element_name(xml_obj));
return FALSE;
}
if (ticket_str == NULL) {
pcmk__config_err("Ignoring constraint '%s' without ticket specified",
id);
return FALSE;
} else {
ticket = g_hash_table_lookup(data_set->tickets, ticket_str);
}
if (ticket == NULL) {
pcmk__config_err("Ignoring constraint '%s' because ticket '%s' "
"does not exist", id, ticket_str);
return FALSE;
}
if (id_lh == NULL) {
pcmk__config_err("Ignoring constraint '%s' without resource", id);
return FALSE;
} else {
rsc_lh = pe_find_constraint_resource(data_set->resources, id_lh);
}
if (rsc_lh == NULL) {
pcmk__config_err("Ignoring constraint '%s' because resource '%s' "
"does not exist", id, id_lh);
return FALSE;
} else if (instance_lh && pe_rsc_is_clone(rsc_lh) == FALSE) {
pcmk__config_err("Ignoring constraint '%s' because resource '%s' "
"is not a clone but instance '%s' was requested",
id, id_lh, instance_lh);
return FALSE;
}
if (instance_lh) {
rsc_lh = find_clone_instance(rsc_lh, instance_lh, data_set);
if (rsc_lh == NULL) {
pcmk__config_warn("Ignoring constraint '%s' because resource '%s' "
"does not have an instance '%s'",
"'%s'", id, id_lh, instance_lh);
return FALSE;
}
}
rsc_ticket_new(id, rsc_lh, ticket, state_lh, loss_policy, data_set);
return TRUE;
}
static gboolean
unpack_rsc_ticket_tags(xmlNode * xml_obj, xmlNode ** expanded_xml, pe_working_set_t * data_set)
{
const char *id = NULL;
const char *id_lh = NULL;
const char *state_lh = NULL;
pe_resource_t *rsc_lh = NULL;
pe_tag_t *tag_lh = NULL;
xmlNode *new_xml = NULL;
xmlNode *rsc_set_lh = NULL;
gboolean any_sets = FALSE;
*expanded_xml = NULL;
CRM_CHECK(xml_obj != NULL, return FALSE);
id = ID(xml_obj);
if (id == NULL) {
pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID,
crm_element_name(xml_obj));
return FALSE;
}
/* Attempt to expand any template/tag references in possible resource sets. */
expand_tags_in_sets(xml_obj, &new_xml, data_set);
if (new_xml) {
/* There are resource sets referencing templates/tags. Return with the expanded XML. */
crm_log_xml_trace(new_xml, "Expanded rsc_ticket...");
*expanded_xml = new_xml;
return TRUE;
}
id_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE);
if (id_lh == NULL) {
return TRUE;
}
if (valid_resource_or_tag(data_set, id_lh, &rsc_lh, &tag_lh) == FALSE) {
pcmk__config_err("Ignoring constraint '%s' because '%s' is not a "
"valid resource or tag", id, id_lh);
return FALSE;
} else if (rsc_lh) {
/* No template/tag is referenced. */
return TRUE;
}
state_lh = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE_ROLE);
new_xml = copy_xml(xml_obj);
/* Convert the template/tag reference in "rsc" into a resource_set under the rsc_ticket constraint. */
if (tag_to_set(new_xml, &rsc_set_lh, XML_COLOC_ATTR_SOURCE, FALSE, data_set) == FALSE) {
free_xml(new_xml);
return FALSE;
}
if (rsc_set_lh) {
if (state_lh) {
/* A "rsc-role" is specified.
Move it into the converted resource_set as a "role"" attribute. */
crm_xml_add(rsc_set_lh, "role", state_lh);
xml_remove_prop(new_xml, XML_COLOC_ATTR_SOURCE_ROLE);
}
any_sets = TRUE;
}
if (any_sets) {
crm_log_xml_trace(new_xml, "Expanded rsc_ticket...");
*expanded_xml = new_xml;
} else {
free_xml(new_xml);
}
return TRUE;
}
gboolean
unpack_rsc_ticket(xmlNode * xml_obj, pe_working_set_t * data_set)
{
xmlNode *set = NULL;
gboolean any_sets = FALSE;
const char *id = NULL;
const char *ticket_str = crm_element_value(xml_obj, XML_TICKET_ATTR_TICKET);
const char *loss_policy = crm_element_value(xml_obj, XML_TICKET_ATTR_LOSS_POLICY);
pe_ticket_t *ticket = NULL;
xmlNode *orig_xml = NULL;
xmlNode *expanded_xml = NULL;
gboolean rc = TRUE;
CRM_CHECK(xml_obj != NULL, return FALSE);
id = ID(xml_obj);
if (id == NULL) {
pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID,
crm_element_name(xml_obj));
return FALSE;
}
if (data_set->tickets == NULL) {
data_set->tickets =
g_hash_table_new_full(crm_str_hash, g_str_equal, free, destroy_ticket);
}
if (ticket_str == NULL) {
pcmk__config_err("Ignoring constraint '%s' without ticket", id);
return FALSE;
} else {
ticket = g_hash_table_lookup(data_set->tickets, ticket_str);
}
if (ticket == NULL) {
ticket = ticket_new(ticket_str, data_set);
if (ticket == NULL) {
return FALSE;
}
}
rc = unpack_rsc_ticket_tags(xml_obj, &expanded_xml, data_set);
if (expanded_xml) {
orig_xml = xml_obj;
xml_obj = expanded_xml;
} else if (rc == FALSE) {
return FALSE;
}
for (set = __xml_first_child_element(xml_obj); set != NULL;
set = __xml_next_element(set)) {
if (crm_str_eq((const char *)set->name, XML_CONS_TAG_RSC_SET, TRUE)) {
any_sets = TRUE;
set = expand_idref(set, data_set->input);
if (unpack_rsc_ticket_set(set, ticket, loss_policy, data_set) == FALSE) {
return FALSE;
}
}
}
if (expanded_xml) {
free_xml(expanded_xml);
xml_obj = orig_xml;
}
if (any_sets == FALSE) {
return unpack_simple_rsc_ticket(xml_obj, data_set);
}
return TRUE;
}
diff --git a/lib/pacemaker/pcmk_sched_messages.c b/lib/pacemaker/pcmk_sched_messages.c
index 0c24cdccdc..1f69e93525 100644
--- a/lib/pacemaker/pcmk_sched_messages.c
+++ b/lib/pacemaker/pcmk_sched_messages.c
@@ -1,140 +1,138 @@
/*
* Copyright 2004-2020 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 <sys/param.h>
#include <crm/crm.h>
#include <crm/cib.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <glib.h>
#include <crm/pengine/status.h>
#include <pacemaker-internal.h>
#include <crm/common/ipcs_internal.h>
gboolean show_scores = FALSE;
gboolean show_utilization = FALSE;
-int utilization_log_level = LOG_TRACE;
static void
log_resource_details(pe_working_set_t *data_set)
{
int rc = pcmk_rc_ok;
pcmk__output_t *out = NULL;
const char* argv[] = { "", NULL };
pcmk__supported_format_t formats[] = {
PCMK__SUPPORTED_FORMAT_LOG,
{ NULL, NULL, NULL }
};
pcmk__register_formats(NULL, formats);
rc = pcmk__output_new(&out, "log", NULL, (char**)argv);
if ((rc != pcmk_rc_ok) || (out == NULL)) {
crm_err("Can't log resource details due to internal error: %s\n",
pcmk_rc_str(rc));
return;
}
pe__register_messages(out);
for (GList *item = data_set->resources; item != NULL; item = item->next) {
pe_resource_t *rsc = (pe_resource_t *) item->data;
// Log all resources except inactive orphans
if (is_not_set(rsc->flags, pe_rsc_orphan)
|| (rsc->role != RSC_ROLE_STOPPED)) {
- out->message(out, crm_map_element_name(rsc->xml), pe_print_log,
- rsc);
+ out->message(out, crm_map_element_name(rsc->xml), 0, rsc);
}
}
pcmk__output_free(out);
}
/*!
* \internal
* \brief Run the scheduler for a given CIB
*
* \param[in,out] data_set Cluster working set
* \param[in] xml_input CIB XML to use as scheduler input
* \param[in] now Time to use for rule evaluation (or NULL for now)
*/
xmlNode *
pcmk__schedule_actions(pe_working_set_t *data_set, xmlNode *xml_input,
crm_time_t *now)
{
GListPtr gIter = NULL;
/* pe_debug_on(); */
CRM_ASSERT(xml_input || is_set(data_set->flags, pe_flag_have_status));
if (is_set(data_set->flags, pe_flag_have_status) == FALSE) {
set_working_set_defaults(data_set);
data_set->input = xml_input;
data_set->now = now;
} else {
crm_trace("Already have status - reusing");
}
if (data_set->now == NULL) {
data_set->now = crm_time_new(NULL);
}
crm_trace("Calculate cluster status");
stage0(data_set);
if (is_not_set(data_set->flags, pe_flag_quick_location)) {
log_resource_details(data_set);
}
crm_trace("Applying placement constraints");
stage2(data_set);
if(is_set(data_set->flags, pe_flag_quick_location)){
return NULL;
}
crm_trace("Create internal constraints");
stage3(data_set);
crm_trace("Check actions");
stage4(data_set);
crm_trace("Allocate resources");
stage5(data_set);
crm_trace("Processing fencing and shutdown cases");
stage6(data_set);
crm_trace("Applying ordering constraints");
stage7(data_set);
crm_trace("Create transition graph");
stage8(data_set);
crm_trace("=#=#=#=#= Summary =#=#=#=#=");
crm_trace("\t========= Set %d (Un-runnable) =========", -1);
if (get_crm_log_level() == LOG_TRACE) {
gIter = data_set->actions;
for (; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
if (is_set(action->flags, pe_action_optional) == FALSE
&& is_set(action->flags, pe_action_runnable) == FALSE
&& is_set(action->flags, pe_action_pseudo) == FALSE) {
log_action(LOG_TRACE, "\t", action, TRUE);
}
}
}
return data_set->graph;
}
diff --git a/lib/pacemaker/pcmk_sched_utils.c b/lib/pacemaker/pcmk_sched_utils.c
index 66032a4709..b7713c22d0 100644
--- a/lib/pacemaker/pcmk_sched_utils.c
+++ b/lib/pacemaker/pcmk_sched_utils.c
@@ -1,697 +1,697 @@
/*
* Copyright 2004-2020 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 <crm/msg_xml.h>
#include <crm/lrmd.h> // lrmd_event_data_t
#include <pacemaker-internal.h>
pe__location_t *
rsc2node_new(const char *id, pe_resource_t *rsc,
int node_weight, const char *discover_mode,
pe_node_t *foo_node, pe_working_set_t *data_set)
{
pe__location_t *new_con = NULL;
if (rsc == NULL || id == NULL) {
pe_err("Invalid constraint %s for rsc=%p", crm_str(id), rsc);
return NULL;
} else if (foo_node == NULL) {
CRM_CHECK(node_weight == 0, return NULL);
}
new_con = calloc(1, sizeof(pe__location_t));
if (new_con != NULL) {
new_con->id = strdup(id);
new_con->rsc_lh = rsc;
new_con->node_list_rh = NULL;
new_con->role_filter = RSC_ROLE_UNKNOWN;
if (discover_mode == NULL || safe_str_eq(discover_mode, "always")) {
new_con->discover_mode = pe_discover_always;
} else if (safe_str_eq(discover_mode, "never")) {
new_con->discover_mode = pe_discover_never;
} else if (safe_str_eq(discover_mode, "exclusive")) {
new_con->discover_mode = pe_discover_exclusive;
rsc->exclusive_discover = TRUE;
} else {
pe_err("Invalid %s value %s in location constraint", XML_LOCATION_ATTR_DISCOVERY, discover_mode);
}
if (foo_node != NULL) {
pe_node_t *copy = node_copy(foo_node);
copy->weight = node_weight;
new_con->node_list_rh = g_list_prepend(NULL, copy);
}
data_set->placement_constraints = g_list_prepend(data_set->placement_constraints, new_con);
rsc->rsc_location = g_list_prepend(rsc->rsc_location, new_con);
}
return new_con;
}
gboolean
can_run_resources(const pe_node_t * node)
{
if (node == NULL) {
return FALSE;
}
#if 0
if (node->weight < 0) {
return FALSE;
}
#endif
if (node->details->online == FALSE
|| node->details->shutdown || node->details->unclean
|| node->details->standby || node->details->maintenance) {
crm_trace("%s: online=%d, unclean=%d, standby=%d, maintenance=%d",
node->details->uname, node->details->online,
node->details->unclean, node->details->standby, node->details->maintenance);
return FALSE;
}
return TRUE;
}
struct node_weight_s {
pe_node_t *active;
pe_working_set_t *data_set;
};
/* return -1 if 'a' is more preferred
* return 1 if 'b' is more preferred
*/
static gint
sort_node_weight(gconstpointer a, gconstpointer b, gpointer data)
{
const pe_node_t *node1 = (const pe_node_t *)a;
const pe_node_t *node2 = (const pe_node_t *)b;
struct node_weight_s *nw = data;
int node1_weight = 0;
int node2_weight = 0;
int result = 0;
if (a == NULL) {
return 1;
}
if (b == NULL) {
return -1;
}
node1_weight = node1->weight;
node2_weight = node2->weight;
if (can_run_resources(node1) == FALSE) {
node1_weight = -INFINITY;
}
if (can_run_resources(node2) == FALSE) {
node2_weight = -INFINITY;
}
if (node1_weight > node2_weight) {
crm_trace("%s (%d) > %s (%d) : weight",
node1->details->uname, node1_weight, node2->details->uname, node2_weight);
return -1;
}
if (node1_weight < node2_weight) {
crm_trace("%s (%d) < %s (%d) : weight",
node1->details->uname, node1_weight, node2->details->uname, node2_weight);
return 1;
}
crm_trace("%s (%d) == %s (%d) : weight",
node1->details->uname, node1_weight, node2->details->uname, node2_weight);
if (safe_str_eq(nw->data_set->placement_strategy, "minimal")) {
goto equal;
}
if (safe_str_eq(nw->data_set->placement_strategy, "balanced")) {
result = compare_capacity(node1, node2);
if (result < 0) {
crm_trace("%s > %s : capacity (%d)",
node1->details->uname, node2->details->uname, result);
return -1;
} else if (result > 0) {
crm_trace("%s < %s : capacity (%d)",
node1->details->uname, node2->details->uname, result);
return 1;
}
}
/* now try to balance resources across the cluster */
if (node1->details->num_resources < node2->details->num_resources) {
crm_trace("%s (%d) > %s (%d) : resources",
node1->details->uname, node1->details->num_resources,
node2->details->uname, node2->details->num_resources);
return -1;
} else if (node1->details->num_resources > node2->details->num_resources) {
crm_trace("%s (%d) < %s (%d) : resources",
node1->details->uname, node1->details->num_resources,
node2->details->uname, node2->details->num_resources);
return 1;
}
if (nw->active && nw->active->details == node1->details) {
crm_trace("%s (%d) > %s (%d) : active",
node1->details->uname, node1->details->num_resources,
node2->details->uname, node2->details->num_resources);
return -1;
} else if (nw->active && nw->active->details == node2->details) {
crm_trace("%s (%d) < %s (%d) : active",
node1->details->uname, node1->details->num_resources,
node2->details->uname, node2->details->num_resources);
return 1;
}
equal:
crm_trace("%s = %s", node1->details->uname, node2->details->uname);
return strcmp(node1->details->uname, node2->details->uname);
}
GList *
sort_nodes_by_weight(GList *nodes, pe_node_t *active_node,
pe_working_set_t *data_set)
{
struct node_weight_s nw = { active_node, data_set };
return g_list_sort_with_data(nodes, sort_node_weight, &nw);
}
void
native_deallocate(pe_resource_t * rsc)
{
if (rsc->allocated_to) {
pe_node_t *old = rsc->allocated_to;
crm_info("Deallocating %s from %s", rsc->id, old->details->uname);
set_bit(rsc->flags, pe_rsc_provisional);
rsc->allocated_to = NULL;
old->details->allocated_rsc = g_list_remove(old->details->allocated_rsc, rsc);
old->details->num_resources--;
/* old->count--; */
calculate_utilization(old->details->utilization, rsc->utilization, TRUE);
free(old);
}
}
gboolean
native_assign_node(pe_resource_t * rsc, GListPtr nodes, pe_node_t * chosen, gboolean force)
{
CRM_ASSERT(rsc->variant == pe_native);
if (force == FALSE && chosen != NULL) {
bool unset = FALSE;
if(chosen->weight < 0) {
unset = TRUE;
// Allow the graph to assume that the remote resource will come up
} else if (!can_run_resources(chosen) && !pe__is_guest_node(chosen)) {
unset = TRUE;
}
if(unset) {
crm_debug("All nodes for resource %s are unavailable"
", unclean or shutting down (%s: %d, %d)",
rsc->id, chosen->details->uname, can_run_resources(chosen), chosen->weight);
rsc->next_role = RSC_ROLE_STOPPED;
chosen = NULL;
}
}
/* todo: update the old node for each resource to reflect its
* new resource count
*/
native_deallocate(rsc);
clear_bit(rsc->flags, pe_rsc_provisional);
if (chosen == NULL) {
GListPtr gIter = NULL;
char *rc_inactive = crm_itoa(PCMK_OCF_NOT_RUNNING);
crm_debug("Could not allocate a node for %s", rsc->id);
rsc->next_role = RSC_ROLE_STOPPED;
for (gIter = rsc->actions; gIter != NULL; gIter = gIter->next) {
pe_action_t *op = (pe_action_t *) gIter->data;
const char *interval_ms_s = g_hash_table_lookup(op->meta, XML_LRM_ATTR_INTERVAL_MS);
crm_debug("Processing %s", op->uuid);
if(safe_str_eq(RSC_STOP, op->task)) {
update_action_flags(op, pe_action_optional | pe_action_clear, __FUNCTION__, __LINE__);
} else if(safe_str_eq(RSC_START, op->task)) {
update_action_flags(op, pe_action_runnable | pe_action_clear, __FUNCTION__, __LINE__);
/* set_bit(rsc->flags, pe_rsc_block); */
} else if (interval_ms_s && safe_str_neq(interval_ms_s, "0")) {
if(safe_str_eq(rc_inactive, g_hash_table_lookup(op->meta, XML_ATTR_TE_TARGET_RC))) {
/* This is a recurring monitor for the stopped state, leave it alone */
} else {
/* Normal monitor operation, cancel it */
update_action_flags(op, pe_action_runnable | pe_action_clear, __FUNCTION__, __LINE__);
}
}
}
free(rc_inactive);
return FALSE;
}
crm_debug("Assigning %s to %s", chosen->details->uname, rsc->id);
rsc->allocated_to = node_copy(chosen);
chosen->details->allocated_rsc = g_list_prepend(chosen->details->allocated_rsc, rsc);
chosen->details->num_resources++;
chosen->count++;
calculate_utilization(chosen->details->utilization, rsc->utilization, FALSE);
- dump_rsc_utilization((show_utilization? LOG_STDOUT : utilization_log_level),
+ dump_rsc_utilization((show_utilization? LOG_STDOUT : LOG_TRACE),
__FUNCTION__, rsc, chosen);
return TRUE;
}
void
log_action(unsigned int log_level, const char *pre_text, pe_action_t * action, gboolean details)
{
const char *node_uname = NULL;
const char *node_uuid = NULL;
if (action == NULL) {
crm_trace("%s%s: <NULL>", pre_text == NULL ? "" : pre_text, pre_text == NULL ? "" : ": ");
return;
}
if (is_set(action->flags, pe_action_pseudo)) {
node_uname = NULL;
node_uuid = NULL;
} else if (action->node != NULL) {
node_uname = action->node->details->uname;
node_uuid = action->node->details->id;
} else {
node_uname = "<none>";
node_uuid = NULL;
}
switch (text2task(action->task)) {
case stonith_node:
case shutdown_crm:
crm_trace("%s%s%sAction %d: %s%s%s%s%s%s",
pre_text == NULL ? "" : pre_text,
pre_text == NULL ? "" : ": ",
is_set(action->flags,
pe_action_pseudo) ? "Pseudo " : is_set(action->flags,
pe_action_optional) ?
"Optional " : is_set(action->flags,
pe_action_runnable) ? is_set(action->flags,
pe_action_processed)
? "" : "(Provisional) " : "!!Non-Startable!! ", action->id,
action->uuid, node_uname ? "\ton " : "",
node_uname ? node_uname : "", node_uuid ? "\t\t(" : "",
node_uuid ? node_uuid : "", node_uuid ? ")" : "");
break;
default:
crm_trace("%s%s%sAction %d: %s %s%s%s%s%s%s",
pre_text == NULL ? "" : pre_text,
pre_text == NULL ? "" : ": ",
is_set(action->flags,
pe_action_optional) ? "Optional " : is_set(action->flags,
pe_action_pseudo)
? "Pseudo " : is_set(action->flags,
pe_action_runnable) ? is_set(action->flags,
pe_action_processed)
? "" : "(Provisional) " : "!!Non-Startable!! ", action->id,
action->uuid, action->rsc ? action->rsc->id : "<none>",
node_uname ? "\ton " : "", node_uname ? node_uname : "",
node_uuid ? "\t\t(" : "", node_uuid ? node_uuid : "", node_uuid ? ")" : "");
break;
}
if (details) {
GListPtr gIter = NULL;
crm_trace("\t\t====== Preceding Actions");
gIter = action->actions_before;
for (; gIter != NULL; gIter = gIter->next) {
pe_action_wrapper_t *other = (pe_action_wrapper_t *) gIter->data;
log_action(log_level + 1, "\t\t", other->action, FALSE);
}
crm_trace("\t\t====== Subsequent Actions");
gIter = action->actions_after;
for (; gIter != NULL; gIter = gIter->next) {
pe_action_wrapper_t *other = (pe_action_wrapper_t *) gIter->data;
log_action(log_level + 1, "\t\t", other->action, FALSE);
}
crm_trace("\t\t====== End");
} else {
crm_trace("\t\t(before=%d, after=%d)",
g_list_length(action->actions_before), g_list_length(action->actions_after));
}
}
gboolean
can_run_any(GHashTable * nodes)
{
GHashTableIter iter;
pe_node_t *node = NULL;
if (nodes == NULL) {
return FALSE;
}
g_hash_table_iter_init(&iter, nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) {
if (can_run_resources(node) && node->weight >= 0) {
return TRUE;
}
}
return FALSE;
}
pe_action_t *
create_pseudo_resource_op(pe_resource_t * rsc, const char *task, bool optional, bool runnable, pe_working_set_t *data_set)
{
pe_action_t *action = custom_action(rsc, pcmk__op_key(rsc->id, task, 0),
task, NULL, optional, TRUE, data_set);
update_action_flags(action, pe_action_pseudo, __FUNCTION__, __LINE__);
update_action_flags(action, pe_action_runnable, __FUNCTION__, __LINE__);
if(runnable) {
update_action_flags(action, pe_action_runnable, __FUNCTION__, __LINE__);
}
return action;
}
/*!
* \internal
* \brief Create an executor cancel op
*
* \param[in] rsc Resource of action to cancel
* \param[in] task Name of action to cancel
* \param[in] interval_ms Interval of action to cancel
* \param[in] node Node of action to cancel
* \param[in] data_set Working set of cluster
*
* \return Created op
*/
pe_action_t *
pe_cancel_op(pe_resource_t *rsc, const char *task, guint interval_ms,
pe_node_t *node, pe_working_set_t *data_set)
{
pe_action_t *cancel_op;
char *interval_ms_s = crm_strdup_printf("%u", interval_ms);
// @TODO dangerous if possible to schedule another action with this key
char *key = pcmk__op_key(rsc->id, task, interval_ms);
cancel_op = custom_action(rsc, key, RSC_CANCEL, node, FALSE, TRUE,
data_set);
free(cancel_op->task);
cancel_op->task = strdup(RSC_CANCEL);
free(cancel_op->cancel_task);
cancel_op->cancel_task = strdup(task);
add_hash_param(cancel_op->meta, XML_LRM_ATTR_TASK, task);
add_hash_param(cancel_op->meta, XML_LRM_ATTR_INTERVAL_MS, interval_ms_s);
free(interval_ms_s);
return cancel_op;
}
/*!
* \internal
* \brief Create a shutdown op for a scheduler transition
*
* \param[in] node Node being shut down
* \param[in] data_set Working set of cluster
*
* \return Created op
*/
pe_action_t *
sched_shutdown_op(pe_node_t *node, pe_working_set_t *data_set)
{
char *shutdown_id = crm_strdup_printf("%s-%s", CRM_OP_SHUTDOWN,
node->details->uname);
pe_action_t *shutdown_op = custom_action(NULL, shutdown_id, CRM_OP_SHUTDOWN,
node, FALSE, TRUE, data_set);
crm_notice("Scheduling shutdown of node %s", node->details->uname);
shutdown_constraints(node, shutdown_op, data_set);
add_hash_param(shutdown_op->meta, XML_ATTR_TE_NOWAIT, XML_BOOLEAN_TRUE);
return shutdown_op;
}
static char *
generate_transition_magic(const char *transition_key, int op_status, int op_rc)
{
CRM_CHECK(transition_key != NULL, return NULL);
return crm_strdup_printf("%d:%d;%s", op_status, op_rc, transition_key);
}
static void
append_digest(lrmd_event_data_t *op, xmlNode *update, const char *version,
const char *magic, int level)
{
/* this will enable us to later determine that the
* resource's parameters have changed and we should force
* a restart
*/
char *digest = NULL;
xmlNode *args_xml = NULL;
if (op->params == NULL) {
return;
}
args_xml = create_xml_node(NULL, XML_TAG_PARAMS);
g_hash_table_foreach(op->params, hash2field, args_xml);
pcmk__filter_op_for_digest(args_xml);
digest = calculate_operation_digest(args_xml, version);
#if 0
if (level < get_crm_log_level()
&& op->interval_ms == 0 && crm_str_eq(op->op_type, CRMD_ACTION_START, TRUE)) {
char *digest_source = dump_xml_unformatted(args_xml);
do_crm_log(level, "Calculated digest %s for %s (%s). Source: %s\n",
digest, ID(update), magic, digest_source);
free(digest_source);
}
#endif
crm_xml_add(update, XML_LRM_ATTR_OP_DIGEST, digest);
free_xml(args_xml);
free(digest);
}
#define FAKE_TE_ID "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx"
/*!
* \internal
* \brief Create XML for resource operation history update
*
* \param[in,out] parent Parent XML node to add to
* \param[in,out] op Operation event data
* \param[in] caller_version DC feature set
* \param[in] target_rc Expected result of operation
* \param[in] node Name of node on which operation was performed
* \param[in] origin Arbitrary description of update source
* \param[in] level A log message will be logged at this level
*
* \return Newly created XML node for history update
*/
xmlNode *
pcmk__create_history_xml(xmlNode *parent, lrmd_event_data_t *op,
const char *caller_version, int target_rc,
const char *node, const char *origin, int level)
{
char *key = NULL;
char *magic = NULL;
char *op_id = NULL;
char *op_id_additional = NULL;
char *local_user_data = NULL;
const char *exit_reason = NULL;
xmlNode *xml_op = NULL;
const char *task = NULL;
CRM_CHECK(op != NULL, return NULL);
do_crm_log(level, "%s: Updating resource %s after %s op %s (interval=%u)",
origin, op->rsc_id, op->op_type, services_lrm_status_str(op->op_status),
op->interval_ms);
crm_trace("DC version: %s", caller_version);
task = op->op_type;
/* Record a successful reload as a start, and a failed reload as a monitor,
* to make life easier for the scheduler when determining the current state.
*/
if (crm_str_eq(task, "reload", TRUE)) {
if (op->op_status == PCMK_LRM_OP_DONE) {
task = CRMD_ACTION_START;
} else {
task = CRMD_ACTION_STATUS;
}
}
key = pcmk__op_key(op->rsc_id, task, op->interval_ms);
if (crm_str_eq(task, CRMD_ACTION_NOTIFY, TRUE)) {
const char *n_type = crm_meta_value(op->params, "notify_type");
const char *n_task = crm_meta_value(op->params, "notify_operation");
CRM_LOG_ASSERT(n_type != NULL);
CRM_LOG_ASSERT(n_task != NULL);
op_id = pcmk__notify_key(op->rsc_id, n_type, n_task);
if (op->op_status != PCMK_LRM_OP_PENDING) {
/* Ignore notify errors.
*
* @TODO It might be better to keep the correct result here, and
* ignore it in process_graph_event().
*/
op->op_status = PCMK_LRM_OP_DONE;
op->rc = 0;
}
} else if (did_rsc_op_fail(op, target_rc)) {
op_id = pcmk__op_key(op->rsc_id, "last_failure", 0);
if (op->interval_ms == 0) {
// Ensure 'last' gets updated, in case record-pending is true
op_id_additional = pcmk__op_key(op->rsc_id, "last", 0);
}
exit_reason = op->exit_reason;
} else if (op->interval_ms > 0) {
op_id = strdup(key);
} else {
op_id = pcmk__op_key(op->rsc_id, "last", 0);
}
again:
xml_op = find_entity(parent, XML_LRM_TAG_RSC_OP, op_id);
if (xml_op == NULL) {
xml_op = create_xml_node(parent, XML_LRM_TAG_RSC_OP);
}
if (op->user_data == NULL) {
crm_debug("Generating fake transition key for: " PCMK__OP_FMT
" %d from %s", op->rsc_id, op->op_type, op->interval_ms,
op->call_id, origin);
local_user_data = pcmk__transition_key(-1, op->call_id, target_rc,
FAKE_TE_ID);
op->user_data = local_user_data;
}
if(magic == NULL) {
magic = generate_transition_magic(op->user_data, op->op_status, op->rc);
}
crm_xml_add(xml_op, XML_ATTR_ID, op_id);
crm_xml_add(xml_op, XML_LRM_ATTR_TASK_KEY, key);
crm_xml_add(xml_op, XML_LRM_ATTR_TASK, task);
crm_xml_add(xml_op, XML_ATTR_ORIGIN, origin);
crm_xml_add(xml_op, XML_ATTR_CRM_VERSION, caller_version);
crm_xml_add(xml_op, XML_ATTR_TRANSITION_KEY, op->user_data);
crm_xml_add(xml_op, XML_ATTR_TRANSITION_MAGIC, magic);
crm_xml_add(xml_op, XML_LRM_ATTR_EXIT_REASON, exit_reason == NULL ? "" : exit_reason);
crm_xml_add(xml_op, XML_LRM_ATTR_TARGET, node); /* For context during triage */
crm_xml_add_int(xml_op, XML_LRM_ATTR_CALLID, op->call_id);
crm_xml_add_int(xml_op, XML_LRM_ATTR_RC, op->rc);
crm_xml_add_int(xml_op, XML_LRM_ATTR_OPSTATUS, op->op_status);
crm_xml_add_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, op->interval_ms);
if (compare_version("2.1", caller_version) <= 0) {
if (op->t_run || op->t_rcchange || op->exec_time || op->queue_time) {
crm_trace("Timing data (" PCMK__OP_FMT
"): last=%u change=%u exec=%u queue=%u",
op->rsc_id, op->op_type, op->interval_ms,
op->t_run, op->t_rcchange, op->exec_time, op->queue_time);
if (op->interval_ms == 0) {
crm_xml_add_ll(xml_op, XML_RSC_OP_LAST_CHANGE,
(long long) op->t_run);
// @COMPAT last-run is deprecated
crm_xml_add_ll(xml_op, XML_RSC_OP_LAST_RUN,
(long long) op->t_run);
} else if(op->t_rcchange) {
/* last-run is not accurate for recurring ops */
crm_xml_add_ll(xml_op, XML_RSC_OP_LAST_CHANGE,
(long long) op->t_rcchange);
} else {
/* ...but is better than nothing otherwise */
crm_xml_add_ll(xml_op, XML_RSC_OP_LAST_CHANGE,
(long long) op->t_run);
}
crm_xml_add_int(xml_op, XML_RSC_OP_T_EXEC, op->exec_time);
crm_xml_add_int(xml_op, XML_RSC_OP_T_QUEUE, op->queue_time);
}
}
if (crm_str_eq(op->op_type, CRMD_ACTION_MIGRATE, TRUE)
|| crm_str_eq(op->op_type, CRMD_ACTION_MIGRATED, TRUE)) {
/*
* Record migrate_source and migrate_target always for migrate ops.
*/
const char *name = XML_LRM_ATTR_MIGRATE_SOURCE;
crm_xml_add(xml_op, name, crm_meta_value(op->params, name));
name = XML_LRM_ATTR_MIGRATE_TARGET;
crm_xml_add(xml_op, name, crm_meta_value(op->params, name));
}
append_digest(op, xml_op, caller_version, magic, LOG_DEBUG);
if (op_id_additional) {
free(op_id);
op_id = op_id_additional;
op_id_additional = NULL;
goto again;
}
if (local_user_data) {
free(local_user_data);
op->user_data = NULL;
}
free(magic);
free(op_id);
free(key);
return xml_op;
}
diff --git a/lib/pengine/Makefile.am b/lib/pengine/Makefile.am
index 4aa81797aa..67e521836c 100644
--- a/lib/pengine/Makefile.am
+++ b/lib/pengine/Makefile.am
@@ -1,51 +1,51 @@
#
# Copyright 2004-2018 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 $(top_srcdir)/mk/common.mk
SUBDIRS = tests
## libraries
lib_LTLIBRARIES = libpe_rules.la libpe_status.la
## SOURCES
-noinst_HEADERS = unpack.h variant.h pe_status_private.h
+noinst_HEADERS = variant.h pe_status_private.h
libpe_rules_la_LDFLAGS = -version-info 27:0:1
libpe_rules_la_CFLAGS = $(CFLAGS_HARDENED_LIB)
libpe_rules_la_LDFLAGS += $(LDFLAGS_HARDENED_LIB)
libpe_rules_la_LIBADD = $(top_builddir)/lib/common/libcrmcommon.la
libpe_rules_la_SOURCES = rules.c rules_alerts.c common.c
libpe_status_la_LDFLAGS = -version-info 29:0:1
libpe_status_la_CFLAGS = $(CFLAGS_HARDENED_LIB)
libpe_status_la_LDFLAGS += $(LDFLAGS_HARDENED_LIB)
libpe_status_la_LIBADD = @CURSESLIBS@ $(top_builddir)/lib/common/libcrmcommon.la
# Use += rather than backlashed continuation lines for parsing by bumplibs.sh
libpe_status_la_SOURCES =
libpe_status_la_SOURCES += bundle.c
libpe_status_la_SOURCES += clone.c
libpe_status_la_SOURCES += common.c
libpe_status_la_SOURCES += complex.c
libpe_status_la_SOURCES += failcounts.c
libpe_status_la_SOURCES += group.c
libpe_status_la_SOURCES += native.c
libpe_status_la_SOURCES += remote.c
libpe_status_la_SOURCES += rules.c
libpe_status_la_SOURCES += status.c
libpe_status_la_SOURCES += tags.c
libpe_status_la_SOURCES += unpack.c
libpe_status_la_SOURCES += utils.c
libpe_status_la_SOURCES += pe_output.c
clean-generic:
rm -f *.log *.debug *~
diff --git a/lib/pengine/bundle.c b/lib/pengine/bundle.c
index 911d3a9ec9..d812f5514c 100644
--- a/lib/pengine/bundle.c
+++ b/lib/pengine/bundle.c
@@ -1,1940 +1,1938 @@
/*
* Copyright 2004-2020 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 <ctype.h>
#include <crm/pengine/rules.h>
#include <crm/pengine/status.h>
#include <crm/pengine/internal.h>
#include <crm/msg_xml.h>
#include <pe_status_private.h>
#define PE__VARIANT_BUNDLE 1
#include "./variant.h"
static char *
next_ip(const char *last_ip)
{
unsigned int oct1 = 0;
unsigned int oct2 = 0;
unsigned int oct3 = 0;
unsigned int oct4 = 0;
int rc = sscanf(last_ip, "%u.%u.%u.%u", &oct1, &oct2, &oct3, &oct4);
if (rc != 4) {
/*@ TODO check for IPv6 */
return NULL;
} else if (oct3 > 253) {
return NULL;
} else if (oct4 > 253) {
++oct3;
oct4 = 1;
} else {
++oct4;
}
return crm_strdup_printf("%u.%u.%u.%u", oct1, oct2, oct3, oct4);
}
static int
allocate_ip(pe__bundle_variant_data_t *data, pe__bundle_replica_t *replica,
char *buffer, int max)
{
if(data->ip_range_start == NULL) {
return 0;
} else if(data->ip_last) {
replica->ipaddr = next_ip(data->ip_last);
} else {
replica->ipaddr = strdup(data->ip_range_start);
}
data->ip_last = replica->ipaddr;
switch (data->agent_type) {
case PE__CONTAINER_AGENT_DOCKER:
case PE__CONTAINER_AGENT_PODMAN:
if (data->add_host) {
return snprintf(buffer, max, " --add-host=%s-%d:%s",
data->prefix, replica->offset,
replica->ipaddr);
}
case PE__CONTAINER_AGENT_RKT:
return snprintf(buffer, max, " --hosts-entry=%s=%s-%d",
replica->ipaddr, data->prefix, replica->offset);
default: // PE__CONTAINER_AGENT_UNKNOWN
break;
}
return 0;
}
static xmlNode *
create_resource(const char *name, const char *provider, const char *kind)
{
xmlNode *rsc = create_xml_node(NULL, XML_CIB_TAG_RESOURCE);
crm_xml_add(rsc, XML_ATTR_ID, name);
crm_xml_add(rsc, XML_AGENT_ATTR_CLASS, PCMK_RESOURCE_CLASS_OCF);
crm_xml_add(rsc, XML_AGENT_ATTR_PROVIDER, provider);
crm_xml_add(rsc, XML_ATTR_TYPE, kind);
return rsc;
}
/*!
* \internal
* \brief Check whether cluster can manage resource inside container
*
* \param[in] data Container variant data
*
* \return TRUE if networking configuration is acceptable, FALSE otherwise
*
* \note The resource is manageable if an IP range or control port has been
* specified. If a control port is used without an IP range, replicas per
* host must be 1.
*/
static bool
valid_network(pe__bundle_variant_data_t *data)
{
if(data->ip_range_start) {
return TRUE;
}
if(data->control_port) {
if(data->nreplicas_per_host > 1) {
pe_err("Specifying the 'control-port' for %s requires 'replicas-per-host=1'", data->prefix);
data->nreplicas_per_host = 1;
/* @TODO to be sure: clear_bit(rsc->flags, pe_rsc_unique); */
}
return TRUE;
}
return FALSE;
}
static bool
create_ip_resource(pe_resource_t *parent, pe__bundle_variant_data_t *data,
pe__bundle_replica_t *replica, pe_working_set_t *data_set)
{
if(data->ip_range_start) {
char *id = NULL;
xmlNode *xml_ip = NULL;
xmlNode *xml_obj = NULL;
id = crm_strdup_printf("%s-ip-%s", data->prefix, replica->ipaddr);
crm_xml_sanitize_id(id);
xml_ip = create_resource(id, "heartbeat", "IPaddr2");
free(id);
xml_obj = create_xml_node(xml_ip, XML_TAG_ATTR_SETS);
crm_xml_set_id(xml_obj, "%s-attributes-%d",
data->prefix, replica->offset);
crm_create_nvpair_xml(xml_obj, NULL, "ip", replica->ipaddr);
if(data->host_network) {
crm_create_nvpair_xml(xml_obj, NULL, "nic", data->host_network);
}
if(data->host_netmask) {
crm_create_nvpair_xml(xml_obj, NULL,
"cidr_netmask", data->host_netmask);
} else {
crm_create_nvpair_xml(xml_obj, NULL, "cidr_netmask", "32");
}
xml_obj = create_xml_node(xml_ip, "operations");
crm_create_op_xml(xml_obj, ID(xml_ip), "monitor", "60s", NULL);
// TODO: Other ops? Timeouts and intervals from underlying resource?
if (!common_unpack(xml_ip, &replica->ip, parent, data_set)) {
return FALSE;
}
parent->children = g_list_append(parent->children, replica->ip);
}
return TRUE;
}
static bool
create_docker_resource(pe_resource_t *parent, pe__bundle_variant_data_t *data,
pe__bundle_replica_t *replica,
pe_working_set_t *data_set)
{
int offset = 0, max = 4096;
char *buffer = calloc(1, max+1);
int doffset = 0, dmax = 1024;
char *dbuffer = calloc(1, dmax+1);
char *id = NULL;
xmlNode *xml_container = NULL;
xmlNode *xml_obj = NULL;
id = crm_strdup_printf("%s-docker-%d", data->prefix, replica->offset);
crm_xml_sanitize_id(id);
xml_container = create_resource(id, "heartbeat",
PE__CONTAINER_AGENT_DOCKER_S);
free(id);
xml_obj = create_xml_node(xml_container, XML_TAG_ATTR_SETS);
crm_xml_set_id(xml_obj, "%s-attributes-%d",
data->prefix, replica->offset);
crm_create_nvpair_xml(xml_obj, NULL, "image", data->image);
crm_create_nvpair_xml(xml_obj, NULL, "allow_pull", XML_BOOLEAN_TRUE);
crm_create_nvpair_xml(xml_obj, NULL, "force_kill", XML_BOOLEAN_FALSE);
crm_create_nvpair_xml(xml_obj, NULL, "reuse", XML_BOOLEAN_FALSE);
offset += snprintf(buffer+offset, max-offset, " --restart=no");
/* Set a container hostname only if we have an IP to map it to.
* The user can set -h or --uts=host themselves if they want a nicer
* name for logs, but this makes applications happy who need their
* hostname to match the IP they bind to.
*/
if (data->ip_range_start != NULL) {
offset += snprintf(buffer+offset, max-offset, " -h %s-%d",
data->prefix, replica->offset);
}
offset += snprintf(buffer+offset, max-offset, " -e PCMK_stderr=1");
if (data->container_network) {
#if 0
offset += snprintf(buffer+offset, max-offset, " --link-local-ip=%s",
replica->ipaddr);
#endif
offset += snprintf(buffer+offset, max-offset, " --net=%s",
data->container_network);
}
if(data->control_port) {
offset += snprintf(buffer+offset, max-offset, " -e PCMK_remote_port=%s", data->control_port);
} else {
offset += snprintf(buffer+offset, max-offset, " -e PCMK_remote_port=%d", DEFAULT_REMOTE_PORT);
}
for(GListPtr pIter = data->mounts; pIter != NULL; pIter = pIter->next) {
pe__bundle_mount_t *mount = pIter->data;
if (is_set(mount->flags, pe__bundle_mount_subdir)) {
char *source = crm_strdup_printf(
"%s/%s-%d", mount->source, data->prefix, replica->offset);
if(doffset > 0) {
doffset += snprintf(dbuffer+doffset, dmax-doffset, ",");
}
doffset += snprintf(dbuffer+doffset, dmax-doffset, "%s", source);
offset += snprintf(buffer+offset, max-offset, " -v %s:%s", source, mount->target);
free(source);
} else {
offset += snprintf(buffer+offset, max-offset, " -v %s:%s", mount->source, mount->target);
}
if(mount->options) {
offset += snprintf(buffer+offset, max-offset, ":%s", mount->options);
}
}
for(GListPtr pIter = data->ports; pIter != NULL; pIter = pIter->next) {
pe__bundle_port_t *port = pIter->data;
if (replica->ipaddr) {
offset += snprintf(buffer+offset, max-offset, " -p %s:%s:%s",
replica->ipaddr, port->source,
port->target);
} else if(safe_str_neq(data->container_network, "host")) {
// No need to do port mapping if net=host
offset += snprintf(buffer+offset, max-offset, " -p %s:%s", port->source, port->target);
}
}
if (data->launcher_options) {
offset += snprintf(buffer+offset, max-offset, " %s",
data->launcher_options);
}
if (data->container_host_options) {
offset += snprintf(buffer + offset, max - offset, " %s",
data->container_host_options);
}
crm_create_nvpair_xml(xml_obj, NULL, "run_opts", buffer);
free(buffer);
crm_create_nvpair_xml(xml_obj, NULL, "mount_points", dbuffer);
free(dbuffer);
if (replica->child) {
if (data->container_command) {
crm_create_nvpair_xml(xml_obj, NULL,
"run_cmd", data->container_command);
} else {
crm_create_nvpair_xml(xml_obj, NULL,
"run_cmd", SBIN_DIR "/pacemaker-remoted");
}
/* TODO: Allow users to specify their own?
*
* We just want to know if the container is alive, we'll
* monitor the child independently
*/
crm_create_nvpair_xml(xml_obj, NULL, "monitor_cmd", "/bin/true");
/* } else if(child && data->untrusted) {
* Support this use-case?
*
* The ability to have resources started/stopped by us, but
* unable to set attributes, etc.
*
* Arguably better to control API access this with ACLs like
* "normal" remote nodes
*
* crm_create_nvpair_xml(xml_obj, NULL,
* "run_cmd",
* "/usr/libexec/pacemaker/pacemaker-execd");
* crm_create_nvpair_xml(xml_obj, NULL, "monitor_cmd",
* "/usr/libexec/pacemaker/lrmd_internal_ctl -c poke");
*/
} else {
if (data->container_command) {
crm_create_nvpair_xml(xml_obj, NULL,
"run_cmd", data->container_command);
}
/* TODO: Allow users to specify their own?
*
* We don't know what's in the container, so we just want
* to know if it is alive
*/
crm_create_nvpair_xml(xml_obj, NULL, "monitor_cmd", "/bin/true");
}
xml_obj = create_xml_node(xml_container, "operations");
crm_create_op_xml(xml_obj, ID(xml_container), "monitor", "60s", NULL);
// TODO: Other ops? Timeouts and intervals from underlying resource?
if (!common_unpack(xml_container, &replica->container, parent, data_set)) {
return FALSE;
}
parent->children = g_list_append(parent->children, replica->container);
return TRUE;
}
static bool
create_podman_resource(pe_resource_t *parent, pe__bundle_variant_data_t *data,
pe__bundle_replica_t *replica,
pe_working_set_t *data_set)
{
int offset = 0, max = 4096;
char *buffer = calloc(1, max+1);
int doffset = 0, dmax = 1024;
char *dbuffer = calloc(1, dmax+1);
char *id = NULL;
xmlNode *xml_container = NULL;
xmlNode *xml_obj = NULL;
id = crm_strdup_printf("%s-podman-%d", data->prefix, replica->offset);
crm_xml_sanitize_id(id);
xml_container = create_resource(id, "heartbeat",
PE__CONTAINER_AGENT_PODMAN_S);
free(id);
xml_obj = create_xml_node(xml_container, XML_TAG_ATTR_SETS);
crm_xml_set_id(xml_obj, "%s-attributes-%d",
data->prefix, replica->offset);
crm_create_nvpair_xml(xml_obj, NULL, "image", data->image);
crm_create_nvpair_xml(xml_obj, NULL, "allow_pull", XML_BOOLEAN_TRUE);
crm_create_nvpair_xml(xml_obj, NULL, "force_kill", XML_BOOLEAN_FALSE);
crm_create_nvpair_xml(xml_obj, NULL, "reuse", XML_BOOLEAN_FALSE);
// FIXME: (bandini 2018-08) podman has no restart policies
//offset += snprintf(buffer+offset, max-offset, " --restart=no");
/* Set a container hostname only if we have an IP to map it to.
* The user can set -h or --uts=host themselves if they want a nicer
* name for logs, but this makes applications happy who need their
* hostname to match the IP they bind to.
*/
if (data->ip_range_start != NULL) {
offset += snprintf(buffer+offset, max-offset, " -h %s-%d",
data->prefix, replica->offset);
}
offset += snprintf(buffer+offset, max-offset, " -e PCMK_stderr=1");
if (data->container_network) {
#if 0
// podman has no support for --link-local-ip
offset += snprintf(buffer+offset, max-offset, " --link-local-ip=%s",
replica->ipaddr);
#endif
offset += snprintf(buffer+offset, max-offset, " --net=%s",
data->container_network);
}
if(data->control_port) {
offset += snprintf(buffer+offset, max-offset, " -e PCMK_remote_port=%s", data->control_port);
} else {
offset += snprintf(buffer+offset, max-offset, " -e PCMK_remote_port=%d", DEFAULT_REMOTE_PORT);
}
for(GListPtr pIter = data->mounts; pIter != NULL; pIter = pIter->next) {
pe__bundle_mount_t *mount = pIter->data;
if (is_set(mount->flags, pe__bundle_mount_subdir)) {
char *source = crm_strdup_printf(
"%s/%s-%d", mount->source, data->prefix, replica->offset);
if(doffset > 0) {
doffset += snprintf(dbuffer+doffset, dmax-doffset, ",");
}
doffset += snprintf(dbuffer+doffset, dmax-doffset, "%s", source);
offset += snprintf(buffer+offset, max-offset, " -v %s:%s", source, mount->target);
free(source);
} else {
offset += snprintf(buffer+offset, max-offset, " -v %s:%s", mount->source, mount->target);
}
if(mount->options) {
offset += snprintf(buffer+offset, max-offset, ":%s", mount->options);
}
}
for(GListPtr pIter = data->ports; pIter != NULL; pIter = pIter->next) {
pe__bundle_port_t *port = pIter->data;
if (replica->ipaddr) {
offset += snprintf(buffer+offset, max-offset, " -p %s:%s:%s",
replica->ipaddr, port->source,
port->target);
} else if(safe_str_neq(data->container_network, "host")) {
// No need to do port mapping if net=host
offset += snprintf(buffer+offset, max-offset, " -p %s:%s", port->source, port->target);
}
}
if (data->launcher_options) {
offset += snprintf(buffer+offset, max-offset, " %s",
data->launcher_options);
}
if (data->container_host_options) {
offset += snprintf(buffer + offset, max - offset, " %s",
data->container_host_options);
}
crm_create_nvpair_xml(xml_obj, NULL, "run_opts", buffer);
free(buffer);
crm_create_nvpair_xml(xml_obj, NULL, "mount_points", dbuffer);
free(dbuffer);
if (replica->child) {
if (data->container_command) {
crm_create_nvpair_xml(xml_obj, NULL,
"run_cmd", data->container_command);
} else {
crm_create_nvpair_xml(xml_obj, NULL,
"run_cmd", SBIN_DIR "/pacemaker-remoted");
}
/* TODO: Allow users to specify their own?
*
* We just want to know if the container is alive, we'll
* monitor the child independently
*/
crm_create_nvpair_xml(xml_obj, NULL, "monitor_cmd", "/bin/true");
/* } else if(child && data->untrusted) {
* Support this use-case?
*
* The ability to have resources started/stopped by us, but
* unable to set attributes, etc.
*
* Arguably better to control API access this with ACLs like
* "normal" remote nodes
*
* crm_create_nvpair_xml(xml_obj, NULL,
* "run_cmd",
* "/usr/libexec/pacemaker/pacemaker-execd");
* crm_create_nvpair_xml(xml_obj, NULL, "monitor_cmd",
* "/usr/libexec/pacemaker/lrmd_internal_ctl -c poke");
*/
} else {
if (data->container_command) {
crm_create_nvpair_xml(xml_obj, NULL,
"run_cmd", data->container_command);
}
/* TODO: Allow users to specify their own?
*
* We don't know what's in the container, so we just want
* to know if it is alive
*/
crm_create_nvpair_xml(xml_obj, NULL, "monitor_cmd", "/bin/true");
}
xml_obj = create_xml_node(xml_container, "operations");
crm_create_op_xml(xml_obj, ID(xml_container), "monitor", "60s", NULL);
// TODO: Other ops? Timeouts and intervals from underlying resource?
if (!common_unpack(xml_container, &replica->container, parent,
data_set)) {
return FALSE;
}
parent->children = g_list_append(parent->children, replica->container);
return TRUE;
}
static bool
create_rkt_resource(pe_resource_t *parent, pe__bundle_variant_data_t *data,
pe__bundle_replica_t *replica, pe_working_set_t *data_set)
{
int offset = 0, max = 4096;
char *buffer = calloc(1, max+1);
int doffset = 0, dmax = 1024;
char *dbuffer = calloc(1, dmax+1);
char *id = NULL;
xmlNode *xml_container = NULL;
xmlNode *xml_obj = NULL;
int volid = 0;
id = crm_strdup_printf("%s-rkt-%d", data->prefix, replica->offset);
crm_xml_sanitize_id(id);
xml_container = create_resource(id, "heartbeat",
PE__CONTAINER_AGENT_RKT_S);
free(id);
xml_obj = create_xml_node(xml_container, XML_TAG_ATTR_SETS);
crm_xml_set_id(xml_obj, "%s-attributes-%d",
data->prefix, replica->offset);
crm_create_nvpair_xml(xml_obj, NULL, "image", data->image);
crm_create_nvpair_xml(xml_obj, NULL, "allow_pull", "true");
crm_create_nvpair_xml(xml_obj, NULL, "force_kill", "false");
crm_create_nvpair_xml(xml_obj, NULL, "reuse", "false");
/* Set a container hostname only if we have an IP to map it to.
* The user can set -h or --uts=host themselves if they want a nicer
* name for logs, but this makes applications happy who need their
* hostname to match the IP they bind to.
*/
if (data->ip_range_start != NULL) {
offset += snprintf(buffer+offset, max-offset, " --hostname=%s-%d",
data->prefix, replica->offset);
}
offset += snprintf(buffer+offset, max-offset, " --environment=PCMK_stderr=1");
if (data->container_network) {
#if 0
offset += snprintf(buffer+offset, max-offset, " --link-local-ip=%s",
replica->ipaddr);
#endif
offset += snprintf(buffer+offset, max-offset, " --net=%s",
data->container_network);
}
if(data->control_port) {
offset += snprintf(buffer+offset, max-offset, " --environment=PCMK_remote_port=%s", data->control_port);
} else {
offset += snprintf(buffer+offset, max-offset, " --environment=PCMK_remote_port=%d", DEFAULT_REMOTE_PORT);
}
for(GListPtr pIter = data->mounts; pIter != NULL; pIter = pIter->next) {
pe__bundle_mount_t *mount = pIter->data;
if (is_set(mount->flags, pe__bundle_mount_subdir)) {
char *source = crm_strdup_printf(
"%s/%s-%d", mount->source, data->prefix, replica->offset);
if(doffset > 0) {
doffset += snprintf(dbuffer+doffset, dmax-doffset, ",");
}
doffset += snprintf(dbuffer+doffset, dmax-doffset, "%s", source);
offset += snprintf(buffer+offset, max-offset, " --volume vol%d,kind=host,source=%s", volid, source);
if(mount->options) {
offset += snprintf(buffer+offset, max-offset, ",%s", mount->options);
}
offset += snprintf(buffer+offset, max-offset, " --mount volume=vol%d,target=%s", volid, mount->target);
free(source);
} else {
offset += snprintf(buffer+offset, max-offset, " --volume vol%d,kind=host,source=%s", volid, mount->source);
if(mount->options) {
offset += snprintf(buffer+offset, max-offset, ",%s", mount->options);
}
offset += snprintf(buffer+offset, max-offset, " --mount volume=vol%d,target=%s", volid, mount->target);
}
volid++;
}
for(GListPtr pIter = data->ports; pIter != NULL; pIter = pIter->next) {
pe__bundle_port_t *port = pIter->data;
if (replica->ipaddr) {
offset += snprintf(buffer+offset, max-offset,
" --port=%s:%s:%s", port->target,
replica->ipaddr, port->source);
} else {
offset += snprintf(buffer+offset, max-offset, " --port=%s:%s", port->target, port->source);
}
}
if (data->launcher_options) {
offset += snprintf(buffer+offset, max-offset, " %s",
data->launcher_options);
}
if (data->container_host_options) {
offset += snprintf(buffer + offset, max - offset, " %s",
data->container_host_options);
}
crm_create_nvpair_xml(xml_obj, NULL, "run_opts", buffer);
free(buffer);
crm_create_nvpair_xml(xml_obj, NULL, "mount_points", dbuffer);
free(dbuffer);
if (replica->child) {
if (data->container_command) {
crm_create_nvpair_xml(xml_obj, NULL, "run_cmd",
data->container_command);
} else {
crm_create_nvpair_xml(xml_obj, NULL, "run_cmd",
SBIN_DIR "/pacemaker-remoted");
}
/* TODO: Allow users to specify their own?
*
* We just want to know if the container is alive, we'll
* monitor the child independently
*/
crm_create_nvpair_xml(xml_obj, NULL, "monitor_cmd", "/bin/true");
/* } else if(child && data->untrusted) {
* Support this use-case?
*
* The ability to have resources started/stopped by us, but
* unable to set attributes, etc.
*
* Arguably better to control API access this with ACLs like
* "normal" remote nodes
*
* crm_create_nvpair_xml(xml_obj, NULL,
* "run_cmd",
* "/usr/libexec/pacemaker/pacemaker-execd");
* crm_create_nvpair_xml(xml_obj, NULL, "monitor_cmd",
* "/usr/libexec/pacemaker/lrmd_internal_ctl -c poke");
*/
} else {
if (data->container_command) {
crm_create_nvpair_xml(xml_obj, NULL, "run_cmd",
data->container_command);
}
/* TODO: Allow users to specify their own?
*
* We don't know what's in the container, so we just want
* to know if it is alive
*/
crm_create_nvpair_xml(xml_obj, NULL, "monitor_cmd", "/bin/true");
}
xml_obj = create_xml_node(xml_container, "operations");
crm_create_op_xml(xml_obj, ID(xml_container), "monitor", "60s", NULL);
// TODO: Other ops? Timeouts and intervals from underlying resource?
if (!common_unpack(xml_container, &replica->container, parent, data_set)) {
return FALSE;
}
parent->children = g_list_append(parent->children, replica->container);
return TRUE;
}
/*!
* \brief Ban a node from a resource's (and its children's) allowed nodes list
*
* \param[in,out] rsc Resource to modify
* \param[in] uname Name of node to ban
*/
static void
disallow_node(pe_resource_t *rsc, const char *uname)
{
gpointer match = g_hash_table_lookup(rsc->allowed_nodes, uname);
if (match) {
((pe_node_t *) match)->weight = -INFINITY;
((pe_node_t *) match)->rsc_discover_mode = pe_discover_never;
}
if (rsc->children) {
GListPtr child;
for (child = rsc->children; child != NULL; child = child->next) {
disallow_node((pe_resource_t *) (child->data), uname);
}
}
}
static bool
create_remote_resource(pe_resource_t *parent, pe__bundle_variant_data_t *data,
pe__bundle_replica_t *replica,
pe_working_set_t *data_set)
{
if (replica->child && valid_network(data)) {
GHashTableIter gIter;
GListPtr rsc_iter = NULL;
pe_node_t *node = NULL;
xmlNode *xml_remote = NULL;
char *id = crm_strdup_printf("%s-%d", data->prefix, replica->offset);
char *port_s = NULL;
const char *uname = NULL;
const char *connect_name = NULL;
if (pe_find_resource(data_set->resources, id) != NULL) {
free(id);
// The biggest hammer we have
id = crm_strdup_printf("pcmk-internal-%s-remote-%d",
replica->child->id, replica->offset);
//@TODO return false instead of asserting?
CRM_ASSERT(pe_find_resource(data_set->resources, id) == NULL);
}
/* REMOTE_CONTAINER_HACK: Using "#uname" as the server name when the
* connection does not have its own IP is a magic string that we use to
* support nested remotes (i.e. a bundle running on a remote node).
*/
connect_name = (replica->ipaddr? replica->ipaddr : "#uname");
if (data->control_port == NULL) {
port_s = crm_itoa(DEFAULT_REMOTE_PORT);
}
/* This sets replica->container as replica->remote's container, which is
* similar to what happens with guest nodes. This is how the scheduler
* knows that the bundle node is fenced by recovering the container, and
* that remote should be ordered relative to the container.
*/
xml_remote = pe_create_remote_xml(NULL, id, replica->container->id,
NULL, NULL, NULL,
connect_name, (data->control_port?
data->control_port : port_s));
free(port_s);
/* Abandon our created ID, and pull the copy from the XML, because we
* need something that will get freed during data set cleanup to use as
* the node ID and uname.
*/
free(id);
id = NULL;
uname = ID(xml_remote);
/* Ensure a node has been created for the guest (it may have already
* been, if it has a permanent node attribute), and ensure its weight is
* -INFINITY so no other resources can run on it.
*/
node = pe_find_node(data_set->nodes, uname);
if (node == NULL) {
node = pe_create_node(uname, uname, "remote", "-INFINITY",
data_set);
} else {
node->weight = -INFINITY;
}
node->rsc_discover_mode = pe_discover_never;
/* unpack_remote_nodes() ensures that each remote node and guest node
* has a pe_node_t entry. Ideally, it would do the same for bundle nodes.
* Unfortunately, a bundle has to be mostly unpacked before it's obvious
* what nodes will be needed, so we do it just above.
*
* Worse, that means that the node may have been utilized while
* unpacking other resources, without our weight correction. The most
* likely place for this to happen is when common_unpack() calls
* resource_location() to set a default score in symmetric clusters.
* This adds a node *copy* to each resource's allowed nodes, and these
* copies will have the wrong weight.
*
* As a hacky workaround, fix those copies here.
*
* @TODO Possible alternative: ensure bundles are unpacked before other
* resources, so the weight is correct before any copies are made.
*/
for (rsc_iter = data_set->resources; rsc_iter; rsc_iter = rsc_iter->next) {
disallow_node((pe_resource_t *) (rsc_iter->data), uname);
}
replica->node = node_copy(node);
replica->node->weight = 500;
replica->node->rsc_discover_mode = pe_discover_exclusive;
/* Ensure the node shows up as allowed and with the correct discovery set */
if (replica->child->allowed_nodes != NULL) {
g_hash_table_destroy(replica->child->allowed_nodes);
}
replica->child->allowed_nodes = g_hash_table_new_full(crm_str_hash,
g_str_equal,
NULL, free);
g_hash_table_insert(replica->child->allowed_nodes,
(gpointer) replica->node->details->id,
node_copy(replica->node));
{
pe_node_t *copy = node_copy(replica->node);
copy->weight = -INFINITY;
g_hash_table_insert(replica->child->parent->allowed_nodes,
(gpointer) replica->node->details->id, copy);
}
if (!common_unpack(xml_remote, &replica->remote, parent, data_set)) {
return FALSE;
}
g_hash_table_iter_init(&gIter, replica->remote->allowed_nodes);
while (g_hash_table_iter_next(&gIter, NULL, (void **)&node)) {
if (pe__is_guest_or_remote_node(node)) {
/* Remote resources can only run on 'normal' cluster node */
node->weight = -INFINITY;
}
}
replica->node->details->remote_rsc = replica->remote;
// Ensure pe__is_guest_node() functions correctly immediately
replica->remote->container = replica->container;
/* A bundle's #kind is closer to "container" (guest node) than the
* "remote" set by pe_create_node().
*/
g_hash_table_insert(replica->node->details->attrs,
strdup(CRM_ATTR_KIND), strdup("container"));
/* One effect of this is that setup_container() will add
* replica->remote to replica->container's fillers, which will make
* pe__resource_contains_guest_node() true for replica->container.
*
* replica->child does NOT get added to replica->container's fillers.
* The only noticeable effect if it did would be for its fail count to
* be taken into account when checking replica->container's migration
* threshold.
*/
parent->children = g_list_append(parent->children, replica->remote);
}
return TRUE;
}
static bool
create_container(pe_resource_t *parent, pe__bundle_variant_data_t *data,
pe__bundle_replica_t *replica, pe_working_set_t *data_set)
{
switch (data->agent_type) {
case PE__CONTAINER_AGENT_DOCKER:
if (!create_docker_resource(parent, data, replica, data_set)) {
return FALSE;
}
break;
case PE__CONTAINER_AGENT_PODMAN:
if (!create_podman_resource(parent, data, replica, data_set)) {
return FALSE;
}
break;
case PE__CONTAINER_AGENT_RKT:
if (!create_rkt_resource(parent, data, replica, data_set)) {
return FALSE;
}
break;
default: // PE__CONTAINER_AGENT_UNKNOWN
return FALSE;
}
if (create_ip_resource(parent, data, replica, data_set) == FALSE) {
return FALSE;
}
if(create_remote_resource(parent, data, replica, data_set) == FALSE) {
return FALSE;
}
if (replica->child && replica->ipaddr) {
add_hash_param(replica->child->meta, "external-ip", replica->ipaddr);
}
if (replica->remote) {
/*
* Allow the remote connection resource to be allocated to a
* different node than the one on which the container is active.
*
* This makes it possible to have Pacemaker Remote nodes running
* containers with pacemaker-remoted inside in order to start
* services inside those containers.
*/
set_bit(replica->remote->flags, pe_rsc_allow_remote_remotes);
}
return TRUE;
}
static void
mount_add(pe__bundle_variant_data_t *bundle_data, const char *source,
const char *target, const char *options, uint32_t flags)
{
pe__bundle_mount_t *mount = calloc(1, sizeof(pe__bundle_mount_t));
mount->source = strdup(source);
mount->target = strdup(target);
if (options) {
mount->options = strdup(options);
}
mount->flags = flags;
bundle_data->mounts = g_list_append(bundle_data->mounts, mount);
}
static void
mount_free(pe__bundle_mount_t *mount)
{
free(mount->source);
free(mount->target);
free(mount->options);
free(mount);
}
static void
port_free(pe__bundle_port_t *port)
{
free(port->source);
free(port->target);
free(port);
}
static pe__bundle_replica_t *
replica_for_remote(pe_resource_t *remote)
{
pe_resource_t *top = remote;
pe__bundle_variant_data_t *bundle_data = NULL;
if (top == NULL) {
return NULL;
}
while (top->parent != NULL) {
top = top->parent;
}
get_bundle_variant_data(bundle_data, top);
for (GList *gIter = bundle_data->replicas; gIter != NULL;
gIter = gIter->next) {
pe__bundle_replica_t *replica = gIter->data;
if (replica->remote == remote) {
return replica;
}
}
CRM_LOG_ASSERT(FALSE);
return NULL;
}
bool
pe__bundle_needs_remote_name(pe_resource_t *rsc)
{
const char *value;
if (rsc == NULL) {
return FALSE;
}
value = g_hash_table_lookup(rsc->parameters, XML_RSC_ATTR_REMOTE_RA_ADDR);
if (safe_str_eq(value, "#uname") == FALSE) {
return FALSE;
} else {
const char *match[3][2] = {
{ XML_ATTR_TYPE, "remote" },
{ XML_AGENT_ATTR_CLASS, PCMK_RESOURCE_CLASS_OCF },
{ XML_AGENT_ATTR_PROVIDER, "pacemaker" },
};
for (int m = 0; m < 3; m++) {
value = crm_element_value(rsc->xml, match[m][0]);
if (safe_str_neq(value, match[m][1])) {
return FALSE;
}
}
}
return TRUE;
}
const char *
pe__add_bundle_remote_name(pe_resource_t *rsc, xmlNode *xml, const char *field)
{
// REMOTE_CONTAINER_HACK: Allow remote nodes that start containers with pacemaker remote inside
pe_node_t *node = NULL;
pe__bundle_replica_t *replica = NULL;
if (!pe__bundle_needs_remote_name(rsc)) {
return NULL;
}
replica = replica_for_remote(rsc);
if (replica == NULL) {
return NULL;
}
node = replica->container->allocated_to;
if (node == NULL) {
/* If it won't be running anywhere after the
* transition, go with where it's running now.
*/
node = pe__current_node(replica->container);
}
if(node == NULL) {
crm_trace("Cannot determine address for bundle connection %s", rsc->id);
return NULL;
}
crm_trace("Setting address for bundle connection %s to bundle host %s",
rsc->id, node->details->uname);
if(xml != NULL && field != NULL) {
crm_xml_add(xml, field, node->details->uname);
}
return node->details->uname;
}
gboolean
pe__unpack_bundle(pe_resource_t *rsc, pe_working_set_t *data_set)
{
const char *value = NULL;
xmlNode *xml_obj = NULL;
xmlNode *xml_resource = NULL;
pe__bundle_variant_data_t *bundle_data = NULL;
bool need_log_mount = TRUE;
CRM_ASSERT(rsc != NULL);
pe_rsc_trace(rsc, "Processing resource %s...", rsc->id);
bundle_data = calloc(1, sizeof(pe__bundle_variant_data_t));
rsc->variant_opaque = bundle_data;
bundle_data->prefix = strdup(rsc->id);
xml_obj = first_named_child(rsc->xml, PE__CONTAINER_AGENT_DOCKER_S);
if (xml_obj != NULL) {
bundle_data->agent_type = PE__CONTAINER_AGENT_DOCKER;
} else {
xml_obj = first_named_child(rsc->xml, PE__CONTAINER_AGENT_RKT_S);
if (xml_obj != NULL) {
bundle_data->agent_type = PE__CONTAINER_AGENT_RKT;
} else {
xml_obj = first_named_child(rsc->xml, PE__CONTAINER_AGENT_PODMAN_S);
if (xml_obj != NULL) {
bundle_data->agent_type = PE__CONTAINER_AGENT_PODMAN;
} else {
return FALSE;
}
}
}
value = crm_element_value(xml_obj, XML_RSC_ATTR_PROMOTED_MAX);
if (value == NULL) {
// @COMPAT deprecated since 2.0.0
value = crm_element_value(xml_obj, "masters");
}
bundle_data->promoted_max = crm_parse_int(value, "0");
if (bundle_data->promoted_max < 0) {
pe_err("%s for %s must be nonnegative integer, using 0",
XML_RSC_ATTR_PROMOTED_MAX, rsc->id);
bundle_data->promoted_max = 0;
}
value = crm_element_value(xml_obj, "replicas");
if ((value == NULL) && bundle_data->promoted_max) {
bundle_data->nreplicas = bundle_data->promoted_max;
} else {
bundle_data->nreplicas = crm_parse_int(value, "1");
}
if (bundle_data->nreplicas < 1) {
pe_err("'replicas' for %s must be positive integer, using 1", rsc->id);
bundle_data->nreplicas = 1;
}
/*
* Communication between containers on the same host via the
* floating IPs only works if the container is started with:
* --userland-proxy=false --ip-masq=false
*/
value = crm_element_value(xml_obj, "replicas-per-host");
bundle_data->nreplicas_per_host = crm_parse_int(value, "1");
if (bundle_data->nreplicas_per_host < 1) {
pe_err("'replicas-per-host' for %s must be positive integer, using 1",
rsc->id);
bundle_data->nreplicas_per_host = 1;
}
if (bundle_data->nreplicas_per_host == 1) {
clear_bit(rsc->flags, pe_rsc_unique);
}
bundle_data->container_command = crm_element_value_copy(xml_obj, "run-command");
bundle_data->launcher_options = crm_element_value_copy(xml_obj, "options");
bundle_data->image = crm_element_value_copy(xml_obj, "image");
bundle_data->container_network = crm_element_value_copy(xml_obj, "network");
xml_obj = first_named_child(rsc->xml, "network");
if(xml_obj) {
bundle_data->ip_range_start = crm_element_value_copy(xml_obj, "ip-range-start");
bundle_data->host_netmask = crm_element_value_copy(xml_obj, "host-netmask");
bundle_data->host_network = crm_element_value_copy(xml_obj, "host-interface");
bundle_data->control_port = crm_element_value_copy(xml_obj, "control-port");
value = crm_element_value(xml_obj, "add-host");
- if (pcmk__valid_boolean(value)) {
- crm_str_to_boolean(value, &bundle_data->add_host);
- } else {
+ if (crm_str_to_boolean(value, &bundle_data->add_host) != 1) {
bundle_data->add_host = TRUE;
}
for (xmlNode *xml_child = __xml_first_child_element(xml_obj); xml_child != NULL;
xml_child = __xml_next_element(xml_child)) {
pe__bundle_port_t *port = calloc(1, sizeof(pe__bundle_port_t));
port->source = crm_element_value_copy(xml_child, "port");
if(port->source == NULL) {
port->source = crm_element_value_copy(xml_child, "range");
} else {
port->target = crm_element_value_copy(xml_child, "internal-port");
}
if(port->source != NULL && strlen(port->source) > 0) {
if(port->target == NULL) {
port->target = strdup(port->source);
}
bundle_data->ports = g_list_append(bundle_data->ports, port);
} else {
pe_err("Invalid port directive %s", ID(xml_child));
port_free(port);
}
}
}
xml_obj = first_named_child(rsc->xml, "storage");
for (xmlNode *xml_child = __xml_first_child_element(xml_obj); xml_child != NULL;
xml_child = __xml_next_element(xml_child)) {
const char *source = crm_element_value(xml_child, "source-dir");
const char *target = crm_element_value(xml_child, "target-dir");
const char *options = crm_element_value(xml_child, "options");
int flags = pe__bundle_mount_none;
if (source == NULL) {
source = crm_element_value(xml_child, "source-dir-root");
set_bit(flags, pe__bundle_mount_subdir);
}
if (source && target) {
mount_add(bundle_data, source, target, options, flags);
if (strcmp(target, "/var/log") == 0) {
need_log_mount = FALSE;
}
} else {
pe_err("Invalid mount directive %s", ID(xml_child));
}
}
xml_obj = first_named_child(rsc->xml, "primitive");
if (xml_obj && valid_network(bundle_data)) {
char *value = NULL;
xmlNode *xml_set = NULL;
xml_resource = create_xml_node(NULL, XML_CIB_TAG_INCARNATION);
/* @COMPAT We no longer use the <master> tag, but we need to keep it as
* part of the resource name, so that bundles don't restart in a rolling
* upgrade. (It also avoids needing to change regression tests.)
*/
crm_xml_set_id(xml_resource, "%s-%s", bundle_data->prefix,
(bundle_data->promoted_max? "master"
: (const char *)xml_resource->name));
xml_set = create_xml_node(xml_resource, XML_TAG_META_SETS);
crm_xml_set_id(xml_set, "%s-%s-meta", bundle_data->prefix, xml_resource->name);
crm_create_nvpair_xml(xml_set, NULL,
XML_RSC_ATTR_ORDERED, XML_BOOLEAN_TRUE);
value = crm_itoa(bundle_data->nreplicas);
crm_create_nvpair_xml(xml_set, NULL,
XML_RSC_ATTR_INCARNATION_MAX, value);
free(value);
value = crm_itoa(bundle_data->nreplicas_per_host);
crm_create_nvpair_xml(xml_set, NULL,
XML_RSC_ATTR_INCARNATION_NODEMAX, value);
free(value);
crm_create_nvpair_xml(xml_set, NULL, XML_RSC_ATTR_UNIQUE,
(bundle_data->nreplicas_per_host > 1)?
XML_BOOLEAN_TRUE : XML_BOOLEAN_FALSE);
if (bundle_data->promoted_max) {
crm_create_nvpair_xml(xml_set, NULL,
XML_RSC_ATTR_PROMOTABLE, XML_BOOLEAN_TRUE);
value = crm_itoa(bundle_data->promoted_max);
crm_create_nvpair_xml(xml_set, NULL,
XML_RSC_ATTR_PROMOTED_MAX, value);
free(value);
}
//crm_xml_add(xml_obj, XML_ATTR_ID, bundle_data->prefix);
add_node_copy(xml_resource, xml_obj);
} else if(xml_obj) {
pe_err("Cannot control %s inside %s without either ip-range-start or control-port",
rsc->id, ID(xml_obj));
return FALSE;
}
if(xml_resource) {
int lpc = 0;
GListPtr childIter = NULL;
pe_resource_t *new_rsc = NULL;
pe__bundle_port_t *port = NULL;
int offset = 0, max = 1024;
char *buffer = NULL;
if (common_unpack(xml_resource, &new_rsc, rsc, data_set) == FALSE) {
pe_err("Failed unpacking resource %s", ID(rsc->xml));
if (new_rsc != NULL && new_rsc->fns != NULL) {
new_rsc->fns->free(new_rsc);
}
return FALSE;
}
bundle_data->child = new_rsc;
/* Currently, we always map the default authentication key location
* into the same location inside the container.
*
* Ideally, we would respect the host's PCMK_authkey_location, but:
* - it may be different on different nodes;
* - the actual connection will do extra checking to make sure the key
* file exists and is readable, that we can't do here on the DC
* - tools such as crm_resource and crm_simulate may not have the same
* environment variables as the cluster, causing operation digests to
* differ
*
* Always using the default location inside the container is fine,
* because we control the pacemaker_remote environment, and it avoids
* having to pass another environment variable to the container.
*
* @TODO A better solution may be to have only pacemaker_remote use the
* environment variable, and have the cluster nodes use a new
* cluster option for key location. This would introduce the limitation
* of the location being the same on all cluster nodes, but that's
* reasonable.
*/
mount_add(bundle_data, DEFAULT_REMOTE_KEY_LOCATION,
DEFAULT_REMOTE_KEY_LOCATION, NULL, pe__bundle_mount_none);
if (need_log_mount) {
mount_add(bundle_data, CRM_BUNDLE_DIR, "/var/log", NULL,
pe__bundle_mount_subdir);
}
port = calloc(1, sizeof(pe__bundle_port_t));
if(bundle_data->control_port) {
port->source = strdup(bundle_data->control_port);
} else {
/* If we wanted to respect PCMK_remote_port, we could use
* crm_default_remote_port() here and elsewhere in this file instead
* of DEFAULT_REMOTE_PORT.
*
* However, it gains nothing, since we control both the container
* environment and the connection resource parameters, and the user
* can use a different port if desired by setting control-port.
*/
port->source = crm_itoa(DEFAULT_REMOTE_PORT);
}
port->target = strdup(port->source);
bundle_data->ports = g_list_append(bundle_data->ports, port);
buffer = calloc(1, max+1);
for (childIter = bundle_data->child->children; childIter != NULL;
childIter = childIter->next) {
pe__bundle_replica_t *replica = calloc(1, sizeof(pe__bundle_replica_t));
replica->child = childIter->data;
replica->child->exclusive_discover = TRUE;
replica->offset = lpc++;
// Ensure the child's notify gets set based on the underlying primitive's value
if (is_set(replica->child->flags, pe_rsc_notify)) {
set_bit(bundle_data->child->flags, pe_rsc_notify);
}
offset += allocate_ip(bundle_data, replica, buffer+offset,
max-offset);
bundle_data->replicas = g_list_append(bundle_data->replicas,
replica);
bundle_data->attribute_target = g_hash_table_lookup(replica->child->meta,
XML_RSC_ATTR_TARGET);
}
bundle_data->container_host_options = buffer;
if (bundle_data->attribute_target) {
g_hash_table_replace(rsc->meta, strdup(XML_RSC_ATTR_TARGET),
strdup(bundle_data->attribute_target));
g_hash_table_replace(bundle_data->child->meta,
strdup(XML_RSC_ATTR_TARGET),
strdup(bundle_data->attribute_target));
}
} else {
// Just a naked container, no pacemaker-remote
int offset = 0, max = 1024;
char *buffer = calloc(1, max+1);
for (int lpc = 0; lpc < bundle_data->nreplicas; lpc++) {
pe__bundle_replica_t *replica = calloc(1, sizeof(pe__bundle_replica_t));
replica->offset = lpc;
offset += allocate_ip(bundle_data, replica, buffer+offset,
max-offset);
bundle_data->replicas = g_list_append(bundle_data->replicas,
replica);
}
bundle_data->container_host_options = buffer;
}
for (GList *gIter = bundle_data->replicas; gIter != NULL;
gIter = gIter->next) {
pe__bundle_replica_t *replica = gIter->data;
if (!create_container(rsc, bundle_data, replica, data_set)) {
pe_err("Failed unpacking resource %s", rsc->id);
rsc->fns->free(rsc);
return FALSE;
}
}
if (bundle_data->child) {
rsc->children = g_list_append(rsc->children, bundle_data->child);
}
return TRUE;
}
static int
replica_resource_active(pe_resource_t *rsc, gboolean all)
{
if (rsc) {
gboolean child_active = rsc->fns->active(rsc, all);
if (child_active && !all) {
return TRUE;
} else if (!child_active && all) {
return FALSE;
}
}
return -1;
}
gboolean
pe__bundle_active(pe_resource_t *rsc, gboolean all)
{
pe__bundle_variant_data_t *bundle_data = NULL;
GListPtr iter = NULL;
get_bundle_variant_data(bundle_data, rsc);
for (iter = bundle_data->replicas; iter != NULL; iter = iter->next) {
pe__bundle_replica_t *replica = iter->data;
int rsc_active;
rsc_active = replica_resource_active(replica->ip, all);
if (rsc_active >= 0) {
return (gboolean) rsc_active;
}
rsc_active = replica_resource_active(replica->child, all);
if (rsc_active >= 0) {
return (gboolean) rsc_active;
}
rsc_active = replica_resource_active(replica->container, all);
if (rsc_active >= 0) {
return (gboolean) rsc_active;
}
rsc_active = replica_resource_active(replica->remote, all);
if (rsc_active >= 0) {
return (gboolean) rsc_active;
}
}
/* If "all" is TRUE, we've already checked that no resources were inactive,
* so return TRUE; if "all" is FALSE, we didn't find any active resources,
* so return FALSE.
*/
return all;
}
/*!
* \internal
* \brief Find the bundle replica corresponding to a given node
*
* \param[in] bundle Top-level bundle resource
* \param[in] node Node to search for
*
* \return Bundle replica if found, NULL otherwise
*/
pe_resource_t *
pe__find_bundle_replica(const pe_resource_t *bundle, const pe_node_t *node)
{
pe__bundle_variant_data_t *bundle_data = NULL;
CRM_ASSERT(bundle && node);
get_bundle_variant_data(bundle_data, bundle);
for (GList *gIter = bundle_data->replicas; gIter != NULL;
gIter = gIter->next) {
pe__bundle_replica_t *replica = gIter->data;
CRM_ASSERT(replica && replica->node);
if (replica->node->details == node->details) {
return replica->child;
}
}
return NULL;
}
static void
print_rsc_in_list(pe_resource_t *rsc, const char *pre_text, long options,
void *print_data)
{
if (rsc != NULL) {
if (options & pe_print_html) {
status_print("<li>");
}
rsc->fns->print(rsc, pre_text, options, print_data);
if (options & pe_print_html) {
status_print("</li>\n");
}
}
}
static const char*
container_agent_str(enum pe__container_agent t)
{
switch (t) {
case PE__CONTAINER_AGENT_DOCKER: return PE__CONTAINER_AGENT_DOCKER_S;
case PE__CONTAINER_AGENT_RKT: return PE__CONTAINER_AGENT_RKT_S;
case PE__CONTAINER_AGENT_PODMAN: return PE__CONTAINER_AGENT_PODMAN_S;
default: // PE__CONTAINER_AGENT_UNKNOWN
break;
}
return PE__CONTAINER_AGENT_UNKNOWN_S;
}
static void
bundle_print_xml(pe_resource_t *rsc, const char *pre_text, long options,
void *print_data)
{
pe__bundle_variant_data_t *bundle_data = NULL;
char *child_text = NULL;
CRM_CHECK(rsc != NULL, return);
if (pre_text == NULL) {
pre_text = "";
}
child_text = crm_strdup_printf("%s ", pre_text);
get_bundle_variant_data(bundle_data, rsc);
status_print("%s<bundle ", pre_text);
status_print("id=\"%s\" ", rsc->id);
status_print("type=\"%s\" ", container_agent_str(bundle_data->agent_type));
status_print("image=\"%s\" ", bundle_data->image);
status_print("unique=\"%s\" ", is_set(rsc->flags, pe_rsc_unique)? "true" : "false");
status_print("managed=\"%s\" ", is_set(rsc->flags, pe_rsc_managed) ? "true" : "false");
status_print("failed=\"%s\" ", is_set(rsc->flags, pe_rsc_failed) ? "true" : "false");
status_print(">\n");
for (GList *gIter = bundle_data->replicas; gIter != NULL;
gIter = gIter->next) {
pe__bundle_replica_t *replica = gIter->data;
CRM_ASSERT(replica);
status_print("%s <replica id=\"%d\">\n", pre_text, replica->offset);
print_rsc_in_list(replica->ip, child_text, options, print_data);
print_rsc_in_list(replica->child, child_text, options, print_data);
print_rsc_in_list(replica->container, child_text, options, print_data);
print_rsc_in_list(replica->remote, child_text, options, print_data);
status_print("%s </replica>\n", pre_text);
}
status_print("%s</bundle>\n", pre_text);
free(child_text);
}
int
pe__bundle_xml(pcmk__output_t *out, va_list args)
{
unsigned int options = va_arg(args, unsigned int);
pe_resource_t *rsc = va_arg(args, pe_resource_t *);
pe__bundle_variant_data_t *bundle_data = NULL;
int rc = pcmk_rc_no_output;
CRM_ASSERT(rsc != NULL);
get_bundle_variant_data(bundle_data, rsc);
rc = pe__name_and_nvpairs_xml(out, true, "bundle", 6
, "id", rsc->id
, "type", container_agent_str(bundle_data->agent_type)
, "image", bundle_data->image
, "unique", BOOL2STR(is_set(rsc->flags, pe_rsc_unique))
, "managed", BOOL2STR(is_set(rsc->flags, pe_rsc_managed))
, "failed", BOOL2STR(is_set(rsc->flags, pe_rsc_failed)));
CRM_ASSERT(rc == pcmk_rc_ok);
for (GList *gIter = bundle_data->replicas; gIter != NULL;
gIter = gIter->next) {
pe__bundle_replica_t *replica = gIter->data;
char *id = crm_itoa(replica->offset);
CRM_ASSERT(replica);
rc = pe__name_and_nvpairs_xml(out, true, "replica", 1, "id", id);
free(id);
CRM_ASSERT(rc == pcmk_rc_ok);
if (replica->ip != NULL) {
out->message(out, crm_map_element_name(replica->ip->xml), options, replica->ip);
}
if (replica->child != NULL) {
out->message(out, crm_map_element_name(replica->child->xml), options, replica->child);
}
out->message(out, crm_map_element_name(replica->container->xml), options, replica->container);
if (replica->remote != NULL) {
out->message(out, crm_map_element_name(replica->remote->xml), options, replica->remote);
}
pcmk__output_xml_pop_parent(out); // replica
}
pcmk__output_xml_pop_parent(out); // bundle
return rc;
}
static void
pe__bundle_replica_output_html(pcmk__output_t *out, pe__bundle_replica_t *replica,
long options)
{
pe_node_t *node = NULL;
pe_resource_t *rsc = replica->child;
int offset = 0;
char buffer[LINE_MAX];
if(rsc == NULL) {
rsc = replica->container;
}
if (replica->remote) {
offset += snprintf(buffer + offset, LINE_MAX - offset, "%s",
rsc_printable_id(replica->remote));
} else {
offset += snprintf(buffer + offset, LINE_MAX - offset, "%s",
rsc_printable_id(replica->container));
}
if (replica->ipaddr) {
offset += snprintf(buffer + offset, LINE_MAX - offset, " (%s)",
replica->ipaddr);
}
node = pe__current_node(replica->container);
pe__common_output_html(out, rsc, buffer, node, options);
}
int
pe__bundle_html(pcmk__output_t *out, va_list args)
{
unsigned int options = va_arg(args, unsigned int);
pe_resource_t *rsc = va_arg(args, pe_resource_t *);
pe__bundle_variant_data_t *bundle_data = NULL;
char buffer[LINE_MAX];
CRM_ASSERT(rsc != NULL);
get_bundle_variant_data(bundle_data, rsc);
pcmk__output_create_xml_node(out, "br");
out->begin_list(out, NULL, NULL, "Container bundle%s: %s [%s]%s%s",
(bundle_data->nreplicas > 1)? " set" : "",
rsc->id, bundle_data->image,
is_set(rsc->flags, pe_rsc_unique) ? " (unique)" : "",
is_set(rsc->flags, pe_rsc_managed) ? "" : " (unmanaged)");
for (GList *gIter = bundle_data->replicas; gIter != NULL;
gIter = gIter->next) {
pe__bundle_replica_t *replica = gIter->data;
CRM_ASSERT(replica);
pcmk__output_xml_create_parent(out, "li");
if (is_set(options, pe_print_implicit)) {
if (pcmk__list_of_multiple(bundle_data->replicas)) {
snprintf(buffer, LINE_MAX, " Replica[%d]", replica->offset);
xmlNodeSetContent(pcmk__output_xml_peek_parent(out), (pcmkXmlStr) buffer);
}
pcmk__output_create_xml_node(out, "br");
out->begin_list(out, NULL, NULL, NULL);
if (replica->ip != NULL) {
out->message(out, crm_map_element_name(replica->ip->xml), options, replica->ip);
}
if (replica->child != NULL) {
out->message(out, crm_map_element_name(replica->child->xml), options, replica->child);
}
out->message(out, crm_map_element_name(replica->container->xml), options, replica->container);
if (replica->remote != NULL) {
out->message(out, crm_map_element_name(replica->remote->xml), options, replica->remote);
}
out->end_list(out);
} else {
pe__bundle_replica_output_html(out, replica, options);
}
pcmk__output_xml_pop_parent(out);
}
out->end_list(out);
return pcmk_rc_ok;
}
static void
pe__bundle_replica_output_text(pcmk__output_t *out, pe__bundle_replica_t *replica,
long options)
{
pe_node_t *node = NULL;
pe_resource_t *rsc = replica->child;
int offset = 0;
char buffer[LINE_MAX];
if(rsc == NULL) {
rsc = replica->container;
}
if (replica->remote) {
offset += snprintf(buffer + offset, LINE_MAX - offset, "%s",
rsc_printable_id(replica->remote));
} else {
offset += snprintf(buffer + offset, LINE_MAX - offset, "%s",
rsc_printable_id(replica->container));
}
if (replica->ipaddr) {
offset += snprintf(buffer + offset, LINE_MAX - offset, " (%s)",
replica->ipaddr);
}
node = pe__current_node(replica->container);
pe__common_output_text(out, rsc, buffer, node, options);
}
int
pe__bundle_text(pcmk__output_t *out, va_list args)
{
unsigned int options = va_arg(args, unsigned int);
pe_resource_t *rsc = va_arg(args, pe_resource_t *);
pe__bundle_variant_data_t *bundle_data = NULL;
CRM_ASSERT(rsc != NULL);
get_bundle_variant_data(bundle_data, rsc);
out->begin_list(out, NULL, NULL, "Container bundle%s: %s [%s]%s%s",
(bundle_data->nreplicas > 1)? " set" : "",
rsc->id, bundle_data->image,
is_set(rsc->flags, pe_rsc_unique) ? " (unique)" : "",
is_set(rsc->flags, pe_rsc_managed) ? "" : " (unmanaged)");
for (GList *gIter = bundle_data->replicas; gIter != NULL;
gIter = gIter->next) {
pe__bundle_replica_t *replica = gIter->data;
CRM_ASSERT(replica);
if (is_set(options, pe_print_implicit)) {
if (pcmk__list_of_multiple(bundle_data->replicas)) {
out->list_item(out, NULL, "Replica[%d]", replica->offset);
}
out->begin_list(out, NULL, NULL, NULL);
if (replica->ip != NULL) {
out->message(out, crm_map_element_name(replica->ip->xml), options, replica->ip);
}
if (replica->child != NULL) {
out->message(out, crm_map_element_name(replica->child->xml), options, replica->child);
}
out->message(out, crm_map_element_name(replica->container->xml), options, replica->container);
if (replica->remote != NULL) {
out->message(out, crm_map_element_name(replica->remote->xml), options, replica->remote);
}
out->end_list(out);
} else {
pe__bundle_replica_output_text(out, replica, options);
}
}
out->end_list(out);
return pcmk_rc_ok;
}
static void
print_bundle_replica(pe__bundle_replica_t *replica, const char *pre_text,
long options, void *print_data)
{
pe_node_t *node = NULL;
pe_resource_t *rsc = replica->child;
int offset = 0;
char buffer[LINE_MAX];
if(rsc == NULL) {
rsc = replica->container;
}
if (replica->remote) {
offset += snprintf(buffer + offset, LINE_MAX - offset, "%s",
rsc_printable_id(replica->remote));
} else {
offset += snprintf(buffer + offset, LINE_MAX - offset, "%s",
rsc_printable_id(replica->container));
}
if (replica->ipaddr) {
offset += snprintf(buffer + offset, LINE_MAX - offset, " (%s)",
replica->ipaddr);
}
node = pe__current_node(replica->container);
common_print(rsc, pre_text, buffer, node, options, print_data);
}
void
pe__print_bundle(pe_resource_t *rsc, const char *pre_text, long options,
void *print_data)
{
pe__bundle_variant_data_t *bundle_data = NULL;
char *child_text = NULL;
CRM_CHECK(rsc != NULL, return);
if (options & pe_print_xml) {
bundle_print_xml(rsc, pre_text, options, print_data);
return;
}
get_bundle_variant_data(bundle_data, rsc);
if (pre_text == NULL) {
pre_text = " ";
}
status_print("%sContainer bundle%s: %s [%s]%s%s\n",
pre_text, ((bundle_data->nreplicas > 1)? " set" : ""),
rsc->id, bundle_data->image,
is_set(rsc->flags, pe_rsc_unique) ? " (unique)" : "",
is_set(rsc->flags, pe_rsc_managed) ? "" : " (unmanaged)");
if (options & pe_print_html) {
status_print("<br />\n<ul>\n");
}
for (GList *gIter = bundle_data->replicas; gIter != NULL;
gIter = gIter->next) {
pe__bundle_replica_t *replica = gIter->data;
CRM_ASSERT(replica);
if (options & pe_print_html) {
status_print("<li>");
}
if (is_set(options, pe_print_implicit)) {
child_text = crm_strdup_printf(" %s", pre_text);
if (pcmk__list_of_multiple(bundle_data->replicas)) {
status_print(" %sReplica[%d]\n", pre_text, replica->offset);
}
if (options & pe_print_html) {
status_print("<br />\n<ul>\n");
}
print_rsc_in_list(replica->ip, child_text, options, print_data);
print_rsc_in_list(replica->container, child_text, options, print_data);
print_rsc_in_list(replica->remote, child_text, options, print_data);
print_rsc_in_list(replica->child, child_text, options, print_data);
if (options & pe_print_html) {
status_print("</ul>\n");
}
} else {
child_text = crm_strdup_printf("%s ", pre_text);
print_bundle_replica(replica, child_text, options, print_data);
}
free(child_text);
if (options & pe_print_html) {
status_print("</li>\n");
}
}
if (options & pe_print_html) {
status_print("</ul>\n");
}
}
static void
free_bundle_replica(pe__bundle_replica_t *replica)
{
if (replica == NULL) {
return;
}
if (replica->node) {
free(replica->node);
replica->node = NULL;
}
if (replica->ip) {
free_xml(replica->ip->xml);
replica->ip->xml = NULL;
replica->ip->fns->free(replica->ip);
replica->ip = NULL;
}
if (replica->container) {
free_xml(replica->container->xml);
replica->container->xml = NULL;
replica->container->fns->free(replica->container);
replica->container = NULL;
}
if (replica->remote) {
free_xml(replica->remote->xml);
replica->remote->xml = NULL;
replica->remote->fns->free(replica->remote);
replica->remote = NULL;
}
free(replica->ipaddr);
free(replica);
}
void
pe__free_bundle(pe_resource_t *rsc)
{
pe__bundle_variant_data_t *bundle_data = NULL;
CRM_CHECK(rsc != NULL, return);
get_bundle_variant_data(bundle_data, rsc);
pe_rsc_trace(rsc, "Freeing %s", rsc->id);
free(bundle_data->prefix);
free(bundle_data->image);
free(bundle_data->control_port);
free(bundle_data->host_network);
free(bundle_data->host_netmask);
free(bundle_data->ip_range_start);
free(bundle_data->container_network);
free(bundle_data->launcher_options);
free(bundle_data->container_command);
free(bundle_data->container_host_options);
g_list_free_full(bundle_data->replicas,
(GDestroyNotify) free_bundle_replica);
g_list_free_full(bundle_data->mounts, (GDestroyNotify)mount_free);
g_list_free_full(bundle_data->ports, (GDestroyNotify)port_free);
g_list_free(rsc->children);
if(bundle_data->child) {
free_xml(bundle_data->child->xml);
bundle_data->child->xml = NULL;
bundle_data->child->fns->free(bundle_data->child);
}
common_free(rsc);
}
enum rsc_role_e
pe__bundle_resource_state(const pe_resource_t *rsc, gboolean current)
{
enum rsc_role_e container_role = RSC_ROLE_UNKNOWN;
return container_role;
}
/*!
* \brief Get the number of configured replicas in a bundle
*
* \param[in] rsc Bundle resource
*
* \return Number of configured replicas, or 0 on error
*/
int
pe_bundle_replicas(const pe_resource_t *rsc)
{
if ((rsc == NULL) || (rsc->variant != pe_container)) {
return 0;
} else {
pe__bundle_variant_data_t *bundle_data = NULL;
get_bundle_variant_data(bundle_data, rsc);
return bundle_data->nreplicas;
}
}
void
pe__count_bundle(pe_resource_t *rsc)
{
pe__bundle_variant_data_t *bundle_data = NULL;
get_bundle_variant_data(bundle_data, rsc);
for (GList *item = bundle_data->replicas; item != NULL; item = item->next) {
pe__bundle_replica_t *replica = item->data;
if (replica->ip) {
replica->ip->fns->count(replica->ip);
}
if (replica->child) {
replica->child->fns->count(replica->child);
}
if (replica->container) {
replica->container->fns->count(replica->container);
}
if (replica->remote) {
replica->remote->fns->count(replica->remote);
}
}
}
diff --git a/lib/pengine/complex.c b/lib/pengine/complex.c
index 496452e4eb..c268fa5444 100644
--- a/lib/pengine/complex.c
+++ b/lib/pengine/complex.c
@@ -1,975 +1,964 @@
/*
* Copyright 2004-2020 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 <crm/pengine/rules.h>
#include <crm/pengine/internal.h>
#include <crm/msg_xml.h>
-#include <unpack.h>
-
void populate_hash(xmlNode * nvpair_list, GHashTable * hash, const char **attrs, int attrs_length);
resource_object_functions_t resource_class_functions[] = {
{
native_unpack,
native_find_rsc,
native_parameter,
native_print,
native_active,
native_resource_state,
native_location,
native_free,
pe__count_common,
},
{
group_unpack,
native_find_rsc,
native_parameter,
group_print,
group_active,
group_resource_state,
native_location,
group_free,
pe__count_common,
},
{
clone_unpack,
native_find_rsc,
native_parameter,
clone_print,
clone_active,
clone_resource_state,
native_location,
clone_free,
pe__count_common,
},
{
pe__unpack_bundle,
native_find_rsc,
native_parameter,
pe__print_bundle,
pe__bundle_active,
pe__bundle_resource_state,
native_location,
pe__free_bundle,
pe__count_bundle,
}
};
static enum pe_obj_types
get_resource_type(const char *name)
{
if (safe_str_eq(name, XML_CIB_TAG_RESOURCE)) {
return pe_native;
} else if (safe_str_eq(name, XML_CIB_TAG_GROUP)) {
return pe_group;
} else if (safe_str_eq(name, XML_CIB_TAG_INCARNATION)) {
return pe_clone;
} else if (safe_str_eq(name, XML_CIB_TAG_MASTER)) {
// @COMPAT deprecated since 2.0.0
return pe_clone;
} else if (safe_str_eq(name, XML_CIB_TAG_CONTAINER)) {
return pe_container;
}
return pe_unknown;
}
static void
dup_attr(gpointer key, gpointer value, gpointer user_data)
{
add_hash_param(user_data, key, value);
}
void
get_meta_attributes(GHashTable * meta_hash, pe_resource_t * rsc,
pe_node_t * node, pe_working_set_t * data_set)
{
GHashTable *node_hash = NULL;
if (node) {
node_hash = node->details->attrs;
}
if (rsc->xml) {
xmlAttrPtr xIter = NULL;
for (xIter = rsc->xml->properties; xIter; xIter = xIter->next) {
const char *prop_name = (const char *)xIter->name;
const char *prop_value = crm_element_value(rsc->xml, prop_name);
add_hash_param(meta_hash, prop_name, prop_value);
}
}
pe__unpack_dataset_nvpairs(rsc->xml, XML_TAG_META_SETS, node_hash,
meta_hash, NULL, FALSE, data_set);
/* set anything else based on the parent */
if (rsc->parent != NULL) {
g_hash_table_foreach(rsc->parent->meta, dup_attr, meta_hash);
}
/* and finally check the defaults */
pe__unpack_dataset_nvpairs(data_set->rsc_defaults, XML_TAG_META_SETS,
node_hash, meta_hash, NULL, FALSE, data_set);
}
void
get_rsc_attributes(GHashTable * meta_hash, pe_resource_t * rsc,
pe_node_t * node, pe_working_set_t * data_set)
{
GHashTable *node_hash = NULL;
if (node) {
node_hash = node->details->attrs;
}
pe__unpack_dataset_nvpairs(rsc->xml, XML_TAG_ATTR_SETS, node_hash,
meta_hash, NULL, FALSE, data_set);
/* set anything else based on the parent */
if (rsc->parent != NULL) {
get_rsc_attributes(meta_hash, rsc->parent, node, data_set);
} else {
/* and finally check the defaults */
pe__unpack_dataset_nvpairs(data_set->rsc_defaults, XML_TAG_ATTR_SETS,
node_hash, meta_hash, NULL, FALSE, data_set);
}
}
#if ENABLE_VERSIONED_ATTRS
void
pe_get_versioned_attributes(xmlNode * meta_hash, pe_resource_t * rsc,
pe_node_t * node, pe_working_set_t * data_set)
{
GHashTable *node_hash = NULL;
if (node) {
node_hash = node->details->attrs;
}
pe_unpack_versioned_attributes(data_set->input, rsc->xml, XML_TAG_ATTR_SETS, node_hash,
meta_hash, data_set->now, NULL);
/* set anything else based on the parent */
if (rsc->parent != NULL) {
pe_get_versioned_attributes(meta_hash, rsc->parent, node, data_set);
} else {
/* and finally check the defaults */
pe_unpack_versioned_attributes(data_set->input, data_set->rsc_defaults, XML_TAG_ATTR_SETS,
node_hash, meta_hash, data_set->now, NULL);
}
}
#endif
static char *
template_op_key(xmlNode * op)
{
const char *name = crm_element_value(op, "name");
const char *role = crm_element_value(op, "role");
char *key = NULL;
if (role == NULL || crm_str_eq(role, RSC_ROLE_STARTED_S, TRUE)
|| crm_str_eq(role, RSC_ROLE_SLAVE_S, TRUE)) {
role = RSC_ROLE_UNKNOWN_S;
}
key = crm_strdup_printf("%s-%s", name, role);
return key;
}
static gboolean
unpack_template(xmlNode * xml_obj, xmlNode ** expanded_xml, pe_working_set_t * data_set)
{
xmlNode *cib_resources = NULL;
xmlNode *template = NULL;
xmlNode *new_xml = NULL;
xmlNode *child_xml = NULL;
xmlNode *rsc_ops = NULL;
xmlNode *template_ops = NULL;
const char *template_ref = NULL;
const char *clone = NULL;
const char *id = NULL;
if (xml_obj == NULL) {
pe_err("No resource object for template unpacking");
return FALSE;
}
template_ref = crm_element_value(xml_obj, XML_CIB_TAG_RSC_TEMPLATE);
if (template_ref == NULL) {
return TRUE;
}
id = ID(xml_obj);
if (id == NULL) {
pe_err("'%s' object must have a id", crm_element_name(xml_obj));
return FALSE;
}
if (crm_str_eq(template_ref, id, TRUE)) {
pe_err("The resource object '%s' should not reference itself", id);
return FALSE;
}
cib_resources = get_xpath_object("//"XML_CIB_TAG_RESOURCES, data_set->input, LOG_TRACE);
if (cib_resources == NULL) {
pe_err("No resources configured");
return FALSE;
}
template = find_entity(cib_resources, XML_CIB_TAG_RSC_TEMPLATE, template_ref);
if (template == NULL) {
pe_err("No template named '%s'", template_ref);
return FALSE;
}
new_xml = copy_xml(template);
xmlNodeSetName(new_xml, xml_obj->name);
crm_xml_replace(new_xml, XML_ATTR_ID, id);
clone = crm_element_value(xml_obj, XML_RSC_ATTR_INCARNATION);
if(clone) {
crm_xml_add(new_xml, XML_RSC_ATTR_INCARNATION, clone);
}
template_ops = find_xml_node(new_xml, "operations", FALSE);
for (child_xml = __xml_first_child_element(xml_obj); child_xml != NULL;
child_xml = __xml_next_element(child_xml)) {
xmlNode *new_child = NULL;
new_child = add_node_copy(new_xml, child_xml);
if (crm_str_eq((const char *)new_child->name, "operations", TRUE)) {
rsc_ops = new_child;
}
}
if (template_ops && rsc_ops) {
xmlNode *op = NULL;
GHashTable *rsc_ops_hash = g_hash_table_new_full(crm_str_hash,
g_str_equal, free,
NULL);
for (op = __xml_first_child_element(rsc_ops); op != NULL;
op = __xml_next_element(op)) {
char *key = template_op_key(op);
g_hash_table_insert(rsc_ops_hash, key, op);
}
for (op = __xml_first_child_element(template_ops); op != NULL;
op = __xml_next_element(op)) {
char *key = template_op_key(op);
if (g_hash_table_lookup(rsc_ops_hash, key) == NULL) {
add_node_copy(rsc_ops, op);
}
free(key);
}
if (rsc_ops_hash) {
g_hash_table_destroy(rsc_ops_hash);
}
free_xml(template_ops);
}
/*free_xml(*expanded_xml); */
*expanded_xml = new_xml;
/* Disable multi-level templates for now */
/*if(unpack_template(new_xml, expanded_xml, data_set) == FALSE) {
free_xml(*expanded_xml);
*expanded_xml = NULL;
return FALSE;
} */
return TRUE;
}
static gboolean
add_template_rsc(xmlNode * xml_obj, pe_working_set_t * data_set)
{
const char *template_ref = NULL;
const char *id = NULL;
if (xml_obj == NULL) {
pe_err("No resource object for processing resource list of template");
return FALSE;
}
template_ref = crm_element_value(xml_obj, XML_CIB_TAG_RSC_TEMPLATE);
if (template_ref == NULL) {
return TRUE;
}
id = ID(xml_obj);
if (id == NULL) {
pe_err("'%s' object must have a id", crm_element_name(xml_obj));
return FALSE;
}
if (crm_str_eq(template_ref, id, TRUE)) {
pe_err("The resource object '%s' should not reference itself", id);
return FALSE;
}
if (add_tag_ref(data_set->template_rsc_sets, template_ref, id) == FALSE) {
return FALSE;
}
return TRUE;
}
static bool
detect_promotable(pe_resource_t *rsc)
{
const char *promotable = g_hash_table_lookup(rsc->meta,
XML_RSC_ATTR_PROMOTABLE);
if (crm_is_true(promotable)) {
return TRUE;
}
// @COMPAT deprecated since 2.0.0
if (safe_str_eq(crm_element_name(rsc->xml), XML_CIB_TAG_MASTER)) {
/* @TODO in some future version, pe_warn_once() here,
* then drop support in even later version
*/
g_hash_table_insert(rsc->meta, strdup(XML_RSC_ATTR_PROMOTABLE),
strdup(XML_BOOLEAN_TRUE));
return TRUE;
}
return FALSE;
}
gboolean
common_unpack(xmlNode * xml_obj, pe_resource_t ** rsc,
pe_resource_t * parent, pe_working_set_t * data_set)
{
bool isdefault = FALSE;
xmlNode *expanded_xml = NULL;
xmlNode *ops = NULL;
const char *value = NULL;
const char *rclass = NULL; /* Look for this after any templates have been expanded */
const char *id = crm_element_value(xml_obj, XML_ATTR_ID);
bool guest_node = FALSE;
bool remote_node = FALSE;
bool has_versioned_params = FALSE;
crm_log_xml_trace(xml_obj, "Processing resource input...");
if (id == NULL) {
pe_err("Must specify id tag in <resource>");
return FALSE;
} else if (rsc == NULL) {
pe_err("Nowhere to unpack resource into");
return FALSE;
}
if (unpack_template(xml_obj, &expanded_xml, data_set) == FALSE) {
return FALSE;
}
*rsc = calloc(1, sizeof(pe_resource_t));
(*rsc)->cluster = data_set;
if (expanded_xml) {
crm_log_xml_trace(expanded_xml, "Expanded resource...");
(*rsc)->xml = expanded_xml;
(*rsc)->orig_xml = xml_obj;
} else {
(*rsc)->xml = xml_obj;
(*rsc)->orig_xml = NULL;
}
/* Do not use xml_obj from here on, use (*rsc)->xml in case templates are involved */
rclass = crm_element_value((*rsc)->xml, XML_AGENT_ATTR_CLASS);
(*rsc)->parent = parent;
ops = find_xml_node((*rsc)->xml, "operations", FALSE);
(*rsc)->ops_xml = expand_idref(ops, data_set->input);
(*rsc)->variant = get_resource_type(crm_element_name((*rsc)->xml));
if ((*rsc)->variant == pe_unknown) {
pe_err("Unknown resource type: %s", crm_element_name((*rsc)->xml));
free(*rsc);
return FALSE;
}
(*rsc)->parameters = crm_str_table_new();
#if ENABLE_VERSIONED_ATTRS
(*rsc)->versioned_parameters = create_xml_node(NULL, XML_TAG_RSC_VER_ATTRS);
#endif
(*rsc)->meta = crm_str_table_new();
(*rsc)->allowed_nodes =
g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, free);
(*rsc)->known_on = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL,
free);
value = crm_element_value((*rsc)->xml, XML_RSC_ATTR_INCARNATION);
if (value) {
(*rsc)->id = crm_strdup_printf("%s:%s", id, value);
add_hash_param((*rsc)->meta, XML_RSC_ATTR_INCARNATION, value);
} else {
(*rsc)->id = strdup(id);
}
(*rsc)->fns = &resource_class_functions[(*rsc)->variant];
pe_rsc_trace((*rsc), "Unpacking resource...");
get_meta_attributes((*rsc)->meta, *rsc, NULL, data_set);
get_rsc_attributes((*rsc)->parameters, *rsc, NULL, data_set);
#if ENABLE_VERSIONED_ATTRS
pe_get_versioned_attributes((*rsc)->versioned_parameters, *rsc, NULL, data_set);
#endif
(*rsc)->flags = 0;
set_bit((*rsc)->flags, pe_rsc_runnable);
set_bit((*rsc)->flags, pe_rsc_provisional);
if (is_not_set(data_set->flags, pe_flag_maintenance_mode)) {
set_bit((*rsc)->flags, pe_rsc_managed);
}
(*rsc)->rsc_cons = NULL;
(*rsc)->rsc_tickets = NULL;
(*rsc)->actions = NULL;
(*rsc)->role = RSC_ROLE_STOPPED;
(*rsc)->next_role = RSC_ROLE_UNKNOWN;
(*rsc)->recovery_type = recovery_stop_start;
(*rsc)->stickiness = 0;
(*rsc)->migration_threshold = INFINITY;
(*rsc)->failure_timeout = 0;
value = g_hash_table_lookup((*rsc)->meta, XML_CIB_ATTR_PRIORITY);
(*rsc)->priority = crm_parse_int(value, "0");
value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_NOTIFY);
if (crm_is_true(value)) {
set_bit((*rsc)->flags, pe_rsc_notify);
}
if (xml_contains_remote_node((*rsc)->xml)) {
(*rsc)->is_remote_node = TRUE;
if (g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_CONTAINER)) {
guest_node = TRUE;
} else {
remote_node = TRUE;
}
}
value = g_hash_table_lookup((*rsc)->meta, XML_OP_ATTR_ALLOW_MIGRATE);
#if ENABLE_VERSIONED_ATTRS
has_versioned_params = xml_has_children((*rsc)->versioned_parameters);
#endif
if (crm_is_true(value) && has_versioned_params) {
pe_rsc_trace((*rsc), "Migration is disabled for resources with versioned parameters");
} else if (crm_is_true(value)) {
set_bit((*rsc)->flags, pe_rsc_allow_migrate);
} else if ((value == NULL) && remote_node && !has_versioned_params) {
/* By default, we want remote nodes to be able
* to float around the cluster without having to stop all the
* resources within the remote-node before moving. Allowing
* migration support enables this feature. If this ever causes
* problems, migration support can be explicitly turned off with
* allow-migrate=false.
* We don't support migration for versioned resources, though. */
set_bit((*rsc)->flags, pe_rsc_allow_migrate);
}
value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_MANAGED);
if (value != NULL && safe_str_neq("default", value)) {
- gboolean bool_value = TRUE;
-
- crm_str_to_boolean(value, &bool_value);
- if (bool_value == FALSE) {
- clear_bit((*rsc)->flags, pe_rsc_managed);
- } else {
+ if (crm_is_true(value)) {
set_bit((*rsc)->flags, pe_rsc_managed);
+ } else {
+ clear_bit((*rsc)->flags, pe_rsc_managed);
}
}
value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_MAINTENANCE);
- if (value != NULL && safe_str_neq("default", value)) {
- gboolean bool_value = FALSE;
-
- crm_str_to_boolean(value, &bool_value);
- if (bool_value == TRUE) {
- clear_bit((*rsc)->flags, pe_rsc_managed);
- set_bit((*rsc)->flags, pe_rsc_maintenance);
- }
-
+ if (crm_is_true(value)) {
+ clear_bit((*rsc)->flags, pe_rsc_managed);
+ set_bit((*rsc)->flags, pe_rsc_maintenance);
}
if (is_set(data_set->flags, pe_flag_maintenance_mode)) {
clear_bit((*rsc)->flags, pe_rsc_managed);
set_bit((*rsc)->flags, pe_rsc_maintenance);
}
if (pe_rsc_is_clone(uber_parent(*rsc))) {
value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_UNIQUE);
if (crm_is_true(value)) {
set_bit((*rsc)->flags, pe_rsc_unique);
}
if (detect_promotable(*rsc)) {
set_bit((*rsc)->flags, pe_rsc_promotable);
}
} else {
set_bit((*rsc)->flags, pe_rsc_unique);
}
pe_rsc_trace((*rsc), "Options for %s", (*rsc)->id);
value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_RESTART);
if (safe_str_eq(value, "restart")) {
(*rsc)->restart_type = pe_restart_restart;
pe_rsc_trace((*rsc), "\tDependency restart handling: restart");
pe_warn_once(pe_wo_restart_type,
"Support for restart-type is deprecated and will be removed in a future release");
} else {
(*rsc)->restart_type = pe_restart_ignore;
pe_rsc_trace((*rsc), "\tDependency restart handling: ignore");
}
value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_MULTIPLE);
if (safe_str_eq(value, "stop_only")) {
(*rsc)->recovery_type = recovery_stop_only;
pe_rsc_trace((*rsc), "\tMultiple running resource recovery: stop only");
} else if (safe_str_eq(value, "block")) {
(*rsc)->recovery_type = recovery_block;
pe_rsc_trace((*rsc), "\tMultiple running resource recovery: block");
} else {
(*rsc)->recovery_type = recovery_stop_start;
pe_rsc_trace((*rsc), "\tMultiple running resource recovery: stop/start");
}
value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_STICKINESS);
if (value != NULL && safe_str_neq("default", value)) {
(*rsc)->stickiness = char2score(value);
}
value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_FAIL_STICKINESS);
if (value != NULL && safe_str_neq("default", value)) {
(*rsc)->migration_threshold = char2score(value);
if ((*rsc)->migration_threshold < 0) {
/* @TODO We use 1 here to preserve previous behavior, but this
* should probably use the default (INFINITY) or 0 (to disable)
* instead.
*/
pe_warn_once(pe_wo_neg_threshold,
XML_RSC_ATTR_FAIL_STICKINESS
" must be non-negative, using 1 instead");
(*rsc)->migration_threshold = 1;
}
}
if (safe_str_eq(rclass, PCMK_RESOURCE_CLASS_STONITH)) {
set_bit(data_set->flags, pe_flag_have_stonith_resource);
set_bit((*rsc)->flags, pe_rsc_fence_device);
}
value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_REQUIRES);
handle_requires_pref:
if (safe_str_eq(value, "nothing")) {
} else if (safe_str_eq(value, "quorum")) {
set_bit((*rsc)->flags, pe_rsc_needs_quorum);
} else if (safe_str_eq(value, "unfencing")) {
if (is_set((*rsc)->flags, pe_rsc_fence_device)) {
pcmk__config_warn("Resetting '" XML_RSC_ATTR_REQUIRES "' for %s "
"to 'quorum' because fencing devices cannot "
"require unfencing", (*rsc)->id);
value = "quorum";
isdefault = TRUE;
goto handle_requires_pref;
} else if (is_not_set(data_set->flags, pe_flag_stonith_enabled)) {
pcmk__config_warn("Resetting '" XML_RSC_ATTR_REQUIRES "' for %s "
"to 'quorum' because fencing is disabled",
(*rsc)->id);
value = "quorum";
isdefault = TRUE;
goto handle_requires_pref;
} else {
set_bit((*rsc)->flags, pe_rsc_needs_fencing);
set_bit((*rsc)->flags, pe_rsc_needs_unfencing);
}
} else if (safe_str_eq(value, "fencing")) {
set_bit((*rsc)->flags, pe_rsc_needs_fencing);
if (is_not_set(data_set->flags, pe_flag_stonith_enabled)) {
pcmk__config_warn("%s requires fencing but fencing is disabled",
(*rsc)->id);
}
} else {
const char *orig_value = value;
isdefault = TRUE;
if(is_set((*rsc)->flags, pe_rsc_fence_device)) {
value = "quorum";
} else if (((*rsc)->variant == pe_native)
&& safe_str_eq(crm_element_value((*rsc)->xml, XML_AGENT_ATTR_CLASS),
PCMK_RESOURCE_CLASS_OCF)
&& safe_str_eq(crm_element_value((*rsc)->xml, XML_AGENT_ATTR_PROVIDER), "pacemaker")
&& safe_str_eq(crm_element_value((*rsc)->xml, XML_ATTR_TYPE), "remote")
) {
value = "quorum";
} else if (is_set(data_set->flags, pe_flag_enable_unfencing)) {
value = "unfencing";
} else if (is_set(data_set->flags, pe_flag_stonith_enabled)) {
value = "fencing";
} else if (data_set->no_quorum_policy == no_quorum_ignore) {
value = "nothing";
} else {
value = "quorum";
}
if (orig_value != NULL) {
pcmk__config_err("Resetting '" XML_RSC_ATTR_REQUIRES "' for %s "
"to '%s' because '%s' is not valid",
(*rsc)->id, value, orig_value);
}
goto handle_requires_pref;
}
pe_rsc_trace((*rsc), "\tRequired to start: %s%s", value, isdefault?" (default)":"");
value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_FAIL_TIMEOUT);
if (value != NULL) {
// Stored as seconds
(*rsc)->failure_timeout = (int) (crm_parse_interval_spec(value) / 1000);
}
if (remote_node) {
value = g_hash_table_lookup((*rsc)->parameters, XML_REMOTE_ATTR_RECONNECT_INTERVAL);
if (value) {
/* reconnect delay works by setting failure_timeout and preventing the
* connection from starting until the failure is cleared. */
(*rsc)->remote_reconnect_ms = crm_parse_interval_spec(value);
/* we want to override any default failure_timeout in use when remote
* reconnect_interval is in use. */
(*rsc)->failure_timeout = (*rsc)->remote_reconnect_ms / 1000;
}
}
get_target_role(*rsc, &((*rsc)->next_role));
pe_rsc_trace((*rsc), "\tDesired next state: %s",
(*rsc)->next_role != RSC_ROLE_UNKNOWN ? role2text((*rsc)->next_role) : "default");
if ((*rsc)->fns->unpack(*rsc, data_set) == FALSE) {
return FALSE;
}
if (is_set(data_set->flags, pe_flag_symmetric_cluster)) {
// This tag must stay exactly the same because it is tested elsewhere
resource_location(*rsc, NULL, 0, "symmetric_default", data_set);
} else if (guest_node) {
/* remote resources tied to a container resource must always be allowed
* to opt-in to the cluster. Whether the connection resource is actually
* allowed to be placed on a node is dependent on the container resource */
resource_location(*rsc, NULL, 0, "remote_connection_default", data_set);
}
pe_rsc_trace((*rsc), "\tAction notification: %s",
is_set((*rsc)->flags, pe_rsc_notify) ? "required" : "not required");
(*rsc)->utilization = crm_str_table_new();
pe__unpack_dataset_nvpairs((*rsc)->xml, XML_TAG_UTILIZATION, NULL,
(*rsc)->utilization, NULL, FALSE, data_set);
/* data_set->resources = g_list_append(data_set->resources, (*rsc)); */
if (expanded_xml) {
if (add_template_rsc(xml_obj, data_set) == FALSE) {
return FALSE;
}
}
return TRUE;
}
void
common_update_score(pe_resource_t * rsc, const char *id, int score)
{
pe_node_t *node = NULL;
node = pe_hash_table_lookup(rsc->allowed_nodes, id);
if (node != NULL) {
pe_rsc_trace(rsc, "Updating score for %s on %s: %d + %d", rsc->id, id, node->weight, score);
node->weight = merge_weights(node->weight, score);
}
if (rsc->children) {
GListPtr gIter = rsc->children;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
common_update_score(child_rsc, id, score);
}
}
}
gboolean
is_parent(pe_resource_t *child, pe_resource_t *rsc)
{
pe_resource_t *parent = child;
if (parent == NULL || rsc == NULL) {
return FALSE;
}
while (parent->parent != NULL) {
if (parent->parent == rsc) {
return TRUE;
}
parent = parent->parent;
}
return FALSE;
}
pe_resource_t *
uber_parent(pe_resource_t * rsc)
{
pe_resource_t *parent = rsc;
if (parent == NULL) {
return NULL;
}
while (parent->parent != NULL && parent->parent->variant != pe_container) {
parent = parent->parent;
}
return parent;
}
void
common_free(pe_resource_t * rsc)
{
if (rsc == NULL) {
return;
}
pe_rsc_trace(rsc, "Freeing %s %d", rsc->id, rsc->variant);
g_list_free(rsc->rsc_cons);
g_list_free(rsc->rsc_cons_lhs);
g_list_free(rsc->rsc_tickets);
g_list_free(rsc->dangling_migrations);
if (rsc->parameters != NULL) {
g_hash_table_destroy(rsc->parameters);
}
#if ENABLE_VERSIONED_ATTRS
if (rsc->versioned_parameters != NULL) {
free_xml(rsc->versioned_parameters);
}
#endif
if (rsc->meta != NULL) {
g_hash_table_destroy(rsc->meta);
}
if (rsc->utilization != NULL) {
g_hash_table_destroy(rsc->utilization);
}
if (rsc->parent == NULL && is_set(rsc->flags, pe_rsc_orphan)) {
free_xml(rsc->xml);
rsc->xml = NULL;
free_xml(rsc->orig_xml);
rsc->orig_xml = NULL;
/* if rsc->orig_xml, then rsc->xml is an expanded xml from a template */
} else if (rsc->orig_xml) {
free_xml(rsc->xml);
rsc->xml = NULL;
}
if (rsc->running_on) {
g_list_free(rsc->running_on);
rsc->running_on = NULL;
}
if (rsc->known_on) {
g_hash_table_destroy(rsc->known_on);
rsc->known_on = NULL;
}
if (rsc->actions) {
g_list_free(rsc->actions);
rsc->actions = NULL;
}
if (rsc->allowed_nodes) {
g_hash_table_destroy(rsc->allowed_nodes);
rsc->allowed_nodes = NULL;
}
g_list_free(rsc->fillers);
g_list_free(rsc->rsc_location);
pe_rsc_trace(rsc, "Resource freed");
free(rsc->id);
free(rsc->clone_name);
free(rsc->allocated_to);
free(rsc->variant_opaque);
free(rsc->pending_task);
free(rsc);
}
/*!
* \brief
* \internal Find a node (and optionally count all) where resource is active
*
* \param[in] rsc Resource to check
* \param[out] count_all If not NULL, will be set to count of active nodes
* \param[out] count_clean If not NULL, will be set to count of clean nodes
*
* \return An active node (or NULL if resource is not active anywhere)
*
* \note The order of preference is: an active node that is the resource's
* partial migration source; if the resource's "requires" is "quorum" or
* "nothing", the first active node in the list that is clean and online;
* the first active node in the list.
*/
pe_node_t *
pe__find_active_on(const pe_resource_t *rsc, unsigned int *count_all,
unsigned int *count_clean)
{
pe_node_t *active = NULL;
pe_node_t *node = NULL;
bool keep_looking = FALSE;
bool is_happy = FALSE;
if (count_all) {
*count_all = 0;
}
if (count_clean) {
*count_clean = 0;
}
if (rsc == NULL) {
return NULL;
}
for (GList *node_iter = rsc->running_on; node_iter != NULL;
node_iter = node_iter->next) {
node = node_iter->data;
keep_looking = FALSE;
is_happy = node->details->online && !node->details->unclean;
if (count_all) {
++*count_all;
}
if (count_clean && is_happy) {
++*count_clean;
}
if (count_all || count_clean) {
// If we're counting, we need to go through entire list
keep_looking = TRUE;
}
if (rsc->partial_migration_source != NULL) {
if (node->details == rsc->partial_migration_source->details) {
// This is the migration source
active = node;
} else {
keep_looking = TRUE;
}
} else if (is_not_set(rsc->flags, pe_rsc_needs_fencing)) {
if (is_happy && (!active || !active->details->online
|| active->details->unclean)) {
// This is the first clean node
active = node;
} else {
keep_looking = TRUE;
}
}
if (active == NULL) {
// This is first node in list
active = node;
}
if (keep_looking == FALSE) {
// Don't waste time iterating if we don't have to
break;
}
}
return active;
}
/*!
* \brief
* \internal Find and count active nodes according to "requires"
*
* \param[in] rsc Resource to check
* \param[out] count If not NULL, will be set to count of active nodes
*
* \return An active node (or NULL if resource is not active anywhere)
*
* \note This is a convenience wrapper for pe__find_active_on() where the count
* of all active nodes or only clean active nodes is desired according to
* the "requires" meta-attribute.
*/
pe_node_t *
pe__find_active_requires(const pe_resource_t *rsc, unsigned int *count)
{
if (rsc && is_not_set(rsc->flags, pe_rsc_needs_fencing)) {
return pe__find_active_on(rsc, NULL, count);
}
return pe__find_active_on(rsc, count, NULL);
}
void
pe__count_common(pe_resource_t *rsc)
{
if (rsc->children != NULL) {
for (GList *item = rsc->children; item != NULL; item = item->next) {
((pe_resource_t *) item->data)->fns->count(item->data);
}
} else if (is_not_set(rsc->flags, pe_rsc_orphan)
|| (rsc->role > RSC_ROLE_STOPPED)) {
rsc->cluster->ninstances++;
if (pe__resource_is_disabled(rsc)) {
rsc->cluster->disabled_resources++;
}
if (is_set(rsc->flags, pe_rsc_block)) {
rsc->cluster->blocked_resources++;
}
}
}
diff --git a/lib/pengine/group.c b/lib/pengine/group.c
index c61b3b2600..24106a0abd 100644
--- a/lib/pengine/group.c
+++ b/lib/pengine/group.c
@@ -1,289 +1,289 @@
/*
* Copyright 2004-2020 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 <crm/pengine/rules.h>
#include <crm/pengine/status.h>
#include <crm/pengine/internal.h>
#include <crm/msg_xml.h>
#include <pe_status_private.h>
#define VARIANT_GROUP 1
#include "./variant.h"
gboolean
group_unpack(pe_resource_t * rsc, pe_working_set_t * data_set)
{
xmlNode *xml_obj = rsc->xml;
xmlNode *xml_native_rsc = NULL;
group_variant_data_t *group_data = NULL;
const char *group_ordered = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_ORDERED);
const char *group_colocated = g_hash_table_lookup(rsc->meta, "collocated");
const char *clone_id = NULL;
pe_rsc_trace(rsc, "Processing resource %s...", rsc->id);
group_data = calloc(1, sizeof(group_variant_data_t));
group_data->num_children = 0;
group_data->first_child = NULL;
group_data->last_child = NULL;
rsc->variant_opaque = group_data;
- group_data->ordered = TRUE;
- group_data->colocated = TRUE;
-
- if (group_ordered != NULL) {
- crm_str_to_boolean(group_ordered, &(group_data->ordered));
+ // We don't actually need the null checks but it speeds up the common case
+ if ((group_ordered == NULL)
+ || (crm_str_to_boolean(group_ordered, &(group_data->ordered)) < 0)) {
+ group_data->ordered = TRUE;
}
- if (group_colocated != NULL) {
- crm_str_to_boolean(group_colocated, &(group_data->colocated));
+ if ((group_colocated == NULL)
+ || (crm_str_to_boolean(group_colocated, &(group_data->colocated)) < 0)) {
+ group_data->colocated = TRUE;
}
clone_id = crm_element_value(rsc->xml, XML_RSC_ATTR_INCARNATION);
for (xml_native_rsc = __xml_first_child_element(xml_obj); xml_native_rsc != NULL;
xml_native_rsc = __xml_next_element(xml_native_rsc)) {
if (crm_str_eq((const char *)xml_native_rsc->name, XML_CIB_TAG_RESOURCE, TRUE)) {
pe_resource_t *new_rsc = NULL;
crm_xml_add(xml_native_rsc, XML_RSC_ATTR_INCARNATION, clone_id);
if (common_unpack(xml_native_rsc, &new_rsc, rsc, data_set) == FALSE) {
pe_err("Failed unpacking resource %s", crm_element_value(xml_obj, XML_ATTR_ID));
if (new_rsc != NULL && new_rsc->fns != NULL) {
new_rsc->fns->free(new_rsc);
}
continue;
}
group_data->num_children++;
rsc->children = g_list_append(rsc->children, new_rsc);
if (group_data->first_child == NULL) {
group_data->first_child = new_rsc;
}
group_data->last_child = new_rsc;
pe_rsc_trace(rsc, "Added %s member %s", rsc->id, new_rsc->id);
}
}
if (group_data->num_children == 0) {
pcmk__config_warn("Group %s does not have any children", rsc->id);
return TRUE; // Allow empty groups, children can be added later
}
pe_rsc_trace(rsc, "Added %d children to resource %s...", group_data->num_children, rsc->id);
return TRUE;
}
gboolean
group_active(pe_resource_t * rsc, gboolean all)
{
gboolean c_all = TRUE;
gboolean c_any = FALSE;
GListPtr gIter = rsc->children;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
if (child_rsc->fns->active(child_rsc, all)) {
c_any = TRUE;
} else {
c_all = FALSE;
}
}
if (c_any == FALSE) {
return FALSE;
} else if (all && c_all == FALSE) {
return FALSE;
}
return TRUE;
}
static void
group_print_xml(pe_resource_t * rsc, const char *pre_text, long options, void *print_data)
{
GListPtr gIter = rsc->children;
char *child_text = crm_strdup_printf("%s ", pre_text);
status_print("%s<group id=\"%s\" ", pre_text, rsc->id);
status_print("number_resources=\"%d\" ", g_list_length(rsc->children));
status_print(">\n");
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
child_rsc->fns->print(child_rsc, child_text, options, print_data);
}
status_print("%s</group>\n", pre_text);
free(child_text);
}
void
group_print(pe_resource_t * rsc, const char *pre_text, long options, void *print_data)
{
char *child_text = NULL;
GListPtr gIter = rsc->children;
if (pre_text == NULL) {
pre_text = " ";
}
if (options & pe_print_xml) {
group_print_xml(rsc, pre_text, options, print_data);
return;
}
child_text = crm_strdup_printf("%s ", pre_text);
status_print("%sResource Group: %s", pre_text ? pre_text : "", rsc->id);
if (options & pe_print_html) {
status_print("\n<ul>\n");
} else if ((options & pe_print_log) == 0) {
status_print("\n");
}
if (options & pe_print_brief) {
print_rscs_brief(rsc->children, child_text, options, print_data, TRUE);
} else {
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
if (options & pe_print_html) {
status_print("<li>\n");
}
child_rsc->fns->print(child_rsc, child_text, options, print_data);
if (options & pe_print_html) {
status_print("</li>\n");
}
}
}
if (options & pe_print_html) {
status_print("</ul>\n");
}
free(child_text);
}
int
pe__group_xml(pcmk__output_t *out, va_list args)
{
unsigned int options = va_arg(args, unsigned int);
pe_resource_t *rsc = va_arg(args, pe_resource_t *);
GListPtr gIter = rsc->children;
char *count = crm_itoa(g_list_length(gIter));
int rc = pe__name_and_nvpairs_xml(out, true, "group", 2
, "id", rsc->id
, "number_resources", count);
free(count);
CRM_ASSERT(rc == pcmk_rc_ok);
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
out->message(out, crm_map_element_name(child_rsc->xml), options, child_rsc);
}
pcmk__output_xml_pop_parent(out);
return rc;
}
int
pe__group_html(pcmk__output_t *out, va_list args)
{
unsigned int options = va_arg(args, unsigned int);
pe_resource_t *rsc = va_arg(args, pe_resource_t *);
out->begin_list(out, NULL, NULL, "Resource Group: %s", rsc->id);
if (options & pe_print_brief) {
pe__rscs_brief_output(out, rsc->children, options, TRUE);
} else {
for (GListPtr gIter = rsc->children; gIter; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
out->message(out, crm_map_element_name(child_rsc->xml), options, child_rsc);
}
}
out->end_list(out);
return pcmk_rc_ok;
}
int
pe__group_text(pcmk__output_t *out, va_list args)
{
unsigned int options = va_arg(args, unsigned int);
pe_resource_t *rsc = va_arg(args, pe_resource_t *);
out->begin_list(out, NULL, NULL, "Resource Group: %s", rsc->id);
if (options & pe_print_brief) {
pe__rscs_brief_output(out, rsc->children, options, TRUE);
} else {
for (GListPtr gIter = rsc->children; gIter; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
out->message(out, crm_map_element_name(child_rsc->xml), options, child_rsc);
}
}
out->end_list(out);
return pcmk_rc_ok;
}
void
group_free(pe_resource_t * rsc)
{
CRM_CHECK(rsc != NULL, return);
pe_rsc_trace(rsc, "Freeing %s", rsc->id);
for (GListPtr gIter = rsc->children; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
CRM_ASSERT(child_rsc);
pe_rsc_trace(child_rsc, "Freeing child %s", child_rsc->id);
child_rsc->fns->free(child_rsc);
}
pe_rsc_trace(rsc, "Freeing child list");
g_list_free(rsc->children);
common_free(rsc);
}
enum rsc_role_e
group_resource_state(const pe_resource_t * rsc, gboolean current)
{
enum rsc_role_e group_role = RSC_ROLE_UNKNOWN;
GListPtr gIter = rsc->children;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
enum rsc_role_e role = child_rsc->fns->state(child_rsc, current);
if (role > group_role) {
group_role = role;
}
}
pe_rsc_trace(rsc, "%s role: %s", rsc->id, role2text(group_role));
return group_role;
}
diff --git a/lib/pengine/pe_output.c b/lib/pengine/pe_output.c
index c585b3b968..6820ac00d3 100644
--- a/lib/pengine/pe_output.c
+++ b/lib/pengine/pe_output.c
@@ -1,1545 +1,1544 @@
/*
* Copyright 2019-2020 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 <crm/common/iso8601_internal.h>
#include <crm/msg_xml.h>
#include <crm/pengine/internal.h>
#define SUMMARY_HEADER(rc, out) do { \
if (rc == pcmk_rc_no_output) { \
out->begin_list(out, NULL, NULL, "Cluster Summary"); \
rc = pcmk_rc_ok; \
} \
} while (0)
static char *
failed_action_string(xmlNodePtr xml_op) {
const char *op_key = crm_element_value(xml_op, XML_LRM_ATTR_TASK_KEY);
int rc = crm_parse_int(crm_element_value(xml_op, XML_LRM_ATTR_RC), "0");
int status = crm_parse_int(crm_element_value(xml_op, XML_LRM_ATTR_OPSTATUS), "0");
const char *exit_reason = crm_element_value(xml_op, XML_LRM_ATTR_EXIT_REASON);
time_t last_change = 0;
if (crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE,
&last_change) == pcmk_ok) {
crm_time_t *crm_when = crm_time_new(NULL);
char *time_s = NULL;
char *buf = NULL;
crm_time_set_timet(crm_when, &last_change);
time_s = crm_time_as_string(crm_when, crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone);
buf = crm_strdup_printf("%s on %s '%s' (%d): call=%s, status='%s', "
"exitreason='%s', " XML_RSC_OP_LAST_CHANGE
"='%s', queued=%sms, exec=%sms",
op_key ? op_key : ID(xml_op),
crm_element_value(xml_op, XML_ATTR_UNAME),
services_ocf_exitcode_str(rc), rc,
crm_element_value(xml_op, XML_LRM_ATTR_CALLID),
services_lrm_status_str(status),
exit_reason ? exit_reason : "none",
time_s,
crm_element_value(xml_op, XML_RSC_OP_T_QUEUE),
crm_element_value(xml_op, XML_RSC_OP_T_EXEC));
crm_time_free(crm_when);
free(time_s);
return buf;
} else {
return crm_strdup_printf("%s on %s '%s' (%d): call=%s, status=%s, exitreason='%s'",
op_key ? op_key : ID(xml_op),
crm_element_value(xml_op, XML_ATTR_UNAME),
services_ocf_exitcode_str(rc), rc,
crm_element_value(xml_op, XML_LRM_ATTR_CALLID),
services_lrm_status_str(status),
exit_reason ? exit_reason : "none");
}
}
static const char *
get_cluster_stack(pe_working_set_t *data_set)
{
xmlNode *stack = get_xpath_object("//nvpair[@name='cluster-infrastructure']",
data_set->input, LOG_DEBUG);
return stack? crm_element_value(stack, XML_NVPAIR_ATTR_VALUE) : "unknown";
}
static char *
last_changed_string(const char *last_written, const char *user,
const char *client, const char *origin) {
if (last_written != NULL || user != NULL || client != NULL || origin != NULL) {
return crm_strdup_printf("%s%s%s%s%s%s%s",
last_written ? last_written : "",
user ? " by " : "",
user ? user : "",
client ? " via " : "",
client ? client : "",
origin ? " on " : "",
origin ? origin : "");
} else {
return strdup("");
}
}
static char *
op_history_string(xmlNode *xml_op, const char *task, const char *interval_ms_s,
int rc, gboolean print_timing) {
const char *call = crm_element_value(xml_op, XML_LRM_ATTR_CALLID);
char *interval_str = NULL;
char *buf = NULL;
if (interval_ms_s && safe_str_neq(interval_ms_s, "0")) {
char *pair = pcmk_format_nvpair("interval", interval_ms_s, "ms");
interval_str = crm_strdup_printf(" %s", pair);
free(pair);
}
if (print_timing) {
char *last_change_str = NULL;
char *last_run_str = NULL;
char *exec_str = NULL;
char *queue_str = NULL;
const char *value = NULL;
time_t epoch = 0;
if ((crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &epoch) == pcmk_ok)
&& (epoch > 0)) {
char *time = pcmk_format_named_time(XML_RSC_OP_LAST_CHANGE, epoch);
last_change_str = crm_strdup_printf(" %s", time);
free(time);
}
if ((crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_RUN, &epoch) == pcmk_ok)
&& (epoch > 0)) {
char *time = pcmk_format_named_time(XML_RSC_OP_LAST_RUN, epoch);
last_run_str = crm_strdup_printf(" %s", time);
free(time);
}
value = crm_element_value(xml_op, XML_RSC_OP_T_EXEC);
if (value) {
char *pair = pcmk_format_nvpair(XML_RSC_OP_T_EXEC, value, "ms");
exec_str = crm_strdup_printf(" %s", pair);
free(pair);
}
value = crm_element_value(xml_op, XML_RSC_OP_T_QUEUE);
if (value) {
char *pair = pcmk_format_nvpair(XML_RSC_OP_T_QUEUE, value, "ms");
queue_str = crm_strdup_printf(" %s", pair);
free(pair);
}
buf = crm_strdup_printf("(%s) %s:%s%s%s%s%s rc=%d (%s)", call, task,
interval_str ? interval_str : "",
last_change_str ? last_change_str : "",
last_run_str ? last_run_str : "",
exec_str ? exec_str : "",
queue_str ? queue_str : "",
rc, services_ocf_exitcode_str(rc));
if (last_change_str) {
free(last_change_str);
}
if (last_run_str) {
free(last_run_str);
}
if (exec_str) {
free(exec_str);
}
if (queue_str) {
free(queue_str);
}
} else {
buf = crm_strdup_printf("(%s) %s%s%s", call, task,
interval_str ? ":" : "",
interval_str ? interval_str : "");
}
if (interval_str) {
free(interval_str);
}
return buf;
}
static char *
resource_history_string(pe_resource_t *rsc, const char *rsc_id, gboolean all,
int failcount, time_t last_failure) {
char *buf = NULL;
if (rsc == NULL) {
buf = crm_strdup_printf("%s: orphan", rsc_id);
} else if (all || failcount || last_failure > 0) {
char *failcount_s = NULL;
char *lastfail_s = NULL;
if (failcount > 0) {
failcount_s = crm_strdup_printf(" %s=%d", PCMK__FAIL_COUNT_PREFIX,
failcount);
} else {
failcount_s = strdup("");
}
if (last_failure > 0) {
lastfail_s = crm_strdup_printf(" %s='%s'",
PCMK__LAST_FAILURE_PREFIX,
pcmk__epoch2str(&last_failure));
}
buf = crm_strdup_printf("%s: migration-threshold=%d%s%s",
rsc_id, rsc->migration_threshold, failcount_s,
lastfail_s? lastfail_s : "");
free(failcount_s);
free(lastfail_s);
} else {
buf = crm_strdup_printf("%s:", rsc_id);
}
return buf;
}
int
pe__cluster_summary(pcmk__output_t *out, va_list args) {
pe_working_set_t *data_set = va_arg(args, pe_working_set_t *);
gboolean print_clone_detail = va_arg(args, gboolean);
gboolean show_stack = va_arg(args, gboolean);
gboolean show_dc = va_arg(args, gboolean);
gboolean show_times = va_arg(args, gboolean);
gboolean show_counts = va_arg(args, gboolean);
gboolean show_options = va_arg(args, gboolean);
int rc = pcmk_rc_no_output;
const char *stack_s = get_cluster_stack(data_set);
if (show_stack) {
SUMMARY_HEADER(rc, out);
out->message(out, "cluster-stack", stack_s);
}
/* Always print DC if none, even if not requested */
if (data_set->dc_node == NULL || show_dc) {
xmlNode *dc_version = get_xpath_object("//nvpair[@name='dc-version']",
data_set->input, LOG_DEBUG);
const char *dc_version_s = dc_version?
crm_element_value(dc_version, XML_NVPAIR_ATTR_VALUE)
: NULL;
const char *quorum = crm_element_value(data_set->input, XML_ATTR_HAVE_QUORUM);
char *dc_name = data_set->dc_node ? pe__node_display_name(data_set->dc_node, print_clone_detail) : NULL;
SUMMARY_HEADER(rc, out);
out->message(out, "cluster-dc", data_set->dc_node, quorum, dc_version_s, dc_name);
free(dc_name);
}
if (show_times) {
const char *last_written = crm_element_value(data_set->input, XML_CIB_ATTR_WRITTEN);
const char *user = crm_element_value(data_set->input, XML_ATTR_UPDATE_USER);
const char *client = crm_element_value(data_set->input, XML_ATTR_UPDATE_CLIENT);
const char *origin = crm_element_value(data_set->input, XML_ATTR_UPDATE_ORIG);
SUMMARY_HEADER(rc, out);
out->message(out, "cluster-times", last_written, user, client, origin);
}
if (show_counts) {
SUMMARY_HEADER(rc, out);
out->message(out, "cluster-counts", g_list_length(data_set->nodes),
data_set->ninstances, data_set->disabled_resources,
data_set->blocked_resources);
}
if (show_options) {
SUMMARY_HEADER(rc, out);
out->message(out, "cluster-options", data_set);
}
if (rc == pcmk_rc_ok) {
out->end_list(out);
}
if (is_set(data_set->flags, pe_flag_maintenance_mode)) {
out->message(out, "maint-mode");
rc = pcmk_rc_ok;
}
return rc;
}
int
pe__cluster_summary_html(pcmk__output_t *out, va_list args) {
pe_working_set_t *data_set = va_arg(args, pe_working_set_t *);
gboolean print_clone_detail = va_arg(args, gboolean);
gboolean show_stack = va_arg(args, gboolean);
gboolean show_dc = va_arg(args, gboolean);
gboolean show_times = va_arg(args, gboolean);
gboolean show_counts = va_arg(args, gboolean);
gboolean show_options = va_arg(args, gboolean);
int rc = pcmk_rc_no_output;
const char *stack_s = get_cluster_stack(data_set);
if (show_stack) {
SUMMARY_HEADER(rc, out);
out->message(out, "cluster-stack", stack_s);
}
/* Always print DC if none, even if not requested */
if (data_set->dc_node == NULL || show_dc) {
xmlNode *dc_version = get_xpath_object("//nvpair[@name='dc-version']",
data_set->input, LOG_DEBUG);
const char *dc_version_s = dc_version?
crm_element_value(dc_version, XML_NVPAIR_ATTR_VALUE)
: NULL;
const char *quorum = crm_element_value(data_set->input, XML_ATTR_HAVE_QUORUM);
char *dc_name = data_set->dc_node ? pe__node_display_name(data_set->dc_node, print_clone_detail) : NULL;
SUMMARY_HEADER(rc, out);
out->message(out, "cluster-dc", data_set->dc_node, quorum, dc_version_s, dc_name);
free(dc_name);
}
if (show_times) {
const char *last_written = crm_element_value(data_set->input, XML_CIB_ATTR_WRITTEN);
const char *user = crm_element_value(data_set->input, XML_ATTR_UPDATE_USER);
const char *client = crm_element_value(data_set->input, XML_ATTR_UPDATE_CLIENT);
const char *origin = crm_element_value(data_set->input, XML_ATTR_UPDATE_ORIG);
SUMMARY_HEADER(rc, out);
out->message(out, "cluster-times", last_written, user, client, origin);
}
if (show_counts) {
SUMMARY_HEADER(rc, out);
out->message(out, "cluster-counts", g_list_length(data_set->nodes),
data_set->ninstances, data_set->disabled_resources,
data_set->blocked_resources);
}
if (show_options) {
/* Kind of a hack - close the list we may have opened earlier in this
* function so we can put all the options into their own list. We
* only want to do this on HTML output, though.
*/
if (rc == pcmk_rc_ok) {
out->end_list(out);
}
out->begin_list(out, NULL, NULL, "Config Options");
out->message(out, "cluster-options", data_set);
}
if (rc == pcmk_rc_ok) {
out->end_list(out);
}
if (is_set(data_set->flags, pe_flag_maintenance_mode)) {
out->message(out, "maint-mode");
rc = pcmk_rc_ok;
}
return rc;
}
char *
pe__node_display_name(pe_node_t *node, bool print_detail)
{
char *node_name;
const char *node_host = NULL;
const char *node_id = NULL;
int name_len;
CRM_ASSERT((node != NULL) && (node->details != NULL) && (node->details->uname != NULL));
/* Host is displayed only if this is a guest node */
if (pe__is_guest_node(node)) {
pe_node_t *host_node = pe__current_node(node->details->remote_rsc);
if (host_node && host_node->details) {
node_host = host_node->details->uname;
}
if (node_host == NULL) {
node_host = ""; /* so we at least get "uname@" to indicate guest */
}
}
/* Node ID is displayed if different from uname and detail is requested */
if (print_detail && safe_str_neq(node->details->uname, node->details->id)) {
node_id = node->details->id;
}
/* Determine name length */
name_len = strlen(node->details->uname) + 1;
if (node_host) {
name_len += strlen(node_host) + 1; /* "@node_host" */
}
if (node_id) {
name_len += strlen(node_id) + 3; /* + " (node_id)" */
}
/* Allocate and populate display name */
node_name = malloc(name_len);
CRM_ASSERT(node_name != NULL);
strcpy(node_name, node->details->uname);
if (node_host) {
strcat(node_name, "@");
strcat(node_name, node_host);
}
if (node_id) {
strcat(node_name, " (");
strcat(node_name, node_id);
strcat(node_name, ")");
}
return node_name;
}
int
pe__name_and_nvpairs_xml(pcmk__output_t *out, bool is_list, const char *tag_name
, size_t pairs_count, ...)
{
xmlNodePtr xml_node = NULL;
va_list args;
CRM_ASSERT(tag_name != NULL);
xml_node = pcmk__output_xml_peek_parent(out);
CRM_ASSERT(xml_node != NULL);
xml_node = is_list
? create_xml_node(xml_node, tag_name)
: xmlNewChild(xml_node, NULL, (pcmkXmlStr) tag_name, NULL);
va_start(args, pairs_count);
while(pairs_count--) {
const char *param_name = va_arg(args, const char *);
const char *param_value = va_arg(args, const char *);
if (param_name && param_value) {
xmlSetProp(xml_node, (pcmkXmlStr)param_name, (pcmkXmlStr)param_value);
}
};
va_end(args);
if (is_list) {
pcmk__output_xml_push_parent(out, xml_node);
}
return pcmk_rc_ok;
}
int
pe__ban_html(pcmk__output_t *out, va_list args) {
pe_node_t *pe_node = va_arg(args, pe_node_t *);
pe__location_t *location = va_arg(args, pe__location_t *);
gboolean print_clone_detail = va_arg(args, gboolean);
char *node_name = pe__node_display_name(pe_node, print_clone_detail);
char *buf = crm_strdup_printf("%s\tprevents %s from running %son %s",
location->id, location->rsc_lh->id,
location->role_filter == RSC_ROLE_MASTER ? "as Master " : "",
node_name);
pcmk__output_create_html_node(out, "li", NULL, NULL, buf);
free(node_name);
free(buf);
return pcmk_rc_ok;
}
int
pe__ban_text(pcmk__output_t *out, va_list args) {
pe_node_t *pe_node = va_arg(args, pe_node_t *);
pe__location_t *location = va_arg(args, pe__location_t *);
gboolean print_clone_detail = va_arg(args, gboolean);
char *node_name = pe__node_display_name(pe_node, print_clone_detail);
out->list_item(out, NULL, "%s\tprevents %s from running %son %s",
location->id, location->rsc_lh->id,
location->role_filter == RSC_ROLE_MASTER ? "as Master " : "",
node_name);
free(node_name);
return pcmk_rc_ok;
}
int
pe__ban_xml(pcmk__output_t *out, va_list args) {
xmlNodePtr node = pcmk__output_create_xml_node(out, "ban");
pe_node_t *pe_node = va_arg(args, pe_node_t *);
pe__location_t *location = va_arg(args, pe__location_t *);
gboolean print_clone_detail G_GNUC_UNUSED = va_arg(args, gboolean);
char *weight_s = crm_itoa(pe_node->weight);
xmlSetProp(node, (pcmkXmlStr) "id", (pcmkXmlStr) location->id);
xmlSetProp(node, (pcmkXmlStr) "resource", (pcmkXmlStr) location->rsc_lh->id);
xmlSetProp(node, (pcmkXmlStr) "node", (pcmkXmlStr) pe_node->details->uname);
xmlSetProp(node, (pcmkXmlStr) "weight", (pcmkXmlStr) weight_s);
xmlSetProp(node, (pcmkXmlStr) "master_only",
(pcmkXmlStr) (location->role_filter == RSC_ROLE_MASTER ? "true" : "false"));
free(weight_s);
return pcmk_rc_ok;
}
int
pe__cluster_counts_html(pcmk__output_t *out, va_list args) {
xmlNodePtr nodes_node = pcmk__output_create_xml_node(out, "li");
xmlNodePtr resources_node = pcmk__output_create_xml_node(out, "li");
unsigned int nnodes = va_arg(args, unsigned int);
int nresources = va_arg(args, int);
int ndisabled = va_arg(args, int);
int nblocked = va_arg(args, int);
char *nnodes_str = crm_strdup_printf("%d node%s configured",
nnodes, pcmk__plural_s(nnodes));
pcmk_create_html_node(nodes_node, "span", NULL, NULL, nnodes_str);
free(nnodes_str);
if (ndisabled && nblocked) {
char *s = crm_strdup_printf("%d resource instance%s configured (%d ",
nresources, pcmk__plural_s(nresources),
ndisabled);
pcmk_create_html_node(resources_node, "span", NULL, NULL, s);
free(s);
pcmk_create_html_node(resources_node, "span", NULL, "bold", "DISABLED");
s = crm_strdup_printf(", %d ", nblocked);
pcmk_create_html_node(resources_node, "span", NULL, NULL, s);
free(s);
pcmk_create_html_node(resources_node, "span", NULL, "bold", "BLOCKED");
pcmk_create_html_node(resources_node, "span", NULL, NULL,
" from further action due to failure)");
} else if (ndisabled && !nblocked) {
char *s = crm_strdup_printf("%d resource instance%s configured (%d ",
nresources, pcmk__plural_s(nresources),
ndisabled);
pcmk_create_html_node(resources_node, "span", NULL, NULL, s);
free(s);
pcmk_create_html_node(resources_node, "span", NULL, "bold", "DISABLED");
pcmk_create_html_node(resources_node, "span", NULL, NULL, ")");
} else if (!ndisabled && nblocked) {
char *s = crm_strdup_printf("%d resource instance%s configured (%d ",
nresources, pcmk__plural_s(nresources),
nblocked);
pcmk_create_html_node(resources_node, "span", NULL, NULL, s);
free(s);
pcmk_create_html_node(resources_node, "span", NULL, "bold", "BLOCKED");
pcmk_create_html_node(resources_node, "span", NULL, NULL,
" from further action due to failure)");
} else {
char *s = crm_strdup_printf("%d resource instance%s configured",
nresources, pcmk__plural_s(nresources));
pcmk_create_html_node(resources_node, "span", NULL, NULL, s);
free(s);
}
return pcmk_rc_ok;
}
int
pe__cluster_counts_text(pcmk__output_t *out, va_list args) {
unsigned int nnodes = va_arg(args, unsigned int);
int nresources = va_arg(args, int);
int ndisabled = va_arg(args, int);
int nblocked = va_arg(args, int);
out->list_item(out, NULL, "%d node%s configured",
nnodes, pcmk__plural_s(nnodes));
if (ndisabled && nblocked) {
out->list_item(out, NULL, "%d resource instance%s configured "
"(%d DISABLED, %d BLOCKED from "
"further action due to failure)",
nresources, pcmk__plural_s(nresources), ndisabled,
nblocked);
} else if (ndisabled && !nblocked) {
out->list_item(out, NULL, "%d resource instance%s configured "
"(%d DISABLED)",
nresources, pcmk__plural_s(nresources), ndisabled);
} else if (!ndisabled && nblocked) {
out->list_item(out, NULL, "%d resource instance%s configured "
"(%d BLOCKED from further action "
"due to failure)",
nresources, pcmk__plural_s(nresources), nblocked);
} else {
out->list_item(out, NULL, "%d resource instance%s configured",
nresources, pcmk__plural_s(nresources));
}
return pcmk_rc_ok;
}
int
pe__cluster_counts_xml(pcmk__output_t *out, va_list args) {
xmlNodePtr nodes_node = pcmk__output_create_xml_node(out, "nodes_configured");
xmlNodePtr resources_node = pcmk__output_create_xml_node(out, "resources_configured");
unsigned int nnodes = va_arg(args, unsigned int);
int nresources = va_arg(args, int);
int ndisabled = va_arg(args, int);
int nblocked = va_arg(args, int);
char *s = crm_itoa(nnodes);
xmlSetProp(nodes_node, (pcmkXmlStr) "number", (pcmkXmlStr) s);
free(s);
s = crm_itoa(nresources);
xmlSetProp(resources_node, (pcmkXmlStr) "number", (pcmkXmlStr) s);
free(s);
s = crm_itoa(ndisabled);
xmlSetProp(resources_node, (pcmkXmlStr) "disabled", (pcmkXmlStr) s);
free(s);
s = crm_itoa(nblocked);
xmlSetProp(resources_node, (pcmkXmlStr) "blocked", (pcmkXmlStr) s);
free(s);
return pcmk_rc_ok;
}
int
pe__cluster_dc_html(pcmk__output_t *out, va_list args) {
xmlNodePtr node = pcmk__output_create_xml_node(out, "li");
pe_node_t *dc = va_arg(args, pe_node_t *);
const char *quorum = va_arg(args, const char *);
const char *dc_version_s = va_arg(args, const char *);
char *dc_name = va_arg(args, char *);
pcmk_create_html_node(node, "span", NULL, "bold", "Current DC: ");
if (dc) {
if (crm_is_true(quorum)) {
char *buf = crm_strdup_printf("%s (version %s) - partition with quorum",
dc_name, dc_version_s ? dc_version_s : "unknown");
pcmk_create_html_node(node, "span", NULL, NULL, buf);
free(buf);
} else {
char *buf = crm_strdup_printf("%s (version %s) - partition",
dc_name, dc_version_s ? dc_version_s : "unknown");
pcmk_create_html_node(node, "span", NULL, NULL, buf);
free(buf);
pcmk_create_html_node(node, "span", NULL, "warning", "WITHOUT");
pcmk_create_html_node(node, "span", NULL, NULL, "quorum");
}
} else {
pcmk_create_html_node(node ,"span", NULL, "warning", "NONE");
}
return pcmk_rc_ok;
}
int
pe__cluster_dc_text(pcmk__output_t *out, va_list args) {
pe_node_t *dc = va_arg(args, pe_node_t *);
const char *quorum = va_arg(args, const char *);
const char *dc_version_s = va_arg(args, const char *);
char *dc_name = va_arg(args, char *);
if (dc) {
out->list_item(out, "Current DC", "%s (version %s) - partition %s quorum",
dc_name, dc_version_s ? dc_version_s : "unknown",
crm_is_true(quorum) ? "with" : "WITHOUT");
} else {
out->list_item(out, "Current DC", "NONE");
}
return pcmk_rc_ok;
}
int
pe__cluster_dc_xml(pcmk__output_t *out, va_list args) {
xmlNodePtr node = pcmk__output_create_xml_node(out, "current_dc");
pe_node_t *dc = va_arg(args, pe_node_t *);
const char *quorum = va_arg(args, const char *);
const char *dc_version_s = va_arg(args, const char *);
char *dc_name G_GNUC_UNUSED = va_arg(args, char *);
if (dc) {
xmlSetProp(node, (pcmkXmlStr) "present", (pcmkXmlStr) "true");
xmlSetProp(node, (pcmkXmlStr) "version", (pcmkXmlStr) (dc_version_s ? dc_version_s : ""));
xmlSetProp(node, (pcmkXmlStr) "name", (pcmkXmlStr) dc->details->uname);
xmlSetProp(node, (pcmkXmlStr) "id", (pcmkXmlStr) dc->details->id);
xmlSetProp(node, (pcmkXmlStr) "with_quorum", (pcmkXmlStr) (crm_is_true(quorum) ? "true" : "false"));
} else {
xmlSetProp(node, (pcmkXmlStr) "present", (pcmkXmlStr) "false");
}
return pcmk_rc_ok;
}
int
pe__cluster_maint_mode_text(pcmk__output_t *out, va_list args) {
fprintf(out->dest, "\n *** Resource management is DISABLED ***");
fprintf(out->dest, "\n The cluster will not attempt to start, stop or recover services");
fprintf(out->dest, "\n");
return pcmk_rc_ok;
}
int
pe__cluster_options_html(pcmk__output_t *out, va_list args) {
pe_working_set_t *data_set = va_arg(args, pe_working_set_t *);
out->list_item(out, NULL, "STONITH of failed nodes %s",
is_set(data_set->flags, pe_flag_stonith_enabled) ? "enabled" : "disabled");
out->list_item(out, NULL, "Cluster is %s",
is_set(data_set->flags, pe_flag_symmetric_cluster) ? "symmetric" : "asymmetric");
switch (data_set->no_quorum_policy) {
case no_quorum_freeze:
out->list_item(out, NULL, "No quorum policy: Freeze resources");
break;
case no_quorum_stop:
out->list_item(out, NULL, "No quorum policy: Stop ALL resources");
break;
case no_quorum_ignore:
out->list_item(out, NULL, "No quorum policy: Ignore");
break;
case no_quorum_suicide:
out->list_item(out, NULL, "No quorum policy: Suicide");
break;
}
if (is_set(data_set->flags, pe_flag_maintenance_mode)) {
xmlNodePtr node = pcmk__output_create_xml_node(out, "li");
pcmk_create_html_node(node, "span", NULL, NULL, "Resource management: ");
pcmk_create_html_node(node, "span", NULL, "bold", "DISABLED");
pcmk_create_html_node(node, "span", NULL, NULL,
" (the cluster will not attempt to start, stop, or recover services)");
} else {
out->list_item(out, NULL, "Resource management: enabled");
}
return pcmk_rc_ok;
}
int
pe__cluster_options_log(pcmk__output_t *out, va_list args) {
pe_working_set_t *data_set = va_arg(args, pe_working_set_t *);
if (is_set(data_set->flags, pe_flag_maintenance_mode)) {
out->info(out, "Resource management is DISABLED. The cluster will not attempt to start, stop or recover services.");
return pcmk_rc_ok;
} else {
return pcmk_rc_no_output;
}
}
int
pe__cluster_options_text(pcmk__output_t *out, va_list args) {
pe_working_set_t *data_set = va_arg(args, pe_working_set_t *);
out->list_item(out, NULL, "STONITH of failed nodes %s",
is_set(data_set->flags, pe_flag_stonith_enabled) ? "enabled" : "disabled");
out->list_item(out, NULL, "Cluster is %s",
is_set(data_set->flags, pe_flag_symmetric_cluster) ? "symmetric" : "asymmetric");
switch (data_set->no_quorum_policy) {
case no_quorum_freeze:
out->list_item(out, NULL, "No quorum policy: Freeze resources");
break;
case no_quorum_stop:
out->list_item(out, NULL, "No quorum policy: Stop ALL resources");
break;
case no_quorum_ignore:
out->list_item(out, NULL, "No quorum policy: Ignore");
break;
case no_quorum_suicide:
out->list_item(out, NULL, "No quorum policy: Suicide");
break;
}
return pcmk_rc_ok;
}
int
pe__cluster_options_xml(pcmk__output_t *out, va_list args) {
xmlNodePtr node = pcmk__output_create_xml_node(out, "cluster_options");
pe_working_set_t *data_set = va_arg(args, pe_working_set_t *);
xmlSetProp(node, (pcmkXmlStr) "stonith-enabled",
(pcmkXmlStr) (is_set(data_set->flags, pe_flag_stonith_enabled) ? "true" : "false"));
xmlSetProp(node, (pcmkXmlStr) "symmetric-cluster",
(pcmkXmlStr) (is_set(data_set->flags, pe_flag_symmetric_cluster) ? "true" : "false"));
switch (data_set->no_quorum_policy) {
case no_quorum_freeze:
xmlSetProp(node, (pcmkXmlStr) "no-quorum-policy", (pcmkXmlStr) "freeze");
break;
case no_quorum_stop:
xmlSetProp(node, (pcmkXmlStr) "no-quorum-policy", (pcmkXmlStr) "stop");
break;
case no_quorum_ignore:
xmlSetProp(node, (pcmkXmlStr) "no-quorum-policy", (pcmkXmlStr) "ignore");
break;
case no_quorum_suicide:
xmlSetProp(node, (pcmkXmlStr) "no-quorum-policy", (pcmkXmlStr) "suicide");
break;
}
xmlSetProp(node, (pcmkXmlStr) "maintenance-mode",
(pcmkXmlStr) (is_set(data_set->flags, pe_flag_maintenance_mode) ? "true" : "false"));
return pcmk_rc_ok;
}
int
pe__cluster_stack_html(pcmk__output_t *out, va_list args) {
xmlNodePtr node = pcmk__output_create_xml_node(out, "li");
const char *stack_s = va_arg(args, const char *);
pcmk_create_html_node(node, "span", NULL, "bold", "Stack: ");
pcmk_create_html_node(node, "span", NULL, NULL, stack_s);
return pcmk_rc_ok;
}
int
pe__cluster_stack_text(pcmk__output_t *out, va_list args) {
const char *stack_s = va_arg(args, const char *);
out->list_item(out, "Stack", "%s", stack_s);
return pcmk_rc_ok;
}
int
pe__cluster_stack_xml(pcmk__output_t *out, va_list args) {
xmlNodePtr node = pcmk__output_create_xml_node(out, "stack");
const char *stack_s = va_arg(args, const char *);
xmlSetProp(node, (pcmkXmlStr) "type", (pcmkXmlStr) stack_s);
return pcmk_rc_ok;
}
int
pe__cluster_times_html(pcmk__output_t *out, va_list args) {
xmlNodePtr updated_node = pcmk__output_create_xml_node(out, "li");
xmlNodePtr changed_node = pcmk__output_create_xml_node(out, "li");
const char *last_written = va_arg(args, const char *);
const char *user = va_arg(args, const char *);
const char *client = va_arg(args, const char *);
const char *origin = va_arg(args, const char *);
char *buf = last_changed_string(last_written, user, client, origin);
pcmk_create_html_node(updated_node, "span", NULL, "bold", "Last updated: ");
pcmk_create_html_node(updated_node, "span", NULL, NULL,
pcmk__epoch2str(NULL));
pcmk_create_html_node(changed_node, "span", NULL, "bold", "Last change: ");
pcmk_create_html_node(changed_node, "span", NULL, NULL, buf);
free(buf);
return pcmk_rc_ok;
}
int
pe__cluster_times_xml(pcmk__output_t *out, va_list args) {
xmlNodePtr updated_node = pcmk__output_create_xml_node(out, "last_update");
xmlNodePtr changed_node = pcmk__output_create_xml_node(out, "last_change");
const char *last_written = va_arg(args, const char *);
const char *user = va_arg(args, const char *);
const char *client = va_arg(args, const char *);
const char *origin = va_arg(args, const char *);
xmlSetProp(updated_node, (pcmkXmlStr) "time",
(pcmkXmlStr) pcmk__epoch2str(NULL));
xmlSetProp(changed_node, (pcmkXmlStr) "time", (pcmkXmlStr) (last_written ? last_written : ""));
xmlSetProp(changed_node, (pcmkXmlStr) "user", (pcmkXmlStr) (user ? user : ""));
xmlSetProp(changed_node, (pcmkXmlStr) "client", (pcmkXmlStr) (client ? client : ""));
xmlSetProp(changed_node, (pcmkXmlStr) "origin", (pcmkXmlStr) (origin ? origin : ""));
return pcmk_rc_ok;
}
int
pe__cluster_times_text(pcmk__output_t *out, va_list args) {
const char *last_written = va_arg(args, const char *);
const char *user = va_arg(args, const char *);
const char *client = va_arg(args, const char *);
const char *origin = va_arg(args, const char *);
char *buf = last_changed_string(last_written, user, client, origin);
out->list_item(out, "Last updated", "%s", pcmk__epoch2str(NULL));
out->list_item(out, "Last change", " %s", buf);
free(buf);
return pcmk_rc_ok;
}
int
pe__failed_action_text(pcmk__output_t *out, va_list args) {
xmlNodePtr xml_op = va_arg(args, xmlNodePtr);
char *s = failed_action_string(xml_op);
out->list_item(out, NULL, "%s", s);
free(s);
return pcmk_rc_ok;
}
int
pe__failed_action_xml(pcmk__output_t *out, va_list args) {
xmlNodePtr xml_op = va_arg(args, xmlNodePtr);
const char *op_key = crm_element_value(xml_op, XML_LRM_ATTR_TASK_KEY);
const char *last = crm_element_value(xml_op, XML_RSC_OP_LAST_CHANGE);
int rc = crm_parse_int(crm_element_value(xml_op, XML_LRM_ATTR_RC), "0");
int status = crm_parse_int(crm_element_value(xml_op, XML_LRM_ATTR_OPSTATUS), "0");
const char *exit_reason = crm_element_value(xml_op, XML_LRM_ATTR_EXIT_REASON);
char *rc_s = crm_itoa(rc);
char *reason_s = crm_xml_escape(exit_reason ? exit_reason : "none");
xmlNodePtr node = pcmk__output_create_xml_node(out, "failure");
xmlSetProp(node, (pcmkXmlStr) (op_key ? "op_key" : "id"),
(pcmkXmlStr) (op_key ? op_key : "id"));
xmlSetProp(node, (pcmkXmlStr) "node",
(pcmkXmlStr) crm_element_value(xml_op, XML_ATTR_UNAME));
xmlSetProp(node, (pcmkXmlStr) "exitstatus",
(pcmkXmlStr) services_ocf_exitcode_str(rc));
xmlSetProp(node, (pcmkXmlStr) "exitreason", (pcmkXmlStr) reason_s);
xmlSetProp(node, (pcmkXmlStr) "exitcode", (pcmkXmlStr) rc_s);
xmlSetProp(node, (pcmkXmlStr) "call",
(pcmkXmlStr) crm_element_value(xml_op, XML_LRM_ATTR_CALLID));
xmlSetProp(node, (pcmkXmlStr) "status",
(pcmkXmlStr) services_lrm_status_str(status));
if (last) {
guint interval_ms = 0;
char *s = NULL;
time_t when = crm_parse_int(last, "0");
crm_time_t *crm_when = crm_time_new(NULL);
char *rc_change = NULL;
crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
s = crm_itoa(interval_ms);
crm_time_set_timet(crm_when, &when);
rc_change = crm_time_as_string(crm_when, crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone);
xmlSetProp(node, (pcmkXmlStr) XML_RSC_OP_LAST_CHANGE, (pcmkXmlStr) rc_change);
xmlSetProp(node, (pcmkXmlStr) "queued",
(pcmkXmlStr) crm_element_value(xml_op, XML_RSC_OP_T_QUEUE));
xmlSetProp(node, (pcmkXmlStr) "exec",
(pcmkXmlStr) crm_element_value(xml_op, XML_RSC_OP_T_EXEC));
xmlSetProp(node, (pcmkXmlStr) "interval", (pcmkXmlStr) s);
xmlSetProp(node, (pcmkXmlStr) "task",
(pcmkXmlStr) crm_element_value(xml_op, XML_LRM_ATTR_TASK));
free(s);
free(rc_change);
crm_time_free(crm_when);
}
free(reason_s);
free(rc_s);
return pcmk_rc_ok;
}
int
pe__node_html(pcmk__output_t *out, va_list args) {
pe_node_t *node = va_arg(args, pe_node_t *);
unsigned int print_opts = va_arg(args, unsigned int);
gboolean full = va_arg(args, gboolean);
const char *node_mode G_GNUC_UNUSED = va_arg(args, const char *);
gboolean print_clone_detail = va_arg(args, gboolean);
gboolean print_brief = va_arg(args, gboolean);
gboolean group_by_node = va_arg(args, gboolean);
char *node_name = pe__node_display_name(node, print_clone_detail);
char *buf = crm_strdup_printf("Node: %s", node_name);
if (full) {
xmlNodePtr item_node = pcmk__output_create_xml_node(out, "li");
pcmk_create_html_node(item_node, "span", NULL, NULL, buf);
if (node->details->standby_onfail && node->details->online) {
pcmk_create_html_node(item_node, "span", NULL, "standby", " standby (on-fail)");
} else if (node->details->standby && node->details->online) {
char *s = crm_strdup_printf(" standby%s", node->details->running_rsc ? " (with active resources)" : "");
pcmk_create_html_node(item_node, "span", NULL, " standby", s);
free(s);
} else if (node->details->standby) {
pcmk_create_html_node(item_node, "span", NULL, "offline", " OFFLINE (standby)");
} else if (node->details->maintenance && node->details->online) {
pcmk_create_html_node(item_node, "span", NULL, "maint", " maintenance");
} else if (node->details->maintenance) {
pcmk_create_html_node(item_node, "span", NULL, "offline", " OFFLINE (maintenance)");
} else if (node->details->online) {
pcmk_create_html_node(item_node, "span", NULL, "online", " online");
} else {
pcmk_create_html_node(item_node, "span", NULL, "offline", " OFFLINE");
}
if (print_brief && group_by_node) {
out->begin_list(out, NULL, NULL, NULL);
pe__rscs_brief_output(out, node->details->running_rsc, print_opts | pe_print_rsconly,
FALSE);
out->end_list(out);
} else if (group_by_node) {
GListPtr lpc2 = NULL;
out->begin_list(out, NULL, NULL, NULL);
for (lpc2 = node->details->running_rsc; lpc2 != NULL; lpc2 = lpc2->next) {
pe_resource_t *rsc = (pe_resource_t *) lpc2->data;
out->message(out, crm_map_element_name(rsc->xml), print_opts | pe_print_rsconly, rsc);
}
out->end_list(out);
}
} else {
out->begin_list(out, NULL, NULL, "%s", buf);
}
free(buf);
free(node_name);
return pcmk_rc_ok;
}
int
pe__node_text(pcmk__output_t *out, va_list args) {
pe_node_t *node = va_arg(args, pe_node_t *);
unsigned int print_opts = va_arg(args, unsigned int);
gboolean full = va_arg(args, gboolean);
const char *node_mode = va_arg(args, const char *);
gboolean print_clone_detail = va_arg(args, gboolean);
gboolean print_brief = va_arg(args, gboolean);
gboolean group_by_node = va_arg(args, gboolean);
if (full) {
char *node_name = pe__node_display_name(node, print_clone_detail);
char *buf = NULL;
/* Print the node name and status */
if (pe__is_guest_node(node)) {
buf = crm_strdup_printf("GuestNode %s: %s", node_name, node_mode);
} else if (pe__is_remote_node(node)) {
buf = crm_strdup_printf("RemoteNode %s: %s", node_name, node_mode);
} else {
buf = crm_strdup_printf("Node %s: %s", node_name, node_mode);
}
/* If we're grouping by node, print its resources */
if (group_by_node) {
out->begin_list(out, NULL, NULL, "%s", buf);
out->begin_list(out, NULL, NULL, "Resources");
if (print_brief) {
pe__rscs_brief_output(out, node->details->running_rsc,
print_opts | pe_print_rsconly, FALSE);
} else {
GListPtr gIter2 = NULL;
for (gIter2 = node->details->running_rsc; gIter2 != NULL; gIter2 = gIter2->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter2->data;
out->message(out, crm_map_element_name(rsc->xml), print_opts | pe_print_rsconly, rsc);
}
}
out->end_list(out);
out->end_list(out);
} else {
out->list_item(out, NULL, "%s", buf);
}
free(buf);
free(node_name);
} else {
out->begin_list(out, NULL, NULL, "Node: %s", pe__node_display_name(node, print_clone_detail));
}
return pcmk_rc_ok;
}
int
pe__node_xml(pcmk__output_t *out, va_list args) {
pe_node_t *node = va_arg(args, pe_node_t *);
unsigned int print_opts = va_arg(args, unsigned int);
gboolean full = va_arg(args, gboolean);
const char *node_mode G_GNUC_UNUSED = va_arg(args, const char *);
gboolean print_clone_detail G_GNUC_UNUSED = va_arg(args, gboolean);
gboolean print_brief G_GNUC_UNUSED = va_arg(args, gboolean);
gboolean group_by_node = va_arg(args, gboolean);
if (full) {
const char *node_type = "unknown";
char *length_s = crm_itoa(g_list_length(node->details->running_rsc));
switch (node->details->type) {
case node_member:
node_type = "member";
break;
case node_remote:
node_type = "remote";
break;
case node_ping:
node_type = "ping";
break;
}
pe__name_and_nvpairs_xml(out, true, "node", 13,
"name", node->details->uname,
"id", node->details->id,
"online", node->details->online ? "true" : "false",
"standby", node->details->standby ? "true" : "false",
"standby_onfail", node->details->standby_onfail ? "true" : "false",
"maintenance", node->details->maintenance ? "true" : "false",
"pending", node->details->pending ? "true" : "false",
"unclean", node->details->unclean ? "true" : "false",
"shutdown", node->details->shutdown ? "true" : "false",
"expected_up", node->details->expected_up ? "true" : "false",
"is_dc", node->details->is_dc ? "true" : "false",
"resources_running", length_s,
"type", node_type);
if (pe__is_guest_node(node)) {
xmlNodePtr xml_node = pcmk__output_xml_peek_parent(out);
xmlSetProp(xml_node, (pcmkXmlStr) "id_as_resource",
(pcmkXmlStr) node->details->remote_rsc->container->id);
}
if (group_by_node) {
GListPtr lpc = NULL;
for (lpc = node->details->running_rsc; lpc != NULL; lpc = lpc->next) {
pe_resource_t *rsc = (pe_resource_t *) lpc->data;
out->message(out, crm_map_element_name(rsc->xml), print_opts | pe_print_rsconly, rsc);
}
}
free(length_s);
out->end_list(out);
} else {
xmlNodePtr parent = pcmk__output_xml_create_parent(out, "node");
xmlSetProp(parent, (pcmkXmlStr) "name", (pcmkXmlStr) node->details->uname);
}
return pcmk_rc_ok;
}
int
pe__node_attribute_text(pcmk__output_t *out, va_list args) {
const char *name = va_arg(args, const char *);
const char *value = va_arg(args, const char *);
gboolean add_extra = va_arg(args, gboolean);
int expected_score = va_arg(args, int);
if (add_extra) {
int v = crm_parse_int(value, "0");
if (v <= 0) {
out->list_item(out, NULL, "%-32s\t: %-10s\t: Connectivity is lost", name, value);
} else if (v < expected_score) {
out->list_item(out, NULL, "%-32s\t: %-10s\t: Connectivity is degraded (Expected=%d)", name, value, expected_score);
} else {
out->list_item(out, NULL, "%-32s\t: %-10s", name, value);
}
} else {
out->list_item(out, NULL, "%-32s\t: %-10s", name, value);
}
return pcmk_rc_ok;
}
int
pe__node_attribute_html(pcmk__output_t *out, va_list args) {
const char *name = va_arg(args, const char *);
const char *value = va_arg(args, const char *);
gboolean add_extra = va_arg(args, gboolean);
int expected_score = va_arg(args, int);
if (add_extra) {
int v = crm_parse_int(value, "0");
char *s = crm_strdup_printf("%s: %s", name, value);
xmlNodePtr item_node = pcmk__output_create_xml_node(out, "li");
pcmk_create_html_node(item_node, "span", NULL, NULL, s);
free(s);
if (v <= 0) {
pcmk_create_html_node(item_node, "span", NULL, "bold", "(connectivity is lost)");
} else if (v < expected_score) {
char *buf = crm_strdup_printf("(connectivity is degraded -- expected %d", expected_score);
pcmk_create_html_node(item_node, "span", NULL, "bold", buf);
free(buf);
}
} else {
out->list_item(out, NULL, "%s: %s", name, value);
}
return pcmk_rc_ok;
}
int
pe__node_attribute_xml(pcmk__output_t *out, va_list args) {
const char *name = va_arg(args, const char *);
const char *value = va_arg(args, const char *);
gboolean add_extra = va_arg(args, gboolean);
int expected_score = va_arg(args, int);
xmlNodePtr node = pcmk__output_create_xml_node(out, "attribute");
xmlSetProp(node, (pcmkXmlStr) "name", (pcmkXmlStr) name);
xmlSetProp(node, (pcmkXmlStr) "value", (pcmkXmlStr) value);
if (add_extra) {
char *buf = crm_itoa(expected_score);
xmlSetProp(node, (pcmkXmlStr) "expected", (pcmkXmlStr) buf);
free(buf);
}
return pcmk_rc_ok;
}
int
pe__op_history_text(pcmk__output_t *out, va_list args) {
xmlNode *xml_op = va_arg(args, xmlNode *);
const char *task = va_arg(args, const char *);
const char *interval_ms_s = va_arg(args, const char *);
int rc = va_arg(args, int);
gboolean print_timing = va_arg(args, gboolean);
char *buf = op_history_string(xml_op, task, interval_ms_s, rc, print_timing);
out->list_item(out, NULL, "%s", buf);
free(buf);
return pcmk_rc_ok;
}
int
pe__op_history_xml(pcmk__output_t *out, va_list args) {
xmlNode *xml_op = va_arg(args, xmlNode *);
const char *task = va_arg(args, const char *);
const char *interval_ms_s = va_arg(args, const char *);
int rc = va_arg(args, int);
gboolean print_timing = va_arg(args, gboolean);
char *rc_s = NULL;
xmlNodePtr node = pcmk__output_create_xml_node(out, "operation_history");
xmlSetProp(node, (pcmkXmlStr) "call",
(pcmkXmlStr) crm_element_value(xml_op, XML_LRM_ATTR_CALLID));
xmlSetProp(node, (pcmkXmlStr) "task", (pcmkXmlStr) task);
if (interval_ms_s && safe_str_neq(interval_ms_s, "0")) {
char *s = crm_strdup_printf("%sms", interval_ms_s);
xmlSetProp(node, (pcmkXmlStr) "interval", (pcmkXmlStr) s);
free(s);
}
if (print_timing) {
const char *value = NULL;
value = crm_element_value(xml_op, XML_RSC_OP_LAST_CHANGE);
if (value) {
time_t int_value = (time_t) crm_parse_int(value, NULL);
if (int_value > 0) {
xmlSetProp(node, (pcmkXmlStr) XML_RSC_OP_LAST_CHANGE,
(pcmkXmlStr) pcmk__epoch2str(&int_value));
}
}
value = crm_element_value(xml_op, XML_RSC_OP_LAST_RUN);
if (value) {
time_t int_value = (time_t) crm_parse_int(value, NULL);
if (int_value > 0) {
xmlSetProp(node, (pcmkXmlStr) XML_RSC_OP_LAST_RUN,
(pcmkXmlStr) pcmk__epoch2str(&int_value));
}
}
value = crm_element_value(xml_op, XML_RSC_OP_T_EXEC);
if (value) {
char *s = crm_strdup_printf("%sms", value);
xmlSetProp(node, (pcmkXmlStr) XML_RSC_OP_T_EXEC, (pcmkXmlStr) s);
free(s);
}
value = crm_element_value(xml_op, XML_RSC_OP_T_QUEUE);
if (value) {
char *s = crm_strdup_printf("%sms", value);
xmlSetProp(node, (pcmkXmlStr) XML_RSC_OP_T_QUEUE, (pcmkXmlStr) s);
free(s);
}
}
rc_s = crm_itoa(rc);
xmlSetProp(node, (pcmkXmlStr) "rc", (pcmkXmlStr) rc_s);
xmlSetProp(node, (pcmkXmlStr) "rc_text", (pcmkXmlStr) services_ocf_exitcode_str(rc));
free(rc_s);
return pcmk_rc_ok;
}
int
pe__resource_history_text(pcmk__output_t *out, va_list args) {
pe_resource_t *rsc = va_arg(args, pe_resource_t *);
const char *rsc_id = va_arg(args, const char *);
gboolean all = va_arg(args, gboolean);
int failcount = va_arg(args, int);
time_t last_failure = va_arg(args, int);
gboolean as_header = va_arg(args, gboolean);
char *buf = resource_history_string(rsc, rsc_id, all, failcount, last_failure);
if (as_header) {
out->begin_list(out, NULL, NULL, "%s", buf);
} else {
out->list_item(out, NULL, "%s", buf);
}
free(buf);
return pcmk_rc_ok;
}
int
pe__resource_history_xml(pcmk__output_t *out, va_list args) {
pe_resource_t *rsc = va_arg(args, pe_resource_t *);
const char *rsc_id = va_arg(args, const char *);
gboolean all = va_arg(args, gboolean);
int failcount = va_arg(args, int);
time_t last_failure = va_arg(args, int);
gboolean as_header = va_arg(args, gboolean);
xmlNodePtr node = pcmk__output_xml_create_parent(out, "resource_history");
xmlSetProp(node, (pcmkXmlStr) "id", (pcmkXmlStr) rsc_id);
if (rsc == NULL) {
xmlSetProp(node, (pcmkXmlStr) "orphan", (pcmkXmlStr) "true");
} else if (all || failcount || last_failure > 0) {
char *migration_s = crm_itoa(rsc->migration_threshold);
xmlSetProp(node, (pcmkXmlStr) "orphan", (pcmkXmlStr) "false");
xmlSetProp(node, (pcmkXmlStr) "migration-threshold",
(pcmkXmlStr) migration_s);
free(migration_s);
if (failcount > 0) {
char *s = crm_itoa(failcount);
xmlSetProp(node, (pcmkXmlStr) PCMK__FAIL_COUNT_PREFIX,
(pcmkXmlStr) s);
free(s);
}
if (last_failure > 0) {
xmlSetProp(node, (pcmkXmlStr) PCMK__LAST_FAILURE_PREFIX,
(pcmkXmlStr) pcmk__epoch2str(&last_failure));
}
}
if (as_header == FALSE) {
pcmk__output_xml_pop_parent(out);
}
return pcmk_rc_ok;
}
int
pe__ticket_html(pcmk__output_t *out, va_list args) {
pe_ticket_t *ticket = va_arg(args, pe_ticket_t *);
if (ticket->last_granted > -1) {
char *time = pcmk_format_named_time("last-granted", ticket->last_granted);
out->list_item(out, NULL, "%s:\t%s%s %s", ticket->id,
ticket->granted ? "granted" : "revoked",
ticket->standby ? " [standby]" : "",
time);
free(time);
} else {
out->list_item(out, NULL, "%s:\t%s%s", ticket->id,
ticket->granted ? "granted" : "revoked",
ticket->standby ? " [standby]" : "");
}
return pcmk_rc_ok;
}
int
pe__ticket_text(pcmk__output_t *out, va_list args) {
pe_ticket_t *ticket = va_arg(args, pe_ticket_t *);
if (ticket->last_granted > -1) {
char *time = pcmk_format_named_time("last-granted", ticket->last_granted);
out->list_item(out, ticket->id, "\t%s%s %s",
ticket->granted ? "granted" : "revoked",
ticket->standby ? " [standby]" : "",
time);
free(time);
} else {
out->list_item(out, ticket->id, "\t%s%s",
ticket->granted ? "granted" : "revoked",
ticket->standby ? " [standby]" : "");
}
return pcmk_rc_ok;
}
int
pe__ticket_xml(pcmk__output_t *out, va_list args) {
xmlNodePtr node = NULL;
pe_ticket_t *ticket = va_arg(args, pe_ticket_t *);
node = pcmk__output_create_xml_node(out, "ticket");
xmlSetProp(node, (pcmkXmlStr) "id", (pcmkXmlStr) ticket->id);
xmlSetProp(node, (pcmkXmlStr) "status", (pcmkXmlStr) (ticket->granted ? "granted" : "revoked"));
xmlSetProp(node, (pcmkXmlStr) "standby", (pcmkXmlStr) (ticket->standby ? "true" : "false"));
if (ticket->last_granted > -1) {
xmlSetProp(node, (pcmkXmlStr) "last-granted",
(pcmkXmlStr) pcmk__epoch2str(&ticket->last_granted));
}
return pcmk_rc_ok;
}
static pcmk__message_entry_t fmt_functions[] = {
{ "ban", "html", pe__ban_html },
{ "ban", "log", pe__ban_text },
{ "ban", "text", pe__ban_text },
{ "ban", "xml", pe__ban_xml },
{ "bundle", "xml", pe__bundle_xml },
{ "bundle", "html", pe__bundle_html },
{ "bundle", "text", pe__bundle_text },
{ "bundle", "log", pe__bundle_text },
{ "clone", "xml", pe__clone_xml },
{ "clone", "html", pe__clone_html },
{ "clone", "text", pe__clone_text },
{ "clone", "log", pe__clone_text },
{ "cluster-counts", "html", pe__cluster_counts_html },
{ "cluster-counts", "log", pe__cluster_counts_text },
{ "cluster-counts", "text", pe__cluster_counts_text },
{ "cluster-counts", "xml", pe__cluster_counts_xml },
{ "cluster-dc", "html", pe__cluster_dc_html },
{ "cluster-dc", "log", pe__cluster_dc_text },
{ "cluster-dc", "text", pe__cluster_dc_text },
{ "cluster-dc", "xml", pe__cluster_dc_xml },
{ "cluster-options", "html", pe__cluster_options_html },
{ "cluster-options", "log", pe__cluster_options_log },
{ "cluster-options", "text", pe__cluster_options_text },
{ "cluster-options", "xml", pe__cluster_options_xml },
{ "cluster-summary", "html", pe__cluster_summary_html },
{ "cluster-summary", "log", pe__cluster_summary },
{ "cluster-summary", "text", pe__cluster_summary },
{ "cluster-summary", "xml", pe__cluster_summary },
{ "cluster-stack", "html", pe__cluster_stack_html },
{ "cluster-stack", "log", pe__cluster_stack_text },
{ "cluster-stack", "text", pe__cluster_stack_text },
{ "cluster-stack", "xml", pe__cluster_stack_xml },
{ "cluster-times", "html", pe__cluster_times_html },
{ "cluster-times", "log", pe__cluster_times_text },
{ "cluster-times", "text", pe__cluster_times_text },
{ "cluster-times", "xml", pe__cluster_times_xml },
{ "failed-action", "html", pe__failed_action_text },
{ "failed-action", "log", pe__failed_action_text },
{ "failed-action", "text", pe__failed_action_text },
{ "failed-action", "xml", pe__failed_action_xml },
{ "group", "xml", pe__group_xml },
{ "group", "html", pe__group_html },
{ "group", "text", pe__group_text },
{ "group", "log", pe__group_text },
/* maint-mode only exists for text and log. Other formatters output it as
* part of the cluster-options handler.
*/
{ "maint-mode", "log", pe__cluster_maint_mode_text },
{ "maint-mode", "text", pe__cluster_maint_mode_text },
{ "node", "html", pe__node_html },
{ "node", "log", pe__node_text },
{ "node", "text", pe__node_text },
{ "node", "xml", pe__node_xml },
{ "node-attribute", "html", pe__node_attribute_html },
{ "node-attribute", "log", pe__node_attribute_text },
{ "node-attribute", "text", pe__node_attribute_text },
{ "node-attribute", "xml", pe__node_attribute_xml },
{ "op-history", "html", pe__op_history_text },
{ "op-history", "log", pe__op_history_text },
{ "op-history", "text", pe__op_history_text },
{ "op-history", "xml", pe__op_history_xml },
{ "primitive", "xml", pe__resource_xml },
{ "primitive", "html", pe__resource_html },
{ "primitive", "text", pe__resource_text },
{ "primitive", "log", pe__resource_text },
{ "resource-history", "html", pe__resource_history_text },
{ "resource-history", "log", pe__resource_history_text },
{ "resource-history", "text", pe__resource_history_text },
{ "resource-history", "xml", pe__resource_history_xml },
{ "ticket", "html", pe__ticket_html },
{ "ticket", "log", pe__ticket_text },
{ "ticket", "text", pe__ticket_text },
{ "ticket", "xml", pe__ticket_xml },
{ NULL, NULL, NULL }
};
void
pe__register_messages(pcmk__output_t *out) {
pcmk__register_messages(out, fmt_functions);
}
void
pe__output_node(pe_node_t *node, gboolean details, pcmk__output_t *out)
{
if (node == NULL) {
crm_trace("<NULL>");
return;
}
CRM_ASSERT(node->details);
crm_trace("%sNode %s: (weight=%d, fixed=%s)",
node->details->online ? "" : "Unavailable/Unclean ",
node->details->uname, node->weight, node->fixed ? "True" : "False");
if (details) {
char *pe_mutable = strdup("\t\t");
GListPtr gIter = node->details->running_rsc;
crm_trace("\t\t===Node Attributes");
g_hash_table_foreach(node->details->attrs, print_str_str, pe_mutable);
free(pe_mutable);
crm_trace("\t\t=== Resources");
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
- // @TODO pe_print_log probably doesn't belong here
out->message(out, crm_map_element_name(rsc->xml),
- pe_print_log|pe_print_pending, rsc);
+ pe_print_pending, rsc);
}
}
}
diff --git a/lib/pengine/pe_status_private.h b/lib/pengine/pe_status_private.h
index 4587db6347..0f3814ff10 100644
--- a/lib/pengine/pe_status_private.h
+++ b/lib/pengine/pe_status_private.h
@@ -1,48 +1,68 @@
/*
- * Copyright 2018 Andrew Beekhof <andrew@beekhof.net>
+ * Copyright 2018-2020 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.
*/
#ifndef PE_STATUS_PRIVATE__H
# define PE_STATUS_PRIVATE__H
/* This header is for the sole use of libpe_status, so that functions can be
* declared with G_GNUC_INTERNAL for efficiency.
*/
# if CURSES_ENABLED
# define status_printw(fmt, args...) printw(fmt, ##args)
# else
# define status_printw(fmt, args...) \
crm_err("printw support requires ncurses to be available during configure"); \
do_crm_log(LOG_WARNING, fmt, ##args);
# endif
# define status_print(fmt, args...) \
if(options & pe_print_html) { \
FILE *stream = print_data; \
fprintf(stream, fmt, ##args); \
} else if(options & pe_print_ncurses) { \
status_printw(fmt, ##args); \
} else if(options & pe_print_printf) { \
FILE *stream = print_data; \
fprintf(stream, fmt, ##args); \
} else if(options & pe_print_xml) { \
FILE *stream = print_data; \
fprintf(stream, fmt, ##args); \
} else if(options & pe_print_log) { \
int log_level = *(int*)print_data; \
do_crm_log(log_level, fmt, ##args); \
}
G_GNUC_INTERNAL
pe_resource_t *pe__create_clone_child(pe_resource_t *rsc,
pe_working_set_t *data_set);
G_GNUC_INTERNAL
void pe__force_anon(const char *standard, pe_resource_t *rsc, const char *rid,
pe_working_set_t *data_set);
+G_GNUC_INTERNAL
+gboolean unpack_remote_nodes(xmlNode *xml_resources, pe_working_set_t *data_set);
+
+G_GNUC_INTERNAL
+gboolean unpack_resources(xmlNode *xml_resources, pe_working_set_t *data_set);
+
+G_GNUC_INTERNAL
+gboolean unpack_config(xmlNode *config, pe_working_set_t *data_set);
+
+G_GNUC_INTERNAL
+gboolean unpack_nodes(xmlNode *xml_nodes, pe_working_set_t *data_set);
+
+G_GNUC_INTERNAL
+gboolean unpack_tags(xmlNode *xml_tags, pe_working_set_t *data_set);
+
+G_GNUC_INTERNAL
+gboolean unpack_status(xmlNode *status, pe_working_set_t *data_set);
+
#endif // PE_STATUS_PRIVATE__H
diff --git a/lib/pengine/status.c b/lib/pengine/status.c
index eabb6f6d78..8dc5095c15 100644
--- a/lib/pengine/status.c
+++ b/lib/pengine/status.c
@@ -1,440 +1,440 @@
/*
- * Copyright 2004-2019 the Pacemaker project contributors
+ * Copyright 2004-2020 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 <sys/param.h>
#include <crm/crm.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <glib.h>
#include <crm/pengine/internal.h>
-#include <unpack.h>
+#include <pe_status_private.h>
/*!
* \brief Create a new working set
*
* \return New, initialized working set on success, else NULL (and set errno)
* \note Only pe_working_set_t objects created with this function (as opposed
* to statically declared or directly allocated) should be used with the
* functions in this library, to allow for future extensions to the
* data type. The caller is responsible for freeing the memory with
* pe_free_working_set() when the instance is no longer needed.
*/
pe_working_set_t *
pe_new_working_set()
{
pe_working_set_t *data_set = calloc(1, sizeof(pe_working_set_t));
if (data_set != NULL) {
set_working_set_defaults(data_set);
}
return data_set;
}
/*!
* \brief Free a working set
*
* \param[in] data_set Working set to free
*/
void
pe_free_working_set(pe_working_set_t *data_set)
{
if (data_set != NULL) {
pe_reset_working_set(data_set);
free(data_set);
}
}
/*
* Unpack everything
* At the end you'll have:
* - A list of nodes
* - A list of resources (each with any dependencies on other resources)
* - A list of constraints between resources and nodes
* - A list of constraints between start/stop actions
* - A list of nodes that need to be stonith'd
* - A list of nodes that need to be shutdown
* - A list of the possible stop/start actions (without dependencies)
*/
gboolean
cluster_status(pe_working_set_t * data_set)
{
xmlNode *config = get_xpath_object("//"XML_CIB_TAG_CRMCONFIG, data_set->input, LOG_TRACE);
xmlNode *cib_nodes = get_xpath_object("//"XML_CIB_TAG_NODES, data_set->input, LOG_TRACE);
xmlNode *cib_resources = get_xpath_object("//"XML_CIB_TAG_RESOURCES, data_set->input, LOG_TRACE);
xmlNode *cib_status = get_xpath_object("//"XML_CIB_TAG_STATUS, data_set->input, LOG_TRACE);
xmlNode *cib_tags = get_xpath_object("//" XML_CIB_TAG_TAGS, data_set->input,
LOG_NEVER);
const char *value = crm_element_value(data_set->input, XML_ATTR_HAVE_QUORUM);
crm_trace("Beginning unpack");
/* reset remaining global variables */
data_set->failed = create_xml_node(NULL, "failed-ops");
if (data_set->input == NULL) {
return FALSE;
}
if (data_set->now == NULL) {
data_set->now = crm_time_new(NULL);
}
if (data_set->dc_uuid == NULL) {
data_set->dc_uuid = crm_element_value_copy(data_set->input,
XML_ATTR_DC_UUID);
}
clear_bit(data_set->flags, pe_flag_have_quorum);
if (crm_is_true(value)) {
set_bit(data_set->flags, pe_flag_have_quorum);
}
data_set->op_defaults = get_xpath_object("//" XML_CIB_TAG_OPCONFIG,
data_set->input, LOG_NEVER);
data_set->rsc_defaults = get_xpath_object("//" XML_CIB_TAG_RSCCONFIG,
data_set->input, LOG_NEVER);
unpack_config(config, data_set);
if (is_not_set(data_set->flags, pe_flag_quick_location)
&& is_not_set(data_set->flags, pe_flag_have_quorum)
&& data_set->no_quorum_policy != no_quorum_ignore) {
crm_warn("Fencing and resource management disabled due to lack of quorum");
}
unpack_nodes(cib_nodes, data_set);
if(is_not_set(data_set->flags, pe_flag_quick_location)) {
unpack_remote_nodes(cib_resources, data_set);
}
unpack_resources(cib_resources, data_set);
unpack_tags(cib_tags, data_set);
if(is_not_set(data_set->flags, pe_flag_quick_location)) {
unpack_status(cib_status, data_set);
}
if (is_not_set(data_set->flags, pe_flag_no_counts)) {
for (GList *item = data_set->resources; item != NULL;
item = item->next) {
((pe_resource_t *) (item->data))->fns->count(item->data);
}
}
set_bit(data_set->flags, pe_flag_have_status);
return TRUE;
}
/*!
* \internal
* \brief Free a list of pe_resource_t
*
* \param[in] resources List to free
*
* \note When a working set's resource list is freed, that includes the original
* storage for the uname and id of any Pacemaker Remote nodes in the
* working set's node list, so take care not to use those afterward.
* \todo Refactor pe_node_t to strdup() the node name.
*/
static void
pe_free_resources(GListPtr resources)
{
pe_resource_t *rsc = NULL;
GListPtr iterator = resources;
while (iterator != NULL) {
rsc = (pe_resource_t *) iterator->data;
iterator = iterator->next;
rsc->fns->free(rsc);
}
if (resources != NULL) {
g_list_free(resources);
}
}
static void
pe_free_actions(GListPtr actions)
{
GListPtr iterator = actions;
while (iterator != NULL) {
pe_free_action(iterator->data);
iterator = iterator->next;
}
if (actions != NULL) {
g_list_free(actions);
}
}
static void
pe_free_nodes(GListPtr nodes)
{
for (GList *iterator = nodes; iterator != NULL; iterator = iterator->next) {
pe_node_t *node = (pe_node_t *) iterator->data;
// Shouldn't be possible, but to be safe ...
if (node == NULL) {
continue;
}
if (node->details == NULL) {
free(node);
continue;
}
/* This is called after pe_free_resources(), which means that we can't
* use node->details->uname for Pacemaker Remote nodes.
*/
crm_trace("Freeing node %s", (pe__is_guest_or_remote_node(node)?
"(guest or remote)" : node->details->uname));
if (node->details->attrs != NULL) {
g_hash_table_destroy(node->details->attrs);
}
if (node->details->utilization != NULL) {
g_hash_table_destroy(node->details->utilization);
}
if (node->details->digest_cache != NULL) {
g_hash_table_destroy(node->details->digest_cache);
}
g_list_free(node->details->running_rsc);
g_list_free(node->details->allocated_rsc);
free(node->details);
free(node);
}
if (nodes != NULL) {
g_list_free(nodes);
}
}
static void
pe__free_ordering(GListPtr constraints)
{
GListPtr iterator = constraints;
while (iterator != NULL) {
pe__ordering_t *order = iterator->data;
iterator = iterator->next;
free(order->lh_action_task);
free(order->rh_action_task);
free(order);
}
if (constraints != NULL) {
g_list_free(constraints);
}
}
static void
pe__free_location(GListPtr constraints)
{
GListPtr iterator = constraints;
while (iterator != NULL) {
pe__location_t *cons = iterator->data;
iterator = iterator->next;
g_list_free_full(cons->node_list_rh, free);
free(cons->id);
free(cons);
}
if (constraints != NULL) {
g_list_free(constraints);
}
}
/*!
* \brief Reset working set to default state without freeing it or constraints
*
* \param[in,out] data_set Working set to reset
*
* \deprecated This function is deprecated as part of the API;
* pe_reset_working_set() should be used instead.
*/
void
cleanup_calculations(pe_working_set_t * data_set)
{
if (data_set == NULL) {
return;
}
clear_bit(data_set->flags, pe_flag_have_status);
if (data_set->config_hash != NULL) {
g_hash_table_destroy(data_set->config_hash);
}
if (data_set->singletons != NULL) {
g_hash_table_destroy(data_set->singletons);
}
if (data_set->tickets) {
g_hash_table_destroy(data_set->tickets);
}
if (data_set->template_rsc_sets) {
g_hash_table_destroy(data_set->template_rsc_sets);
}
if (data_set->tags) {
g_hash_table_destroy(data_set->tags);
}
free(data_set->dc_uuid);
crm_trace("deleting resources");
pe_free_resources(data_set->resources);
crm_trace("deleting actions");
pe_free_actions(data_set->actions);
crm_trace("deleting nodes");
pe_free_nodes(data_set->nodes);
pe__free_param_checks(data_set);
g_list_free(data_set->stop_needed);
free_xml(data_set->graph);
crm_time_free(data_set->now);
free_xml(data_set->input);
free_xml(data_set->failed);
set_working_set_defaults(data_set);
CRM_CHECK(data_set->ordering_constraints == NULL,;
);
CRM_CHECK(data_set->placement_constraints == NULL,;
);
}
/*!
* \brief Reset a working set to default state without freeing it
*
* \param[in,out] data_set Working set to reset
*/
void
pe_reset_working_set(pe_working_set_t *data_set)
{
if (data_set == NULL) {
return;
}
crm_trace("Deleting %d ordering constraints",
g_list_length(data_set->ordering_constraints));
pe__free_ordering(data_set->ordering_constraints);
data_set->ordering_constraints = NULL;
crm_trace("Deleting %d location constraints",
g_list_length(data_set->placement_constraints));
pe__free_location(data_set->placement_constraints);
data_set->placement_constraints = NULL;
crm_trace("Deleting %d colocation constraints",
g_list_length(data_set->colocation_constraints));
g_list_free_full(data_set->colocation_constraints, free);
data_set->colocation_constraints = NULL;
crm_trace("Deleting %d ticket constraints",
g_list_length(data_set->ticket_constraints));
g_list_free_full(data_set->ticket_constraints, free);
data_set->ticket_constraints = NULL;
cleanup_calculations(data_set);
}
void
set_working_set_defaults(pe_working_set_t * data_set)
{
memset(data_set, 0, sizeof(pe_working_set_t));
data_set->order_id = 1;
data_set->action_id = 1;
data_set->no_quorum_policy = no_quorum_freeze;
data_set->flags = 0x0ULL;
set_bit(data_set->flags, pe_flag_stop_rsc_orphans);
set_bit(data_set->flags, pe_flag_symmetric_cluster);
set_bit(data_set->flags, pe_flag_stop_action_orphans);
#ifdef DEFAULT_CONCURRENT_FENCING_TRUE
set_bit(data_set->flags, pe_flag_concurrent_fencing);
#endif
}
pe_resource_t *
pe_find_resource(GListPtr rsc_list, const char *id)
{
return pe_find_resource_with_flags(rsc_list, id, pe_find_renamed);
}
pe_resource_t *
pe_find_resource_with_flags(GListPtr rsc_list, const char *id, enum pe_find flags)
{
GListPtr rIter = NULL;
for (rIter = rsc_list; id && rIter; rIter = rIter->next) {
pe_resource_t *parent = rIter->data;
pe_resource_t *match =
parent->fns->find_rsc(parent, id, NULL, flags);
if (match != NULL) {
return match;
}
}
crm_trace("No match for %s", id);
return NULL;
}
pe_node_t *
pe_find_node_any(GListPtr nodes, const char *id, const char *uname)
{
pe_node_t *match = pe_find_node_id(nodes, id);
if (match) {
return match;
}
crm_trace("Looking up %s via its uname instead", uname);
return pe_find_node(nodes, uname);
}
pe_node_t *
pe_find_node_id(GListPtr nodes, const char *id)
{
GListPtr gIter = nodes;
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
if (node && safe_str_eq(node->details->id, id)) {
return node;
}
}
/* error */
return NULL;
}
pe_node_t *
pe_find_node(GListPtr nodes, const char *uname)
{
GListPtr gIter = nodes;
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
if (node && safe_str_eq(node->details->uname, uname)) {
return node;
}
}
/* error */
return NULL;
}
diff --git a/lib/pengine/unpack.c b/lib/pengine/unpack.c
index aa51f9e02d..37138cedd8 100644
--- a/lib/pengine/unpack.c
+++ b/lib/pengine/unpack.c
@@ -1,3873 +1,3871 @@
/*
* Copyright 2004-2020 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 <stdio.h>
#include <string.h>
#include <glib.h>
#include <time.h>
#include <crm/crm.h>
#include <crm/services.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <crm/common/util.h>
#include <crm/pengine/rules.h>
#include <crm/pengine/internal.h>
#include <crm/common/iso8601_internal.h>
-#include <unpack.h>
#include <pe_status_private.h>
CRM_TRACE_INIT_DATA(pe_status);
#define set_config_flag(data_set, option, flag) do { \
const char *tmp = pe_pref(data_set->config_hash, option); \
if(tmp) { \
if(crm_is_true(tmp)) { \
set_bit(data_set->flags, flag); \
} else { \
clear_bit(data_set->flags, flag); \
} \
} \
} while(0)
static void unpack_rsc_op(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op,
xmlNode **last_failure,
enum action_fail_response *failed,
pe_working_set_t *data_set);
static void determine_remote_online_status(pe_working_set_t *data_set,
pe_node_t *this_node);
static void add_node_attrs(xmlNode *attrs, pe_node_t *node, bool overwrite,
pe_working_set_t *data_set);
static void determine_online_status(xmlNode *node_state, pe_node_t *this_node,
pe_working_set_t *data_set);
+static void unpack_lrm_resources(pe_node_t *node, xmlNode *lrm_state,
+ pe_working_set_t *data_set);
+
// Bitmask for warnings we only want to print once
uint32_t pe_wo = 0;
static gboolean
is_dangling_guest_node(pe_node_t *node)
{
/* we are looking for a remote-node that was supposed to be mapped to a
* container resource, but all traces of that container have disappeared
* from both the config and the status section. */
if (pe__is_guest_or_remote_node(node) &&
node->details->remote_rsc &&
node->details->remote_rsc->container == NULL &&
is_set(node->details->remote_rsc->flags, pe_rsc_orphan_container_filler)) {
return TRUE;
}
return FALSE;
}
/*!
* \brief Schedule a fence action for a node
*
* \param[in,out] data_set Current working set of cluster
* \param[in,out] node Node to fence
* \param[in] reason Text description of why fencing is needed
* \param[in] priority_delay Whether to consider `priority-fencing-delay`
*/
void
pe_fence_node(pe_working_set_t * data_set, pe_node_t * node,
const char *reason, bool priority_delay)
{
CRM_CHECK(node, return);
/* A guest node is fenced by marking its container as failed */
if (pe__is_guest_node(node)) {
pe_resource_t *rsc = node->details->remote_rsc->container;
if (is_set(rsc->flags, pe_rsc_failed) == FALSE) {
if (!is_set(rsc->flags, pe_rsc_managed)) {
crm_notice("Not fencing guest node %s "
"(otherwise would because %s): "
"its guest resource %s is unmanaged",
node->details->uname, reason, rsc->id);
} else {
crm_warn("Guest node %s will be fenced "
"(by recovering its guest resource %s): %s",
node->details->uname, rsc->id, reason);
/* We don't mark the node as unclean because that would prevent the
* node from running resources. We want to allow it to run resources
* in this transition if the recovery succeeds.
*/
node->details->remote_requires_reset = TRUE;
set_bit(rsc->flags, pe_rsc_failed);
}
}
} else if (is_dangling_guest_node(node)) {
crm_info("Cleaning up dangling connection for guest node %s: "
"fencing was already done because %s, "
"and guest resource no longer exists",
node->details->uname, reason);
set_bit(node->details->remote_rsc->flags, pe_rsc_failed);
} else if (pe__is_remote_node(node)) {
pe_resource_t *rsc = node->details->remote_rsc;
if (rsc && (!is_set(rsc->flags, pe_rsc_managed))) {
crm_notice("Not fencing remote node %s "
"(otherwise would because %s): connection is unmanaged",
node->details->uname, reason);
} else if(node->details->remote_requires_reset == FALSE) {
node->details->remote_requires_reset = TRUE;
crm_warn("Remote node %s %s: %s",
node->details->uname,
pe_can_fence(data_set, node)? "will be fenced" : "is unclean",
reason);
}
node->details->unclean = TRUE;
// No need to apply `priority-fencing-delay` for remote nodes
pe_fence_op(node, NULL, TRUE, reason, FALSE, data_set);
} else if (node->details->unclean) {
crm_trace("Cluster node %s %s because %s",
node->details->uname,
pe_can_fence(data_set, node)? "would also be fenced" : "also is unclean",
reason);
} else {
crm_warn("Cluster node %s %s: %s",
node->details->uname,
pe_can_fence(data_set, node)? "will be fenced" : "is unclean",
reason);
node->details->unclean = TRUE;
pe_fence_op(node, NULL, TRUE, reason, priority_delay, data_set);
}
}
// @TODO xpaths can't handle templates, rules, or id-refs
// nvpair with provides or requires set to unfencing
#define XPATH_UNFENCING_NVPAIR XML_CIB_TAG_NVPAIR \
"[(@" XML_NVPAIR_ATTR_NAME "='" XML_RSC_ATTR_PROVIDES "'" \
"or @" XML_NVPAIR_ATTR_NAME "='" XML_RSC_ATTR_REQUIRES "') " \
"and @" XML_NVPAIR_ATTR_VALUE "='unfencing']"
// unfencing in rsc_defaults or any resource
#define XPATH_ENABLE_UNFENCING \
"/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_RESOURCES \
"//" XML_TAG_META_SETS "/" XPATH_UNFENCING_NVPAIR \
"|/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_RSCCONFIG \
"/" XML_TAG_META_SETS "/" XPATH_UNFENCING_NVPAIR
static
void set_if_xpath(unsigned long long flag, const char *xpath,
pe_working_set_t *data_set)
{
xmlXPathObjectPtr result = NULL;
if (is_not_set(data_set->flags, flag)) {
result = xpath_search(data_set->input, xpath);
if (result && (numXpathResults(result) > 0)) {
set_bit(data_set->flags, flag);
}
freeXpathObject(result);
}
}
gboolean
unpack_config(xmlNode * config, pe_working_set_t * data_set)
{
const char *value = NULL;
GHashTable *config_hash = crm_str_table_new();
data_set->config_hash = config_hash;
pe__unpack_dataset_nvpairs(config, XML_CIB_TAG_PROPSET, NULL, config_hash,
CIB_OPTIONS_FIRST, FALSE, data_set);
verify_pe_options(data_set->config_hash);
set_config_flag(data_set, "enable-startup-probes", pe_flag_startup_probes);
if(is_not_set(data_set->flags, pe_flag_startup_probes)) {
crm_info("Startup probes: disabled (dangerous)");
}
value = pe_pref(data_set->config_hash, XML_ATTR_HAVE_WATCHDOG);
if (value && crm_is_true(value)) {
crm_notice("Watchdog will be used via SBD if fencing is required "
"and stonith-watchdog-timeout is nonzero");
set_bit(data_set->flags, pe_flag_have_stonith_resource);
}
/* Set certain flags via xpath here, so they can be used before the relevant
* configuration sections are unpacked.
*/
set_if_xpath(pe_flag_enable_unfencing, XPATH_ENABLE_UNFENCING, data_set);
value = pe_pref(data_set->config_hash, "stonith-timeout");
data_set->stonith_timeout = (int) crm_parse_interval_spec(value);
crm_debug("STONITH timeout: %d", data_set->stonith_timeout);
set_config_flag(data_set, "stonith-enabled", pe_flag_stonith_enabled);
crm_debug("STONITH of failed nodes is %s",
is_set(data_set->flags, pe_flag_stonith_enabled) ? "enabled" : "disabled");
data_set->stonith_action = pe_pref(data_set->config_hash, "stonith-action");
if (!strcmp(data_set->stonith_action, "poweroff")) {
pe_warn_once(pe_wo_poweroff,
"Support for stonith-action of 'poweroff' is deprecated "
"and will be removed in a future release (use 'off' instead)");
data_set->stonith_action = "off";
}
crm_trace("STONITH will %s nodes", data_set->stonith_action);
set_config_flag(data_set, "concurrent-fencing", pe_flag_concurrent_fencing);
crm_debug("Concurrent fencing is %s",
is_set(data_set->flags, pe_flag_concurrent_fencing) ? "enabled" : "disabled");
// Default value -1 means `priority-fencing-delay` is disabled
data_set->priority_fencing_delay = -1;
value = pe_pref(data_set->config_hash,
XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY);
if (value) {
data_set->priority_fencing_delay = crm_parse_interval_spec(value) / 1000;
crm_trace("Priority fencing delay is %ds", data_set->priority_fencing_delay);
}
set_config_flag(data_set, "stop-all-resources", pe_flag_stop_everything);
crm_debug("Stop all active resources: %s",
is_set(data_set->flags, pe_flag_stop_everything) ? "true" : "false");
set_config_flag(data_set, "symmetric-cluster", pe_flag_symmetric_cluster);
if (is_set(data_set->flags, pe_flag_symmetric_cluster)) {
crm_debug("Cluster is symmetric" " - resources can run anywhere by default");
}
value = pe_pref(data_set->config_hash, "no-quorum-policy");
if (safe_str_eq(value, "ignore")) {
data_set->no_quorum_policy = no_quorum_ignore;
} else if (safe_str_eq(value, "freeze")) {
data_set->no_quorum_policy = no_quorum_freeze;
} else if (safe_str_eq(value, "suicide")) {
if (is_set(data_set->flags, pe_flag_stonith_enabled)) {
int do_panic = 0;
crm_element_value_int(data_set->input, XML_ATTR_QUORUM_PANIC,
&do_panic);
if (do_panic || is_set(data_set->flags, pe_flag_have_quorum)) {
data_set->no_quorum_policy = no_quorum_suicide;
} else {
crm_notice("Resetting no-quorum-policy to 'stop': cluster has never had quorum");
data_set->no_quorum_policy = no_quorum_stop;
}
} else {
pcmk__config_err("Resetting no-quorum-policy to 'stop' because "
"fencing is disabled");
data_set->no_quorum_policy = no_quorum_stop;
}
} else {
data_set->no_quorum_policy = no_quorum_stop;
}
switch (data_set->no_quorum_policy) {
case no_quorum_freeze:
crm_debug("On loss of quorum: Freeze resources");
break;
case no_quorum_stop:
crm_debug("On loss of quorum: Stop ALL resources");
break;
case no_quorum_suicide:
crm_notice("On loss of quorum: Fence all remaining nodes");
break;
case no_quorum_ignore:
crm_notice("On loss of quorum: Ignore");
break;
}
set_config_flag(data_set, "stop-orphan-resources", pe_flag_stop_rsc_orphans);
crm_trace("Orphan resources are %s",
is_set(data_set->flags, pe_flag_stop_rsc_orphans) ? "stopped" : "ignored");
set_config_flag(data_set, "stop-orphan-actions", pe_flag_stop_action_orphans);
crm_trace("Orphan resource actions are %s",
is_set(data_set->flags, pe_flag_stop_action_orphans) ? "stopped" : "ignored");
set_config_flag(data_set, "remove-after-stop", pe_flag_remove_after_stop);
crm_trace("Stopped resources are removed from the status section: %s",
is_set(data_set->flags, pe_flag_remove_after_stop) ? "true" : "false");
set_config_flag(data_set, "maintenance-mode", pe_flag_maintenance_mode);
crm_trace("Maintenance mode: %s",
is_set(data_set->flags, pe_flag_maintenance_mode) ? "true" : "false");
set_config_flag(data_set, "start-failure-is-fatal", pe_flag_start_failure_fatal);
crm_trace("Start failures are %s",
is_set(data_set->flags,
pe_flag_start_failure_fatal) ? "always fatal" : "handled by failcount");
if (is_set(data_set->flags, pe_flag_stonith_enabled)) {
set_config_flag(data_set, "startup-fencing", pe_flag_startup_fencing);
}
if (is_set(data_set->flags, pe_flag_startup_fencing)) {
crm_trace("Unseen nodes will be fenced");
} else {
pe_warn_once(pe_wo_blind, "Blind faith: not fencing unseen nodes");
}
pcmk__score_red = char2score(pe_pref(data_set->config_hash, "node-health-red"));
pcmk__score_green = char2score(pe_pref(data_set->config_hash, "node-health-green"));
pcmk__score_yellow = char2score(pe_pref(data_set->config_hash, "node-health-yellow"));
crm_debug("Node scores: 'red' = %s, 'yellow' = %s, 'green' = %s",
pe_pref(data_set->config_hash, "node-health-red"),
pe_pref(data_set->config_hash, "node-health-yellow"),
pe_pref(data_set->config_hash, "node-health-green"));
data_set->placement_strategy = pe_pref(data_set->config_hash, "placement-strategy");
crm_trace("Placement strategy: %s", data_set->placement_strategy);
set_config_flag(data_set, "shutdown-lock", pe_flag_shutdown_lock);
crm_trace("Resources will%s be locked to cleanly shut down nodes",
(is_set(data_set->flags, pe_flag_shutdown_lock)? "" : " not"));
if (is_set(data_set->flags, pe_flag_shutdown_lock)) {
value = pe_pref(data_set->config_hash,
XML_CONFIG_ATTR_SHUTDOWN_LOCK_LIMIT);
data_set->shutdown_lock = crm_parse_interval_spec(value) / 1000;
crm_trace("Shutdown locks expire after %us", data_set->shutdown_lock);
}
return TRUE;
}
static void
destroy_digest_cache(gpointer ptr)
{
op_digest_cache_t *data = ptr;
free_xml(data->params_all);
free_xml(data->params_secure);
free_xml(data->params_restart);
free(data->digest_all_calc);
free(data->digest_restart_calc);
free(data->digest_secure_calc);
free(data);
}
pe_node_t *
pe_create_node(const char *id, const char *uname, const char *type,
const char *score, pe_working_set_t * data_set)
{
pe_node_t *new_node = NULL;
if (pe_find_node(data_set->nodes, uname) != NULL) {
pcmk__config_warn("More than one node entry has name '%s'", uname);
}
new_node = calloc(1, sizeof(pe_node_t));
if (new_node == NULL) {
return NULL;
}
new_node->weight = char2score(score);
new_node->fixed = FALSE;
new_node->details = calloc(1, sizeof(struct pe_node_shared_s));
if (new_node->details == NULL) {
free(new_node);
return NULL;
}
crm_trace("Creating node for entry %s/%s", uname, id);
new_node->details->id = id;
new_node->details->uname = uname;
new_node->details->online = FALSE;
new_node->details->shutdown = FALSE;
new_node->details->rsc_discovery_enabled = TRUE;
new_node->details->running_rsc = NULL;
new_node->details->type = node_ping;
if (safe_str_eq(type, "remote")) {
new_node->details->type = node_remote;
set_bit(data_set->flags, pe_flag_have_remote_nodes);
} else if ((type == NULL) || safe_str_eq(type, "member")) {
new_node->details->type = node_member;
}
new_node->details->attrs = crm_str_table_new();
if (pe__is_guest_or_remote_node(new_node)) {
g_hash_table_insert(new_node->details->attrs, strdup(CRM_ATTR_KIND),
strdup("remote"));
} else {
g_hash_table_insert(new_node->details->attrs, strdup(CRM_ATTR_KIND),
strdup("cluster"));
}
new_node->details->utilization = crm_str_table_new();
new_node->details->digest_cache = g_hash_table_new_full(crm_str_hash,
g_str_equal, free,
destroy_digest_cache);
data_set->nodes = g_list_insert_sorted(data_set->nodes, new_node, sort_node_uname);
return new_node;
}
static const char *
expand_remote_rsc_meta(xmlNode *xml_obj, xmlNode *parent, pe_working_set_t *data)
{
xmlNode *attr_set = NULL;
xmlNode *attr = NULL;
const char *container_id = ID(xml_obj);
const char *remote_name = NULL;
const char *remote_server = NULL;
const char *remote_port = NULL;
const char *connect_timeout = "60s";
const char *remote_allow_migrate=NULL;
const char *is_managed = NULL;
for (attr_set = __xml_first_child_element(xml_obj); attr_set != NULL;
attr_set = __xml_next_element(attr_set)) {
if (safe_str_neq((const char *)attr_set->name, XML_TAG_META_SETS)) {
continue;
}
for (attr = __xml_first_child_element(attr_set); attr != NULL;
attr = __xml_next_element(attr)) {
const char *value = crm_element_value(attr, XML_NVPAIR_ATTR_VALUE);
const char *name = crm_element_value(attr, XML_NVPAIR_ATTR_NAME);
if (safe_str_eq(name, XML_RSC_ATTR_REMOTE_NODE)) {
remote_name = value;
} else if (safe_str_eq(name, "remote-addr")) {
remote_server = value;
} else if (safe_str_eq(name, "remote-port")) {
remote_port = value;
} else if (safe_str_eq(name, "remote-connect-timeout")) {
connect_timeout = value;
} else if (safe_str_eq(name, "remote-allow-migrate")) {
remote_allow_migrate=value;
} else if (safe_str_eq(name, XML_RSC_ATTR_MANAGED)) {
is_managed = value;
}
}
}
if (remote_name == NULL) {
return NULL;
}
if (pe_find_resource(data->resources, remote_name) != NULL) {
return NULL;
}
pe_create_remote_xml(parent, remote_name, container_id,
remote_allow_migrate, is_managed,
connect_timeout, remote_server, remote_port);
return remote_name;
}
static void
handle_startup_fencing(pe_working_set_t *data_set, pe_node_t *new_node)
{
if ((new_node->details->type == node_remote) && (new_node->details->remote_rsc == NULL)) {
/* Ignore fencing for remote nodes that don't have a connection resource
* associated with them. This happens when remote node entries get left
* in the nodes section after the connection resource is removed.
*/
return;
}
if (is_set(data_set->flags, pe_flag_startup_fencing)) {
// All nodes are unclean until we've seen their status entry
new_node->details->unclean = TRUE;
} else {
// Blind faith ...
new_node->details->unclean = FALSE;
}
/* We need to be able to determine if a node's status section
* exists or not separate from whether the node is unclean. */
new_node->details->unseen = TRUE;
}
gboolean
unpack_nodes(xmlNode * xml_nodes, pe_working_set_t * data_set)
{
xmlNode *xml_obj = NULL;
pe_node_t *new_node = NULL;
const char *id = NULL;
const char *uname = NULL;
const char *type = NULL;
const char *score = NULL;
for (xml_obj = __xml_first_child_element(xml_nodes); xml_obj != NULL;
xml_obj = __xml_next_element(xml_obj)) {
if (crm_str_eq((const char *)xml_obj->name, XML_CIB_TAG_NODE, TRUE)) {
new_node = NULL;
id = crm_element_value(xml_obj, XML_ATTR_ID);
uname = crm_element_value(xml_obj, XML_ATTR_UNAME);
type = crm_element_value(xml_obj, XML_ATTR_TYPE);
score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE);
crm_trace("Processing node %s/%s", uname, id);
if (id == NULL) {
pcmk__config_err("Ignoring <" XML_CIB_TAG_NODE
"> entry in configuration without id");
continue;
}
new_node = pe_create_node(id, uname, type, score, data_set);
if (new_node == NULL) {
return FALSE;
}
/* if(data_set->have_quorum == FALSE */
/* && data_set->no_quorum_policy == no_quorum_stop) { */
/* /\* start shutting resources down *\/ */
/* new_node->weight = -INFINITY; */
/* } */
handle_startup_fencing(data_set, new_node);
add_node_attrs(xml_obj, new_node, FALSE, data_set);
pe__unpack_dataset_nvpairs(xml_obj, XML_TAG_UTILIZATION, NULL,
new_node->details->utilization, NULL,
FALSE, data_set);
crm_trace("Done with node %s", crm_element_value(xml_obj, XML_ATTR_UNAME));
}
}
if (data_set->localhost && pe_find_node(data_set->nodes, data_set->localhost) == NULL) {
crm_info("Creating a fake local node");
pe_create_node(data_set->localhost, data_set->localhost, NULL, 0,
data_set);
}
return TRUE;
}
static void
setup_container(pe_resource_t * rsc, pe_working_set_t * data_set)
{
const char *container_id = NULL;
if (rsc->children) {
GListPtr gIter = rsc->children;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
setup_container(child_rsc, data_set);
}
return;
}
container_id = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_CONTAINER);
if (container_id && safe_str_neq(container_id, rsc->id)) {
pe_resource_t *container = pe_find_resource(data_set->resources, container_id);
if (container) {
rsc->container = container;
set_bit(container->flags, pe_rsc_is_container);
container->fillers = g_list_append(container->fillers, rsc);
pe_rsc_trace(rsc, "Resource %s's container is %s", rsc->id, container_id);
} else {
pe_err("Resource %s: Unknown resource container (%s)", rsc->id, container_id);
}
}
}
gboolean
unpack_remote_nodes(xmlNode * xml_resources, pe_working_set_t * data_set)
{
xmlNode *xml_obj = NULL;
/* Create remote nodes and guest nodes from the resource configuration
* before unpacking resources.
*/
for (xml_obj = __xml_first_child_element(xml_resources); xml_obj != NULL;
xml_obj = __xml_next_element(xml_obj)) {
const char *new_node_id = NULL;
/* Check for remote nodes, which are defined by ocf:pacemaker:remote
* primitives.
*/
if (xml_contains_remote_node(xml_obj)) {
new_node_id = ID(xml_obj);
/* The "pe_find_node" check is here to make sure we don't iterate over
* an expanded node that has already been added to the node list. */
if (new_node_id && pe_find_node(data_set->nodes, new_node_id) == NULL) {
crm_trace("Found remote node %s defined by resource %s",
new_node_id, ID(xml_obj));
pe_create_node(new_node_id, new_node_id, "remote", NULL,
data_set);
}
continue;
}
/* Check for guest nodes, which are defined by special meta-attributes
* of a primitive of any type (for example, VirtualDomain or Xen).
*/
if (crm_str_eq((const char *)xml_obj->name, XML_CIB_TAG_RESOURCE, TRUE)) {
/* This will add an ocf:pacemaker:remote primitive to the
* configuration for the guest node's connection, to be unpacked
* later.
*/
new_node_id = expand_remote_rsc_meta(xml_obj, xml_resources, data_set);
if (new_node_id && pe_find_node(data_set->nodes, new_node_id) == NULL) {
crm_trace("Found guest node %s in resource %s",
new_node_id, ID(xml_obj));
pe_create_node(new_node_id, new_node_id, "remote", NULL,
data_set);
}
continue;
}
/* Check for guest nodes inside a group. Clones are currently not
* supported as guest nodes.
*/
if (crm_str_eq((const char *)xml_obj->name, XML_CIB_TAG_GROUP, TRUE)) {
xmlNode *xml_obj2 = NULL;
for (xml_obj2 = __xml_first_child_element(xml_obj); xml_obj2 != NULL;
xml_obj2 = __xml_next_element(xml_obj2)) {
new_node_id = expand_remote_rsc_meta(xml_obj2, xml_resources, data_set);
if (new_node_id && pe_find_node(data_set->nodes, new_node_id) == NULL) {
crm_trace("Found guest node %s in resource %s inside group %s",
new_node_id, ID(xml_obj2), ID(xml_obj));
pe_create_node(new_node_id, new_node_id, "remote", NULL,
data_set);
}
}
}
}
return TRUE;
}
/* Call this after all the nodes and resources have been
* unpacked, but before the status section is read.
*
* A remote node's online status is reflected by the state
* of the remote node's connection resource. We need to link
* the remote node to this connection resource so we can have
* easy access to the connection resource during the scheduler calculations.
*/
static void
link_rsc2remotenode(pe_working_set_t *data_set, pe_resource_t *new_rsc)
{
pe_node_t *remote_node = NULL;
if (new_rsc->is_remote_node == FALSE) {
return;
}
if (is_set(data_set->flags, pe_flag_quick_location)) {
/* remote_nodes and remote_resources are not linked in quick location calculations */
return;
}
remote_node = pe_find_node(data_set->nodes, new_rsc->id);
CRM_CHECK(remote_node != NULL, return;);
pe_rsc_trace(new_rsc, "Linking remote connection resource %s to node %s",
new_rsc->id, remote_node->details->uname);
remote_node->details->remote_rsc = new_rsc;
if (new_rsc->container == NULL) {
/* Handle start-up fencing for remote nodes (as opposed to guest nodes)
* the same as is done for cluster nodes.
*/
handle_startup_fencing(data_set, remote_node);
} else {
/* pe_create_node() marks the new node as "remote" or "cluster"; now
* that we know the node is a guest node, update it correctly.
*/
g_hash_table_replace(remote_node->details->attrs, strdup(CRM_ATTR_KIND),
strdup("container"));
}
}
static void
destroy_tag(gpointer data)
{
pe_tag_t *tag = data;
if (tag) {
free(tag->id);
g_list_free_full(tag->refs, free);
free(tag);
}
}
/*!
* \internal
* \brief Parse configuration XML for resource information
*
* \param[in] xml_resources Top of resource configuration XML
* \param[in,out] data_set Where to put resource information
*
* \return TRUE
*
* \note unpack_remote_nodes() MUST be called before this, so that the nodes can
* be used when common_unpack() calls resource_location()
*/
gboolean
unpack_resources(xmlNode * xml_resources, pe_working_set_t * data_set)
{
xmlNode *xml_obj = NULL;
GListPtr gIter = NULL;
data_set->template_rsc_sets = g_hash_table_new_full(crm_str_hash,
g_str_equal, free,
destroy_tag);
for (xml_obj = __xml_first_child_element(xml_resources); xml_obj != NULL;
xml_obj = __xml_next_element(xml_obj)) {
pe_resource_t *new_rsc = NULL;
if (crm_str_eq((const char *)xml_obj->name, XML_CIB_TAG_RSC_TEMPLATE, TRUE)) {
const char *template_id = ID(xml_obj);
if (template_id && g_hash_table_lookup_extended(data_set->template_rsc_sets,
template_id, NULL, NULL) == FALSE) {
/* Record the template's ID for the knowledge of its existence anyway. */
g_hash_table_insert(data_set->template_rsc_sets, strdup(template_id), NULL);
}
continue;
}
crm_trace("Beginning unpack... <%s id=%s... >", crm_element_name(xml_obj), ID(xml_obj));
if (common_unpack(xml_obj, &new_rsc, NULL, data_set)) {
data_set->resources = g_list_append(data_set->resources, new_rsc);
pe_rsc_trace(new_rsc, "Added resource %s", new_rsc->id);
} else {
pcmk__config_err("Ignoring <%s> resource '%s' "
"because configuration is invalid",
crm_element_name(xml_obj), crm_str(ID(xml_obj)));
if (new_rsc != NULL && new_rsc->fns != NULL) {
new_rsc->fns->free(new_rsc);
}
}
}
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
setup_container(rsc, data_set);
link_rsc2remotenode(data_set, rsc);
}
data_set->resources = g_list_sort(data_set->resources, sort_rsc_priority);
if (is_set(data_set->flags, pe_flag_quick_location)) {
/* Ignore */
} else if (is_set(data_set->flags, pe_flag_stonith_enabled)
&& is_set(data_set->flags, pe_flag_have_stonith_resource) == FALSE) {
pcmk__config_err("Resource start-up disabled since no STONITH resources have been defined");
pcmk__config_err("Either configure some or disable STONITH with the stonith-enabled option");
pcmk__config_err("NOTE: Clusters with shared data need STONITH to ensure data integrity");
}
return TRUE;
}
gboolean
unpack_tags(xmlNode * xml_tags, pe_working_set_t * data_set)
{
xmlNode *xml_tag = NULL;
data_set->tags = g_hash_table_new_full(crm_str_hash, g_str_equal, free,
destroy_tag);
for (xml_tag = __xml_first_child_element(xml_tags); xml_tag != NULL;
xml_tag = __xml_next_element(xml_tag)) {
xmlNode *xml_obj_ref = NULL;
const char *tag_id = ID(xml_tag);
if (crm_str_eq((const char *)xml_tag->name, XML_CIB_TAG_TAG, TRUE) == FALSE) {
continue;
}
if (tag_id == NULL) {
pcmk__config_err("Ignoring <%s> without " XML_ATTR_ID,
crm_element_name(xml_tag));
continue;
}
for (xml_obj_ref = __xml_first_child_element(xml_tag); xml_obj_ref != NULL;
xml_obj_ref = __xml_next_element(xml_obj_ref)) {
const char *obj_ref = ID(xml_obj_ref);
if (crm_str_eq((const char *)xml_obj_ref->name, XML_CIB_TAG_OBJ_REF, TRUE) == FALSE) {
continue;
}
if (obj_ref == NULL) {
pcmk__config_err("Ignoring <%s> for tag '%s' without " XML_ATTR_ID,
crm_element_name(xml_obj_ref), tag_id);
continue;
}
if (add_tag_ref(data_set->tags, tag_id, obj_ref) == FALSE) {
return FALSE;
}
}
}
return TRUE;
}
/* The ticket state section:
* "/cib/status/tickets/ticket_state" */
static gboolean
unpack_ticket_state(xmlNode * xml_ticket, pe_working_set_t * data_set)
{
const char *ticket_id = NULL;
const char *granted = NULL;
const char *last_granted = NULL;
const char *standby = NULL;
xmlAttrPtr xIter = NULL;
pe_ticket_t *ticket = NULL;
ticket_id = ID(xml_ticket);
if (ticket_id == NULL || strlen(ticket_id) == 0) {
return FALSE;
}
crm_trace("Processing ticket state for %s", ticket_id);
ticket = g_hash_table_lookup(data_set->tickets, ticket_id);
if (ticket == NULL) {
ticket = ticket_new(ticket_id, data_set);
if (ticket == NULL) {
return FALSE;
}
}
for (xIter = xml_ticket->properties; xIter; xIter = xIter->next) {
const char *prop_name = (const char *)xIter->name;
const char *prop_value = crm_element_value(xml_ticket, prop_name);
if (crm_str_eq(prop_name, XML_ATTR_ID, TRUE)) {
continue;
}
g_hash_table_replace(ticket->state, strdup(prop_name), strdup(prop_value));
}
granted = g_hash_table_lookup(ticket->state, "granted");
if (granted && crm_is_true(granted)) {
ticket->granted = TRUE;
crm_info("We have ticket '%s'", ticket->id);
} else {
ticket->granted = FALSE;
crm_info("We do not have ticket '%s'", ticket->id);
}
last_granted = g_hash_table_lookup(ticket->state, "last-granted");
if (last_granted) {
ticket->last_granted = crm_parse_int(last_granted, 0);
}
standby = g_hash_table_lookup(ticket->state, "standby");
if (standby && crm_is_true(standby)) {
ticket->standby = TRUE;
if (ticket->granted) {
crm_info("Granted ticket '%s' is in standby-mode", ticket->id);
}
} else {
ticket->standby = FALSE;
}
crm_trace("Done with ticket state for %s", ticket_id);
return TRUE;
}
static gboolean
unpack_tickets_state(xmlNode * xml_tickets, pe_working_set_t * data_set)
{
xmlNode *xml_obj = NULL;
for (xml_obj = __xml_first_child_element(xml_tickets); xml_obj != NULL;
xml_obj = __xml_next_element(xml_obj)) {
if (crm_str_eq((const char *)xml_obj->name, XML_CIB_TAG_TICKET_STATE, TRUE) == FALSE) {
continue;
}
unpack_ticket_state(xml_obj, data_set);
}
return TRUE;
}
static void
unpack_handle_remote_attrs(pe_node_t *this_node, xmlNode *state, pe_working_set_t * data_set)
{
const char *resource_discovery_enabled = NULL;
xmlNode *attrs = NULL;
pe_resource_t *rsc = NULL;
if (crm_str_eq((const char *)state->name, XML_CIB_TAG_STATE, TRUE) == FALSE) {
return;
}
if ((this_node == NULL) || !pe__is_guest_or_remote_node(this_node)) {
return;
}
crm_trace("Processing remote node id=%s, uname=%s", this_node->details->id, this_node->details->uname);
this_node->details->remote_maintenance =
crm_atoi(crm_element_value(state, XML_NODE_IS_MAINTENANCE), "0");
rsc = this_node->details->remote_rsc;
if (this_node->details->remote_requires_reset == FALSE) {
this_node->details->unclean = FALSE;
this_node->details->unseen = FALSE;
}
attrs = find_xml_node(state, XML_TAG_TRANSIENT_NODEATTRS, FALSE);
add_node_attrs(attrs, this_node, TRUE, data_set);
if (pe__shutdown_requested(this_node)) {
crm_info("Node %s is shutting down", this_node->details->uname);
this_node->details->shutdown = TRUE;
if (rsc) {
rsc->next_role = RSC_ROLE_STOPPED;
}
}
if (crm_is_true(pe_node_attribute_raw(this_node, "standby"))) {
crm_info("Node %s is in standby-mode", this_node->details->uname);
this_node->details->standby = TRUE;
}
if (crm_is_true(pe_node_attribute_raw(this_node, "maintenance")) ||
(rsc && !is_set(rsc->flags, pe_rsc_managed))) {
crm_info("Node %s is in maintenance-mode", this_node->details->uname);
this_node->details->maintenance = TRUE;
}
resource_discovery_enabled = pe_node_attribute_raw(this_node, XML_NODE_ATTR_RSC_DISCOVERY);
if (resource_discovery_enabled && !crm_is_true(resource_discovery_enabled)) {
if (pe__is_remote_node(this_node)
&& is_not_set(data_set->flags, pe_flag_stonith_enabled)) {
crm_warn("Ignoring %s attribute on remote node %s because stonith is disabled",
XML_NODE_ATTR_RSC_DISCOVERY, this_node->details->uname);
} else {
/* This is either a remote node with fencing enabled, or a guest
* node. We don't care whether fencing is enabled when fencing guest
* nodes, because they are "fenced" by recovering their containing
* resource.
*/
crm_info("Node %s has resource discovery disabled", this_node->details->uname);
this_node->details->rsc_discovery_enabled = FALSE;
}
}
}
static bool
unpack_node_loop(xmlNode * status, bool fence, pe_working_set_t * data_set)
{
bool changed = false;
xmlNode *lrm_rsc = NULL;
for (xmlNode *state = __xml_first_child_element(status); state != NULL;
state = __xml_next_element(state)) {
const char *id = NULL;
const char *uname = NULL;
pe_node_t *this_node = NULL;
bool process = FALSE;
if (crm_str_eq((const char *)state->name, XML_CIB_TAG_STATE, TRUE) == FALSE) {
continue;
}
id = crm_element_value(state, XML_ATTR_ID);
uname = crm_element_value(state, XML_ATTR_UNAME);
this_node = pe_find_node_any(data_set->nodes, id, uname);
if (this_node == NULL) {
crm_info("Node %s is unknown", id);
continue;
} else if (this_node->details->unpacked) {
crm_trace("Node %s was already processed", id);
continue;
} else if (!pe__is_guest_or_remote_node(this_node)
&& is_set(data_set->flags, pe_flag_stonith_enabled)) {
// A redundant test, but preserves the order for regression tests
process = TRUE;
} else if (pe__is_guest_or_remote_node(this_node)) {
bool check = FALSE;
pe_resource_t *rsc = this_node->details->remote_rsc;
if(fence) {
check = TRUE;
} else if(rsc == NULL) {
/* Not ready yet */
} else if (pe__is_guest_node(this_node)
&& rsc->role == RSC_ROLE_STARTED
&& rsc->container->role == RSC_ROLE_STARTED) {
/* Both the connection and its containing resource need to be
* known to be up before we process resources running in it.
*/
check = TRUE;
crm_trace("Checking node %s/%s/%s status %d/%d/%d", id, rsc->id, rsc->container->id, fence, rsc->role, RSC_ROLE_STARTED);
} else if (!pe__is_guest_node(this_node)
&& ((rsc->role == RSC_ROLE_STARTED)
|| is_set(data_set->flags, pe_flag_shutdown_lock))) {
check = TRUE;
crm_trace("Checking node %s/%s status %d/%d/%d", id, rsc->id, fence, rsc->role, RSC_ROLE_STARTED);
}
if (check) {
determine_remote_online_status(data_set, this_node);
unpack_handle_remote_attrs(this_node, state, data_set);
process = TRUE;
}
} else if (this_node->details->online) {
process = TRUE;
} else if (fence) {
process = TRUE;
} else if (is_set(data_set->flags, pe_flag_shutdown_lock)) {
process = TRUE;
}
if(process) {
crm_trace("Processing lrm resource entries on %shealthy%s node: %s",
fence?"un":"",
(pe__is_guest_or_remote_node(this_node)? " remote" : ""),
this_node->details->uname);
changed = TRUE;
this_node->details->unpacked = TRUE;
lrm_rsc = find_xml_node(state, XML_CIB_TAG_LRM, FALSE);
lrm_rsc = find_xml_node(lrm_rsc, XML_LRM_TAG_RESOURCES, FALSE);
unpack_lrm_resources(this_node, lrm_rsc, data_set);
}
}
return changed;
}
/* remove nodes that are down, stopping */
/* create positive rsc_to_node constraints between resources and the nodes they are running on */
/* anything else? */
gboolean
unpack_status(xmlNode * status, pe_working_set_t * data_set)
{
const char *id = NULL;
const char *uname = NULL;
xmlNode *state = NULL;
pe_node_t *this_node = NULL;
crm_trace("Beginning unpack");
if (data_set->tickets == NULL) {
data_set->tickets = g_hash_table_new_full(crm_str_hash, g_str_equal,
free, destroy_ticket);
}
for (state = __xml_first_child_element(status); state != NULL;
state = __xml_next_element(state)) {
if (crm_str_eq((const char *)state->name, XML_CIB_TAG_TICKETS, TRUE)) {
unpack_tickets_state((xmlNode *) state, data_set);
} else if (crm_str_eq((const char *)state->name, XML_CIB_TAG_STATE, TRUE)) {
xmlNode *attrs = NULL;
const char *resource_discovery_enabled = NULL;
id = crm_element_value(state, XML_ATTR_ID);
uname = crm_element_value(state, XML_ATTR_UNAME);
this_node = pe_find_node_any(data_set->nodes, id, uname);
if (uname == NULL) {
/* error */
continue;
} else if (this_node == NULL) {
pcmk__config_warn("Ignoring recorded node status for '%s' "
"because no longer in configuration", uname);
continue;
} else if (pe__is_guest_or_remote_node(this_node)) {
/* online state for remote nodes is determined by the
* rsc state after all the unpacking is done. we do however
* need to mark whether or not the node has been fenced as this plays
* a role during unpacking cluster node resource state */
this_node->details->remote_was_fenced =
crm_atoi(crm_element_value(state, XML_NODE_IS_FENCED), "0");
continue;
}
crm_trace("Processing node id=%s, uname=%s", id, uname);
/* Mark the node as provisionally clean
* - at least we have seen it in the current cluster's lifetime
*/
this_node->details->unclean = FALSE;
this_node->details->unseen = FALSE;
attrs = find_xml_node(state, XML_TAG_TRANSIENT_NODEATTRS, FALSE);
add_node_attrs(attrs, this_node, TRUE, data_set);
if (crm_is_true(pe_node_attribute_raw(this_node, "standby"))) {
crm_info("Node %s is in standby-mode", this_node->details->uname);
this_node->details->standby = TRUE;
}
if (crm_is_true(pe_node_attribute_raw(this_node, "maintenance"))) {
crm_info("Node %s is in maintenance-mode", this_node->details->uname);
this_node->details->maintenance = TRUE;
}
resource_discovery_enabled = pe_node_attribute_raw(this_node, XML_NODE_ATTR_RSC_DISCOVERY);
if (resource_discovery_enabled && !crm_is_true(resource_discovery_enabled)) {
crm_warn("ignoring %s attribute on node %s, disabling resource discovery is not allowed on cluster nodes",
XML_NODE_ATTR_RSC_DISCOVERY, this_node->details->uname);
}
crm_trace("determining node state");
determine_online_status(state, this_node, data_set);
if (is_not_set(data_set->flags, pe_flag_have_quorum)
&& this_node->details->online
&& (data_set->no_quorum_policy == no_quorum_suicide)) {
/* Everything else should flow from this automatically
* (at least until the scheduler becomes able to migrate off
* healthy resources)
*/
pe_fence_node(data_set, this_node, "cluster does not have quorum", FALSE);
}
}
}
while(unpack_node_loop(status, FALSE, data_set)) {
crm_trace("Start another loop");
}
// Now catch any nodes we didn't see
unpack_node_loop(status, is_set(data_set->flags, pe_flag_stonith_enabled), data_set);
/* Now that we know where resources are, we can schedule stops of containers
* with failed bundle connections
*/
if (data_set->stop_needed != NULL) {
for (GList *item = data_set->stop_needed; item; item = item->next) {
pe_resource_t *container = item->data;
pe_node_t *node = pe__current_node(container);
if (node) {
stop_action(container, node, FALSE);
}
}
g_list_free(data_set->stop_needed);
data_set->stop_needed = NULL;
}
for (GListPtr gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
pe_node_t *this_node = gIter->data;
if (this_node == NULL) {
continue;
} else if (!pe__is_guest_or_remote_node(this_node)) {
continue;
} else if(this_node->details->unpacked) {
continue;
}
determine_remote_online_status(data_set, this_node);
}
return TRUE;
}
static gboolean
determine_online_status_no_fencing(pe_working_set_t * data_set, xmlNode * node_state,
pe_node_t * this_node)
{
gboolean online = FALSE;
const char *join = crm_element_value(node_state, XML_NODE_JOIN_STATE);
const char *is_peer = crm_element_value(node_state, XML_NODE_IS_PEER);
const char *in_cluster = crm_element_value(node_state, XML_NODE_IN_CLUSTER);
const char *exp_state = crm_element_value(node_state, XML_NODE_EXPECTED);
if (!crm_is_true(in_cluster)) {
crm_trace("Node is down: in_cluster=%s", crm_str(in_cluster));
} else if (safe_str_eq(is_peer, ONLINESTATUS)) {
if (safe_str_eq(join, CRMD_JOINSTATE_MEMBER)) {
online = TRUE;
} else {
crm_debug("Node is not ready to run resources: %s", join);
}
} else if (this_node->details->expected_up == FALSE) {
crm_trace("Controller is down: in_cluster=%s", crm_str(in_cluster));
crm_trace("\tis_peer=%s, join=%s, expected=%s",
crm_str(is_peer), crm_str(join), crm_str(exp_state));
} else {
/* mark it unclean */
pe_fence_node(data_set, this_node, "peer is unexpectedly down", FALSE);
crm_info("\tin_cluster=%s, is_peer=%s, join=%s, expected=%s",
crm_str(in_cluster), crm_str(is_peer), crm_str(join), crm_str(exp_state));
}
return online;
}
static gboolean
determine_online_status_fencing(pe_working_set_t * data_set, xmlNode * node_state,
pe_node_t * this_node)
{
gboolean online = FALSE;
gboolean do_terminate = FALSE;
bool crmd_online = FALSE;
const char *join = crm_element_value(node_state, XML_NODE_JOIN_STATE);
const char *is_peer = crm_element_value(node_state, XML_NODE_IS_PEER);
const char *in_cluster = crm_element_value(node_state, XML_NODE_IN_CLUSTER);
const char *exp_state = crm_element_value(node_state, XML_NODE_EXPECTED);
const char *terminate = pe_node_attribute_raw(this_node, "terminate");
/*
- XML_NODE_IN_CLUSTER ::= true|false
- XML_NODE_IS_PEER ::= online|offline
- XML_NODE_JOIN_STATE ::= member|down|pending|banned
- XML_NODE_EXPECTED ::= member|down
*/
if (crm_is_true(terminate)) {
do_terminate = TRUE;
} else if (terminate != NULL && strlen(terminate) > 0) {
/* could be a time() value */
char t = terminate[0];
if (t != '0' && isdigit(t)) {
do_terminate = TRUE;
}
}
crm_trace("%s: in_cluster=%s, is_peer=%s, join=%s, expected=%s, term=%d",
this_node->details->uname, crm_str(in_cluster), crm_str(is_peer),
crm_str(join), crm_str(exp_state), do_terminate);
online = crm_is_true(in_cluster);
crmd_online = safe_str_eq(is_peer, ONLINESTATUS);
if (exp_state == NULL) {
exp_state = CRMD_JOINSTATE_DOWN;
}
if (this_node->details->shutdown) {
crm_debug("%s is shutting down", this_node->details->uname);
/* Slightly different criteria since we can't shut down a dead peer */
online = crmd_online;
} else if (in_cluster == NULL) {
pe_fence_node(data_set, this_node, "peer has not been seen by the cluster", FALSE);
} else if (safe_str_eq(join, CRMD_JOINSTATE_NACK)) {
pe_fence_node(data_set, this_node, "peer failed the pacemaker membership criteria", FALSE);
} else if (do_terminate == FALSE && safe_str_eq(exp_state, CRMD_JOINSTATE_DOWN)) {
if (crm_is_true(in_cluster) || crmd_online) {
crm_info("- Node %s is not ready to run resources", this_node->details->uname);
this_node->details->standby = TRUE;
this_node->details->pending = TRUE;
} else {
crm_trace("%s is down or still coming up", this_node->details->uname);
}
} else if (do_terminate && safe_str_eq(join, CRMD_JOINSTATE_DOWN)
&& crm_is_true(in_cluster) == FALSE && !crmd_online) {
crm_info("Node %s was just shot", this_node->details->uname);
online = FALSE;
} else if (crm_is_true(in_cluster) == FALSE) {
// Consider `priority-fencing-delay` for lost nodes
pe_fence_node(data_set, this_node, "peer is no longer part of the cluster", TRUE);
} else if (!crmd_online) {
pe_fence_node(data_set, this_node, "peer process is no longer available", FALSE);
/* Everything is running at this point, now check join state */
} else if (do_terminate) {
pe_fence_node(data_set, this_node, "termination was requested", FALSE);
} else if (safe_str_eq(join, CRMD_JOINSTATE_MEMBER)) {
crm_info("Node %s is active", this_node->details->uname);
} else if (safe_str_eq(join, CRMD_JOINSTATE_PENDING)
|| safe_str_eq(join, CRMD_JOINSTATE_DOWN)) {
crm_info("Node %s is not ready to run resources", this_node->details->uname);
this_node->details->standby = TRUE;
this_node->details->pending = TRUE;
} else {
pe_fence_node(data_set, this_node, "peer was in an unknown state", FALSE);
crm_warn("%s: in-cluster=%s, is-peer=%s, join=%s, expected=%s, term=%d, shutdown=%d",
this_node->details->uname, crm_str(in_cluster), crm_str(is_peer),
crm_str(join), crm_str(exp_state), do_terminate, this_node->details->shutdown);
}
return online;
}
static void
determine_remote_online_status(pe_working_set_t * data_set, pe_node_t * this_node)
{
pe_resource_t *rsc = this_node->details->remote_rsc;
pe_resource_t *container = NULL;
pe_node_t *host = NULL;
/* If there is a node state entry for a (former) Pacemaker Remote node
* but no resource creating that node, the node's connection resource will
* be NULL. Consider it an offline remote node in that case.
*/
if (rsc == NULL) {
this_node->details->online = FALSE;
goto remote_online_done;
}
container = rsc->container;
if (container && pcmk__list_of_1(rsc->running_on)) {
host = rsc->running_on->data;
}
/* If the resource is currently started, mark it online. */
if (rsc->role == RSC_ROLE_STARTED) {
crm_trace("%s node %s presumed ONLINE because connection resource is started",
(container? "Guest" : "Remote"), this_node->details->id);
this_node->details->online = TRUE;
}
/* consider this node shutting down if transitioning start->stop */
if (rsc->role == RSC_ROLE_STARTED && rsc->next_role == RSC_ROLE_STOPPED) {
crm_trace("%s node %s shutting down because connection resource is stopping",
(container? "Guest" : "Remote"), this_node->details->id);
this_node->details->shutdown = TRUE;
}
/* Now check all the failure conditions. */
if(container && is_set(container->flags, pe_rsc_failed)) {
crm_trace("Guest node %s UNCLEAN because guest resource failed",
this_node->details->id);
this_node->details->online = FALSE;
this_node->details->remote_requires_reset = TRUE;
} else if(is_set(rsc->flags, pe_rsc_failed)) {
crm_trace("%s node %s OFFLINE because connection resource failed",
(container? "Guest" : "Remote"), this_node->details->id);
this_node->details->online = FALSE;
} else if (rsc->role == RSC_ROLE_STOPPED
|| (container && container->role == RSC_ROLE_STOPPED)) {
crm_trace("%s node %s OFFLINE because its resource is stopped",
(container? "Guest" : "Remote"), this_node->details->id);
this_node->details->online = FALSE;
this_node->details->remote_requires_reset = FALSE;
} else if (host && (host->details->online == FALSE)
&& host->details->unclean) {
crm_trace("Guest node %s UNCLEAN because host is unclean",
this_node->details->id);
this_node->details->online = FALSE;
this_node->details->remote_requires_reset = TRUE;
}
remote_online_done:
crm_trace("Remote node %s online=%s",
this_node->details->id, this_node->details->online ? "TRUE" : "FALSE");
}
static void
determine_online_status(xmlNode * node_state, pe_node_t * this_node, pe_working_set_t * data_set)
{
gboolean online = FALSE;
const char *exp_state = crm_element_value(node_state, XML_NODE_EXPECTED);
CRM_CHECK(this_node != NULL, return);
this_node->details->shutdown = FALSE;
this_node->details->expected_up = FALSE;
if (pe__shutdown_requested(this_node)) {
this_node->details->shutdown = TRUE;
} else if (safe_str_eq(exp_state, CRMD_JOINSTATE_MEMBER)) {
this_node->details->expected_up = TRUE;
}
if (this_node->details->type == node_ping) {
this_node->details->unclean = FALSE;
online = FALSE; /* As far as resource management is concerned,
* the node is safely offline.
* Anyone caught abusing this logic will be shot
*/
} else if (is_set(data_set->flags, pe_flag_stonith_enabled) == FALSE) {
online = determine_online_status_no_fencing(data_set, node_state, this_node);
} else {
online = determine_online_status_fencing(data_set, node_state, this_node);
}
if (online) {
this_node->details->online = TRUE;
} else {
/* remove node from contention */
this_node->fixed = TRUE;
this_node->weight = -INFINITY;
}
if (online && this_node->details->shutdown) {
/* don't run resources here */
this_node->fixed = TRUE;
this_node->weight = -INFINITY;
}
if (this_node->details->type == node_ping) {
crm_info("Node %s is not a pacemaker node", this_node->details->uname);
} else if (this_node->details->unclean) {
pe_proc_warn("Node %s is unclean", this_node->details->uname);
} else if (this_node->details->online) {
crm_info("Node %s is %s", this_node->details->uname,
this_node->details->shutdown ? "shutting down" :
this_node->details->pending ? "pending" :
this_node->details->standby ? "standby" :
this_node->details->maintenance ? "maintenance" : "online");
} else {
crm_trace("Node %s is offline", this_node->details->uname);
}
}
/*!
* \internal
* \brief Find the end of a resource's name, excluding any clone suffix
*
* \param[in] id Resource ID to check
*
* \return Pointer to last character of resource's base name
*/
const char *
pe_base_name_end(const char *id)
{
if (!pcmk__str_empty(id)) {
const char *end = id + strlen(id) - 1;
for (const char *s = end; s > id; --s) {
switch (*s) {
case '0':
case '1':
case '2':
case '3':
case '4':
case '5':
case '6':
case '7':
case '8':
case '9':
break;
case ':':
return (s == end)? s : (s - 1);
default:
return end;
}
}
return end;
}
return NULL;
}
/*!
* \internal
* \brief Get a resource name excluding any clone suffix
*
* \param[in] last_rsc_id Resource ID to check
*
* \return Pointer to newly allocated string with resource's base name
* \note It is the caller's responsibility to free() the result.
* This asserts on error, so callers can assume result is not NULL.
*/
char *
clone_strip(const char *last_rsc_id)
{
const char *end = pe_base_name_end(last_rsc_id);
char *basename = NULL;
CRM_ASSERT(end);
basename = strndup(last_rsc_id, end - last_rsc_id + 1);
CRM_ASSERT(basename);
return basename;
}
/*!
* \internal
* \brief Get the name of the first instance of a cloned resource
*
* \param[in] last_rsc_id Resource ID to check
*
* \return Pointer to newly allocated string with resource's base name plus :0
* \note It is the caller's responsibility to free() the result.
* This asserts on error, so callers can assume result is not NULL.
*/
char *
clone_zero(const char *last_rsc_id)
{
const char *end = pe_base_name_end(last_rsc_id);
size_t base_name_len = end - last_rsc_id + 1;
char *zero = NULL;
CRM_ASSERT(end);
zero = calloc(base_name_len + 3, sizeof(char));
CRM_ASSERT(zero);
memcpy(zero, last_rsc_id, base_name_len);
zero[base_name_len] = ':';
zero[base_name_len + 1] = '0';
return zero;
}
static pe_resource_t *
create_fake_resource(const char *rsc_id, xmlNode * rsc_entry, pe_working_set_t * data_set)
{
pe_resource_t *rsc = NULL;
xmlNode *xml_rsc = create_xml_node(NULL, XML_CIB_TAG_RESOURCE);
copy_in_properties(xml_rsc, rsc_entry);
crm_xml_add(xml_rsc, XML_ATTR_ID, rsc_id);
crm_log_xml_debug(xml_rsc, "Orphan resource");
if (!common_unpack(xml_rsc, &rsc, NULL, data_set)) {
return NULL;
}
if (xml_contains_remote_node(xml_rsc)) {
pe_node_t *node;
crm_debug("Detected orphaned remote node %s", rsc_id);
node = pe_find_node(data_set->nodes, rsc_id);
if (node == NULL) {
node = pe_create_node(rsc_id, rsc_id, "remote", NULL, data_set);
}
link_rsc2remotenode(data_set, rsc);
if (node) {
crm_trace("Setting node %s as shutting down due to orphaned connection resource", rsc_id);
node->details->shutdown = TRUE;
}
}
if (crm_element_value(rsc_entry, XML_RSC_ATTR_CONTAINER)) {
/* This orphaned rsc needs to be mapped to a container. */
crm_trace("Detected orphaned container filler %s", rsc_id);
set_bit(rsc->flags, pe_rsc_orphan_container_filler);
}
set_bit(rsc->flags, pe_rsc_orphan);
data_set->resources = g_list_append(data_set->resources, rsc);
return rsc;
}
/*!
* \internal
* \brief Create orphan instance for anonymous clone resource history
*/
static pe_resource_t *
create_anonymous_orphan(pe_resource_t *parent, const char *rsc_id,
pe_node_t *node, pe_working_set_t *data_set)
{
pe_resource_t *top = pe__create_clone_child(parent, data_set);
// find_rsc() because we might be a cloned group
pe_resource_t *orphan = top->fns->find_rsc(top, rsc_id, NULL, pe_find_clone);
pe_rsc_debug(parent, "Created orphan %s for %s: %s on %s",
top->id, parent->id, rsc_id, node->details->uname);
return orphan;
}
/*!
* \internal
* \brief Check a node for an instance of an anonymous clone
*
* Return a child instance of the specified anonymous clone, in order of
* preference: (1) the instance running on the specified node, if any;
* (2) an inactive instance (i.e. within the total of clone-max instances);
* (3) a newly created orphan (i.e. clone-max instances are already active).
*
* \param[in] data_set Cluster information
* \param[in] node Node on which to check for instance
* \param[in] parent Clone to check
* \param[in] rsc_id Name of cloned resource in history (without instance)
*/
static pe_resource_t *
find_anonymous_clone(pe_working_set_t * data_set, pe_node_t * node, pe_resource_t * parent,
const char *rsc_id)
{
GListPtr rIter = NULL;
pe_resource_t *rsc = NULL;
pe_resource_t *inactive_instance = NULL;
gboolean skip_inactive = FALSE;
CRM_ASSERT(parent != NULL);
CRM_ASSERT(pe_rsc_is_clone(parent));
CRM_ASSERT(is_not_set(parent->flags, pe_rsc_unique));
// Check for active (or partially active, for cloned groups) instance
pe_rsc_trace(parent, "Looking for %s on %s in %s", rsc_id, node->details->uname, parent->id);
for (rIter = parent->children; rsc == NULL && rIter; rIter = rIter->next) {
GListPtr locations = NULL;
pe_resource_t *child = rIter->data;
/* Check whether this instance is already known to be active or pending
* anywhere, at this stage of unpacking. Because this function is called
* for a resource before the resource's individual operation history
* entries are unpacked, locations will generally not contain the
* desired node.
*
* However, there are three exceptions:
* (1) when child is a cloned group and we have already unpacked the
* history of another member of the group on the same node;
* (2) when we've already unpacked the history of another numbered
* instance on the same node (which can happen if globally-unique
* was flipped from true to false); and
* (3) when we re-run calculations on the same data set as part of a
* simulation.
*/
child->fns->location(child, &locations, 2);
if (locations) {
/* We should never associate the same numbered anonymous clone
* instance with multiple nodes, and clone instances can't migrate,
* so there must be only one location, regardless of history.
*/
CRM_LOG_ASSERT(locations->next == NULL);
if (((pe_node_t *)locations->data)->details == node->details) {
/* This child instance is active on the requested node, so check
* for a corresponding configured resource. We use find_rsc()
* instead of child because child may be a cloned group, and we
* need the particular member corresponding to rsc_id.
*
* If the history entry is orphaned, rsc will be NULL.
*/
rsc = parent->fns->find_rsc(child, rsc_id, NULL, pe_find_clone);
if (rsc) {
/* If there are multiple instance history entries for an
* anonymous clone in a single node's history (which can
* happen if globally-unique is switched from true to
* false), we want to consider the instances beyond the
* first as orphans, even if there are inactive instance
* numbers available.
*/
if (rsc->running_on) {
crm_notice("Active (now-)anonymous clone %s has "
"multiple (orphan) instance histories on %s",
parent->id, node->details->uname);
skip_inactive = TRUE;
rsc = NULL;
} else {
pe_rsc_trace(parent, "Resource %s, active", rsc->id);
}
}
}
g_list_free(locations);
} else {
pe_rsc_trace(parent, "Resource %s, skip inactive", child->id);
if (!skip_inactive && !inactive_instance
&& is_not_set(child->flags, pe_rsc_block)) {
// Remember one inactive instance in case we don't find active
inactive_instance = parent->fns->find_rsc(child, rsc_id, NULL,
pe_find_clone);
/* ... but don't use it if it was already associated with a
* pending action on another node
*/
if (inactive_instance && inactive_instance->pending_node
&& (inactive_instance->pending_node->details != node->details)) {
inactive_instance = NULL;
}
}
}
}
if ((rsc == NULL) && !skip_inactive && (inactive_instance != NULL)) {
pe_rsc_trace(parent, "Resource %s, empty slot", inactive_instance->id);
rsc = inactive_instance;
}
/* If the resource has "requires" set to "quorum" or "nothing", and we don't
* have a clone instance for every node, we don't want to consume a valid
* instance number for unclean nodes. Such instances may appear to be active
* according to the history, but should be considered inactive, so we can
* start an instance elsewhere. Treat such instances as orphans.
*
* An exception is instances running on guest nodes -- since guest node
* "fencing" is actually just a resource stop, requires shouldn't apply.
*
* @TODO Ideally, we'd use an inactive instance number if it is not needed
* for any clean instances. However, we don't know that at this point.
*/
if ((rsc != NULL) && is_not_set(rsc->flags, pe_rsc_needs_fencing)
&& (!node->details->online || node->details->unclean)
&& !pe__is_guest_node(node)
&& !pe__is_universal_clone(parent, data_set)) {
rsc = NULL;
}
if (rsc == NULL) {
rsc = create_anonymous_orphan(parent, rsc_id, node, data_set);
pe_rsc_trace(parent, "Resource %s, orphan", rsc->id);
}
return rsc;
}
static pe_resource_t *
unpack_find_resource(pe_working_set_t * data_set, pe_node_t * node, const char *rsc_id,
xmlNode * rsc_entry)
{
pe_resource_t *rsc = NULL;
pe_resource_t *parent = NULL;
crm_trace("looking for %s", rsc_id);
rsc = pe_find_resource(data_set->resources, rsc_id);
if (rsc == NULL) {
/* If we didn't find the resource by its name in the operation history,
* check it again as a clone instance. Even when clone-max=0, we create
* a single :0 orphan to match against here.
*/
char *clone0_id = clone_zero(rsc_id);
pe_resource_t *clone0 = pe_find_resource(data_set->resources, clone0_id);
if (clone0 && is_not_set(clone0->flags, pe_rsc_unique)) {
rsc = clone0;
parent = uber_parent(clone0);
crm_trace("%s found as %s (%s)", rsc_id, clone0_id, parent->id);
} else {
crm_trace("%s is not known as %s either (orphan)",
rsc_id, clone0_id);
}
free(clone0_id);
} else if (rsc->variant > pe_native) {
crm_trace("Resource history for %s is orphaned because it is no longer primitive",
rsc_id);
return NULL;
} else {
parent = uber_parent(rsc);
}
if (pe_rsc_is_anon_clone(parent)) {
if (pe_rsc_is_bundled(parent)) {
rsc = pe__find_bundle_replica(parent->parent, node);
} else {
char *base = clone_strip(rsc_id);
rsc = find_anonymous_clone(data_set, node, parent, base);
free(base);
CRM_ASSERT(rsc != NULL);
}
}
if (rsc && safe_str_neq(rsc_id, rsc->id)
&& safe_str_neq(rsc_id, rsc->clone_name)) {
free(rsc->clone_name);
rsc->clone_name = strdup(rsc_id);
pe_rsc_debug(rsc, "Internally renamed %s on %s to %s%s",
rsc_id, node->details->uname, rsc->id,
(is_set(rsc->flags, pe_rsc_orphan)? " (ORPHAN)" : ""));
}
return rsc;
}
static pe_resource_t *
process_orphan_resource(xmlNode * rsc_entry, pe_node_t * node, pe_working_set_t * data_set)
{
pe_resource_t *rsc = NULL;
const char *rsc_id = crm_element_value(rsc_entry, XML_ATTR_ID);
crm_debug("Detected orphan resource %s on %s", rsc_id, node->details->uname);
rsc = create_fake_resource(rsc_id, rsc_entry, data_set);
if (is_set(data_set->flags, pe_flag_stop_rsc_orphans) == FALSE) {
clear_bit(rsc->flags, pe_rsc_managed);
} else {
CRM_CHECK(rsc != NULL, return NULL);
pe_rsc_trace(rsc, "Added orphan %s", rsc->id);
resource_location(rsc, NULL, -INFINITY, "__orphan_do_not_run__", data_set);
}
return rsc;
}
static void
process_rsc_state(pe_resource_t * rsc, pe_node_t * node,
enum action_fail_response on_fail,
xmlNode * migrate_op, pe_working_set_t * data_set)
{
pe_node_t *tmpnode = NULL;
char *reason = NULL;
CRM_ASSERT(rsc);
pe_rsc_trace(rsc, "Resource %s is %s on %s: on_fail=%s",
rsc->id, role2text(rsc->role), node->details->uname, fail2text(on_fail));
/* process current state */
if (rsc->role != RSC_ROLE_UNKNOWN) {
pe_resource_t *iter = rsc;
while (iter) {
if (g_hash_table_lookup(iter->known_on, node->details->id) == NULL) {
pe_node_t *n = node_copy(node);
pe_rsc_trace(rsc, "%s (aka. %s) known on %s", rsc->id, rsc->clone_name,
n->details->uname);
g_hash_table_insert(iter->known_on, (gpointer) n->details->id, n);
}
if (is_set(iter->flags, pe_rsc_unique)) {
break;
}
iter = iter->parent;
}
}
/* If a managed resource is believed to be running, but node is down ... */
if (rsc->role > RSC_ROLE_STOPPED
&& node->details->online == FALSE
&& node->details->maintenance == FALSE
&& is_set(rsc->flags, pe_rsc_managed)) {
gboolean should_fence = FALSE;
/* If this is a guest node, fence it (regardless of whether fencing is
* enabled, because guest node fencing is done by recovery of the
* container resource rather than by the fencer). Mark the resource
* we're processing as failed. When the guest comes back up, its
* operation history in the CIB will be cleared, freeing the affected
* resource to run again once we are sure we know its state.
*/
if (pe__is_guest_node(node)) {
set_bit(rsc->flags, pe_rsc_failed);
should_fence = TRUE;
} else if (is_set(data_set->flags, pe_flag_stonith_enabled)) {
if (pe__is_remote_node(node) && node->details->remote_rsc
&& is_not_set(node->details->remote_rsc->flags, pe_rsc_failed)) {
/* Setting unseen means that fencing of the remote node will
* occur only if the connection resource is not going to start
* somewhere. This allows connection resources on a failed
* cluster node to move to another node without requiring the
* remote nodes to be fenced as well.
*/
node->details->unseen = TRUE;
reason = crm_strdup_printf("%s is active there (fencing will be"
" revoked if remote connection can "
"be re-established elsewhere)",
rsc->id);
}
should_fence = TRUE;
}
if (should_fence) {
if (reason == NULL) {
reason = crm_strdup_printf("%s is thought to be active there", rsc->id);
}
pe_fence_node(data_set, node, reason, FALSE);
}
free(reason);
}
if (node->details->unclean) {
/* No extra processing needed
* Also allows resources to be started again after a node is shot
*/
on_fail = action_fail_ignore;
}
switch (on_fail) {
case action_fail_ignore:
/* nothing to do */
break;
case action_fail_fence:
/* treat it as if it is still running
* but also mark the node as unclean
*/
reason = crm_strdup_printf("%s failed there", rsc->id);
pe_fence_node(data_set, node, reason, FALSE);
free(reason);
break;
case action_fail_standby:
node->details->standby = TRUE;
node->details->standby_onfail = TRUE;
break;
case action_fail_block:
/* is_managed == FALSE will prevent any
* actions being sent for the resource
*/
clear_bit(rsc->flags, pe_rsc_managed);
set_bit(rsc->flags, pe_rsc_block);
break;
case action_fail_migrate:
/* make sure it comes up somewhere else
* or not at all
*/
resource_location(rsc, node, -INFINITY, "__action_migration_auto__", data_set);
break;
case action_fail_stop:
rsc->next_role = RSC_ROLE_STOPPED;
break;
case action_fail_recover:
if (rsc->role != RSC_ROLE_STOPPED && rsc->role != RSC_ROLE_UNKNOWN) {
set_bit(rsc->flags, pe_rsc_failed);
stop_action(rsc, node, FALSE);
}
break;
case action_fail_restart_container:
set_bit(rsc->flags, pe_rsc_failed);
if (rsc->container && pe_rsc_is_bundled(rsc)) {
/* A bundle's remote connection can run on a different node than
* the bundle's container. We don't necessarily know where the
* container is running yet, so remember it and add a stop
* action for it later.
*/
data_set->stop_needed = g_list_prepend(data_set->stop_needed,
rsc->container);
} else if (rsc->container) {
stop_action(rsc->container, node, FALSE);
} else if (rsc->role != RSC_ROLE_STOPPED && rsc->role != RSC_ROLE_UNKNOWN) {
stop_action(rsc, node, FALSE);
}
break;
case action_fail_reset_remote:
set_bit(rsc->flags, pe_rsc_failed);
if (is_set(data_set->flags, pe_flag_stonith_enabled)) {
tmpnode = NULL;
if (rsc->is_remote_node) {
tmpnode = pe_find_node(data_set->nodes, rsc->id);
}
if (tmpnode &&
pe__is_remote_node(tmpnode) &&
tmpnode->details->remote_was_fenced == 0) {
/* The remote connection resource failed in a way that
* should result in fencing the remote node.
*/
pe_fence_node(data_set, tmpnode,
"remote connection is unrecoverable", FALSE);
}
}
/* require the stop action regardless if fencing is occurring or not. */
if (rsc->role > RSC_ROLE_STOPPED) {
stop_action(rsc, node, FALSE);
}
/* if reconnect delay is in use, prevent the connection from exiting the
* "STOPPED" role until the failure is cleared by the delay timeout. */
if (rsc->remote_reconnect_ms) {
rsc->next_role = RSC_ROLE_STOPPED;
}
break;
}
/* ensure a remote-node connection failure forces an unclean remote-node
* to be fenced. By setting unseen = FALSE, the remote-node failure will
* result in a fencing operation regardless if we're going to attempt to
* reconnect to the remote-node in this transition or not. */
if (is_set(rsc->flags, pe_rsc_failed) && rsc->is_remote_node) {
tmpnode = pe_find_node(data_set->nodes, rsc->id);
if (tmpnode && tmpnode->details->unclean) {
tmpnode->details->unseen = FALSE;
}
}
if (rsc->role != RSC_ROLE_STOPPED && rsc->role != RSC_ROLE_UNKNOWN) {
if (is_set(rsc->flags, pe_rsc_orphan)) {
if (is_set(rsc->flags, pe_rsc_managed)) {
pcmk__config_warn("Detected active orphan %s running on %s",
rsc->id, node->details->uname);
} else {
pcmk__config_warn("Resource '%s' must be stopped manually on "
"%s because cluster is configured not to "
"stop active orphans",
rsc->id, node->details->uname);
}
}
native_add_running(rsc, node, data_set);
if (on_fail != action_fail_ignore) {
set_bit(rsc->flags, pe_rsc_failed);
}
} else if (rsc->clone_name && strchr(rsc->clone_name, ':') != NULL) {
/* Only do this for older status sections that included instance numbers
* Otherwise stopped instances will appear as orphans
*/
pe_rsc_trace(rsc, "Resetting clone_name %s for %s (stopped)", rsc->clone_name, rsc->id);
free(rsc->clone_name);
rsc->clone_name = NULL;
} else {
GList *possible_matches = pe__resource_actions(rsc, node, RSC_STOP,
FALSE);
GListPtr gIter = possible_matches;
for (; gIter != NULL; gIter = gIter->next) {
pe_action_t *stop = (pe_action_t *) gIter->data;
stop->flags |= pe_action_optional;
}
g_list_free(possible_matches);
}
}
/* create active recurring operations as optional */
static void
process_recurring(pe_node_t * node, pe_resource_t * rsc,
int start_index, int stop_index,
GListPtr sorted_op_list, pe_working_set_t * data_set)
{
int counter = -1;
const char *task = NULL;
const char *status = NULL;
GListPtr gIter = sorted_op_list;
CRM_ASSERT(rsc);
pe_rsc_trace(rsc, "%s: Start index %d, stop index = %d", rsc->id, start_index, stop_index);
for (; gIter != NULL; gIter = gIter->next) {
xmlNode *rsc_op = (xmlNode *) gIter->data;
guint interval_ms = 0;
char *key = NULL;
const char *id = ID(rsc_op);
counter++;
if (node->details->online == FALSE) {
pe_rsc_trace(rsc, "Skipping %s/%s: node is offline", rsc->id, node->details->uname);
break;
/* Need to check if there's a monitor for role="Stopped" */
} else if (start_index < stop_index && counter <= stop_index) {
pe_rsc_trace(rsc, "Skipping %s/%s: resource is not active", id, node->details->uname);
continue;
} else if (counter < start_index) {
pe_rsc_trace(rsc, "Skipping %s/%s: old %d", id, node->details->uname, counter);
continue;
}
crm_element_value_ms(rsc_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
if (interval_ms == 0) {
pe_rsc_trace(rsc, "Skipping %s/%s: non-recurring", id, node->details->uname);
continue;
}
status = crm_element_value(rsc_op, XML_LRM_ATTR_OPSTATUS);
if (safe_str_eq(status, "-1")) {
pe_rsc_trace(rsc, "Skipping %s/%s: status", id, node->details->uname);
continue;
}
task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK);
/* create the action */
key = pcmk__op_key(rsc->id, task, interval_ms);
pe_rsc_trace(rsc, "Creating %s/%s", key, node->details->uname);
custom_action(rsc, key, task, node, TRUE, TRUE, data_set);
}
}
void
calculate_active_ops(GListPtr sorted_op_list, int *start_index, int *stop_index)
{
int counter = -1;
int implied_monitor_start = -1;
int implied_clone_start = -1;
const char *task = NULL;
const char *status = NULL;
GListPtr gIter = sorted_op_list;
*stop_index = -1;
*start_index = -1;
for (; gIter != NULL; gIter = gIter->next) {
xmlNode *rsc_op = (xmlNode *) gIter->data;
counter++;
task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK);
status = crm_element_value(rsc_op, XML_LRM_ATTR_OPSTATUS);
if (safe_str_eq(task, CRMD_ACTION_STOP)
&& safe_str_eq(status, "0")) {
*stop_index = counter;
} else if (safe_str_eq(task, CRMD_ACTION_START) || safe_str_eq(task, CRMD_ACTION_MIGRATED)) {
*start_index = counter;
} else if ((implied_monitor_start <= *stop_index) && safe_str_eq(task, CRMD_ACTION_STATUS)) {
const char *rc = crm_element_value(rsc_op, XML_LRM_ATTR_RC);
if (safe_str_eq(rc, "0") || safe_str_eq(rc, "8")) {
implied_monitor_start = counter;
}
} else if (safe_str_eq(task, CRMD_ACTION_PROMOTE) || safe_str_eq(task, CRMD_ACTION_DEMOTE)) {
implied_clone_start = counter;
}
}
if (*start_index == -1) {
if (implied_clone_start != -1) {
*start_index = implied_clone_start;
} else if (implied_monitor_start != -1) {
*start_index = implied_monitor_start;
}
}
}
// If resource history entry has shutdown lock, remember lock node and time
static void
unpack_shutdown_lock(xmlNode *rsc_entry, pe_resource_t *rsc, pe_node_t *node,
pe_working_set_t *data_set)
{
time_t lock_time = 0; // When lock started (i.e. node shutdown time)
if ((crm_element_value_epoch(rsc_entry, XML_CONFIG_ATTR_SHUTDOWN_LOCK,
&lock_time) == pcmk_ok) && (lock_time != 0)) {
if ((data_set->shutdown_lock > 0)
&& (get_effective_time(data_set)
> (lock_time + data_set->shutdown_lock))) {
pe_rsc_info(rsc, "Shutdown lock for %s on %s expired",
rsc->id, node->details->uname);
pe__clear_resource_history(rsc, node, data_set);
} else {
rsc->lock_node = node;
rsc->lock_time = lock_time;
}
}
}
static pe_resource_t *
unpack_lrm_rsc_state(pe_node_t * node, xmlNode * rsc_entry, pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
int stop_index = -1;
int start_index = -1;
enum rsc_role_e req_role = RSC_ROLE_UNKNOWN;
const char *task = NULL;
const char *rsc_id = crm_element_value(rsc_entry, XML_ATTR_ID);
pe_resource_t *rsc = NULL;
GListPtr op_list = NULL;
GListPtr sorted_op_list = NULL;
xmlNode *migrate_op = NULL;
xmlNode *rsc_op = NULL;
xmlNode *last_failure = NULL;
enum action_fail_response on_fail = FALSE;
enum rsc_role_e saved_role = RSC_ROLE_UNKNOWN;
crm_trace("[%s] Processing %s on %s",
crm_element_name(rsc_entry), rsc_id, node->details->uname);
/* extract operations */
op_list = NULL;
sorted_op_list = NULL;
for (rsc_op = __xml_first_child_element(rsc_entry); rsc_op != NULL;
rsc_op = __xml_next_element(rsc_op)) {
if (crm_str_eq((const char *)rsc_op->name, XML_LRM_TAG_RSC_OP, TRUE)) {
op_list = g_list_prepend(op_list, rsc_op);
}
}
if (is_not_set(data_set->flags, pe_flag_shutdown_lock)) {
if (op_list == NULL) {
// If there are no operations, there is nothing to do
return NULL;
}
}
/* find the resource */
rsc = unpack_find_resource(data_set, node, rsc_id, rsc_entry);
if (rsc == NULL) {
if (op_list == NULL) {
// If there are no operations, there is nothing to do
return NULL;
} else {
rsc = process_orphan_resource(rsc_entry, node, data_set);
}
}
CRM_ASSERT(rsc != NULL);
// Check whether the resource is "shutdown-locked" to this node
if (is_set(data_set->flags, pe_flag_shutdown_lock)) {
unpack_shutdown_lock(rsc_entry, rsc, node, data_set);
}
/* process operations */
saved_role = rsc->role;
on_fail = action_fail_ignore;
rsc->role = RSC_ROLE_UNKNOWN;
sorted_op_list = g_list_sort(op_list, sort_op_by_callid);
for (gIter = sorted_op_list; gIter != NULL; gIter = gIter->next) {
xmlNode *rsc_op = (xmlNode *) gIter->data;
task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK);
if (safe_str_eq(task, CRMD_ACTION_MIGRATED)) {
migrate_op = rsc_op;
}
unpack_rsc_op(rsc, node, rsc_op, &last_failure, &on_fail, data_set);
}
/* create active recurring operations as optional */
calculate_active_ops(sorted_op_list, &start_index, &stop_index);
process_recurring(node, rsc, start_index, stop_index, sorted_op_list, data_set);
/* no need to free the contents */
g_list_free(sorted_op_list);
process_rsc_state(rsc, node, on_fail, migrate_op, data_set);
if (get_target_role(rsc, &req_role)) {
if (rsc->next_role == RSC_ROLE_UNKNOWN || req_role < rsc->next_role) {
pe_rsc_debug(rsc, "%s: Overwriting calculated next role %s"
" with requested next role %s",
rsc->id, role2text(rsc->next_role), role2text(req_role));
rsc->next_role = req_role;
} else if (req_role > rsc->next_role) {
pe_rsc_info(rsc, "%s: Not overwriting calculated next role %s"
" with requested next role %s",
rsc->id, role2text(rsc->next_role), role2text(req_role));
}
}
if (saved_role > rsc->role) {
rsc->role = saved_role;
}
return rsc;
}
static void
handle_orphaned_container_fillers(xmlNode * lrm_rsc_list, pe_working_set_t * data_set)
{
xmlNode *rsc_entry = NULL;
for (rsc_entry = __xml_first_child_element(lrm_rsc_list); rsc_entry != NULL;
rsc_entry = __xml_next_element(rsc_entry)) {
pe_resource_t *rsc;
pe_resource_t *container;
const char *rsc_id;
const char *container_id;
if (safe_str_neq((const char *)rsc_entry->name, XML_LRM_TAG_RESOURCE)) {
continue;
}
container_id = crm_element_value(rsc_entry, XML_RSC_ATTR_CONTAINER);
rsc_id = crm_element_value(rsc_entry, XML_ATTR_ID);
if (container_id == NULL || rsc_id == NULL) {
continue;
}
container = pe_find_resource(data_set->resources, container_id);
if (container == NULL) {
continue;
}
rsc = pe_find_resource(data_set->resources, rsc_id);
if (rsc == NULL ||
is_set(rsc->flags, pe_rsc_orphan_container_filler) == FALSE ||
rsc->container != NULL) {
continue;
}
pe_rsc_trace(rsc, "Mapped container of orphaned resource %s to %s",
rsc->id, container_id);
rsc->container = container;
container->fillers = g_list_append(container->fillers, rsc);
}
}
-gboolean
-unpack_lrm_resources(pe_node_t * node, xmlNode * lrm_rsc_list, pe_working_set_t * data_set)
+static void
+unpack_lrm_resources(pe_node_t *node, xmlNode *lrm_rsc_list,
+ pe_working_set_t *data_set)
{
xmlNode *rsc_entry = NULL;
gboolean found_orphaned_container_filler = FALSE;
- CRM_CHECK(node != NULL, return FALSE);
-
- crm_trace("Unpacking resources on %s", node->details->uname);
-
for (rsc_entry = __xml_first_child_element(lrm_rsc_list); rsc_entry != NULL;
rsc_entry = __xml_next_element(rsc_entry)) {
if (crm_str_eq((const char *)rsc_entry->name, XML_LRM_TAG_RESOURCE, TRUE)) {
pe_resource_t *rsc = unpack_lrm_rsc_state(node, rsc_entry, data_set);
if (!rsc) {
continue;
}
if (is_set(rsc->flags, pe_rsc_orphan_container_filler)) {
found_orphaned_container_filler = TRUE;
}
}
}
/* now that all the resource state has been unpacked for this node
* we have to go back and map any orphaned container fillers to their
* container resource */
if (found_orphaned_container_filler) {
handle_orphaned_container_fillers(lrm_rsc_list, data_set);
}
- return TRUE;
}
static void
set_active(pe_resource_t * rsc)
{
pe_resource_t *top = uber_parent(rsc);
if (top && is_set(top->flags, pe_rsc_promotable)) {
rsc->role = RSC_ROLE_SLAVE;
} else {
rsc->role = RSC_ROLE_STARTED;
}
}
static void
set_node_score(gpointer key, gpointer value, gpointer user_data)
{
pe_node_t *node = value;
int *score = user_data;
node->weight = *score;
}
#define STATUS_PATH_MAX 1024
static xmlNode *
find_lrm_op(const char *resource, const char *op, const char *node, const char *source,
bool success_only, pe_working_set_t *data_set)
{
int offset = 0;
char xpath[STATUS_PATH_MAX];
xmlNode *xml = NULL;
offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "//node_state[@uname='%s']", node);
offset +=
snprintf(xpath + offset, STATUS_PATH_MAX - offset, "//" XML_LRM_TAG_RESOURCE "[@id='%s']",
resource);
/* Need to check against transition_magic too? */
if (source && safe_str_eq(op, CRMD_ACTION_MIGRATE)) {
offset +=
snprintf(xpath + offset, STATUS_PATH_MAX - offset,
"/" XML_LRM_TAG_RSC_OP "[@operation='%s' and @migrate_target='%s']", op,
source);
} else if (source && safe_str_eq(op, CRMD_ACTION_MIGRATED)) {
offset +=
snprintf(xpath + offset, STATUS_PATH_MAX - offset,
"/" XML_LRM_TAG_RSC_OP "[@operation='%s' and @migrate_source='%s']", op,
source);
} else {
offset +=
snprintf(xpath + offset, STATUS_PATH_MAX - offset,
"/" XML_LRM_TAG_RSC_OP "[@operation='%s']", op);
}
CRM_LOG_ASSERT(offset > 0);
xml = get_xpath_object(xpath, data_set->input, LOG_DEBUG);
if (xml && success_only) {
int rc = PCMK_OCF_UNKNOWN_ERROR;
int status = PCMK_LRM_OP_ERROR;
crm_element_value_int(xml, XML_LRM_ATTR_RC, &rc);
crm_element_value_int(xml, XML_LRM_ATTR_OPSTATUS, &status);
if ((rc != PCMK_OCF_OK) || (status != PCMK_LRM_OP_DONE)) {
return NULL;
}
}
return xml;
}
static int
pe__call_id(xmlNode *op_xml)
{
int id = 0;
if (op_xml) {
crm_element_value_int(op_xml, XML_LRM_ATTR_CALLID, &id);
}
return id;
}
/*!
* \brief Check whether a stop happened on the same node after some event
*
* \param[in] rsc Resource being checked
* \param[in] node Node being checked
* \param[in] xml_op Event that stop is being compared to
* \param[in] data_set Cluster working set
*
* \return TRUE if stop happened after event, FALSE otherwise
*
* \note This is really unnecessary, but kept as a safety mechanism. We
* currently don't save more than one successful event in history, so this
* only matters when processing really old CIB files that we don't
* technically support anymore, or as preparation for logging an extended
* history in the future.
*/
static bool
stop_happened_after(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op,
pe_working_set_t *data_set)
{
xmlNode *stop_op = find_lrm_op(rsc->id, CRMD_ACTION_STOP,
node->details->uname, NULL, TRUE, data_set);
return (stop_op && (pe__call_id(stop_op) > pe__call_id(xml_op)));
}
static void
unpack_migrate_to_success(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op,
pe_working_set_t *data_set)
{
/* A successful migration sequence is:
* migrate_to on source node
* migrate_from on target node
* stop on source node
*
* If a migrate_to is followed by a stop, the entire migration (successful
* or failed) is complete, and we don't care what happened on the target.
*
* If no migrate_from has happened, the migration is considered to be
* "partial". If the migrate_from failed, make sure the resource gets
* stopped on both source and target (if up).
*
* If the migrate_to and migrate_from both succeeded (which also implies the
* resource is no longer running on the source), but there is no stop, the
* migration is considered to be "dangling". Schedule a stop on the source
* in this case.
*/
int from_rc = 0;
int from_status = 0;
pe_node_t *target_node = NULL;
pe_node_t *source_node = NULL;
xmlNode *migrate_from = NULL;
const char *source = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_SOURCE);
const char *target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET);
// Sanity check
CRM_CHECK(source && target && !strcmp(source, node->details->uname), return);
if (stop_happened_after(rsc, node, xml_op, data_set)) {
return;
}
// Clones are not allowed to migrate, so role can't be master
rsc->role = RSC_ROLE_STARTED;
target_node = pe_find_node(data_set->nodes, target);
source_node = pe_find_node(data_set->nodes, source);
// Check whether there was a migrate_from action on the target
migrate_from = find_lrm_op(rsc->id, CRMD_ACTION_MIGRATED, target,
source, FALSE, data_set);
if (migrate_from) {
crm_element_value_int(migrate_from, XML_LRM_ATTR_RC, &from_rc);
crm_element_value_int(migrate_from, XML_LRM_ATTR_OPSTATUS, &from_status);
pe_rsc_trace(rsc, "%s op on %s exited with status=%d, rc=%d",
ID(migrate_from), target, from_status, from_rc);
}
if (migrate_from && from_rc == PCMK_OCF_OK
&& from_status == PCMK_LRM_OP_DONE) {
/* The migrate_to and migrate_from both succeeded, so mark the migration
* as "dangling". This will be used to schedule a stop action on the
* source without affecting the target.
*/
pe_rsc_trace(rsc, "Detected dangling migration op: %s on %s", ID(xml_op),
source);
rsc->role = RSC_ROLE_STOPPED;
rsc->dangling_migrations = g_list_prepend(rsc->dangling_migrations, node);
} else if (migrate_from && (from_status != PCMK_LRM_OP_PENDING)) { // Failed
if (target_node && target_node->details->online) {
pe_rsc_trace(rsc, "Marking active on %s %p %d", target, target_node,
target_node->details->online);
native_add_running(rsc, target_node, data_set);
}
} else { // Pending, or complete but erased
if (target_node && target_node->details->online) {
pe_rsc_trace(rsc, "Marking active on %s %p %d", target, target_node,
target_node->details->online);
native_add_running(rsc, target_node, data_set);
if (source_node && source_node->details->online) {
/* This is a partial migration: the migrate_to completed
* successfully on the source, but the migrate_from has not
* completed. Remember the source and target; if the newly
* chosen target remains the same when we schedule actions
* later, we may continue with the migration.
*/
rsc->partial_migration_target = target_node;
rsc->partial_migration_source = source_node;
}
} else {
/* Consider it failed here - forces a restart, prevents migration */
set_bit(rsc->flags, pe_rsc_failed);
clear_bit(rsc->flags, pe_rsc_allow_migrate);
}
}
}
static void
unpack_migrate_to_failure(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op,
pe_working_set_t *data_set)
{
int target_stop_id = 0;
int target_migrate_from_id = 0;
xmlNode *target_stop = NULL;
xmlNode *target_migrate_from = NULL;
const char *source = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_SOURCE);
const char *target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET);
// Sanity check
CRM_CHECK(source && target && !strcmp(source, node->details->uname), return);
/* If a migration failed, we have to assume the resource is active. Clones
* are not allowed to migrate, so role can't be master.
*/
rsc->role = RSC_ROLE_STARTED;
// Check for stop on the target
target_stop = find_lrm_op(rsc->id, CRMD_ACTION_STOP, target, NULL,
TRUE, data_set);
target_stop_id = pe__call_id(target_stop);
// Check for migrate_from on the target
target_migrate_from = find_lrm_op(rsc->id, CRMD_ACTION_MIGRATED, target,
source, TRUE, data_set);
target_migrate_from_id = pe__call_id(target_migrate_from);
if ((target_stop == NULL) || (target_stop_id < target_migrate_from_id)) {
/* There was no stop on the source, or a stop that happened before a
* migrate_from, so assume the resource is still active on the target
* (if it is up).
*/
pe_node_t *target_node = pe_find_node(data_set->nodes, target);
pe_rsc_trace(rsc, "stop (%d) + migrate_from (%d)",
target_stop_id, target_migrate_from_id);
if (target_node && target_node->details->online) {
native_add_running(rsc, target_node, data_set);
}
} else if (target_migrate_from == NULL) {
/* We know there was a stop on the target, but there may not have been a
* migrate_from (the stop could have happened before migrate_from was
* scheduled or attempted).
*
* That means this could be a "dangling" migration. But first, check
* whether there is a newer migrate_from or start on the source node --
* it's possible the failed migration was followed by a successful
* full restart or migration in the reverse direction, in which case we
* don't want to force it to stop.
*/
xmlNode *source_migrate_from = NULL;
xmlNode *source_start = NULL;
int source_migrate_to_id = pe__call_id(xml_op);
source_migrate_from = find_lrm_op(rsc->id, CRMD_ACTION_MIGRATED, source,
NULL, TRUE, data_set);
if (pe__call_id(source_migrate_from) > source_migrate_to_id) {
return;
}
source_start = find_lrm_op(rsc->id, CRMD_ACTION_START, source, NULL,
TRUE, data_set);
if (pe__call_id(source_start) > source_migrate_to_id) {
return;
}
// Mark node as having dangling migration so we can force a stop later
rsc->dangling_migrations = g_list_prepend(rsc->dangling_migrations, node);
}
}
static void
unpack_migrate_from_failure(pe_resource_t *rsc, pe_node_t *node,
xmlNode *xml_op, pe_working_set_t *data_set)
{
xmlNode *source_stop = NULL;
xmlNode *source_migrate_to = NULL;
const char *source = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_SOURCE);
const char *target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET);
// Sanity check
CRM_CHECK(source && target && !strcmp(target, node->details->uname), return);
/* If a migration failed, we have to assume the resource is active. Clones
* are not allowed to migrate, so role can't be master.
*/
rsc->role = RSC_ROLE_STARTED;
// Check for a stop on the source
source_stop = find_lrm_op(rsc->id, CRMD_ACTION_STOP, source, NULL,
TRUE, data_set);
// Check for a migrate_to on the source
source_migrate_to = find_lrm_op(rsc->id, CRMD_ACTION_MIGRATE,
source, target, TRUE, data_set);
if ((source_stop == NULL)
|| (pe__call_id(source_stop) < pe__call_id(source_migrate_to))) {
/* There was no stop on the source, or a stop that happened before
* migrate_to, so assume the resource is still active on the source (if
* it is up).
*/
pe_node_t *source_node = pe_find_node(data_set->nodes, source);
if (source_node && source_node->details->online) {
native_add_running(rsc, source_node, data_set);
}
}
}
static void
record_failed_op(xmlNode *op, const pe_node_t *node,
const pe_resource_t *rsc, pe_working_set_t *data_set)
{
xmlNode *xIter = NULL;
const char *op_key = crm_element_value(op, XML_LRM_ATTR_TASK_KEY);
if (node->details->online == FALSE) {
return;
}
for (xIter = data_set->failed->children; xIter; xIter = xIter->next) {
const char *key = crm_element_value(xIter, XML_LRM_ATTR_TASK_KEY);
const char *uname = crm_element_value(xIter, XML_ATTR_UNAME);
if(safe_str_eq(op_key, key) && safe_str_eq(uname, node->details->uname)) {
crm_trace("Skipping duplicate entry %s on %s", op_key, node->details->uname);
return;
}
}
crm_trace("Adding entry %s on %s", op_key, node->details->uname);
crm_xml_add(op, XML_ATTR_UNAME, node->details->uname);
crm_xml_add(op, XML_LRM_ATTR_RSCID, rsc->id);
add_node_copy(data_set->failed, op);
}
static const char *get_op_key(xmlNode *xml_op)
{
const char *key = crm_element_value(xml_op, XML_LRM_ATTR_TASK_KEY);
if(key == NULL) {
key = ID(xml_op);
}
return key;
}
static const char *
last_change_str(xmlNode *xml_op)
{
time_t when;
const char *when_s = NULL;
if (crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE,
&when) == pcmk_ok) {
when_s = pcmk__epoch2str(&when);
if (when_s) {
// Skip day of week to make message shorter
when_s = strchr(when_s, ' ');
if (when_s) {
++when_s;
}
}
}
return ((when_s && *when_s)? when_s : "unknown time");
}
static void
unpack_rsc_op_failure(pe_resource_t * rsc, pe_node_t * node, int rc, xmlNode * xml_op, xmlNode ** last_failure,
enum action_fail_response * on_fail, pe_working_set_t * data_set)
{
guint interval_ms = 0;
bool is_probe = false;
pe_action_t *action = NULL;
const char *key = get_op_key(xml_op);
const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
const char *exit_reason = crm_element_value(xml_op,
XML_LRM_ATTR_EXIT_REASON);
CRM_ASSERT(rsc);
CRM_CHECK(task != NULL, return);
*last_failure = xml_op;
crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
if ((interval_ms == 0) && !strcmp(task, CRMD_ACTION_STATUS)) {
is_probe = true;
}
if (exit_reason == NULL) {
exit_reason = "";
}
if (is_not_set(data_set->flags, pe_flag_symmetric_cluster)
&& (rc == PCMK_OCF_NOT_INSTALLED)) {
crm_trace("Unexpected result (%s%s%s) was recorded for "
"%s of %s on %s at %s " CRM_XS " rc=%d id=%s",
services_ocf_exitcode_str(rc),
(*exit_reason? ": " : ""), exit_reason,
(is_probe? "probe" : task), rsc->id, node->details->uname,
last_change_str(xml_op), rc, ID(xml_op));
} else {
crm_warn("Unexpected result (%s%s%s) was recorded for "
"%s of %s on %s at %s " CRM_XS " rc=%d id=%s",
services_ocf_exitcode_str(rc),
(*exit_reason? ": " : ""), exit_reason,
(is_probe? "probe" : task), rsc->id, node->details->uname,
last_change_str(xml_op), rc, ID(xml_op));
if (is_probe && (rc != PCMK_OCF_OK)
&& (rc != PCMK_OCF_NOT_RUNNING)
&& (rc != PCMK_OCF_RUNNING_MASTER)) {
/* A failed (not just unexpected) probe result could mean the user
* didn't know resources will be probed even where they can't run.
*/
crm_notice("If it is not possible for %s to run on %s, see "
"the resource-discovery option for location constraints",
rsc->id, node->details->uname);
}
record_failed_op(xml_op, node, rsc, data_set);
}
action = custom_action(rsc, strdup(key), task, NULL, TRUE, FALSE, data_set);
if ((action->on_fail <= action_fail_fence && *on_fail < action->on_fail) ||
(action->on_fail == action_fail_reset_remote && *on_fail <= action_fail_recover) ||
(action->on_fail == action_fail_restart_container && *on_fail <= action_fail_recover) ||
(*on_fail == action_fail_restart_container && action->on_fail >= action_fail_migrate)) {
pe_rsc_trace(rsc, "on-fail %s -> %s for %s (%s)", fail2text(*on_fail),
fail2text(action->on_fail), action->uuid, key);
*on_fail = action->on_fail;
}
if (!strcmp(task, CRMD_ACTION_STOP)) {
resource_location(rsc, node, -INFINITY, "__stop_fail__", data_set);
} else if (!strcmp(task, CRMD_ACTION_MIGRATE)) {
unpack_migrate_to_failure(rsc, node, xml_op, data_set);
} else if (!strcmp(task, CRMD_ACTION_MIGRATED)) {
unpack_migrate_from_failure(rsc, node, xml_op, data_set);
} else if (!strcmp(task, CRMD_ACTION_PROMOTE)) {
rsc->role = RSC_ROLE_MASTER;
} else if (!strcmp(task, CRMD_ACTION_DEMOTE)) {
if (action->on_fail == action_fail_block) {
rsc->role = RSC_ROLE_MASTER;
rsc->next_role = RSC_ROLE_STOPPED;
} else if(rc == PCMK_OCF_NOT_RUNNING) {
rsc->role = RSC_ROLE_STOPPED;
} else {
/* Staying in master role would put the scheduler and controller
* into a loop. Setting slave role is not dangerous because the
* resource will be stopped as part of recovery, and any master
* promotion will be ordered after that stop.
*/
rsc->role = RSC_ROLE_SLAVE;
}
}
if(is_probe && rc == PCMK_OCF_NOT_INSTALLED) {
/* leave stopped */
pe_rsc_trace(rsc, "Leaving %s stopped", rsc->id);
rsc->role = RSC_ROLE_STOPPED;
} else if (rsc->role < RSC_ROLE_STARTED) {
pe_rsc_trace(rsc, "Setting %s active", rsc->id);
set_active(rsc);
}
pe_rsc_trace(rsc, "Resource %s: role=%s, unclean=%s, on_fail=%s, fail_role=%s",
rsc->id, role2text(rsc->role),
node->details->unclean ? "true" : "false",
fail2text(action->on_fail), role2text(action->fail_role));
if (action->fail_role != RSC_ROLE_STARTED && rsc->next_role < action->fail_role) {
rsc->next_role = action->fail_role;
}
if (action->fail_role == RSC_ROLE_STOPPED) {
int score = -INFINITY;
pe_resource_t *fail_rsc = rsc;
if (fail_rsc->parent) {
pe_resource_t *parent = uber_parent(fail_rsc);
if (pe_rsc_is_clone(parent)
&& is_not_set(parent->flags, pe_rsc_unique)) {
/* For clone resources, if a child fails on an operation
* with on-fail = stop, all the resources fail. Do this by preventing
* the parent from coming up again. */
fail_rsc = parent;
}
}
crm_notice("%s will not be started under current conditions",
fail_rsc->id);
/* make sure it doesn't come up again */
if (fail_rsc->allowed_nodes != NULL) {
g_hash_table_destroy(fail_rsc->allowed_nodes);
}
fail_rsc->allowed_nodes = node_hash_from_list(data_set->nodes);
g_hash_table_foreach(fail_rsc->allowed_nodes, set_node_score, &score);
}
pe_free_action(action);
}
/*!
* \internal
* \brief Remap operation status based on action result
*
* Given an action result, determine an appropriate operation status for the
* purposes of responding to the action (the status provided by the executor is
* not directly usable since the executor does not know what was expected).
*
* \param[in,out] rsc Resource that operation history entry is for
* \param[in] rc Actual return code of operation
* \param[in] target_rc Expected return code of operation
* \param[in] node Node where operation was executed
* \param[in] xml_op Operation history entry XML from CIB status
* \param[in,out] on_fail What should be done about the result
* \param[in] data_set Current cluster working set
*
* \return Operation status based on return code and action info
* \note This may update the resource's current and next role.
*/
static int
determine_op_status(
pe_resource_t *rsc, int rc, int target_rc, pe_node_t * node, xmlNode * xml_op, enum action_fail_response * on_fail, pe_working_set_t * data_set)
{
guint interval_ms = 0;
bool is_probe = false;
int result = PCMK_LRM_OP_DONE;
const char *key = get_op_key(xml_op);
const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
const char *exit_reason = crm_element_value(xml_op,
XML_LRM_ATTR_EXIT_REASON);
CRM_ASSERT(rsc);
CRM_CHECK(task != NULL, return PCMK_LRM_OP_ERROR);
if (exit_reason == NULL) {
exit_reason = "";
}
crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
if ((interval_ms == 0) && !strcmp(task, CRMD_ACTION_STATUS)) {
is_probe = true;
task = "probe";
}
if (target_rc < 0) {
/* Pre-1.0 Pacemaker versions, and Pacemaker 1.1.6 or earlier with
* Heartbeat 2.0.7 or earlier as the cluster layer, did not include the
* target_rc in the transition key, which (along with the similar case
* of a corrupted transition key in the CIB) will be reported to this
* function as -1. Pacemaker 2.0+ does not support rolling upgrades from
* those versions or processing of saved CIB files from those versions,
* so we do not need to care much about this case.
*/
result = PCMK_LRM_OP_ERROR;
crm_warn("Expected result not found for %s on %s (corrupt or obsolete CIB?)",
key, node->details->uname);
} else if (target_rc != rc) {
result = PCMK_LRM_OP_ERROR;
pe_rsc_debug(rsc, "%s on %s: expected %d (%s), got %d (%s%s%s)",
key, node->details->uname,
target_rc, services_ocf_exitcode_str(target_rc),
rc, services_ocf_exitcode_str(rc),
(*exit_reason? ": " : ""), exit_reason);
}
switch (rc) {
case PCMK_OCF_OK:
if (is_probe && (target_rc == PCMK_OCF_NOT_RUNNING)) {
result = PCMK_LRM_OP_DONE;
pe_rsc_info(rsc, "Probe found %s active on %s at %s",
rsc->id, node->details->uname,
last_change_str(xml_op));
}
break;
case PCMK_OCF_NOT_RUNNING:
if (is_probe || target_rc == rc || is_not_set(rsc->flags, pe_rsc_managed)) {
result = PCMK_LRM_OP_DONE;
rsc->role = RSC_ROLE_STOPPED;
/* clear any previous failure actions */
*on_fail = action_fail_ignore;
rsc->next_role = RSC_ROLE_UNKNOWN;
}
break;
case PCMK_OCF_RUNNING_MASTER:
if (is_probe && (rc != target_rc)) {
result = PCMK_LRM_OP_DONE;
pe_rsc_info(rsc,
"Probe found %s active and promoted on %s at %s",
rsc->id, node->details->uname,
last_change_str(xml_op));
}
rsc->role = RSC_ROLE_MASTER;
break;
case PCMK_OCF_DEGRADED_MASTER:
case PCMK_OCF_FAILED_MASTER:
rsc->role = RSC_ROLE_MASTER;
result = PCMK_LRM_OP_ERROR;
break;
case PCMK_OCF_NOT_CONFIGURED:
result = PCMK_LRM_OP_ERROR_FATAL;
break;
case PCMK_OCF_UNIMPLEMENT_FEATURE:
if (interval_ms > 0) {
result = PCMK_LRM_OP_NOTSUPPORTED;
break;
}
// fall through
case PCMK_OCF_NOT_INSTALLED:
case PCMK_OCF_INVALID_PARAM:
case PCMK_OCF_INSUFFICIENT_PRIV:
if (!pe_can_fence(data_set, node)
&& !strcmp(task, CRMD_ACTION_STOP)) {
/* If a stop fails and we can't fence, there's nothing else we can do */
pe_proc_err("No further recovery can be attempted for %s "
"because %s on %s failed (%s%s%s) at %s "
CRM_XS " rc=%d id=%s", rsc->id, task,
node->details->uname, services_ocf_exitcode_str(rc),
(*exit_reason? ": " : ""), exit_reason,
last_change_str(xml_op), rc, ID(xml_op));
clear_bit(rsc->flags, pe_rsc_managed);
set_bit(rsc->flags, pe_rsc_block);
}
result = PCMK_LRM_OP_ERROR_HARD;
break;
default:
if (result == PCMK_LRM_OP_DONE) {
crm_info("Treating unknown exit status %d from %s of %s "
"on %s at %s as failure",
rc, task, rsc->id, node->details->uname,
last_change_str(xml_op));
result = PCMK_LRM_OP_ERROR;
}
break;
}
return result;
}
// return TRUE if start or monitor last failure but parameters changed
static bool
should_clear_for_param_change(xmlNode *xml_op, const char *task,
pe_resource_t *rsc, pe_node_t *node,
pe_working_set_t *data_set)
{
if (!strcmp(task, "start") || !strcmp(task, "monitor")) {
if (pe__bundle_needs_remote_name(rsc)) {
/* We haven't allocated resources yet, so we can't reliably
* substitute addr parameters for the REMOTE_CONTAINER_HACK.
* When that's needed, defer the check until later.
*/
pe__add_param_check(xml_op, rsc, node, pe_check_last_failure,
data_set);
} else {
op_digest_cache_t *digest_data = NULL;
digest_data = rsc_action_digest_cmp(rsc, xml_op, node, data_set);
switch (digest_data->rc) {
case RSC_DIGEST_UNKNOWN:
crm_trace("Resource %s history entry %s on %s"
" has no digest to compare",
rsc->id, get_op_key(xml_op), node->details->id);
break;
case RSC_DIGEST_MATCH:
break;
default:
return TRUE;
}
}
}
return FALSE;
}
// Order action after fencing of remote node, given connection rsc
static void
order_after_remote_fencing(pe_action_t *action, pe_resource_t *remote_conn,
pe_working_set_t *data_set)
{
pe_node_t *remote_node = pe_find_node(data_set->nodes, remote_conn->id);
if (remote_node) {
pe_action_t *fence = pe_fence_op(remote_node, NULL, TRUE, NULL,
FALSE, data_set);
order_actions(fence, action, pe_order_implies_then);
}
}
static bool
should_ignore_failure_timeout(pe_resource_t *rsc, xmlNode *xml_op,
const char *task, guint interval_ms,
bool is_last_failure, pe_working_set_t *data_set)
{
/* Clearing failures of recurring monitors has special concerns. The
* executor reports only changes in the monitor result, so if the
* monitor is still active and still getting the same failure result,
* that will go undetected after the failure is cleared.
*
* Also, the operation history will have the time when the recurring
* monitor result changed to the given code, not the time when the
* result last happened.
*
* @TODO We probably should clear such failures only when the failure
* timeout has passed since the last occurrence of the failed result.
* However we don't record that information. We could maybe approximate
* that by clearing only if there is a more recent successful monitor or
* stop result, but we don't even have that information at this point
* since we are still unpacking the resource's operation history.
*
* This is especially important for remote connection resources with a
* reconnect interval, so in that case, we skip clearing failures
* if the remote node hasn't been fenced.
*/
if (rsc->remote_reconnect_ms
&& is_set(data_set->flags, pe_flag_stonith_enabled)
&& (interval_ms != 0) && safe_str_eq(task, CRMD_ACTION_STATUS)) {
pe_node_t *remote_node = pe_find_node(data_set->nodes, rsc->id);
if (remote_node && !remote_node->details->remote_was_fenced) {
if (is_last_failure) {
crm_info("Waiting to clear monitor failure for remote node %s"
" until fencing has occurred", rsc->id);
}
return TRUE;
}
}
return FALSE;
}
/*!
* \internal
* \brief Check operation age and schedule failure clearing when appropriate
*
* This function has two distinct purposes. The first is to check whether an
* operation history entry is expired (i.e. the resource has a failure timeout,
* the entry is older than the timeout, and the resource either has no fail
* count or its fail count is entirely older than the timeout). The second is to
* schedule fail count clearing when appropriate (i.e. the operation is expired
* and either the resource has an expired fail count or the operation is a
* last_failure for a remote connection resource with a reconnect interval,
* or the operation is a last_failure for a start or monitor operation and the
* resource's parameters have changed since the operation).
*
* \param[in] rsc Resource that operation happened to
* \param[in] node Node that operation happened on
* \param[in] rc Actual result of operation
* \param[in] xml_op Operation history entry XML
* \param[in] data_set Current working set
*
* \return TRUE if operation history entry is expired, FALSE otherwise
*/
static bool
check_operation_expiry(pe_resource_t *rsc, pe_node_t *node, int rc,
xmlNode *xml_op, pe_working_set_t *data_set)
{
bool expired = FALSE;
bool is_last_failure = pcmk__ends_with(ID(xml_op), "_last_failure_0");
time_t last_run = 0;
guint interval_ms = 0;
int unexpired_fail_count = 0;
const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
const char *clear_reason = NULL;
crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
if ((rsc->failure_timeout > 0)
&& (crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE,
&last_run) == 0)) {
// Resource has a failure-timeout, and history entry has a timestamp
time_t now = get_effective_time(data_set);
time_t last_failure = 0;
// Is this particular operation history older than the failure timeout?
if ((now >= (last_run + rsc->failure_timeout))
&& !should_ignore_failure_timeout(rsc, xml_op, task, interval_ms,
is_last_failure, data_set)) {
expired = TRUE;
}
// Does the resource as a whole have an unexpired fail count?
unexpired_fail_count = pe_get_failcount(node, rsc, &last_failure,
pe_fc_effective, xml_op,
data_set);
// Update scheduler recheck time according to *last* failure
crm_trace("%s@%lld is %sexpired @%lld with unexpired_failures=%d timeout=%ds"
" last-failure@%lld",
ID(xml_op), (long long) last_run, (expired? "" : "not "),
(long long) now, unexpired_fail_count, rsc->failure_timeout,
(long long) last_failure);
last_failure += rsc->failure_timeout + 1;
if (unexpired_fail_count && (now < last_failure)) {
pe__update_recheck_time(last_failure, data_set);
}
}
if (expired) {
if (pe_get_failcount(node, rsc, NULL, pe_fc_default, xml_op, data_set)) {
// There is a fail count ignoring timeout
if (unexpired_fail_count == 0) {
// There is no fail count considering timeout
clear_reason = "it expired";
} else {
/* This operation is old, but there is an unexpired fail count.
* In a properly functioning cluster, this should only be
* possible if this operation is not a failure (otherwise the
* fail count should be expired too), so this is really just a
* failsafe.
*/
expired = FALSE;
}
} else if (is_last_failure && rsc->remote_reconnect_ms) {
/* Clear any expired last failure when reconnect interval is set,
* even if there is no fail count.
*/
clear_reason = "reconnect interval is set";
}
}
if (!expired && is_last_failure
&& should_clear_for_param_change(xml_op, task, rsc, node, data_set)) {
clear_reason = "resource parameters have changed";
}
if (clear_reason != NULL) {
// Schedule clearing of the fail count
pe_action_t *clear_op = pe__clear_failcount(rsc, node, clear_reason,
data_set);
if (is_set(data_set->flags, pe_flag_stonith_enabled)
&& rsc->remote_reconnect_ms) {
/* If we're clearing a remote connection due to a reconnect
* interval, we want to wait until any scheduled fencing
* completes.
*
* We could limit this to remote_node->details->unclean, but at
* this point, that's always true (it won't be reliable until
* after unpack_node_loop() is done).
*/
crm_info("Clearing %s failure will wait until any scheduled "
"fencing of %s completes", task, rsc->id);
order_after_remote_fencing(clear_op, rsc, data_set);
}
}
if (expired && (interval_ms == 0) && safe_str_eq(task, CRMD_ACTION_STATUS)) {
switch(rc) {
case PCMK_OCF_OK:
case PCMK_OCF_NOT_RUNNING:
case PCMK_OCF_RUNNING_MASTER:
case PCMK_OCF_DEGRADED:
case PCMK_OCF_DEGRADED_MASTER:
// Don't expire probes that return these values
expired = FALSE;
break;
}
}
return expired;
}
int pe__target_rc_from_xml(xmlNode *xml_op)
{
int target_rc = 0;
const char *key = crm_element_value(xml_op, XML_ATTR_TRANSITION_KEY);
if (key == NULL) {
return -1;
}
decode_transition_key(key, NULL, NULL, NULL, &target_rc);
return target_rc;
}
static enum action_fail_response
get_action_on_fail(pe_resource_t *rsc, const char *key, const char *task, pe_working_set_t * data_set)
{
int result = action_fail_recover;
pe_action_t *action = custom_action(rsc, strdup(key), task, NULL, TRUE, FALSE, data_set);
result = action->on_fail;
pe_free_action(action);
return result;
}
static void
update_resource_state(pe_resource_t * rsc, pe_node_t * node, xmlNode * xml_op, const char * task, int rc,
xmlNode * last_failure, enum action_fail_response * on_fail, pe_working_set_t * data_set)
{
gboolean clear_past_failure = FALSE;
CRM_ASSERT(rsc);
CRM_ASSERT(xml_op);
if (rc == PCMK_OCF_NOT_RUNNING) {
clear_past_failure = TRUE;
} else if (rc == PCMK_OCF_NOT_INSTALLED) {
rsc->role = RSC_ROLE_STOPPED;
} else if (safe_str_eq(task, CRMD_ACTION_STATUS)) {
if (last_failure) {
const char *op_key = get_op_key(xml_op);
const char *last_failure_key = get_op_key(last_failure);
if (safe_str_eq(op_key, last_failure_key)) {
clear_past_failure = TRUE;
}
}
if (rsc->role < RSC_ROLE_STARTED) {
set_active(rsc);
}
} else if (safe_str_eq(task, CRMD_ACTION_START)) {
rsc->role = RSC_ROLE_STARTED;
clear_past_failure = TRUE;
} else if (safe_str_eq(task, CRMD_ACTION_STOP)) {
rsc->role = RSC_ROLE_STOPPED;
clear_past_failure = TRUE;
} else if (safe_str_eq(task, CRMD_ACTION_PROMOTE)) {
rsc->role = RSC_ROLE_MASTER;
clear_past_failure = TRUE;
} else if (safe_str_eq(task, CRMD_ACTION_DEMOTE)) {
/* Demote from Master does not clear an error */
rsc->role = RSC_ROLE_SLAVE;
} else if (safe_str_eq(task, CRMD_ACTION_MIGRATED)) {
rsc->role = RSC_ROLE_STARTED;
clear_past_failure = TRUE;
} else if (safe_str_eq(task, CRMD_ACTION_MIGRATE)) {
unpack_migrate_to_success(rsc, node, xml_op, data_set);
} else if (rsc->role < RSC_ROLE_STARTED) {
pe_rsc_trace(rsc, "%s active on %s", rsc->id, node->details->uname);
set_active(rsc);
}
/* clear any previous failure actions */
if (clear_past_failure) {
switch (*on_fail) {
case action_fail_stop:
case action_fail_fence:
case action_fail_migrate:
case action_fail_standby:
pe_rsc_trace(rsc, "%s.%s is not cleared by a completed stop",
rsc->id, fail2text(*on_fail));
break;
case action_fail_block:
case action_fail_ignore:
case action_fail_recover:
case action_fail_restart_container:
*on_fail = action_fail_ignore;
rsc->next_role = RSC_ROLE_UNKNOWN;
break;
case action_fail_reset_remote:
if (rsc->remote_reconnect_ms == 0) {
/* With no reconnect interval, the connection is allowed to
* start again after the remote node is fenced and
* completely stopped. (With a reconnect interval, we wait
* for the failure to be cleared entirely before attempting
* to reconnect.)
*/
*on_fail = action_fail_ignore;
rsc->next_role = RSC_ROLE_UNKNOWN;
}
break;
}
}
}
/*!
* \internal
* \brief Remap informational monitor results to usual values
*
* Certain OCF result codes are for providing extended information to the
* user about services that aren't yet failed but not entirely healthy either.
* These must be treated as the "normal" result by pacemaker.
*
* \param[in] rc Actual result of a monitor action
* \param[in] xml_op Operation history XML
* \param[in] node Node that operation happened on
* \param[in] rsc Resource that operation happened to
* \param[in] data_set Cluster working set
*
* \return Result code that pacemaker should use
*
* \note If the result is remapped, and the node is not shutting down or failed,
* the operation will be recorded in the data set's list of failed
* operations, to highlight it for the user.
*/
static int
remap_monitor_rc(int rc, xmlNode *xml_op, const pe_node_t *node,
const pe_resource_t *rsc, pe_working_set_t *data_set)
{
int remapped_rc = rc;
switch (rc) {
case PCMK_OCF_DEGRADED:
remapped_rc = PCMK_OCF_OK;
break;
case PCMK_OCF_DEGRADED_MASTER:
remapped_rc = PCMK_OCF_RUNNING_MASTER;
break;
default:
break;
}
if (rc != remapped_rc) {
crm_trace("Remapping monitor result %d to %d", rc, remapped_rc);
if (!node->details->shutdown || node->details->online) {
record_failed_op(xml_op, node, rsc, data_set);
}
}
return remapped_rc;
}
static void
unpack_rsc_op(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op,
xmlNode **last_failure, enum action_fail_response *on_fail,
pe_working_set_t *data_set)
{
int rc = 0;
int task_id = 0;
int target_rc = 0;
int status = PCMK_LRM_OP_UNKNOWN;
guint interval_ms = 0;
const char *task = NULL;
const char *task_key = NULL;
const char *exit_reason = NULL;
bool expired = FALSE;
pe_resource_t *parent = rsc;
enum action_fail_response failure_strategy = action_fail_recover;
CRM_CHECK(rsc && node && xml_op, return);
target_rc = pe__target_rc_from_xml(xml_op);
task_key = get_op_key(xml_op);
task = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
exit_reason = crm_element_value(xml_op, XML_LRM_ATTR_EXIT_REASON);
if (exit_reason == NULL) {
exit_reason = "";
}
crm_element_value_int(xml_op, XML_LRM_ATTR_RC, &rc);
crm_element_value_int(xml_op, XML_LRM_ATTR_CALLID, &task_id);
crm_element_value_int(xml_op, XML_LRM_ATTR_OPSTATUS, &status);
crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
CRM_CHECK(task != NULL, return);
CRM_CHECK(status <= PCMK_LRM_OP_INVALID, return);
CRM_CHECK(status >= PCMK_LRM_OP_PENDING, return);
if (!strcmp(task, CRMD_ACTION_NOTIFY) ||
!strcmp(task, CRMD_ACTION_METADATA)) {
/* safe to ignore these */
return;
}
if (is_not_set(rsc->flags, pe_rsc_unique)) {
parent = uber_parent(rsc);
}
pe_rsc_trace(rsc, "Unpacking task %s/%s (call_id=%d, status=%d, rc=%d) on %s (role=%s)",
task_key, task, task_id, status, rc, node->details->uname, role2text(rsc->role));
if (node->details->unclean) {
pe_rsc_trace(rsc, "Node %s (where %s is running) is unclean."
" Further action depends on the value of the stop's on-fail attribute",
node->details->uname, rsc->id);
}
/* It should be possible to call remap_monitor_rc() first then call
* check_operation_expiry() only if rc != target_rc, because there should
* never be a fail count without at least one unexpected result in the
* resource history. That would be more efficient by avoiding having to call
* check_operation_expiry() for expected results.
*
* However, we do have such configurations in the scheduler regression
* tests, even if it shouldn't be possible with the current code. It's
* probably a good idea anyway, but that would require updating the test
* inputs to something currently possible.
*/
if ((status != PCMK_LRM_OP_NOT_INSTALLED)
&& check_operation_expiry(rsc, node, rc, xml_op, data_set)) {
expired = TRUE;
}
if (!strcmp(task, CRMD_ACTION_STATUS)) {
rc = remap_monitor_rc(rc, xml_op, node, rsc, data_set);
}
if (expired && (rc != target_rc)) {
const char *magic = crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC);
if (interval_ms == 0) {
crm_notice("Ignoring expired %s failure on %s "
CRM_XS " actual=%d expected=%d magic=%s",
task_key, node->details->uname, rc, target_rc, magic);
goto done;
} else if(node->details->online && node->details->unclean == FALSE) {
/* Reschedule the recurring monitor. CancelXmlOp() won't work at
* this stage, so as a hacky workaround, forcibly change the restart
* digest so check_action_definition() does what we want later.
*
* @TODO We should skip this if there is a newer successful monitor.
* Also, this causes rescheduling only if the history entry
* has an op-digest (which the expire-non-blocked-failure
* scheduler regression test doesn't, but that may not be a
* realistic scenario in production).
*/
crm_notice("Rescheduling %s after failure expired on %s "
CRM_XS " actual=%d expected=%d magic=%s",
task_key, node->details->uname, rc, target_rc, magic);
crm_xml_add(xml_op, XML_LRM_ATTR_RESTART_DIGEST, "calculated-failure-timeout");
goto done;
}
}
/* If the executor reported an operation status of anything but done or
* error, consider that final. But for done or error, we know better whether
* it should be treated as a failure or not, because we know the expected
* result.
*/
if(status == PCMK_LRM_OP_DONE || status == PCMK_LRM_OP_ERROR) {
status = determine_op_status(rsc, rc, target_rc, node, xml_op, on_fail, data_set);
pe_rsc_trace(rsc, "Remapped %s status to %d", task_key, status);
}
switch (status) {
case PCMK_LRM_OP_CANCELLED:
// Should never happen
pe_err("Resource history contains cancellation '%s' "
"(%s of %s on %s at %s)",
ID(xml_op), task, rsc->id, node->details->uname,
last_change_str(xml_op));
break;
case PCMK_LRM_OP_PENDING:
if (!strcmp(task, CRMD_ACTION_START)) {
set_bit(rsc->flags, pe_rsc_start_pending);
set_active(rsc);
} else if (!strcmp(task, CRMD_ACTION_PROMOTE)) {
rsc->role = RSC_ROLE_MASTER;
} else if (!strcmp(task, CRMD_ACTION_MIGRATE) && node->details->unclean) {
/* If a pending migrate_to action is out on a unclean node,
* we have to force the stop action on the target. */
const char *migrate_target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET);
pe_node_t *target = pe_find_node(data_set->nodes, migrate_target);
if (target) {
stop_action(rsc, target, FALSE);
}
}
if (rsc->pending_task == NULL) {
if ((interval_ms != 0) || strcmp(task, CRMD_ACTION_STATUS)) {
rsc->pending_task = strdup(task);
rsc->pending_node = node;
} else {
/* Pending probes are not printed, even if pending
* operations are requested. If someone ever requests that
* behavior, enable the below and the corresponding part of
* native.c:native_pending_task().
*/
#if 0
rsc->pending_task = strdup("probe");
rsc->pending_node = node;
#endif
}
}
break;
case PCMK_LRM_OP_DONE:
pe_rsc_trace(rsc, "%s of %s on %s completed at %s " CRM_XS " id=%s",
task, rsc->id, node->details->uname,
last_change_str(xml_op), ID(xml_op));
update_resource_state(rsc, node, xml_op, task, rc, *last_failure, on_fail, data_set);
break;
case PCMK_LRM_OP_NOT_INSTALLED:
failure_strategy = get_action_on_fail(rsc, task_key, task, data_set);
if (failure_strategy == action_fail_ignore) {
crm_warn("Cannot ignore failed %s of %s on %s: "
"Resource agent doesn't exist "
CRM_XS " status=%d rc=%d id=%s",
task, rsc->id, node->details->uname, status, rc,
ID(xml_op));
/* Also for printing it as "FAILED" by marking it as pe_rsc_failed later */
*on_fail = action_fail_migrate;
}
resource_location(parent, node, -INFINITY, "hard-error", data_set);
unpack_rsc_op_failure(rsc, node, rc, xml_op, last_failure, on_fail, data_set);
break;
case PCMK_LRM_OP_NOT_CONNECTED:
if (pe__is_guest_or_remote_node(node)
&& is_set(node->details->remote_rsc->flags, pe_rsc_managed)) {
/* We should never get into a situation where a managed remote
* connection resource is considered OK but a resource action
* behind the connection gets a "not connected" status. But as a
* fail-safe in case a bug or unusual circumstances do lead to
* that, ensure the remote connection is considered failed.
*/
set_bit(node->details->remote_rsc->flags, pe_rsc_failed);
}
// fall through
case PCMK_LRM_OP_ERROR:
case PCMK_LRM_OP_ERROR_HARD:
case PCMK_LRM_OP_ERROR_FATAL:
case PCMK_LRM_OP_TIMEOUT:
case PCMK_LRM_OP_NOTSUPPORTED:
case PCMK_LRM_OP_INVALID:
failure_strategy = get_action_on_fail(rsc, task_key, task, data_set);
if ((failure_strategy == action_fail_ignore)
|| (failure_strategy == action_fail_restart_container
&& !strcmp(task, CRMD_ACTION_STOP))) {
crm_warn("Pretending failed %s (%s%s%s) of %s on %s at %s "
"succeeded " CRM_XS " rc=%d id=%s",
task, services_ocf_exitcode_str(rc),
(*exit_reason? ": " : ""), exit_reason, rsc->id,
node->details->uname, last_change_str(xml_op), rc,
ID(xml_op));
update_resource_state(rsc, node, xml_op, task, target_rc, *last_failure, on_fail, data_set);
crm_xml_add(xml_op, XML_ATTR_UNAME, node->details->uname);
set_bit(rsc->flags, pe_rsc_failure_ignored);
record_failed_op(xml_op, node, rsc, data_set);
if (failure_strategy == action_fail_restart_container && *on_fail <= action_fail_recover) {
*on_fail = failure_strategy;
}
} else {
unpack_rsc_op_failure(rsc, node, rc, xml_op, last_failure, on_fail, data_set);
if(status == PCMK_LRM_OP_ERROR_HARD) {
do_crm_log(rc != PCMK_OCF_NOT_INSTALLED?LOG_ERR:LOG_NOTICE,
"Preventing %s from restarting on %s because "
"of hard failure (%s%s%s)" CRM_XS " rc=%d id=%s",
parent->id, node->details->uname,
services_ocf_exitcode_str(rc),
(*exit_reason? ": " : ""), exit_reason,
rc, ID(xml_op));
resource_location(parent, node, -INFINITY, "hard-error", data_set);
} else if(status == PCMK_LRM_OP_ERROR_FATAL) {
crm_err("Preventing %s from restarting anywhere because "
"of fatal failure (%s%s%s) " CRM_XS " rc=%d id=%s",
parent->id, services_ocf_exitcode_str(rc),
(*exit_reason? ": " : ""), exit_reason,
rc, ID(xml_op));
resource_location(parent, NULL, -INFINITY, "fatal-error", data_set);
}
}
break;
}
done:
pe_rsc_trace(rsc, "Resource %s after %s: role=%s, next=%s",
rsc->id, task, role2text(rsc->role),
role2text(rsc->next_role));
}
static void
add_node_attrs(xmlNode *xml_obj, pe_node_t *node, bool overwrite,
pe_working_set_t *data_set)
{
const char *cluster_name = NULL;
g_hash_table_insert(node->details->attrs,
strdup(CRM_ATTR_UNAME), strdup(node->details->uname));
g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_ID),
strdup(node->details->id));
if (safe_str_eq(node->details->id, data_set->dc_uuid)) {
data_set->dc_node = node;
node->details->is_dc = TRUE;
g_hash_table_insert(node->details->attrs,
strdup(CRM_ATTR_IS_DC), strdup(XML_BOOLEAN_TRUE));
} else {
g_hash_table_insert(node->details->attrs,
strdup(CRM_ATTR_IS_DC), strdup(XML_BOOLEAN_FALSE));
}
cluster_name = g_hash_table_lookup(data_set->config_hash, "cluster-name");
if (cluster_name) {
g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_CLUSTER_NAME),
strdup(cluster_name));
}
pe__unpack_dataset_nvpairs(xml_obj, XML_TAG_ATTR_SETS, NULL,
node->details->attrs, NULL, overwrite, data_set);
if (pe_node_attribute_raw(node, CRM_ATTR_SITE_NAME) == NULL) {
const char *site_name = pe_node_attribute_raw(node, "site-name");
if (site_name) {
g_hash_table_insert(node->details->attrs,
strdup(CRM_ATTR_SITE_NAME),
strdup(site_name));
} else if (cluster_name) {
/* Default to cluster-name if unset */
g_hash_table_insert(node->details->attrs,
strdup(CRM_ATTR_SITE_NAME),
strdup(cluster_name));
}
}
}
static GListPtr
extract_operations(const char *node, const char *rsc, xmlNode * rsc_entry, gboolean active_filter)
{
int counter = -1;
int stop_index = -1;
int start_index = -1;
xmlNode *rsc_op = NULL;
GListPtr gIter = NULL;
GListPtr op_list = NULL;
GListPtr sorted_op_list = NULL;
/* extract operations */
op_list = NULL;
sorted_op_list = NULL;
for (rsc_op = __xml_first_child_element(rsc_entry);
rsc_op != NULL; rsc_op = __xml_next_element(rsc_op)) {
if (crm_str_eq((const char *)rsc_op->name, XML_LRM_TAG_RSC_OP, TRUE)) {
crm_xml_add(rsc_op, "resource", rsc);
crm_xml_add(rsc_op, XML_ATTR_UNAME, node);
op_list = g_list_prepend(op_list, rsc_op);
}
}
if (op_list == NULL) {
/* if there are no operations, there is nothing to do */
return NULL;
}
sorted_op_list = g_list_sort(op_list, sort_op_by_callid);
/* create active recurring operations as optional */
if (active_filter == FALSE) {
return sorted_op_list;
}
op_list = NULL;
calculate_active_ops(sorted_op_list, &start_index, &stop_index);
for (gIter = sorted_op_list; gIter != NULL; gIter = gIter->next) {
xmlNode *rsc_op = (xmlNode *) gIter->data;
counter++;
if (start_index < stop_index) {
crm_trace("Skipping %s: not active", ID(rsc_entry));
break;
} else if (counter < start_index) {
crm_trace("Skipping %s: old", ID(rsc_op));
continue;
}
op_list = g_list_append(op_list, rsc_op);
}
g_list_free(sorted_op_list);
return op_list;
}
GListPtr
find_operations(const char *rsc, const char *node, gboolean active_filter,
pe_working_set_t * data_set)
{
GListPtr output = NULL;
GListPtr intermediate = NULL;
xmlNode *tmp = NULL;
xmlNode *status = find_xml_node(data_set->input, XML_CIB_TAG_STATUS, TRUE);
pe_node_t *this_node = NULL;
xmlNode *node_state = NULL;
for (node_state = __xml_first_child_element(status); node_state != NULL;
node_state = __xml_next_element(node_state)) {
if (crm_str_eq((const char *)node_state->name, XML_CIB_TAG_STATE, TRUE)) {
const char *uname = crm_element_value(node_state, XML_ATTR_UNAME);
if (node != NULL && safe_str_neq(uname, node)) {
continue;
}
this_node = pe_find_node(data_set->nodes, uname);
if(this_node == NULL) {
CRM_LOG_ASSERT(this_node != NULL);
continue;
} else if (pe__is_guest_or_remote_node(this_node)) {
determine_remote_online_status(data_set, this_node);
} else {
determine_online_status(node_state, this_node, data_set);
}
if (this_node->details->online || is_set(data_set->flags, pe_flag_stonith_enabled)) {
/* offline nodes run no resources...
* unless stonith is enabled in which case we need to
* make sure rsc start events happen after the stonith
*/
xmlNode *lrm_rsc = NULL;
tmp = find_xml_node(node_state, XML_CIB_TAG_LRM, FALSE);
tmp = find_xml_node(tmp, XML_LRM_TAG_RESOURCES, FALSE);
for (lrm_rsc = __xml_first_child_element(tmp); lrm_rsc != NULL;
lrm_rsc = __xml_next_element(lrm_rsc)) {
if (crm_str_eq((const char *)lrm_rsc->name, XML_LRM_TAG_RESOURCE, TRUE)) {
const char *rsc_id = crm_element_value(lrm_rsc, XML_ATTR_ID);
if (rsc != NULL && safe_str_neq(rsc_id, rsc)) {
continue;
}
intermediate = extract_operations(uname, rsc_id, lrm_rsc, active_filter);
output = g_list_concat(output, intermediate);
}
}
}
}
}
return output;
}
diff --git a/lib/pengine/unpack.h b/lib/pengine/unpack.h
deleted file mode 100644
index 6f7226a05e..0000000000
--- a/lib/pengine/unpack.h
+++ /dev/null
@@ -1,55 +0,0 @@
-/*
- * Copyright 2004-2020 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.
- */
-
-#ifndef PENGINE_UNPACK__H
-# define PENGINE_UNPACK__H
-
-extern gboolean unpack_remote_nodes(xmlNode * xml_resources, pe_working_set_t * data_set);
-
-extern gboolean unpack_resources(xmlNode * xml_resources, pe_working_set_t * data_set);
-
-extern gboolean unpack_config(xmlNode * config, pe_working_set_t * data_set);
-
-extern gboolean unpack_nodes(xmlNode * xml_nodes, pe_working_set_t * data_set);
-
-extern gboolean unpack_tags(xmlNode * xml_tags, pe_working_set_t * data_set);
-
-extern gboolean unpack_status(xmlNode * status, pe_working_set_t * data_set);
-
-extern gint sort_op_by_callid(gconstpointer a, gconstpointer b);
-
-extern gboolean unpack_lrm_resources(pe_node_t * node, xmlNode * lrm_state,
- pe_working_set_t * data_set);
-
-// Some warnings we don't want to print every transition
-
-enum pe_warn_once_e {
- pe_wo_blind = 0x0001,
- pe_wo_restart_type = 0x0002,
- pe_wo_role_after = 0x0004,
- pe_wo_poweroff = 0x0008,
- pe_wo_require_all = 0x0010,
- pe_wo_order_score = 0x0020,
- pe_wo_neg_threshold = 0x0040,
-};
-
-extern uint32_t pe_wo;
-
-#define pe_warn_once(pe_wo_bit, fmt...) do { \
- if (is_not_set(pe_wo, pe_wo_bit)) { \
- if (pe_wo_bit == pe_wo_blind) { \
- crm_warn(fmt); \
- } else { \
- pe_warn(fmt); \
- } \
- set_bit(pe_wo, pe_wo_bit); \
- } \
- } while (0);
-
-#endif
diff --git a/lib/pengine/utils.c b/lib/pengine/utils.c
index dbc73b8d4f..d26a1dfe89 100644
--- a/lib/pengine/utils.c
+++ b/lib/pengine/utils.c
@@ -1,2766 +1,2764 @@
/*
* Copyright 2004-2020 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 <crm/crm.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <crm/common/util.h>
#include <ctype.h>
#include <glib.h>
#include <stdbool.h>
#include <crm/pengine/rules.h>
#include <crm/pengine/internal.h>
-#include <unpack.h>
-
extern xmlNode *get_object_root(const char *object_type, xmlNode * the_root);
void print_str_str(gpointer key, gpointer value, gpointer user_data);
gboolean ghash_free_str_str(gpointer key, gpointer value, gpointer user_data);
void unpack_operation(pe_action_t * action, xmlNode * xml_obj, pe_resource_t * container,
pe_working_set_t * data_set);
static xmlNode *find_rsc_op_entry_helper(pe_resource_t * rsc, const char *key,
gboolean include_disabled);
#if ENABLE_VERSIONED_ATTRS
pe_rsc_action_details_t *
pe_rsc_action_details(pe_action_t *action)
{
pe_rsc_action_details_t *details;
CRM_CHECK(action != NULL, return NULL);
if (action->action_details == NULL) {
action->action_details = calloc(1, sizeof(pe_rsc_action_details_t));
CRM_CHECK(action->action_details != NULL, return NULL);
}
details = (pe_rsc_action_details_t *) action->action_details;
if (details->versioned_parameters == NULL) {
details->versioned_parameters = create_xml_node(NULL,
XML_TAG_OP_VER_ATTRS);
}
if (details->versioned_meta == NULL) {
details->versioned_meta = create_xml_node(NULL, XML_TAG_OP_VER_META);
}
return details;
}
static void
pe_free_rsc_action_details(pe_action_t *action)
{
pe_rsc_action_details_t *details;
if ((action == NULL) || (action->action_details == NULL)) {
return;
}
details = (pe_rsc_action_details_t *) action->action_details;
if (details->versioned_parameters) {
free_xml(details->versioned_parameters);
}
if (details->versioned_meta) {
free_xml(details->versioned_meta);
}
action->action_details = NULL;
}
#endif
/*!
* \internal
* \brief Check whether we can fence a particular node
*
* \param[in] data_set Working set for cluster
* \param[in] node Name of node to check
*
* \return true if node can be fenced, false otherwise
*/
bool
pe_can_fence(pe_working_set_t *data_set, pe_node_t *node)
{
if (pe__is_guest_node(node)) {
/* Guest nodes are fenced by stopping their container resource. We can
* do that if the container's host is either online or fenceable.
*/
pe_resource_t *rsc = node->details->remote_rsc->container;
for (GList *n = rsc->running_on; n != NULL; n = n->next) {
pe_node_t *container_node = n->data;
if (!container_node->details->online
&& !pe_can_fence(data_set, container_node)) {
return false;
}
}
return true;
} else if(is_not_set(data_set->flags, pe_flag_stonith_enabled)) {
return false; /* Turned off */
} else if (is_not_set(data_set->flags, pe_flag_have_stonith_resource)) {
return false; /* No devices */
} else if (is_set(data_set->flags, pe_flag_have_quorum)) {
return true;
} else if (data_set->no_quorum_policy == no_quorum_ignore) {
return true;
} else if(node == NULL) {
return false;
} else if(node->details->online) {
crm_notice("We can fence %s without quorum because they're in our membership", node->details->uname);
return true;
}
crm_trace("Cannot fence %s", node->details->uname);
return false;
}
pe_node_t *
node_copy(const pe_node_t *this_node)
{
pe_node_t *new_node = NULL;
CRM_CHECK(this_node != NULL, return NULL);
new_node = calloc(1, sizeof(pe_node_t));
CRM_ASSERT(new_node != NULL);
crm_trace("Copying %p (%s) to %p", this_node, this_node->details->uname, new_node);
new_node->rsc_discover_mode = this_node->rsc_discover_mode;
new_node->weight = this_node->weight;
new_node->fixed = this_node->fixed;
new_node->details = this_node->details;
return new_node;
}
/* any node in list1 or list2 and not in the other gets a score of -INFINITY */
void
node_list_exclude(GHashTable * hash, GListPtr list, gboolean merge_scores)
{
GHashTable *result = hash;
pe_node_t *other_node = NULL;
GListPtr gIter = list;
GHashTableIter iter;
pe_node_t *node = NULL;
g_hash_table_iter_init(&iter, hash);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) {
other_node = pe_find_node_id(list, node->details->id);
if (other_node == NULL) {
node->weight = -INFINITY;
} else if (merge_scores) {
node->weight = merge_weights(node->weight, other_node->weight);
}
}
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
other_node = pe_hash_table_lookup(result, node->details->id);
if (other_node == NULL) {
pe_node_t *new_node = node_copy(node);
new_node->weight = -INFINITY;
g_hash_table_insert(result, (gpointer) new_node->details->id, new_node);
}
}
}
GHashTable *
node_hash_from_list(GListPtr list)
{
GListPtr gIter = list;
GHashTable *result = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL,
free);
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
pe_node_t *n = node_copy(node);
g_hash_table_insert(result, (gpointer) n->details->id, n);
}
return result;
}
GListPtr
node_list_dup(GListPtr list1, gboolean reset, gboolean filter)
{
GListPtr result = NULL;
GListPtr gIter = list1;
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *new_node = NULL;
pe_node_t *this_node = (pe_node_t *) gIter->data;
if (filter && this_node->weight < 0) {
continue;
}
new_node = node_copy(this_node);
if (reset) {
new_node->weight = 0;
}
if (new_node != NULL) {
result = g_list_prepend(result, new_node);
}
}
return result;
}
gint
sort_node_uname(gconstpointer a, gconstpointer b)
{
const char *name_a = ((const pe_node_t *) a)->details->uname;
const char *name_b = ((const pe_node_t *) b)->details->uname;
while (*name_a && *name_b) {
if (isdigit(*name_a) && isdigit(*name_b)) {
// If node names contain a number, sort numerically
char *end_a = NULL;
char *end_b = NULL;
long num_a = strtol(name_a, &end_a, 10);
long num_b = strtol(name_b, &end_b, 10);
// allow ordering e.g. 007 > 7
size_t len_a = end_a - name_a;
size_t len_b = end_b - name_b;
if (num_a < num_b) {
return -1;
} else if (num_a > num_b) {
return 1;
} else if (len_a < len_b) {
return -1;
} else if (len_a > len_b) {
return 1;
}
name_a = end_a;
name_b = end_b;
} else {
// Compare non-digits case-insensitively
int lower_a = tolower(*name_a);
int lower_b = tolower(*name_b);
if (lower_a < lower_b) {
return -1;
} else if (lower_a > lower_b) {
return 1;
}
++name_a;
++name_b;
}
}
if (!*name_a && *name_b) {
return -1;
} else if (*name_a && !*name_b) {
return 1;
}
return 0;
}
/*!
* \internal
* \brief Output node weights to stdout
*
* \param[in] rsc Use allowed nodes for this resource
* \param[in] comment Text description to prefix lines with
* \param[in] nodes If rsc is not specified, use these nodes
*/
static void
pe__output_node_weights(pe_resource_t *rsc, const char *comment,
GHashTable *nodes)
{
char score[128]; // Stack-allocated since this is called frequently
// Sort the nodes so the output is consistent for regression tests
GList *list = g_list_sort(g_hash_table_get_values(nodes), sort_node_uname);
for (GList *gIter = list; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
score2char_stack(node->weight, score, sizeof(score));
if (rsc) {
printf("%s: %s allocation score on %s: %s\n",
comment, rsc->id, node->details->uname, score);
} else {
printf("%s: %s = %s\n", comment, node->details->uname, score);
}
}
g_list_free(list);
}
/*!
* \internal
* \brief Log node weights at trace level
*
* \param[in] file Caller's filename
* \param[in] function Caller's function name
* \param[in] line Caller's line number
* \param[in] rsc Use allowed nodes for this resource
* \param[in] comment Text description to prefix lines with
* \param[in] nodes If rsc is not specified, use these nodes
*/
static void
pe__log_node_weights(const char *file, const char *function, int line,
pe_resource_t *rsc, const char *comment, GHashTable *nodes)
{
GHashTableIter iter;
pe_node_t *node = NULL;
char score[128]; // Stack-allocated since this is called frequently
// Don't waste time if we're not tracing at this point
pcmk__log_else(LOG_TRACE, return);
g_hash_table_iter_init(&iter, nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) {
score2char_stack(node->weight, score, sizeof(score));
if (rsc) {
qb_log_from_external_source(function, file,
"%s: %s allocation score on %s: %s",
LOG_TRACE, line, 0,
comment, rsc->id,
node->details->uname, score);
} else {
qb_log_from_external_source(function, file, "%s: %s = %s",
LOG_TRACE, line, 0,
comment, node->details->uname,
score);
}
}
}
/*!
* \internal
* \brief Log or output node weights
*
* \param[in] file Caller's filename
* \param[in] function Caller's function name
* \param[in] line Caller's line number
* \param[in] to_log Log if true, otherwise output
* \param[in] rsc Use allowed nodes for this resource
* \param[in] comment Text description to prefix lines with
* \param[in] nodes If rsc is not specified, use these nodes
*/
void
pe__show_node_weights_as(const char *file, const char *function, int line,
bool to_log, pe_resource_t *rsc, const char *comment,
GHashTable *nodes)
{
if (rsc != NULL) {
if (is_set(rsc->flags, pe_rsc_orphan)) {
// Don't show allocation scores for orphans
return;
}
nodes = rsc->allowed_nodes;
}
if (nodes == NULL) {
// Nothing to show
return;
}
if (to_log) {
pe__log_node_weights(file, function, line, rsc, comment, nodes);
} else {
pe__output_node_weights(rsc, comment, nodes);
}
// If this resource has children, repeat recursively for each
if (rsc && rsc->children) {
for (GList *gIter = rsc->children; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child = (pe_resource_t *) gIter->data;
pe__show_node_weights_as(file, function, line, to_log, child,
comment, nodes);
}
}
}
static void
append_dump_text(gpointer key, gpointer value, gpointer user_data)
{
char **dump_text = user_data;
char *new_text = crm_strdup_printf("%s %s=%s",
*dump_text, (char *)key, (char *)value);
free(*dump_text);
*dump_text = new_text;
}
void
dump_node_capacity(int level, const char *comment, pe_node_t * node)
{
char *dump_text = crm_strdup_printf("%s: %s capacity:",
comment, node->details->uname);
g_hash_table_foreach(node->details->utilization, append_dump_text, &dump_text);
if (level == LOG_STDOUT) {
fprintf(stdout, "%s\n", dump_text);
} else {
crm_trace("%s", dump_text);
}
free(dump_text);
}
void
dump_rsc_utilization(int level, const char *comment, pe_resource_t * rsc, pe_node_t * node)
{
char *dump_text = crm_strdup_printf("%s: %s utilization on %s:",
comment, rsc->id, node->details->uname);
g_hash_table_foreach(rsc->utilization, append_dump_text, &dump_text);
switch (level) {
case LOG_STDOUT:
fprintf(stdout, "%s\n", dump_text);
break;
case LOG_NEVER:
break;
default:
crm_trace("%s", dump_text);
}
free(dump_text);
}
gint
sort_rsc_index(gconstpointer a, gconstpointer b)
{
const pe_resource_t *resource1 = (const pe_resource_t *)a;
const pe_resource_t *resource2 = (const pe_resource_t *)b;
if (a == NULL && b == NULL) {
return 0;
}
if (a == NULL) {
return 1;
}
if (b == NULL) {
return -1;
}
if (resource1->sort_index > resource2->sort_index) {
return -1;
}
if (resource1->sort_index < resource2->sort_index) {
return 1;
}
return 0;
}
gint
sort_rsc_priority(gconstpointer a, gconstpointer b)
{
const pe_resource_t *resource1 = (const pe_resource_t *)a;
const pe_resource_t *resource2 = (const pe_resource_t *)b;
if (a == NULL && b == NULL) {
return 0;
}
if (a == NULL) {
return 1;
}
if (b == NULL) {
return -1;
}
if (resource1->priority > resource2->priority) {
return -1;
}
if (resource1->priority < resource2->priority) {
return 1;
}
return 0;
}
pe_action_t *
custom_action(pe_resource_t * rsc, char *key, const char *task,
pe_node_t * on_node, gboolean optional, gboolean save_action,
pe_working_set_t * data_set)
{
pe_action_t *action = NULL;
GListPtr possible_matches = NULL;
CRM_CHECK(key != NULL, return NULL);
CRM_CHECK(task != NULL, free(key); return NULL);
if (save_action && rsc != NULL) {
possible_matches = find_actions(rsc->actions, key, on_node);
} else if(save_action) {
#if 0
action = g_hash_table_lookup(data_set->singletons, key);
#else
/* More expensive but takes 'node' into account */
possible_matches = find_actions(data_set->actions, key, on_node);
#endif
}
if(data_set->singletons == NULL) {
data_set->singletons = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, NULL);
}
if (possible_matches != NULL) {
if (pcmk__list_of_multiple(possible_matches)) {
pe_warn("Action %s for %s on %s exists %d times",
task, rsc ? rsc->id : "<NULL>",
on_node ? on_node->details->uname : "<NULL>", g_list_length(possible_matches));
}
action = g_list_nth_data(possible_matches, 0);
pe_rsc_trace(rsc, "Found existing action %d (%s) for %s (%s) on %s",
action->id, action->uuid,
(rsc? rsc->id : "no resource"), task,
(on_node? on_node->details->uname : "no node"));
g_list_free(possible_matches);
}
if (action == NULL) {
if (save_action) {
pe_rsc_trace(rsc, "Creating %s action %d: %s for %s (%s) on %s",
(optional? "optional" : "mandatory"),
data_set->action_id, key,
(rsc? rsc->id : "no resource"), task,
(on_node? on_node->details->uname : "no node"));
}
action = calloc(1, sizeof(pe_action_t));
if (save_action) {
action->id = data_set->action_id++;
} else {
action->id = 0;
}
action->rsc = rsc;
CRM_ASSERT(task != NULL);
action->task = strdup(task);
if (on_node) {
action->node = node_copy(on_node);
}
action->uuid = strdup(key);
if (safe_str_eq(task, CRM_OP_LRM_DELETE)) {
// Resource history deletion for a node can be done on the DC
pe_set_action_bit(action, pe_action_dc);
}
pe_set_action_bit(action, pe_action_runnable);
if (optional) {
pe_set_action_bit(action, pe_action_optional);
} else {
pe_clear_action_bit(action, pe_action_optional);
}
action->extra = crm_str_table_new();
action->meta = crm_str_table_new();
if (save_action) {
data_set->actions = g_list_prepend(data_set->actions, action);
if(rsc == NULL) {
g_hash_table_insert(data_set->singletons, action->uuid, action);
}
}
if (rsc != NULL) {
action->op_entry = find_rsc_op_entry_helper(rsc, key, TRUE);
unpack_operation(action, action->op_entry, rsc->container, data_set);
if (save_action) {
rsc->actions = g_list_prepend(rsc->actions, action);
}
}
if (save_action) {
pe_rsc_trace(rsc, "Action %d created", action->id);
}
}
if (!optional && is_set(action->flags, pe_action_optional)) {
pe_rsc_trace(rsc, "Unset optional on action %d", action->id);
pe_clear_action_bit(action, pe_action_optional);
}
if (rsc != NULL) {
enum action_tasks a_task = text2task(action->task);
int warn_level = LOG_TRACE;
if (save_action) {
warn_level = LOG_WARNING;
}
if (is_set(action->flags, pe_action_have_node_attrs) == FALSE
&& action->node != NULL && action->op_entry != NULL) {
pe_set_action_bit(action, pe_action_have_node_attrs);
pe__unpack_dataset_nvpairs(action->op_entry, XML_TAG_ATTR_SETS,
action->node->details->attrs,
action->extra, NULL, FALSE, data_set);
}
if (is_set(action->flags, pe_action_pseudo)) {
/* leave untouched */
} else if (action->node == NULL) {
pe_rsc_trace(rsc, "Unset runnable on %s", action->uuid);
pe_clear_action_bit(action, pe_action_runnable);
} else if (is_not_set(rsc->flags, pe_rsc_managed)
&& g_hash_table_lookup(action->meta,
XML_LRM_ATTR_INTERVAL_MS) == NULL) {
crm_debug("Action %s (unmanaged)", action->uuid);
pe_rsc_trace(rsc, "Set optional on %s", action->uuid);
pe_set_action_bit(action, pe_action_optional);
/* action->runnable = FALSE; */
} else if (is_not_set(action->flags, pe_action_dc)
&& !(action->node->details->online)
&& (!pe__is_guest_node(action->node)
|| action->node->details->remote_requires_reset)) {
pe_clear_action_bit(action, pe_action_runnable);
do_crm_log(warn_level, "Action %s on %s is unrunnable (offline)",
action->uuid, action->node->details->uname);
if (is_set(action->rsc->flags, pe_rsc_managed)
&& save_action && a_task == stop_rsc
&& action->node->details->unclean == FALSE) {
pe_fence_node(data_set, action->node, "resource actions are unrunnable", FALSE);
}
} else if (is_not_set(action->flags, pe_action_dc)
&& action->node->details->pending) {
pe_clear_action_bit(action, pe_action_runnable);
do_crm_log(warn_level, "Action %s on %s is unrunnable (pending)",
action->uuid, action->node->details->uname);
} else if (action->needs == rsc_req_nothing) {
pe_rsc_trace(rsc, "Action %s does not require anything", action->uuid);
pe_action_set_reason(action, NULL, TRUE);
if (pe__is_guest_node(action->node)
&& !pe_can_fence(data_set, action->node)) {
/* An action that requires nothing usually does not require any
* fencing in order to be runnable. However, there is an
* exception: an action cannot be completed if it is on a guest
* node whose host is unclean and cannot be fenced.
*/
pe_clear_action_bit(action, pe_action_runnable);
crm_debug("%s\t%s (cancelled : host cannot be fenced)",
action->node->details->uname, action->uuid);
} else {
pe_set_action_bit(action, pe_action_runnable);
}
#if 0
/*
* No point checking this
* - if we don't have quorum we can't stonith anyway
*/
} else if (action->needs == rsc_req_stonith) {
crm_trace("Action %s requires only stonith", action->uuid);
action->runnable = TRUE;
#endif
} else if (is_set(data_set->flags, pe_flag_have_quorum) == FALSE
&& data_set->no_quorum_policy == no_quorum_stop) {
pe_action_set_flag_reason(__FUNCTION__, __LINE__, action, NULL, "no quorum", pe_action_runnable, TRUE);
crm_debug("%s\t%s (cancelled : quorum)", action->node->details->uname, action->uuid);
} else if (is_set(data_set->flags, pe_flag_have_quorum) == FALSE
&& data_set->no_quorum_policy == no_quorum_freeze) {
pe_rsc_trace(rsc, "Check resource is already active: %s %s %s %s", rsc->id, action->uuid, role2text(rsc->next_role), role2text(rsc->role));
if (rsc->fns->active(rsc, TRUE) == FALSE || rsc->next_role > rsc->role) {
pe_action_set_flag_reason(__FUNCTION__, __LINE__, action, NULL, "quorum freeze", pe_action_runnable, TRUE);
pe_rsc_debug(rsc, "%s\t%s (cancelled : quorum freeze)",
action->node->details->uname, action->uuid);
}
} else if(is_not_set(action->flags, pe_action_runnable)) {
pe_rsc_trace(rsc, "Action %s is runnable", action->uuid);
//pe_action_set_reason(action, NULL, TRUE);
pe_set_action_bit(action, pe_action_runnable);
}
if (save_action) {
switch (a_task) {
case stop_rsc:
set_bit(rsc->flags, pe_rsc_stopping);
break;
case start_rsc:
clear_bit(rsc->flags, pe_rsc_starting);
if (is_set(action->flags, pe_action_runnable)) {
set_bit(rsc->flags, pe_rsc_starting);
}
break;
default:
break;
}
}
}
free(key);
return action;
}
static const char *
unpack_operation_on_fail(pe_action_t * action)
{
const char *value = g_hash_table_lookup(action->meta, XML_OP_ATTR_ON_FAIL);
if (safe_str_eq(action->task, CRMD_ACTION_STOP) && safe_str_eq(value, "standby")) {
pcmk__config_err("Resetting '" XML_OP_ATTR_ON_FAIL "' for %s stop "
"action to default value because 'standby' is not "
"allowed for stop", action->rsc->id);
return NULL;
} else if (safe_str_eq(action->task, CRMD_ACTION_DEMOTE) && !value) {
/* demote on_fail defaults to master monitor value if present */
xmlNode *operation = NULL;
const char *name = NULL;
const char *role = NULL;
const char *on_fail = NULL;
const char *interval_spec = NULL;
const char *enabled = NULL;
CRM_CHECK(action->rsc != NULL, return NULL);
for (operation = __xml_first_child_element(action->rsc->ops_xml);
operation && !value; operation = __xml_next_element(operation)) {
if (!crm_str_eq((const char *)operation->name, "op", TRUE)) {
continue;
}
name = crm_element_value(operation, "name");
role = crm_element_value(operation, "role");
on_fail = crm_element_value(operation, XML_OP_ATTR_ON_FAIL);
enabled = crm_element_value(operation, "enabled");
interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL);
if (!on_fail) {
continue;
} else if (enabled && !crm_is_true(enabled)) {
continue;
} else if (safe_str_neq(name, "monitor") || safe_str_neq(role, "Master")) {
continue;
} else if (crm_parse_interval_spec(interval_spec) == 0) {
continue;
}
value = on_fail;
}
} else if (safe_str_eq(action->task, CRM_OP_LRM_DELETE)) {
value = "ignore";
}
return value;
}
static xmlNode *
find_min_interval_mon(pe_resource_t * rsc, gboolean include_disabled)
{
guint interval_ms = 0;
guint min_interval_ms = G_MAXUINT;
const char *name = NULL;
const char *value = NULL;
const char *interval_spec = NULL;
xmlNode *op = NULL;
xmlNode *operation = NULL;
for (operation = __xml_first_child_element(rsc->ops_xml); operation != NULL;
operation = __xml_next_element(operation)) {
if (crm_str_eq((const char *)operation->name, "op", TRUE)) {
name = crm_element_value(operation, "name");
interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL);
value = crm_element_value(operation, "enabled");
if (!include_disabled && value && crm_is_true(value) == FALSE) {
continue;
}
if (safe_str_neq(name, RSC_STATUS)) {
continue;
}
interval_ms = crm_parse_interval_spec(interval_spec);
if (interval_ms && (interval_ms < min_interval_ms)) {
min_interval_ms = interval_ms;
op = operation;
}
}
}
return op;
}
static int
unpack_start_delay(const char *value, GHashTable *meta)
{
int start_delay = 0;
if (value != NULL) {
start_delay = crm_get_msec(value);
if (start_delay < 0) {
start_delay = 0;
}
if (meta) {
g_hash_table_replace(meta, strdup(XML_OP_ATTR_START_DELAY), crm_itoa(start_delay));
}
}
return start_delay;
}
// true if value contains valid, non-NULL interval origin for recurring op
static bool
unpack_interval_origin(const char *value, xmlNode *xml_obj, guint interval_ms,
crm_time_t *now, long long *start_delay)
{
long long result = 0;
guint interval_sec = interval_ms / 1000;
crm_time_t *origin = NULL;
// Ignore unspecified values and non-recurring operations
if ((value == NULL) || (interval_ms == 0) || (now == NULL)) {
return false;
}
// Parse interval origin from text
origin = crm_time_new(value);
if (origin == NULL) {
pcmk__config_err("Ignoring '" XML_OP_ATTR_ORIGIN "' for operation "
"'%s' because '%s' is not valid",
(ID(xml_obj)? ID(xml_obj) : "(missing ID)"), value);
return false;
}
// Get seconds since origin (negative if origin is in the future)
result = crm_time_get_seconds(now) - crm_time_get_seconds(origin);
crm_time_free(origin);
// Calculate seconds from closest interval to now
result = result % interval_sec;
// Calculate seconds remaining until next interval
result = ((result <= 0)? 0 : interval_sec) - result;
crm_info("Calculated a start delay of %llds for operation '%s'",
result,
(ID(xml_obj)? ID(xml_obj) : "(unspecified)"));
if (start_delay != NULL) {
*start_delay = result * 1000; // milliseconds
}
return true;
}
static int
unpack_timeout(const char *value)
{
int timeout = crm_get_msec(value);
if (timeout < 0) {
timeout = crm_get_msec(CRM_DEFAULT_OP_TIMEOUT_S);
}
return timeout;
}
int
pe_get_configured_timeout(pe_resource_t *rsc, const char *action, pe_working_set_t *data_set)
{
xmlNode *child = NULL;
const char *timeout = NULL;
int timeout_ms = 0;
for (child = first_named_child(rsc->ops_xml, XML_ATTR_OP);
child != NULL; child = crm_next_same_xml(child)) {
if (safe_str_eq(action, crm_element_value(child, XML_NVPAIR_ATTR_NAME))) {
timeout = crm_element_value(child, XML_ATTR_TIMEOUT);
break;
}
}
if (timeout == NULL && data_set->op_defaults) {
GHashTable *action_meta = crm_str_table_new();
pe__unpack_dataset_nvpairs(data_set->op_defaults, XML_TAG_META_SETS,
NULL, action_meta, NULL, FALSE, data_set);
timeout = g_hash_table_lookup(action_meta, XML_ATTR_TIMEOUT);
}
// @TODO check meta-attributes (including versioned meta-attributes)
// @TODO maybe use min-interval monitor timeout as default for monitors
timeout_ms = crm_get_msec(timeout);
if (timeout_ms < 0) {
timeout_ms = crm_get_msec(CRM_DEFAULT_OP_TIMEOUT_S);
}
return timeout_ms;
}
#if ENABLE_VERSIONED_ATTRS
static void
unpack_versioned_meta(xmlNode *versioned_meta, xmlNode *xml_obj,
guint interval_ms, crm_time_t *now)
{
xmlNode *attrs = NULL;
xmlNode *attr = NULL;
for (attrs = __xml_first_child_element(versioned_meta); attrs != NULL;
attrs = __xml_next_element(attrs)) {
for (attr = __xml_first_child_element(attrs); attr != NULL;
attr = __xml_next_element(attr)) {
const char *name = crm_element_value(attr, XML_NVPAIR_ATTR_NAME);
const char *value = crm_element_value(attr, XML_NVPAIR_ATTR_VALUE);
if (safe_str_eq(name, XML_OP_ATTR_START_DELAY)) {
int start_delay = unpack_start_delay(value, NULL);
crm_xml_add_int(attr, XML_NVPAIR_ATTR_VALUE, start_delay);
} else if (safe_str_eq(name, XML_OP_ATTR_ORIGIN)) {
long long start_delay = 0;
if (unpack_interval_origin(value, xml_obj, interval_ms, now,
&start_delay)) {
crm_xml_add(attr, XML_NVPAIR_ATTR_NAME,
XML_OP_ATTR_START_DELAY);
crm_xml_add_ll(attr, XML_NVPAIR_ATTR_VALUE, start_delay);
}
} else if (safe_str_eq(name, XML_ATTR_TIMEOUT)) {
int timeout = unpack_timeout(value);
crm_xml_add_int(attr, XML_NVPAIR_ATTR_VALUE, timeout);
}
}
}
}
#endif
/*!
* \brief Unpack operation XML into an action structure
*
* Unpack an operation's meta-attributes (normalizing the interval, timeout,
* and start delay values as integer milliseconds), requirements, and
* failure policy.
*
* \param[in,out] action Action to unpack into
* \param[in] xml_obj Operation XML (or NULL if all defaults)
* \param[in] container Resource that contains affected resource, if any
* \param[in] data_set Cluster state
*/
void
unpack_operation(pe_action_t * action, xmlNode * xml_obj, pe_resource_t * container,
pe_working_set_t * data_set)
{
guint interval_ms = 0;
int timeout = 0;
char *value_ms = NULL;
const char *value = NULL;
const char *field = NULL;
char *default_timeout = NULL;
#if ENABLE_VERSIONED_ATTRS
pe_rsc_action_details_t *rsc_details = NULL;
#endif
CRM_CHECK(action && action->rsc, return);
// Cluster-wide <op_defaults> <meta_attributes>
pe__unpack_dataset_nvpairs(data_set->op_defaults, XML_TAG_META_SETS, NULL,
action->meta, NULL, FALSE, data_set);
// Probe timeouts default differently, so handle timeout default later
default_timeout = g_hash_table_lookup(action->meta, XML_ATTR_TIMEOUT);
if (default_timeout) {
default_timeout = strdup(default_timeout);
g_hash_table_remove(action->meta, XML_ATTR_TIMEOUT);
}
if (xml_obj) {
xmlAttrPtr xIter = NULL;
// <op> <meta_attributes> take precedence over defaults
pe__unpack_dataset_nvpairs(xml_obj, XML_TAG_META_SETS, NULL,
action->meta, NULL, TRUE, data_set);
#if ENABLE_VERSIONED_ATTRS
rsc_details = pe_rsc_action_details(action);
pe_unpack_versioned_attributes(data_set->input, xml_obj,
XML_TAG_ATTR_SETS, NULL,
rsc_details->versioned_parameters,
data_set->now, NULL);
pe_unpack_versioned_attributes(data_set->input, xml_obj,
XML_TAG_META_SETS, NULL,
rsc_details->versioned_meta,
data_set->now, NULL);
#endif
/* Anything set as an <op> XML property has highest precedence.
* This ensures we use the name and interval from the <op> tag.
*/
for (xIter = xml_obj->properties; xIter; xIter = xIter->next) {
const char *prop_name = (const char *)xIter->name;
const char *prop_value = crm_element_value(xml_obj, prop_name);
g_hash_table_replace(action->meta, strdup(prop_name), strdup(prop_value));
}
}
g_hash_table_remove(action->meta, "id");
// Normalize interval to milliseconds
field = XML_LRM_ATTR_INTERVAL;
value = g_hash_table_lookup(action->meta, field);
if (value != NULL) {
interval_ms = crm_parse_interval_spec(value);
} else if ((xml_obj == NULL) && !strcmp(action->task, RSC_STATUS)) {
/* An orphaned recurring monitor will not have any XML. However, we
* want the interval to be set, so the action can be properly detected
* as a recurring monitor. Parse it from the key in this case.
*/
parse_op_key(action->uuid, NULL, NULL, &interval_ms);
}
if (interval_ms > 0) {
value_ms = crm_strdup_printf("%u", interval_ms);
g_hash_table_replace(action->meta, strdup(field), value_ms);
} else if (value) {
g_hash_table_remove(action->meta, field);
}
// Handle timeout default, now that we know the interval
if (g_hash_table_lookup(action->meta, XML_ATTR_TIMEOUT)) {
free(default_timeout);
} else {
// Probe timeouts default to minimum-interval monitor's
if (safe_str_eq(action->task, RSC_STATUS) && (interval_ms == 0)) {
xmlNode *min_interval_mon = find_min_interval_mon(action->rsc, FALSE);
if (min_interval_mon) {
value = crm_element_value(min_interval_mon, XML_ATTR_TIMEOUT);
if (value) {
crm_trace("\t%s defaults to minimum-interval monitor's timeout '%s'",
action->uuid, value);
free(default_timeout);
default_timeout = strdup(value);
}
}
}
if (default_timeout) {
g_hash_table_insert(action->meta, strdup(XML_ATTR_TIMEOUT),
default_timeout);
}
}
if (safe_str_neq(action->task, RSC_START)
&& safe_str_neq(action->task, RSC_PROMOTE)) {
action->needs = rsc_req_nothing;
value = "nothing (not start/promote)";
} else if (is_set(action->rsc->flags, pe_rsc_needs_fencing)) {
action->needs = rsc_req_stonith;
value = "fencing (resource)";
} else if (is_set(action->rsc->flags, pe_rsc_needs_quorum)) {
action->needs = rsc_req_quorum;
value = "quorum (resource)";
} else {
action->needs = rsc_req_nothing;
value = "nothing (resource)";
}
pe_rsc_trace(action->rsc, "\tAction %s requires: %s", action->uuid, value);
value = unpack_operation_on_fail(action);
if (value == NULL) {
} else if (safe_str_eq(value, "block")) {
action->on_fail = action_fail_block;
g_hash_table_insert(action->meta, strdup(XML_OP_ATTR_ON_FAIL), strdup("block"));
value = "block"; // The above could destroy the original string
} else if (safe_str_eq(value, "fence")) {
action->on_fail = action_fail_fence;
value = "node fencing";
if (is_set(data_set->flags, pe_flag_stonith_enabled) == FALSE) {
pcmk__config_err("Resetting '" XML_OP_ATTR_ON_FAIL "' for "
"operation '%s' to 'stop' because 'fence' is not "
"valid when fencing is disabled", action->uuid);
action->on_fail = action_fail_stop;
action->fail_role = RSC_ROLE_STOPPED;
value = "stop resource";
}
} else if (safe_str_eq(value, "standby")) {
action->on_fail = action_fail_standby;
value = "node standby";
} else if (safe_str_eq(value, "ignore")
|| safe_str_eq(value, "nothing")) {
action->on_fail = action_fail_ignore;
value = "ignore";
} else if (safe_str_eq(value, "migrate")) {
action->on_fail = action_fail_migrate;
value = "force migration";
} else if (safe_str_eq(value, "stop")) {
action->on_fail = action_fail_stop;
action->fail_role = RSC_ROLE_STOPPED;
value = "stop resource";
} else if (safe_str_eq(value, "restart")) {
action->on_fail = action_fail_recover;
value = "restart (and possibly migrate)";
} else if (safe_str_eq(value, "restart-container")) {
if (container) {
action->on_fail = action_fail_restart_container;
value = "restart container (and possibly migrate)";
} else {
value = NULL;
}
} else {
pe_err("Resource %s: Unknown failure type (%s)", action->rsc->id, value);
value = NULL;
}
/* defaults */
if (value == NULL && container) {
action->on_fail = action_fail_restart_container;
value = "restart container (and possibly migrate) (default)";
/* For remote nodes, ensure that any failure that results in dropping an
* active connection to the node results in fencing of the node.
*
* There are only two action failures that don't result in fencing.
* 1. probes - probe failures are expected.
* 2. start - a start failure indicates that an active connection does not already
* exist. The user can set op on-fail=fence if they really want to fence start
* failures. */
} else if (((value == NULL) || !is_set(action->rsc->flags, pe_rsc_managed)) &&
(pe__resource_is_remote_conn(action->rsc, data_set) &&
!(safe_str_eq(action->task, CRMD_ACTION_STATUS) && (interval_ms == 0)) &&
(safe_str_neq(action->task, CRMD_ACTION_START)))) {
if (!is_set(action->rsc->flags, pe_rsc_managed)) {
action->on_fail = action_fail_stop;
action->fail_role = RSC_ROLE_STOPPED;
value = "stop unmanaged remote node (enforcing default)";
} else {
if (is_set(data_set->flags, pe_flag_stonith_enabled)) {
value = "fence remote node (default)";
} else {
value = "recover remote node connection (default)";
}
if (action->rsc->remote_reconnect_ms) {
action->fail_role = RSC_ROLE_STOPPED;
}
action->on_fail = action_fail_reset_remote;
}
} else if (value == NULL && safe_str_eq(action->task, CRMD_ACTION_STOP)) {
if (is_set(data_set->flags, pe_flag_stonith_enabled)) {
action->on_fail = action_fail_fence;
value = "resource fence (default)";
} else {
action->on_fail = action_fail_block;
value = "resource block (default)";
}
} else if (value == NULL) {
action->on_fail = action_fail_recover;
value = "restart (and possibly migrate) (default)";
}
pe_rsc_trace(action->rsc, "\t%s failure handling: %s", action->task, value);
value = NULL;
if (xml_obj != NULL) {
value = g_hash_table_lookup(action->meta, "role_after_failure");
if (value) {
pe_warn_once(pe_wo_role_after,
"Support for role_after_failure is deprecated and will be removed in a future release");
}
}
if (value != NULL && action->fail_role == RSC_ROLE_UNKNOWN) {
action->fail_role = text2role(value);
}
/* defaults */
if (action->fail_role == RSC_ROLE_UNKNOWN) {
if (safe_str_eq(action->task, CRMD_ACTION_PROMOTE)) {
action->fail_role = RSC_ROLE_SLAVE;
} else {
action->fail_role = RSC_ROLE_STARTED;
}
}
pe_rsc_trace(action->rsc, "\t%s failure results in: %s", action->task,
role2text(action->fail_role));
value = g_hash_table_lookup(action->meta, XML_OP_ATTR_START_DELAY);
if (value) {
unpack_start_delay(value, action->meta);
} else {
long long start_delay = 0;
value = g_hash_table_lookup(action->meta, XML_OP_ATTR_ORIGIN);
if (unpack_interval_origin(value, xml_obj, interval_ms, data_set->now,
&start_delay)) {
g_hash_table_replace(action->meta, strdup(XML_OP_ATTR_START_DELAY),
crm_strdup_printf("%lld", start_delay));
}
}
value = g_hash_table_lookup(action->meta, XML_ATTR_TIMEOUT);
timeout = unpack_timeout(value);
g_hash_table_replace(action->meta, strdup(XML_ATTR_TIMEOUT), crm_itoa(timeout));
#if ENABLE_VERSIONED_ATTRS
unpack_versioned_meta(rsc_details->versioned_meta, xml_obj, interval_ms,
data_set->now);
#endif
}
static xmlNode *
find_rsc_op_entry_helper(pe_resource_t * rsc, const char *key, gboolean include_disabled)
{
guint interval_ms = 0;
gboolean do_retry = TRUE;
char *local_key = NULL;
const char *name = NULL;
const char *value = NULL;
const char *interval_spec = NULL;
char *match_key = NULL;
xmlNode *op = NULL;
xmlNode *operation = NULL;
retry:
for (operation = __xml_first_child_element(rsc->ops_xml); operation != NULL;
operation = __xml_next_element(operation)) {
if (crm_str_eq((const char *)operation->name, "op", TRUE)) {
name = crm_element_value(operation, "name");
interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL);
value = crm_element_value(operation, "enabled");
if (!include_disabled && value && crm_is_true(value) == FALSE) {
continue;
}
interval_ms = crm_parse_interval_spec(interval_spec);
match_key = pcmk__op_key(rsc->id, name, interval_ms);
if (safe_str_eq(key, match_key)) {
op = operation;
}
free(match_key);
if (rsc->clone_name) {
match_key = pcmk__op_key(rsc->clone_name, name, interval_ms);
if (safe_str_eq(key, match_key)) {
op = operation;
}
free(match_key);
}
if (op != NULL) {
free(local_key);
return op;
}
}
}
free(local_key);
if (do_retry == FALSE) {
return NULL;
}
do_retry = FALSE;
if (strstr(key, CRMD_ACTION_MIGRATE) || strstr(key, CRMD_ACTION_MIGRATED)) {
local_key = pcmk__op_key(rsc->id, "migrate", 0);
key = local_key;
goto retry;
} else if (strstr(key, "_notify_")) {
local_key = pcmk__op_key(rsc->id, "notify", 0);
key = local_key;
goto retry;
}
return NULL;
}
xmlNode *
find_rsc_op_entry(pe_resource_t * rsc, const char *key)
{
return find_rsc_op_entry_helper(rsc, key, FALSE);
}
void
print_node(const char *pre_text, pe_node_t * node, gboolean details)
{
if (node == NULL) {
crm_trace("%s%s: <NULL>", pre_text == NULL ? "" : pre_text, pre_text == NULL ? "" : ": ");
return;
}
CRM_ASSERT(node->details);
crm_trace("%s%s%sNode %s: (weight=%d, fixed=%s)",
pre_text == NULL ? "" : pre_text,
pre_text == NULL ? "" : ": ",
node->details->online ? "" : "Unavailable/Unclean ",
node->details->uname, node->weight, node->fixed ? "True" : "False");
if (details) {
int log_level = LOG_TRACE;
char *pe_mutable = strdup("\t\t");
GListPtr gIter = node->details->running_rsc;
crm_trace("\t\t===Node Attributes");
g_hash_table_foreach(node->details->attrs, print_str_str, pe_mutable);
free(pe_mutable);
crm_trace("\t\t=== Resources");
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
rsc->fns->print(rsc, "\t\t", pe_print_log|pe_print_pending,
&log_level);
}
}
}
/*
* Used by the HashTable for-loop
*/
void
print_str_str(gpointer key, gpointer value, gpointer user_data)
{
crm_trace("%s%s %s ==> %s",
user_data == NULL ? "" : (char *)user_data,
user_data == NULL ? "" : ": ", (char *)key, (char *)value);
}
void
pe_free_action(pe_action_t * action)
{
if (action == NULL) {
return;
}
g_list_free_full(action->actions_before, free); /* pe_action_wrapper_t* */
g_list_free_full(action->actions_after, free); /* pe_action_wrapper_t* */
if (action->extra) {
g_hash_table_destroy(action->extra);
}
if (action->meta) {
g_hash_table_destroy(action->meta);
}
#if ENABLE_VERSIONED_ATTRS
if (action->rsc) {
pe_free_rsc_action_details(action);
}
#endif
free(action->cancel_task);
free(action->reason);
free(action->task);
free(action->uuid);
free(action->node);
free(action);
}
GListPtr
find_recurring_actions(GListPtr input, pe_node_t * not_on_node)
{
const char *value = NULL;
GListPtr result = NULL;
GListPtr gIter = input;
CRM_CHECK(input != NULL, return NULL);
for (; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
value = g_hash_table_lookup(action->meta, XML_LRM_ATTR_INTERVAL_MS);
if (value == NULL) {
/* skip */
} else if (safe_str_eq(value, "0")) {
/* skip */
} else if (safe_str_eq(CRMD_ACTION_CANCEL, action->task)) {
/* skip */
} else if (not_on_node == NULL) {
crm_trace("(null) Found: %s", action->uuid);
result = g_list_prepend(result, action);
} else if (action->node == NULL) {
/* skip */
} else if (action->node->details != not_on_node->details) {
crm_trace("Found: %s", action->uuid);
result = g_list_prepend(result, action);
}
}
return result;
}
enum action_tasks
get_complex_task(pe_resource_t * rsc, const char *name, gboolean allow_non_atomic)
{
enum action_tasks task = text2task(name);
if (rsc == NULL) {
return task;
} else if (allow_non_atomic == FALSE || rsc->variant == pe_native) {
switch (task) {
case stopped_rsc:
case started_rsc:
case action_demoted:
case action_promoted:
crm_trace("Folding %s back into its atomic counterpart for %s", name, rsc->id);
return task - 1;
break;
default:
break;
}
}
return task;
}
pe_action_t *
find_first_action(GListPtr input, const char *uuid, const char *task, pe_node_t * on_node)
{
GListPtr gIter = NULL;
CRM_CHECK(uuid || task, return NULL);
for (gIter = input; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
if (uuid != NULL && safe_str_neq(uuid, action->uuid)) {
continue;
} else if (task != NULL && safe_str_neq(task, action->task)) {
continue;
} else if (on_node == NULL) {
return action;
} else if (action->node == NULL) {
continue;
} else if (on_node->details == action->node->details) {
return action;
}
}
return NULL;
}
GListPtr
find_actions(GListPtr input, const char *key, const pe_node_t *on_node)
{
GListPtr gIter = input;
GListPtr result = NULL;
CRM_CHECK(key != NULL, return NULL);
for (; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
if (safe_str_neq(key, action->uuid)) {
crm_trace("%s does not match action %s", key, action->uuid);
continue;
} else if (on_node == NULL) {
crm_trace("Action %s matches (ignoring node)", key);
result = g_list_prepend(result, action);
} else if (action->node == NULL) {
crm_trace("Action %s matches (unallocated, assigning to %s)",
key, on_node->details->uname);
action->node = node_copy(on_node);
result = g_list_prepend(result, action);
} else if (on_node->details == action->node->details) {
crm_trace("Action %s on %s matches", key, on_node->details->uname);
result = g_list_prepend(result, action);
} else {
crm_trace("Action %s on node %s does not match requested node %s",
key, action->node->details->uname,
on_node->details->uname);
}
}
return result;
}
GList *
find_actions_exact(GList *input, const char *key, const pe_node_t *on_node)
{
GList *result = NULL;
CRM_CHECK(key != NULL, return NULL);
if (on_node == NULL) {
crm_trace("Not searching for action %s because node not specified",
key);
return NULL;
}
for (GList *gIter = input; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
if (action->node == NULL) {
crm_trace("Skipping comparison of %s vs action %s without node",
key, action->uuid);
} else if (safe_str_neq(key, action->uuid)) {
crm_trace("Desired action %s doesn't match %s", key, action->uuid);
} else if (safe_str_neq(on_node->details->id,
action->node->details->id)) {
crm_trace("Action %s desired node ID %s doesn't match %s",
key, on_node->details->id, action->node->details->id);
} else {
crm_trace("Action %s matches", key);
result = g_list_prepend(result, action);
}
}
return result;
}
/*!
* \brief Find all actions of given type for a resource
*
* \param[in] rsc Resource to search
* \param[in] node Find only actions scheduled on this node
* \param[in] task Action name to search for
* \param[in] require_node If TRUE, NULL node or action node will not match
*
* \return List of actions found (or NULL if none)
* \note If node is not NULL and require_node is FALSE, matching actions
* without a node will be assigned to node.
*/
GList *
pe__resource_actions(const pe_resource_t *rsc, const pe_node_t *node,
const char *task, bool require_node)
{
GList *result = NULL;
char *key = pcmk__op_key(rsc->id, task, 0);
if (require_node) {
result = find_actions_exact(rsc->actions, key, node);
} else {
result = find_actions(rsc->actions, key, node);
}
free(key);
return result;
}
static void
resource_node_score(pe_resource_t * rsc, pe_node_t * node, int score, const char *tag)
{
pe_node_t *match = NULL;
if ((rsc->exclusive_discover || (node->rsc_discover_mode == pe_discover_never))
&& safe_str_eq(tag, "symmetric_default")) {
/* This string comparision may be fragile, but exclusive resources and
* exclusive nodes should not have the symmetric_default constraint
* applied to them.
*/
return;
} else if (rsc->children) {
GListPtr gIter = rsc->children;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
resource_node_score(child_rsc, node, score, tag);
}
}
pe_rsc_trace(rsc, "Setting %s for %s on %s: %d", tag, rsc->id, node->details->uname, score);
match = pe_hash_table_lookup(rsc->allowed_nodes, node->details->id);
if (match == NULL) {
match = node_copy(node);
g_hash_table_insert(rsc->allowed_nodes, (gpointer) match->details->id, match);
}
match->weight = merge_weights(match->weight, score);
}
void
resource_location(pe_resource_t * rsc, pe_node_t * node, int score, const char *tag,
pe_working_set_t * data_set)
{
if (node != NULL) {
resource_node_score(rsc, node, score, tag);
} else if (data_set != NULL) {
GListPtr gIter = data_set->nodes;
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *node_iter = (pe_node_t *) gIter->data;
resource_node_score(rsc, node_iter, score, tag);
}
} else {
GHashTableIter iter;
pe_node_t *node_iter = NULL;
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node_iter)) {
resource_node_score(rsc, node_iter, score, tag);
}
}
if (node == NULL && score == -INFINITY) {
if (rsc->allocated_to) {
crm_info("Deallocating %s from %s", rsc->id, rsc->allocated_to->details->uname);
free(rsc->allocated_to);
rsc->allocated_to = NULL;
}
}
}
#define sort_return(an_int, why) do { \
free(a_uuid); \
free(b_uuid); \
crm_trace("%s (%d) %c %s (%d) : %s", \
a_xml_id, a_call_id, an_int>0?'>':an_int<0?'<':'=', \
b_xml_id, b_call_id, why); \
return an_int; \
} while(0)
gint
sort_op_by_callid(gconstpointer a, gconstpointer b)
{
int a_call_id = -1;
int b_call_id = -1;
char *a_uuid = NULL;
char *b_uuid = NULL;
const xmlNode *xml_a = a;
const xmlNode *xml_b = b;
const char *a_xml_id = crm_element_value(xml_a, XML_ATTR_ID);
const char *b_xml_id = crm_element_value(xml_b, XML_ATTR_ID);
if (safe_str_eq(a_xml_id, b_xml_id)) {
/* We have duplicate lrm_rsc_op entries in the status
* section which is unlikely to be a good thing
* - we can handle it easily enough, but we need to get
* to the bottom of why it's happening.
*/
pe_err("Duplicate lrm_rsc_op entries named %s", a_xml_id);
sort_return(0, "duplicate");
}
crm_element_value_int(xml_a, XML_LRM_ATTR_CALLID, &a_call_id);
crm_element_value_int(xml_b, XML_LRM_ATTR_CALLID, &b_call_id);
if (a_call_id == -1 && b_call_id == -1) {
/* both are pending ops so it doesn't matter since
* stops are never pending
*/
sort_return(0, "pending");
} else if (a_call_id >= 0 && a_call_id < b_call_id) {
sort_return(-1, "call id");
} else if (b_call_id >= 0 && a_call_id > b_call_id) {
sort_return(1, "call id");
} else if (b_call_id >= 0 && a_call_id == b_call_id) {
/*
* The op and last_failed_op are the same
* Order on last-rc-change
*/
time_t last_a = -1;
time_t last_b = -1;
crm_element_value_epoch(xml_a, XML_RSC_OP_LAST_CHANGE, &last_a);
crm_element_value_epoch(xml_b, XML_RSC_OP_LAST_CHANGE, &last_b);
crm_trace("rc-change: %lld vs %lld",
(long long) last_a, (long long) last_b);
if (last_a >= 0 && last_a < last_b) {
sort_return(-1, "rc-change");
} else if (last_b >= 0 && last_a > last_b) {
sort_return(1, "rc-change");
}
sort_return(0, "rc-change");
} else {
/* One of the inputs is a pending operation
* Attempt to use XML_ATTR_TRANSITION_MAGIC to determine its age relative to the other
*/
int a_id = -1;
int b_id = -1;
const char *a_magic = crm_element_value(xml_a, XML_ATTR_TRANSITION_MAGIC);
const char *b_magic = crm_element_value(xml_b, XML_ATTR_TRANSITION_MAGIC);
CRM_CHECK(a_magic != NULL && b_magic != NULL, sort_return(0, "No magic"));
if (!decode_transition_magic(a_magic, &a_uuid, &a_id, NULL, NULL, NULL,
NULL)) {
sort_return(0, "bad magic a");
}
if (!decode_transition_magic(b_magic, &b_uuid, &b_id, NULL, NULL, NULL,
NULL)) {
sort_return(0, "bad magic b");
}
/* try to determine the relative age of the operation...
* some pending operations (e.g. a start) may have been superseded
* by a subsequent stop
*
* [a|b]_id == -1 means it's a shutdown operation and _always_ comes last
*/
if (safe_str_neq(a_uuid, b_uuid) || a_id == b_id) {
/*
* some of the logic in here may be redundant...
*
* if the UUID from the TE doesn't match then one better
* be a pending operation.
* pending operations don't survive between elections and joins
* because we query the LRM directly
*/
if (b_call_id == -1) {
sort_return(-1, "transition + call");
} else if (a_call_id == -1) {
sort_return(1, "transition + call");
}
} else if ((a_id >= 0 && a_id < b_id) || b_id == -1) {
sort_return(-1, "transition");
} else if ((b_id >= 0 && a_id > b_id) || a_id == -1) {
sort_return(1, "transition");
}
}
/* we should never end up here */
CRM_CHECK(FALSE, sort_return(0, "default"));
}
time_t
get_effective_time(pe_working_set_t * data_set)
{
if(data_set) {
if (data_set->now == NULL) {
crm_trace("Recording a new 'now'");
data_set->now = crm_time_new(NULL);
}
return crm_time_get_seconds_since_epoch(data_set->now);
}
crm_trace("Defaulting to 'now'");
return time(NULL);
}
gboolean
get_target_role(pe_resource_t * rsc, enum rsc_role_e * role)
{
enum rsc_role_e local_role = RSC_ROLE_UNKNOWN;
const char *value = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE);
CRM_CHECK(role != NULL, return FALSE);
if (value == NULL || safe_str_eq("started", value)
|| safe_str_eq("default", value)) {
return FALSE;
}
local_role = text2role(value);
if (local_role == RSC_ROLE_UNKNOWN) {
pcmk__config_err("Ignoring '" XML_RSC_ATTR_TARGET_ROLE "' for %s "
"because '%s' is not valid", rsc->id, value);
return FALSE;
} else if (local_role > RSC_ROLE_STARTED) {
if (is_set(uber_parent(rsc)->flags, pe_rsc_promotable)) {
if (local_role > RSC_ROLE_SLAVE) {
/* This is what we'd do anyway, just leave the default to avoid messing up the placement algorithm */
return FALSE;
}
} else {
pcmk__config_err("Ignoring '" XML_RSC_ATTR_TARGET_ROLE "' for %s "
"because '%s' only makes sense for promotable "
"clones", rsc->id, value);
return FALSE;
}
}
*role = local_role;
return TRUE;
}
gboolean
order_actions(pe_action_t * lh_action, pe_action_t * rh_action, enum pe_ordering order)
{
GListPtr gIter = NULL;
pe_action_wrapper_t *wrapper = NULL;
GListPtr list = NULL;
if (order == pe_order_none) {
return FALSE;
}
if (lh_action == NULL || rh_action == NULL) {
return FALSE;
}
crm_trace("Ordering Action %s before %s", lh_action->uuid, rh_action->uuid);
/* Ensure we never create a dependency on ourselves... it's happened */
CRM_ASSERT(lh_action != rh_action);
/* Filter dups, otherwise update_action_states() has too much work to do */
gIter = lh_action->actions_after;
for (; gIter != NULL; gIter = gIter->next) {
pe_action_wrapper_t *after = (pe_action_wrapper_t *) gIter->data;
if (after->action == rh_action && (after->type & order)) {
return FALSE;
}
}
wrapper = calloc(1, sizeof(pe_action_wrapper_t));
wrapper->action = rh_action;
wrapper->type = order;
list = lh_action->actions_after;
list = g_list_prepend(list, wrapper);
lh_action->actions_after = list;
wrapper = NULL;
/* order |= pe_order_implies_then; */
/* order ^= pe_order_implies_then; */
wrapper = calloc(1, sizeof(pe_action_wrapper_t));
wrapper->action = lh_action;
wrapper->type = order;
list = rh_action->actions_before;
list = g_list_prepend(list, wrapper);
rh_action->actions_before = list;
return TRUE;
}
pe_action_t *
get_pseudo_op(const char *name, pe_working_set_t * data_set)
{
pe_action_t *op = NULL;
if(data_set->singletons) {
op = g_hash_table_lookup(data_set->singletons, name);
}
if (op == NULL) {
op = custom_action(NULL, strdup(name), name, NULL, TRUE, TRUE, data_set);
set_bit(op->flags, pe_action_pseudo);
set_bit(op->flags, pe_action_runnable);
}
return op;
}
void
destroy_ticket(gpointer data)
{
pe_ticket_t *ticket = data;
if (ticket->state) {
g_hash_table_destroy(ticket->state);
}
free(ticket->id);
free(ticket);
}
pe_ticket_t *
ticket_new(const char *ticket_id, pe_working_set_t * data_set)
{
pe_ticket_t *ticket = NULL;
if (ticket_id == NULL || strlen(ticket_id) == 0) {
return NULL;
}
if (data_set->tickets == NULL) {
data_set->tickets =
g_hash_table_new_full(crm_str_hash, g_str_equal, free,
destroy_ticket);
}
ticket = g_hash_table_lookup(data_set->tickets, ticket_id);
if (ticket == NULL) {
ticket = calloc(1, sizeof(pe_ticket_t));
if (ticket == NULL) {
crm_err("Cannot allocate ticket '%s'", ticket_id);
return NULL;
}
crm_trace("Creaing ticket entry for %s", ticket_id);
ticket->id = strdup(ticket_id);
ticket->granted = FALSE;
ticket->last_granted = -1;
ticket->standby = FALSE;
ticket->state = crm_str_table_new();
g_hash_table_insert(data_set->tickets, strdup(ticket->id), ticket);
}
return ticket;
}
static void
filter_parameters(xmlNode * param_set, const char *param_string, bool need_present)
{
if (param_set && param_string) {
xmlAttrPtr xIter = param_set->properties;
while (xIter) {
const char *prop_name = (const char *)xIter->name;
char *name = crm_strdup_printf(" %s ", prop_name);
char *match = strstr(param_string, name);
free(name);
// Do now, because current entry might get removed below
xIter = xIter->next;
if (need_present && match == NULL) {
crm_trace("%s not found in %s", prop_name, param_string);
xml_remove_prop(param_set, prop_name);
} else if (need_present == FALSE && match) {
crm_trace("%s found in %s", prop_name, param_string);
xml_remove_prop(param_set, prop_name);
}
}
}
}
#if ENABLE_VERSIONED_ATTRS
static void
append_versioned_params(xmlNode *versioned_params, const char *ra_version, xmlNode *params)
{
GHashTable *hash = pe_unpack_versioned_parameters(versioned_params, ra_version);
char *key = NULL;
char *value = NULL;
GHashTableIter iter;
g_hash_table_iter_init(&iter, hash);
while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) {
crm_xml_add(params, key, value);
}
g_hash_table_destroy(hash);
}
#endif
/*!
* \internal
* \brief Calculate action digests and store in node's digest cache
*
* \param[in] rsc Resource that action was for
* \param[in] task Name of action performed
* \param[in] key Action's task key
* \param[in] node Node action was performed on
* \param[in] xml_op XML of operation in CIB status (if available)
* \param[in] calc_secure Whether to calculate secure digest
* \param[in] data_set Cluster working set
*
* \return Pointer to node's digest cache entry
*/
static op_digest_cache_t *
rsc_action_digest(pe_resource_t *rsc, const char *task, const char *key,
pe_node_t *node, xmlNode *xml_op, bool calc_secure,
pe_working_set_t *data_set)
{
op_digest_cache_t *data = NULL;
data = g_hash_table_lookup(node->details->digest_cache, key);
if (data == NULL) {
GHashTable *local_rsc_params = crm_str_table_new();
pe_action_t *action = custom_action(rsc, strdup(key), task, node, TRUE, FALSE, data_set);
#if ENABLE_VERSIONED_ATTRS
xmlNode *local_versioned_params = create_xml_node(NULL, XML_TAG_RSC_VER_ATTRS);
const char *ra_version = NULL;
#endif
const char *op_version;
const char *restart_list = NULL;
const char *secure_list = " passwd password ";
data = calloc(1, sizeof(op_digest_cache_t));
CRM_ASSERT(data != NULL);
get_rsc_attributes(local_rsc_params, rsc, node, data_set);
#if ENABLE_VERSIONED_ATTRS
pe_get_versioned_attributes(local_versioned_params, rsc, node, data_set);
#endif
data->params_all = create_xml_node(NULL, XML_TAG_PARAMS);
// REMOTE_CONTAINER_HACK: Allow remote nodes that start containers with pacemaker remote inside
if (pe__add_bundle_remote_name(rsc, data->params_all,
XML_RSC_ATTR_REMOTE_RA_ADDR)) {
crm_trace("Set address for bundle connection %s (on %s)",
rsc->id, node->details->uname);
}
g_hash_table_foreach(local_rsc_params, hash2field, data->params_all);
g_hash_table_foreach(action->extra, hash2field, data->params_all);
g_hash_table_foreach(rsc->parameters, hash2field, data->params_all);
g_hash_table_foreach(action->meta, hash2metafield, data->params_all);
if(xml_op) {
secure_list = crm_element_value(xml_op, XML_LRM_ATTR_OP_SECURE);
restart_list = crm_element_value(xml_op, XML_LRM_ATTR_OP_RESTART);
op_version = crm_element_value(xml_op, XML_ATTR_CRM_VERSION);
#if ENABLE_VERSIONED_ATTRS
ra_version = crm_element_value(xml_op, XML_ATTR_RA_VERSION);
#endif
} else {
op_version = CRM_FEATURE_SET;
}
#if ENABLE_VERSIONED_ATTRS
append_versioned_params(local_versioned_params, ra_version, data->params_all);
append_versioned_params(rsc->versioned_parameters, ra_version, data->params_all);
{
pe_rsc_action_details_t *details = pe_rsc_action_details(action);
append_versioned_params(details->versioned_parameters, ra_version, data->params_all);
}
#endif
pcmk__filter_op_for_digest(data->params_all);
g_hash_table_destroy(local_rsc_params);
pe_free_action(action);
data->digest_all_calc = calculate_operation_digest(data->params_all, op_version);
if (calc_secure) {
data->params_secure = copy_xml(data->params_all);
if(secure_list) {
filter_parameters(data->params_secure, secure_list, FALSE);
}
data->digest_secure_calc = calculate_operation_digest(data->params_secure, op_version);
}
if(xml_op && crm_element_value(xml_op, XML_LRM_ATTR_RESTART_DIGEST) != NULL) {
data->params_restart = copy_xml(data->params_all);
if (restart_list) {
filter_parameters(data->params_restart, restart_list, TRUE);
}
data->digest_restart_calc = calculate_operation_digest(data->params_restart, op_version);
}
g_hash_table_insert(node->details->digest_cache, strdup(key), data);
}
return data;
}
op_digest_cache_t *
rsc_action_digest_cmp(pe_resource_t * rsc, xmlNode * xml_op, pe_node_t * node,
pe_working_set_t * data_set)
{
op_digest_cache_t *data = NULL;
char *key = NULL;
guint interval_ms = 0;
const char *op_version;
const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
const char *digest_all;
const char *digest_restart;
CRM_ASSERT(node != NULL);
op_version = crm_element_value(xml_op, XML_ATTR_CRM_VERSION);
digest_all = crm_element_value(xml_op, XML_LRM_ATTR_OP_DIGEST);
digest_restart = crm_element_value(xml_op, XML_LRM_ATTR_RESTART_DIGEST);
crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
key = pcmk__op_key(rsc->id, task, interval_ms);
data = rsc_action_digest(rsc, task, key, node, xml_op,
is_set(data_set->flags, pe_flag_sanitized),
data_set);
data->rc = RSC_DIGEST_MATCH;
if (digest_restart && data->digest_restart_calc && strcmp(data->digest_restart_calc, digest_restart) != 0) {
pe_rsc_info(rsc, "Parameters to %s on %s changed: was %s vs. now %s (restart:%s) %s",
key, node->details->uname,
crm_str(digest_restart), data->digest_restart_calc,
op_version, crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC));
data->rc = RSC_DIGEST_RESTART;
} else if (digest_all == NULL) {
/* it is unknown what the previous op digest was */
data->rc = RSC_DIGEST_UNKNOWN;
} else if (strcmp(digest_all, data->digest_all_calc) != 0) {
pe_rsc_info(rsc, "Parameters to %s on %s changed: was %s vs. now %s (%s:%s) %s",
key, node->details->uname,
crm_str(digest_all), data->digest_all_calc,
(interval_ms > 0)? "reschedule" : "reload",
op_version, crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC));
data->rc = RSC_DIGEST_ALL;
}
free(key);
return data;
}
/*!
* \internal
* \brief Create an unfencing summary for use in special node attribute
*
* Create a string combining a fence device's resource ID, agent type, and
* parameter digest (whether for all parameters or just non-private parameters).
* This can be stored in a special node attribute, allowing us to detect changes
* in either the agent type or parameters, to know whether unfencing must be
* redone or can be safely skipped when the device's history is cleaned.
*
* \param[in] rsc_id Fence device resource ID
* \param[in] agent_type Fence device agent
* \param[in] param_digest Fence device parameter digest
*
* \return Newly allocated string with unfencing digest
* \note The caller is responsible for freeing the result.
*/
static inline char *
create_unfencing_summary(const char *rsc_id, const char *agent_type,
const char *param_digest)
{
return crm_strdup_printf("%s:%s:%s", rsc_id, agent_type, param_digest);
}
/*!
* \internal
* \brief Check whether a node can skip unfencing
*
* Check whether a fence device's current definition matches a node's
* stored summary of when it was last unfenced by the device.
*
* \param[in] rsc_id Fence device's resource ID
* \param[in] agent Fence device's agent type
* \param[in] digest_calc Fence device's current parameter digest
* \param[in] node_summary Value of node's special unfencing node attribute
* (a comma-separated list of unfencing summaries for
* all devices that have unfenced this node)
*
* \return TRUE if digest matches, FALSE otherwise
*/
static bool
unfencing_digest_matches(const char *rsc_id, const char *agent,
const char *digest_calc, const char *node_summary)
{
bool matches = FALSE;
if (rsc_id && agent && digest_calc && node_summary) {
char *search_secure = create_unfencing_summary(rsc_id, agent,
digest_calc);
/* The digest was calculated including the device ID and agent,
* so there is no risk of collision using strstr().
*/
matches = (strstr(node_summary, search_secure) != NULL);
crm_trace("Calculated unfencing digest '%s' %sfound in '%s'",
search_secure, matches? "" : "not ", node_summary);
free(search_secure);
}
return matches;
}
/* Magic string to use as action name for digest cache entries used for
* unfencing checks. This is not a real action name (i.e. "on"), so
* check_action_definition() won't confuse these entries with real actions.
*/
#define STONITH_DIGEST_TASK "stonith-on"
/*!
* \internal
* \brief Calculate fence device digests and digest comparison result
*
* \param[in] rsc Fence device resource
* \param[in] agent Fence device's agent type
* \param[in] node Node with digest cache to use
* \param[in] data_set Cluster working set
*
* \return Node's digest cache entry
*/
static op_digest_cache_t *
fencing_action_digest_cmp(pe_resource_t *rsc, const char *agent,
pe_node_t *node, pe_working_set_t *data_set)
{
const char *node_summary = NULL;
// Calculate device's current parameter digests
char *key = pcmk__op_key(rsc->id, STONITH_DIGEST_TASK, 0);
op_digest_cache_t *data = rsc_action_digest(rsc, STONITH_DIGEST_TASK, key,
node, NULL, TRUE, data_set);
free(key);
// Check whether node has special unfencing summary node attribute
node_summary = pe_node_attribute_raw(node, CRM_ATTR_DIGESTS_ALL);
if (node_summary == NULL) {
data->rc = RSC_DIGEST_UNKNOWN;
return data;
}
// Check whether full parameter digest matches
if (unfencing_digest_matches(rsc->id, agent, data->digest_all_calc,
node_summary)) {
data->rc = RSC_DIGEST_MATCH;
return data;
}
// Check whether secure parameter digest matches
node_summary = pe_node_attribute_raw(node, CRM_ATTR_DIGESTS_SECURE);
if (unfencing_digest_matches(rsc->id, agent, data->digest_secure_calc,
node_summary)) {
data->rc = RSC_DIGEST_MATCH;
if (is_set(data_set->flags, pe_flag_stdout)) {
printf("Only 'private' parameters to %s for unfencing %s changed\n",
rsc->id, node->details->uname);
}
return data;
}
// Parameters don't match
data->rc = RSC_DIGEST_ALL;
if (is_set(data_set->flags, (pe_flag_sanitized|pe_flag_stdout))
&& data->digest_secure_calc) {
char *digest = create_unfencing_summary(rsc->id, agent,
data->digest_secure_calc);
printf("Parameters to %s for unfencing %s changed, try '%s'\n",
rsc->id, node->details->uname, digest);
free(digest);
}
return data;
}
const char *rsc_printable_id(pe_resource_t *rsc)
{
if (is_not_set(rsc->flags, pe_rsc_unique)) {
return ID(rsc->xml);
}
return rsc->id;
}
void
clear_bit_recursive(pe_resource_t * rsc, unsigned long long flag)
{
GListPtr gIter = rsc->children;
clear_bit(rsc->flags, flag);
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
clear_bit_recursive(child_rsc, flag);
}
}
void
set_bit_recursive(pe_resource_t * rsc, unsigned long long flag)
{
GListPtr gIter = rsc->children;
set_bit(rsc->flags, flag);
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
set_bit_recursive(child_rsc, flag);
}
}
static GListPtr
find_unfencing_devices(GListPtr candidates, GListPtr matches)
{
for (GListPtr gIter = candidates; gIter != NULL; gIter = gIter->next) {
pe_resource_t *candidate = gIter->data;
const char *provides = g_hash_table_lookup(candidate->meta, XML_RSC_ATTR_PROVIDES);
const char *requires = g_hash_table_lookup(candidate->meta, XML_RSC_ATTR_REQUIRES);
if(candidate->children) {
matches = find_unfencing_devices(candidate->children, matches);
} else if (is_not_set(candidate->flags, pe_rsc_fence_device)) {
continue;
} else if (crm_str_eq(provides, "unfencing", FALSE) || crm_str_eq(requires, "unfencing", FALSE)) {
matches = g_list_prepend(matches, candidate);
}
}
return matches;
}
static int
node_priority_fencing_delay(pe_node_t * node, pe_working_set_t * data_set)
{
int member_count = 0;
int online_count = 0;
int top_priority = 0;
int lowest_priority = 0;
GListPtr gIter = NULL;
// `priority-fencing-delay` is disabled
if (data_set->priority_fencing_delay < 0) {
return -1;
}
/* No need to delay fencing if the fencing target is not a normal cluster
* member, for example if it's a remote node or a guest node. */
if (node->details->type != node_member) {
return 0;
}
// No need to delay fencing if the fencing target is in our partition
if (node->details->online) {
return 0;
}
for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
pe_node_t *n = gIter->data;
if (n->details->type != node_member) {
continue;
}
member_count ++;
if (n->details->online) {
online_count++;
}
if (member_count == 1
|| n->details->priority > top_priority) {
top_priority = n->details->priority;
}
if (member_count == 1
|| n->details->priority < lowest_priority) {
lowest_priority = n->details->priority;
}
}
// No need to delay if we have more than half of the cluster members
if (online_count > member_count / 2) {
return 0;
}
/* All the nodes have equal priority.
* Any configured corresponding `pcmk_delay_base/max` will be applied. */
if (lowest_priority == top_priority) {
return -1;
}
if (node->details->priority < top_priority) {
return 0;
}
return data_set->priority_fencing_delay;
}
pe_action_t *
pe_fence_op(pe_node_t * node, const char *op, bool optional, const char *reason,
bool priority_delay, pe_working_set_t * data_set)
{
char *op_key = NULL;
pe_action_t *stonith_op = NULL;
if(op == NULL) {
op = data_set->stonith_action;
}
op_key = crm_strdup_printf("%s-%s-%s", CRM_OP_FENCE, node->details->uname, op);
if(data_set->singletons) {
stonith_op = g_hash_table_lookup(data_set->singletons, op_key);
}
if(stonith_op == NULL) {
stonith_op = custom_action(NULL, op_key, CRM_OP_FENCE, node, TRUE, TRUE, data_set);
add_hash_param(stonith_op->meta, XML_LRM_ATTR_TARGET, node->details->uname);
add_hash_param(stonith_op->meta, XML_LRM_ATTR_TARGET_UUID, node->details->id);
add_hash_param(stonith_op->meta, "stonith_action", op);
if (pe__is_guest_or_remote_node(node)
&& is_set(data_set->flags, pe_flag_enable_unfencing)) {
/* Extra work to detect device changes on remotes
*
* We may do this for all nodes in the future, but for now
* the check_action_definition() based stuff works fine.
*/
long max = 1024;
long digests_all_offset = 0;
long digests_secure_offset = 0;
char *digests_all = calloc(max, sizeof(char));
char *digests_secure = calloc(max, sizeof(char));
GListPtr matches = find_unfencing_devices(data_set->resources, NULL);
for (GListPtr gIter = matches; gIter != NULL; gIter = gIter->next) {
pe_resource_t *match = gIter->data;
const char *agent = g_hash_table_lookup(match->meta,
XML_ATTR_TYPE);
op_digest_cache_t *data = NULL;
data = fencing_action_digest_cmp(match, agent, node, data_set);
if(data->rc == RSC_DIGEST_ALL) {
optional = FALSE;
crm_notice("Unfencing %s (remote): because the definition of %s changed", node->details->uname, match->id);
if (is_set(data_set->flags, pe_flag_stdout)) {
fprintf(stdout, " notice: Unfencing %s (remote): because the definition of %s changed\n", node->details->uname, match->id);
}
}
digests_all_offset += snprintf(
digests_all+digests_all_offset, max-digests_all_offset,
"%s:%s:%s,", match->id, agent, data->digest_all_calc);
digests_secure_offset += snprintf(
digests_secure+digests_secure_offset, max-digests_secure_offset,
"%s:%s:%s,", match->id, agent, data->digest_secure_calc);
}
g_hash_table_insert(stonith_op->meta,
strdup(XML_OP_ATTR_DIGESTS_ALL),
digests_all);
g_hash_table_insert(stonith_op->meta,
strdup(XML_OP_ATTR_DIGESTS_SECURE),
digests_secure);
}
} else {
free(op_key);
}
if (data_set->priority_fencing_delay >= 0
/* It's a suitable case where `priority-fencing-delay` applies.
* At least add `priority-fencing-delay` field as an indicator. */
&& (priority_delay
/* Re-calculate priority delay for the suitable case when
* pe_fence_op() is called again by stage6() after node priority has
* been actually calculated with native_add_running() */
|| g_hash_table_lookup(stonith_op->meta,
XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY) != NULL)) {
/* Add `priority-fencing-delay` to the fencing op even if it's 0 for
* the targeting node. So that it takes precedence over any possible
* `pcmk_delay_base/max`.
*/
char *delay_s = crm_itoa(node_priority_fencing_delay(node, data_set));
g_hash_table_insert(stonith_op->meta,
strdup(XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY),
delay_s);
}
if(optional == FALSE && pe_can_fence(data_set, node)) {
pe_action_required(stonith_op, NULL, reason);
} else if(reason && stonith_op->reason == NULL) {
stonith_op->reason = strdup(reason);
}
return stonith_op;
}
void
trigger_unfencing(
pe_resource_t * rsc, pe_node_t *node, const char *reason, pe_action_t *dependency, pe_working_set_t * data_set)
{
if(is_not_set(data_set->flags, pe_flag_enable_unfencing)) {
/* No resources require it */
return;
} else if (rsc != NULL && is_not_set(rsc->flags, pe_rsc_fence_device)) {
/* Wasn't a stonith device */
return;
} else if(node
&& node->details->online
&& node->details->unclean == FALSE
&& node->details->shutdown == FALSE) {
pe_action_t *unfence = pe_fence_op(node, "on", FALSE, reason, FALSE, data_set);
if(dependency) {
order_actions(unfence, dependency, pe_order_optional);
}
} else if(rsc) {
GHashTableIter iter;
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) {
if(node->details->online && node->details->unclean == FALSE && node->details->shutdown == FALSE) {
trigger_unfencing(rsc, node, reason, dependency, data_set);
}
}
}
}
gboolean
add_tag_ref(GHashTable * tags, const char * tag_name, const char * obj_ref)
{
pe_tag_t *tag = NULL;
GListPtr gIter = NULL;
gboolean is_existing = FALSE;
CRM_CHECK(tags && tag_name && obj_ref, return FALSE);
tag = g_hash_table_lookup(tags, tag_name);
if (tag == NULL) {
tag = calloc(1, sizeof(pe_tag_t));
if (tag == NULL) {
return FALSE;
}
tag->id = strdup(tag_name);
tag->refs = NULL;
g_hash_table_insert(tags, strdup(tag_name), tag);
}
for (gIter = tag->refs; gIter != NULL; gIter = gIter->next) {
const char *existing_ref = (const char *) gIter->data;
if (crm_str_eq(existing_ref, obj_ref, TRUE)){
is_existing = TRUE;
break;
}
}
if (is_existing == FALSE) {
tag->refs = g_list_append(tag->refs, strdup(obj_ref));
crm_trace("Added: tag=%s ref=%s", tag->id, obj_ref);
}
return TRUE;
}
void pe_action_set_flag_reason(const char *function, long line,
pe_action_t *action, pe_action_t *reason, const char *text,
enum pe_action_flags flags, bool overwrite)
{
bool unset = FALSE;
bool update = FALSE;
const char *change = NULL;
if(is_set(flags, pe_action_runnable)) {
unset = TRUE;
change = "unrunnable";
} else if(is_set(flags, pe_action_optional)) {
unset = TRUE;
change = "required";
} else if(is_set(flags, pe_action_migrate_runnable)) {
unset = TRUE;
overwrite = TRUE;
change = "unrunnable";
} else if(is_set(flags, pe_action_dangle)) {
change = "dangling";
} else if(is_set(flags, pe_action_requires_any)) {
change = "required";
} else {
crm_err("Unknown flag change to %x by %s: 0x%s",
flags, action->uuid, (reason? reason->uuid : "0"));
}
if(unset) {
if(is_set(action->flags, flags)) {
action->flags = crm_clear_bit(function, line, action->uuid, action->flags, flags);
update = TRUE;
}
} else {
if(is_not_set(action->flags, flags)) {
action->flags = crm_set_bit(function, line, action->uuid, action->flags, flags);
update = TRUE;
}
}
if((change && update) || text) {
char *reason_text = NULL;
if(reason == NULL) {
pe_action_set_reason(action, text, overwrite);
} else if(reason->rsc == NULL) {
reason_text = crm_strdup_printf("%s %s%c %s", change, reason->task, text?':':0, text?text:"");
} else {
reason_text = crm_strdup_printf("%s %s %s%c %s", change, reason->rsc->id, reason->task, text?':':0, text?text:"NA");
}
if(reason_text && action->rsc != reason->rsc) {
pe_action_set_reason(action, reason_text, overwrite);
}
free(reason_text);
}
}
void pe_action_set_reason(pe_action_t *action, const char *reason, bool overwrite)
{
if (action->reason != NULL && overwrite) {
pe_rsc_trace(action->rsc, "Changing %s reason from '%s' to '%s'",
action->uuid, action->reason, crm_str(reason));
free(action->reason);
} else if (action->reason == NULL) {
pe_rsc_trace(action->rsc, "Set %s reason to '%s'",
action->uuid, crm_str(reason));
} else {
// crm_assert(action->reason != NULL && !overwrite);
return;
}
if (reason != NULL) {
action->reason = strdup(reason);
} else {
action->reason = NULL;
}
}
/*!
* \internal
* \brief Check whether shutdown has been requested for a node
*
* \param[in] node Node to check
*
* \return TRUE if node has shutdown attribute set and nonzero, FALSE otherwise
* \note This differs from simply using node->details->shutdown in that it can
* be used before that has been determined (and in fact to determine it),
* and it can also be used to distinguish requested shutdown from implicit
* shutdown of remote nodes by virtue of their connection stopping.
*/
bool
pe__shutdown_requested(pe_node_t *node)
{
const char *shutdown = pe_node_attribute_raw(node, XML_CIB_ATTR_SHUTDOWN);
return shutdown && strcmp(shutdown, "0");
}
/*!
* \internal
* \brief Update a data set's "recheck by" time
*
* \param[in] recheck Epoch time when recheck should happen
* \param[in,out] data_set Current working set
*/
void
pe__update_recheck_time(time_t recheck, pe_working_set_t *data_set)
{
if ((recheck > get_effective_time(data_set))
&& ((data_set->recheck_by == 0)
|| (data_set->recheck_by > recheck))) {
data_set->recheck_by = recheck;
}
}
/*!
* \internal
* \brief Wrapper for pe_unpack_nvpairs() using a cluster working set
*/
void
pe__unpack_dataset_nvpairs(xmlNode *xml_obj, const char *set_name,
GHashTable *node_hash, GHashTable *hash,
const char *always_first, gboolean overwrite,
pe_working_set_t *data_set)
{
crm_time_t *next_change = crm_time_new_undefined();
pe_unpack_nvpairs(data_set->input, xml_obj, set_name, node_hash, hash,
always_first, overwrite, data_set->now, next_change);
if (crm_time_is_defined(next_change)) {
time_t recheck = (time_t) crm_time_get_seconds_since_epoch(next_change);
pe__update_recheck_time(recheck, data_set);
}
crm_time_free(next_change);
}
bool
pe__resource_is_disabled(pe_resource_t *rsc)
{
const char *target_role = NULL;
CRM_CHECK(rsc != NULL, return false);
target_role = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE);
if (target_role) {
enum rsc_role_e target_role_e = text2role(target_role);
if ((target_role_e == RSC_ROLE_STOPPED)
|| ((target_role_e == RSC_ROLE_SLAVE)
&& is_set(uber_parent(rsc)->flags, pe_rsc_promotable))) {
return true;
}
}
return false;
}
/*!
* \internal
* \brief Create an action to clear a resource's history from CIB
*
* \param[in] rsc Resource to clear
* \param[in] node Node to clear history on
*
* \return New action to clear resource history
*/
pe_action_t *
pe__clear_resource_history(pe_resource_t *rsc, pe_node_t *node,
pe_working_set_t *data_set)
{
char *key = NULL;
CRM_ASSERT(rsc && node);
key = pcmk__op_key(rsc->id, CRM_OP_LRM_DELETE, 0);
return custom_action(rsc, key, CRM_OP_LRM_DELETE, node, FALSE, TRUE,
data_set);
}
diff --git a/lib/services/services_linux.c b/lib/services/services_linux.c
index 375a391ce5..841d756960 100644
--- a/lib/services/services_linux.c
+++ b/lib/services/services_linux.c
@@ -1,1186 +1,1186 @@
/*
* Copyright 2010-2020 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>
#ifndef _GNU_SOURCE
# define _GNU_SOURCE
#endif
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/wait.h>
#include <errno.h>
#include <unistd.h>
#include <dirent.h>
#include <grp.h>
#include <string.h>
#include <sys/time.h>
#include <sys/resource.h>
#include "crm/crm.h"
#include "crm/common/mainloop.h"
#include "crm/services.h"
#include "services_private.h"
static void close_pipe(int fildes[]);
/* We have two alternative ways of handling SIGCHLD when synchronously waiting
* for spawned processes to complete. Both rely on polling a file descriptor to
* discover SIGCHLD events.
*
* If sys/signalfd.h is available (e.g. on Linux), we call signalfd() to
* generate the file descriptor. Otherwise, we use the "self-pipe trick"
* (opening a pipe and writing a byte to it when SIGCHLD is received).
*/
#ifdef HAVE_SYS_SIGNALFD_H
// signalfd() implementation
#include <sys/signalfd.h>
// Everything needed to manage SIGCHLD handling
struct sigchld_data_s {
sigset_t mask; // Signals to block now (including SIGCHLD)
sigset_t old_mask; // Previous set of blocked signals
};
// Initialize SIGCHLD data and prepare for use
static bool
sigchld_setup(struct sigchld_data_s *data)
{
sigemptyset(&(data->mask));
sigaddset(&(data->mask), SIGCHLD);
sigemptyset(&(data->old_mask));
// Block SIGCHLD (saving previous set of blocked signals to restore later)
if (sigprocmask(SIG_BLOCK, &(data->mask), &(data->old_mask)) < 0) {
crm_err("Wait for child process completion failed: %s "
CRM_XS " source=sigprocmask", pcmk_strerror(errno));
return false;
}
return true;
}
// Get a file descriptor suitable for polling for SIGCHLD events
static int
sigchld_open(struct sigchld_data_s *data)
{
int fd;
CRM_CHECK(data != NULL, return -1);
fd = signalfd(-1, &(data->mask), SFD_NONBLOCK);
if (fd < 0) {
crm_err("Wait for child process completion failed: %s "
CRM_XS " source=signalfd", pcmk_strerror(errno));
}
return fd;
}
// Close a file descriptor returned by sigchld_open()
static void
sigchld_close(int fd)
{
if (fd > 0) {
close(fd);
}
}
// Return true if SIGCHLD was received from polled fd
static bool
sigchld_received(int fd)
{
struct signalfd_siginfo fdsi;
ssize_t s;
if (fd < 0) {
return false;
}
s = read(fd, &fdsi, sizeof(struct signalfd_siginfo));
if (s != sizeof(struct signalfd_siginfo)) {
crm_err("Wait for child process completion failed: %s "
CRM_XS " source=read", pcmk_strerror(errno));
} else if (fdsi.ssi_signo == SIGCHLD) {
return true;
}
return false;
}
// Do anything needed after done waiting for SIGCHLD
static void
sigchld_cleanup(struct sigchld_data_s *data)
{
// Restore the original set of blocked signals
if ((sigismember(&(data->old_mask), SIGCHLD) == 0)
&& (sigprocmask(SIG_UNBLOCK, &(data->mask), NULL) < 0)) {
crm_warn("Could not clean up after child process completion: %s",
pcmk_strerror(errno));
}
}
#else // HAVE_SYS_SIGNALFD_H not defined
// Self-pipe implementation (see above for function descriptions)
struct sigchld_data_s {
int pipe_fd[2]; // Pipe file descriptors
struct sigaction sa; // Signal handling info (with SIGCHLD)
struct sigaction old_sa; // Previous signal handling info
};
// We need a global to use in the signal handler
volatile struct sigchld_data_s *last_sigchld_data = NULL;
static void
sigchld_handler()
{
// We received a SIGCHLD, so trigger pipe polling
if ((last_sigchld_data != NULL)
&& (last_sigchld_data->pipe_fd[1] >= 0)
&& (write(last_sigchld_data->pipe_fd[1], "", 1) == -1)) {
crm_err("Wait for child process completion failed: %s "
CRM_XS " source=write", pcmk_strerror(errno));
}
}
static bool
sigchld_setup(struct sigchld_data_s *data)
{
int rc;
data->pipe_fd[0] = data->pipe_fd[1] = -1;
if (pipe(data->pipe_fd) == -1) {
crm_err("Wait for child process completion failed: %s "
CRM_XS " source=pipe", pcmk_strerror(errno));
return false;
}
rc = pcmk__set_nonblocking(data->pipe_fd[0]);
if (rc != pcmk_rc_ok) {
crm_warn("Could not set pipe input non-blocking: %s " CRM_XS " rc=%d",
pcmk_rc_str(rc), rc);
}
rc = pcmk__set_nonblocking(data->pipe_fd[1]);
if (rc != pcmk_rc_ok) {
crm_warn("Could not set pipe output non-blocking: %s " CRM_XS " rc=%d",
pcmk_rc_str(rc), rc);
}
// Set SIGCHLD handler
data->sa.sa_handler = sigchld_handler;
data->sa.sa_flags = 0;
sigemptyset(&(data->sa.sa_mask));
if (sigaction(SIGCHLD, &(data->sa), &(data->old_sa)) < 0) {
crm_err("Wait for child process completion failed: %s "
CRM_XS " source=sigaction", pcmk_strerror(errno));
}
// Remember data for use in signal handler
last_sigchld_data = data;
return true;
}
static int
sigchld_open(struct sigchld_data_s *data)
{
CRM_CHECK(data != NULL, return -1);
return data->pipe_fd[0];
}
static void
sigchld_close(int fd)
{
// Pipe will be closed in sigchld_cleanup()
return;
}
static bool
sigchld_received(int fd)
{
char ch;
if (fd < 0) {
return false;
}
// Clear out the self-pipe
while (read(fd, &ch, 1) == 1) /*omit*/;
return true;
}
static void
sigchld_cleanup(struct sigchld_data_s *data)
{
// Restore the previous SIGCHLD handler
if (sigaction(SIGCHLD, &(data->old_sa), NULL) < 0) {
crm_warn("Could not clean up after child process completion: %s",
pcmk_strerror(errno));
}
close_pipe(data->pipe_fd);
}
#endif
/*!
* \internal
* \brief Close the two file descriptors of a pipe
*
* \param[in] fildes Array of file descriptors opened by pipe()
*/
static void
close_pipe(int fildes[])
{
if (fildes[0] >= 0) {
close(fildes[0]);
fildes[0] = -1;
}
if (fildes[1] >= 0) {
close(fildes[1]);
fildes[1] = -1;
}
}
static gboolean
svc_read_output(int fd, svc_action_t * op, bool is_stderr)
{
char *data = NULL;
int rc = 0, len = 0;
char buf[500];
static const size_t buf_read_len = sizeof(buf) - 1;
if (fd < 0) {
crm_trace("No fd for %s", op->id);
return FALSE;
}
if (is_stderr && op->stderr_data) {
len = strlen(op->stderr_data);
data = op->stderr_data;
crm_trace("Reading %s stderr into offset %d", op->id, len);
} else if (is_stderr == FALSE && op->stdout_data) {
len = strlen(op->stdout_data);
data = op->stdout_data;
crm_trace("Reading %s stdout into offset %d", op->id, len);
} else {
crm_trace("Reading %s %s into offset %d", op->id, is_stderr?"stderr":"stdout", len);
}
do {
rc = read(fd, buf, buf_read_len);
if (rc > 0) {
buf[rc] = 0;
crm_trace("Got %d chars: %.80s", rc, buf);
data = realloc_safe(data, len + rc + 1);
len += sprintf(data + len, "%s", buf);
} else if (errno != EINTR) {
/* error or EOF
* Cleanup happens in pipe_done()
*/
rc = FALSE;
break;
}
} while (rc == buf_read_len || rc < 0);
if (is_stderr) {
op->stderr_data = data;
} else {
op->stdout_data = data;
}
return rc;
}
static int
dispatch_stdout(gpointer userdata)
{
svc_action_t *op = (svc_action_t *) userdata;
return svc_read_output(op->opaque->stdout_fd, op, FALSE);
}
static int
dispatch_stderr(gpointer userdata)
{
svc_action_t *op = (svc_action_t *) userdata;
return svc_read_output(op->opaque->stderr_fd, op, TRUE);
}
static void
pipe_out_done(gpointer user_data)
{
svc_action_t *op = (svc_action_t *) user_data;
crm_trace("%p", op);
op->opaque->stdout_gsource = NULL;
if (op->opaque->stdout_fd > STDOUT_FILENO) {
close(op->opaque->stdout_fd);
}
op->opaque->stdout_fd = -1;
}
static void
pipe_err_done(gpointer user_data)
{
svc_action_t *op = (svc_action_t *) user_data;
op->opaque->stderr_gsource = NULL;
if (op->opaque->stderr_fd > STDERR_FILENO) {
close(op->opaque->stderr_fd);
}
op->opaque->stderr_fd = -1;
}
static struct mainloop_fd_callbacks stdout_callbacks = {
.dispatch = dispatch_stdout,
.destroy = pipe_out_done,
};
static struct mainloop_fd_callbacks stderr_callbacks = {
.dispatch = dispatch_stderr,
.destroy = pipe_err_done,
};
static void
set_ocf_env(const char *key, const char *value, gpointer user_data)
{
if (setenv(key, value, 1) != 0) {
crm_perror(LOG_ERR, "setenv failed for key:%s and value:%s", key, value);
}
}
static void
set_ocf_env_with_prefix(gpointer key, gpointer value, gpointer user_data)
{
char buffer[500];
snprintf(buffer, sizeof(buffer), "OCF_RESKEY_%s", (char *)key);
set_ocf_env(buffer, value, user_data);
}
static void
set_alert_env(gpointer key, gpointer value, gpointer user_data)
{
int rc;
if (value != NULL) {
rc = setenv(key, value, 1);
} else {
rc = unsetenv(key);
}
if (rc < 0) {
crm_perror(LOG_ERR, "setenv %s=%s",
(char*)key, (value? (char*)value : ""));
} else {
crm_trace("setenv %s=%s", (char*)key, (value? (char*)value : ""));
}
}
/*!
* \internal
* \brief Add environment variables suitable for an action
*
* \param[in] op Action to use
*/
static void
add_action_env_vars(const svc_action_t *op)
{
void (*env_setter)(gpointer, gpointer, gpointer) = NULL;
if (op->agent == NULL) {
env_setter = set_alert_env; /* we deal with alert handler */
} else if (safe_str_eq(op->standard, PCMK_RESOURCE_CLASS_OCF)) {
env_setter = set_ocf_env_with_prefix;
}
if (env_setter != NULL && op->params != NULL) {
g_hash_table_foreach(op->params, env_setter, NULL);
}
if (env_setter == NULL || env_setter == set_alert_env) {
return;
}
set_ocf_env("OCF_RA_VERSION_MAJOR", "1", NULL);
set_ocf_env("OCF_RA_VERSION_MINOR", "0", NULL);
set_ocf_env("OCF_ROOT", OCF_ROOT_DIR, NULL);
set_ocf_env("OCF_EXIT_REASON_PREFIX", PCMK_OCF_REASON_PREFIX, NULL);
if (op->rsc) {
set_ocf_env("OCF_RESOURCE_INSTANCE", op->rsc, NULL);
}
if (op->agent != NULL) {
set_ocf_env("OCF_RESOURCE_TYPE", op->agent, NULL);
}
/* Notes: this is not added to specification yet. Sept 10,2004 */
if (op->provider != NULL) {
set_ocf_env("OCF_RESOURCE_PROVIDER", op->provider, NULL);
}
}
static void
pipe_in_single_parameter(gpointer key, gpointer value, gpointer user_data)
{
svc_action_t *op = user_data;
char *buffer = crm_strdup_printf("%s=%s\n", (char *)key, (char *) value);
int ret, total = 0, len = strlen(buffer);
do {
errno = 0;
ret = write(op->opaque->stdin_fd, buffer + total, len - total);
if (ret > 0) {
total += ret;
}
} while ((errno == EINTR) && (total < len));
free(buffer);
}
/*!
* \internal
* \brief Pipe parameters in via stdin for action
*
* \param[in] op Action to use
*/
static void
pipe_in_action_stdin_parameters(const svc_action_t *op)
{
crm_debug("sending args");
if (op->params) {
g_hash_table_foreach(op->params, pipe_in_single_parameter, (gpointer) op);
}
}
gboolean
recurring_action_timer(gpointer data)
{
svc_action_t *op = data;
crm_debug("Scheduling another invocation of %s", op->id);
/* Clean out the old result */
free(op->stdout_data);
op->stdout_data = NULL;
free(op->stderr_data);
op->stderr_data = NULL;
op->opaque->repeat_timer = 0;
services_action_async(op, NULL);
return FALSE;
}
/* Returns FALSE if 'op' should be free'd by the caller */
gboolean
operation_finalize(svc_action_t * op)
{
int recurring = 0;
if (op->interval_ms) {
if (op->cancel) {
op->status = PCMK_LRM_OP_CANCELLED;
cancel_recurring_action(op);
} else {
recurring = 1;
op->opaque->repeat_timer = g_timeout_add(op->interval_ms,
recurring_action_timer, (void *)op);
}
}
if (op->opaque->callback) {
op->opaque->callback(op);
}
op->pid = 0;
services_untrack_op(op);
if (!recurring && op->synchronous == FALSE) {
/*
* If this is a recurring action, do not free explicitly.
* It will get freed whenever the action gets cancelled.
*/
services_action_free(op);
return TRUE;
}
services_action_cleanup(op);
return FALSE;
}
static void
close_op_input(svc_action_t *op)
{
if (op->opaque->stdin_fd >= 0) {
close(op->opaque->stdin_fd);
}
}
static void
finish_op_output(svc_action_t *op, bool is_stderr)
{
mainloop_io_t **source;
int fd;
if (is_stderr) {
source = &(op->opaque->stderr_gsource);
fd = op->opaque->stderr_fd;
} else {
source = &(op->opaque->stdout_gsource);
fd = op->opaque->stdout_fd;
}
if (op->synchronous || *source) {
crm_trace("Finish reading %s[%d] %s",
op->id, op->pid, (is_stderr? "stdout" : "stderr"));
svc_read_output(fd, op, is_stderr);
if (op->synchronous) {
close(fd);
} else {
mainloop_del_fd(*source);
*source = NULL;
}
}
}
// Log an operation's stdout and stderr
static void
log_op_output(svc_action_t *op)
{
char *prefix = crm_strdup_printf("%s[%d] error output", op->id, op->pid);
crm_log_output(LOG_NOTICE, prefix, op->stderr_data);
strcpy(prefix + strlen(prefix) - strlen("error output"), "output");
crm_log_output(LOG_DEBUG, prefix, op->stdout_data);
free(prefix);
}
static void
operation_finished(mainloop_child_t * p, pid_t pid, int core, int signo, int exitcode)
{
svc_action_t *op = mainloop_child_userdata(p);
mainloop_clear_child_userdata(p);
CRM_ASSERT(op->pid == pid);
/* Depending on the priority the mainloop gives the stdout and stderr
* file descriptors, this function could be called before everything has
* been read from them, so force a final read now.
*/
finish_op_output(op, true);
finish_op_output(op, false);
close_op_input(op);
if (signo == 0) {
crm_debug("%s[%d] exited with status %d", op->id, op->pid, exitcode);
op->status = PCMK_LRM_OP_DONE;
op->rc = exitcode;
} else if (mainloop_child_timeout(p)) {
crm_warn("%s[%d] timed out after %dms", op->id, op->pid, op->timeout);
op->status = PCMK_LRM_OP_TIMEOUT;
op->rc = PCMK_OCF_TIMEOUT;
} else if (op->cancel) {
/* If an in-flight recurring operation was killed because it was
* cancelled, don't treat that as a failure.
*/
crm_info("%s[%d] terminated with signal: %s " CRM_XS " (%d)",
op->id, op->pid, strsignal(signo), signo);
op->status = PCMK_LRM_OP_CANCELLED;
op->rc = PCMK_OCF_OK;
} else {
crm_warn("%s[%d] terminated with signal: %s " CRM_XS " (%d)",
op->id, op->pid, strsignal(signo), signo);
op->status = PCMK_LRM_OP_ERROR;
op->rc = PCMK_OCF_SIGNAL;
}
log_op_output(op);
operation_finalize(op);
}
/*!
* \internal
* \brief Set operation rc and status per errno from stat(), fork() or execvp()
*
* \param[in,out] op Operation to set rc and status for
* \param[in] error Value of errno after system call
*
* \return void
*/
static void
services_handle_exec_error(svc_action_t * op, int error)
{
int rc_not_installed, rc_insufficient_priv, rc_exec_error;
/* Mimic the return codes for each standard as that's what we'll convert back from in get_uniform_rc() */
if (safe_str_eq(op->standard, PCMK_RESOURCE_CLASS_LSB)
&& safe_str_eq(op->action, "status")) {
rc_not_installed = PCMK_LSB_STATUS_NOT_INSTALLED;
rc_insufficient_priv = PCMK_LSB_STATUS_INSUFFICIENT_PRIV;
rc_exec_error = PCMK_LSB_STATUS_UNKNOWN;
#if SUPPORT_NAGIOS
} else if (safe_str_eq(op->standard, PCMK_RESOURCE_CLASS_NAGIOS)) {
rc_not_installed = NAGIOS_NOT_INSTALLED;
rc_insufficient_priv = NAGIOS_INSUFFICIENT_PRIV;
rc_exec_error = PCMK_OCF_EXEC_ERROR;
#endif
} else {
rc_not_installed = PCMK_OCF_NOT_INSTALLED;
rc_insufficient_priv = PCMK_OCF_INSUFFICIENT_PRIV;
rc_exec_error = PCMK_OCF_EXEC_ERROR;
}
switch (error) { /* see execve(2), stat(2) and fork(2) */
case ENOENT: /* No such file or directory */
case EISDIR: /* Is a directory */
case ENOTDIR: /* Path component is not a directory */
case EINVAL: /* Invalid executable format */
case ENOEXEC: /* Invalid executable format */
op->rc = rc_not_installed;
op->status = PCMK_LRM_OP_NOT_INSTALLED;
break;
case EACCES: /* permission denied (various errors) */
case EPERM: /* permission denied (various errors) */
op->rc = rc_insufficient_priv;
op->status = PCMK_LRM_OP_ERROR;
break;
default:
op->rc = rc_exec_error;
op->status = PCMK_LRM_OP_ERROR;
}
}
static void
action_launch_child(svc_action_t *op)
{
/* SIGPIPE is ignored (which is different from signal blocking) by the gnutls library.
* Depending on the libqb version in use, libqb may set SIGPIPE to be ignored as well.
* We do not want this to be inherited by the child process. By resetting this the signal
* to the default behavior, we avoid some potential odd problems that occur during OCF
* scripts when SIGPIPE is ignored by the environment. */
signal(SIGPIPE, SIG_DFL);
#if defined(HAVE_SCHED_SETSCHEDULER)
if (sched_getscheduler(0) != SCHED_OTHER) {
struct sched_param sp;
memset(&sp, 0, sizeof(sp));
sp.sched_priority = 0;
if (sched_setscheduler(0, SCHED_OTHER, &sp) == -1) {
crm_perror(LOG_ERR, "Could not reset scheduling policy to SCHED_OTHER for %s", op->id);
}
}
#endif
if (setpriority(PRIO_PROCESS, 0, 0) == -1) {
crm_perror(LOG_ERR, "Could not reset process priority to 0 for %s", op->id);
}
/* Man: The call setpgrp() is equivalent to setpgid(0,0)
* _and_ compiles on BSD variants too
* need to investigate if it works the same too.
*/
setpgid(0, 0);
pcmk__close_fds_in_child(false);
#if SUPPORT_CIBSECRETS
if (pcmk__substitute_secrets(op->rsc, op->params) != pcmk_rc_ok) {
/* replacing secrets failed! */
if (safe_str_eq(op->action,"stop")) {
/* don't fail on stop! */
crm_info("proceeding with the stop operation for %s", op->rsc);
} else {
crm_err("failed to get secrets for %s, "
"considering resource not configured", op->rsc);
_exit(PCMK_OCF_NOT_CONFIGURED);
}
}
#endif
add_action_env_vars(op);
/* Become the desired user */
if (op->opaque->uid && (geteuid() == 0)) {
// If requested, set effective group
if (op->opaque->gid && (setgid(op->opaque->gid) < 0)) {
crm_perror(LOG_ERR, "Could not set child group to %d", op->opaque->gid);
_exit(PCMK_OCF_NOT_CONFIGURED);
}
// Erase supplementary group list
// (We could do initgroups() if we kept a copy of the username)
if (setgroups(0, NULL) < 0) {
crm_perror(LOG_ERR, "Could not set child groups");
_exit(PCMK_OCF_NOT_CONFIGURED);
}
// Set effective user
if (setuid(op->opaque->uid) < 0) {
crm_perror(LOG_ERR, "setting user to %d", op->opaque->uid);
_exit(PCMK_OCF_NOT_CONFIGURED);
}
}
/* execute the RA */
execvp(op->opaque->exec, op->opaque->args);
/* Most cases should have been already handled by stat() */
services_handle_exec_error(op, errno);
_exit(op->rc);
}
static void
action_synced_wait(svc_action_t *op, struct sigchld_data_s *data)
{
int status = 0;
int timeout = op->timeout;
time_t start = -1;
struct pollfd fds[3];
int wait_rc = 0;
fds[0].fd = op->opaque->stdout_fd;
fds[0].events = POLLIN;
fds[0].revents = 0;
fds[1].fd = op->opaque->stderr_fd;
fds[1].events = POLLIN;
fds[1].revents = 0;
fds[2].fd = sigchld_open(data);
fds[2].events = POLLIN;
fds[2].revents = 0;
crm_trace("Waiting for %s[%d]", op->id, op->pid);
start = time(NULL);
do {
int poll_rc = poll(fds, 3, timeout);
if (poll_rc > 0) {
if (fds[0].revents & POLLIN) {
svc_read_output(op->opaque->stdout_fd, op, FALSE);
}
if (fds[1].revents & POLLIN) {
svc_read_output(op->opaque->stderr_fd, op, TRUE);
}
if ((fds[2].revents & POLLIN) && sigchld_received(fds[2].fd)) {
wait_rc = waitpid(op->pid, &status, WNOHANG);
if ((wait_rc > 0) || ((wait_rc < 0) && (errno == ECHILD))) {
// Child process exited or doesn't exist
break;
} else if (wait_rc < 0) {
crm_warn("Wait for completion of %s[%d] failed: %s "
CRM_XS " source=waitpid",
op->id, op->pid, pcmk_strerror(errno));
wait_rc = 0; // Act as if process is still running
}
}
} else if (poll_rc == 0) {
// Poll timed out with no descriptors ready
timeout = 0;
break;
} else if ((poll_rc < 0) && (errno != EINTR)) {
crm_err("Wait for completion of %s[%d] failed: %s "
CRM_XS " source=poll",
op->id, op->pid, pcmk_strerror(errno));
break;
}
timeout = op->timeout - (time(NULL) - start) * 1000;
} while ((op->timeout < 0 || timeout > 0));
crm_trace("Stopped waiting for %s[%d]", op->id, op->pid);
if (wait_rc <= 0) {
op->rc = PCMK_OCF_UNKNOWN_ERROR;
if (op->timeout > 0 && timeout <= 0) {
op->status = PCMK_LRM_OP_TIMEOUT;
crm_warn("%s[%d] timed out after %dms",
op->id, op->pid, op->timeout);
} else {
op->status = PCMK_LRM_OP_ERROR;
}
/* If only child hasn't been successfully waited for, yet.
This is to limit killing wrong target a bit more. */
if (wait_rc == 0 && waitpid(op->pid, &status, WNOHANG) == 0) {
if (kill(op->pid, SIGKILL)) {
crm_warn("Could not kill rogue child %s[%d]: %s",
op->id, op->pid, pcmk_strerror(errno));
}
/* Safe to skip WNOHANG here as we sent non-ignorable signal. */
while (waitpid(op->pid, &status, 0) == (pid_t) -1 && errno == EINTR) /*omit*/;
}
} else if (WIFEXITED(status)) {
op->status = PCMK_LRM_OP_DONE;
op->rc = WEXITSTATUS(status);
crm_info("%s[%d] exited with status %d", op->id, op->pid, op->rc);
} else if (WIFSIGNALED(status)) {
int signo = WTERMSIG(status);
op->status = PCMK_LRM_OP_ERROR;
crm_err("%s[%d] terminated with signal: %s " CRM_XS " (%d)",
op->id, op->pid, strsignal(signo), signo);
}
#ifdef WCOREDUMP
if (WCOREDUMP(status)) {
crm_err("%s[%d] dumped core", op->id, op->pid);
}
#endif
finish_op_output(op, true);
finish_op_output(op, false);
close_op_input(op);
sigchld_close(fds[2].fd);
}
/* For an asynchronous 'op', returns FALSE if 'op' should be free'd by the caller */
/* For a synchronous 'op', returns FALSE if 'op' fails */
gboolean
services_os_action_execute(svc_action_t * op)
{
int stdout_fd[2];
int stderr_fd[2];
int stdin_fd[2] = {-1, -1};
int rc;
struct stat st;
struct sigchld_data_s data;
/* Fail fast */
if(stat(op->opaque->exec, &st) != 0) {
rc = errno;
crm_warn("Cannot execute '%s': %s " CRM_XS " stat rc=%d",
op->opaque->exec, pcmk_strerror(rc), rc);
services_handle_exec_error(op, rc);
if (!op->synchronous) {
return operation_finalize(op);
}
return FALSE;
}
if (pipe(stdout_fd) < 0) {
rc = errno;
crm_err("Cannot execute '%s': %s " CRM_XS " pipe(stdout) rc=%d",
op->opaque->exec, pcmk_strerror(rc), rc);
services_handle_exec_error(op, rc);
if (!op->synchronous) {
return operation_finalize(op);
}
return FALSE;
}
if (pipe(stderr_fd) < 0) {
rc = errno;
close_pipe(stdout_fd);
crm_err("Cannot execute '%s': %s " CRM_XS " pipe(stderr) rc=%d",
op->opaque->exec, pcmk_strerror(rc), rc);
services_handle_exec_error(op, rc);
if (!op->synchronous) {
return operation_finalize(op);
}
return FALSE;
}
- if (safe_str_eq(op->standard, PCMK_RESOURCE_CLASS_STONITH)) {
+ if (is_set(pcmk_get_ra_caps(op->standard), pcmk_ra_cap_stdin)) {
if (pipe(stdin_fd) < 0) {
rc = errno;
close_pipe(stdout_fd);
close_pipe(stderr_fd);
crm_err("Cannot execute '%s': %s " CRM_XS " pipe(stdin) rc=%d",
op->opaque->exec, pcmk_strerror(rc), rc);
services_handle_exec_error(op, rc);
if (!op->synchronous) {
return operation_finalize(op);
}
return FALSE;
}
}
if (op->synchronous && !sigchld_setup(&data)) {
close_pipe(stdin_fd);
close_pipe(stdout_fd);
close_pipe(stderr_fd);
sigchld_cleanup(&data);
return FALSE;
}
op->pid = fork();
switch (op->pid) {
case -1:
rc = errno;
close_pipe(stdin_fd);
close_pipe(stdout_fd);
close_pipe(stderr_fd);
crm_err("Cannot execute '%s': %s " CRM_XS " fork rc=%d",
op->opaque->exec, pcmk_strerror(rc), rc);
services_handle_exec_error(op, rc);
if (!op->synchronous) {
return operation_finalize(op);
}
sigchld_cleanup(&data);
return FALSE;
case 0: /* Child */
close(stdout_fd[0]);
close(stderr_fd[0]);
if (stdin_fd[1] >= 0) {
close(stdin_fd[1]);
}
if (STDOUT_FILENO != stdout_fd[1]) {
if (dup2(stdout_fd[1], STDOUT_FILENO) != STDOUT_FILENO) {
crm_warn("Can't redirect output from '%s': %s "
CRM_XS " errno=%d",
op->opaque->exec, pcmk_strerror(errno), errno);
}
close(stdout_fd[1]);
}
if (STDERR_FILENO != stderr_fd[1]) {
if (dup2(stderr_fd[1], STDERR_FILENO) != STDERR_FILENO) {
crm_warn("Can't redirect error output from '%s': %s "
CRM_XS " errno=%d",
op->opaque->exec, pcmk_strerror(errno), errno);
}
close(stderr_fd[1]);
}
if ((stdin_fd[0] >= 0) &&
(STDIN_FILENO != stdin_fd[0])) {
if (dup2(stdin_fd[0], STDIN_FILENO) != STDIN_FILENO) {
crm_warn("Can't redirect input to '%s': %s "
CRM_XS " errno=%d",
op->opaque->exec, pcmk_strerror(errno), errno);
}
close(stdin_fd[0]);
}
if (op->synchronous) {
sigchld_cleanup(&data);
}
action_launch_child(op);
CRM_ASSERT(0); /* action_launch_child is effectively noreturn */
}
/* Only the parent reaches here */
close(stdout_fd[1]);
close(stderr_fd[1]);
if (stdin_fd[0] >= 0) {
close(stdin_fd[0]);
}
op->opaque->stdout_fd = stdout_fd[0];
rc = pcmk__set_nonblocking(op->opaque->stdout_fd);
if (rc != pcmk_rc_ok) {
crm_warn("Could not set '%s' output non-blocking: %s "
CRM_XS " rc=%d",
op->opaque->exec, pcmk_rc_str(rc), rc);
}
op->opaque->stderr_fd = stderr_fd[0];
rc = pcmk__set_nonblocking(op->opaque->stderr_fd);
if (rc != pcmk_rc_ok) {
crm_warn("Could not set '%s' error output non-blocking: %s "
CRM_XS " rc=%d",
op->opaque->exec, pcmk_rc_str(rc), rc);
}
op->opaque->stdin_fd = stdin_fd[1];
if (op->opaque->stdin_fd >= 0) {
// using buffer behind non-blocking-fd here - that could be improved
// as long as no other standard uses stdin_fd assume stonith
rc = pcmk__set_nonblocking(op->opaque->stdin_fd);
if (rc != pcmk_rc_ok) {
crm_warn("Could not set '%s' input non-blocking: %s "
CRM_XS " fd=%d,rc=%d", op->opaque->exec,
pcmk_rc_str(rc), op->opaque->stdin_fd, rc);
}
pipe_in_action_stdin_parameters(op);
// as long as we are handling parameters directly in here just close
close(op->opaque->stdin_fd);
op->opaque->stdin_fd = -1;
}
// after fds are setup properly and before we plug anything into mainloop
if (op->opaque->fork_callback) {
op->opaque->fork_callback(op);
}
if (op->synchronous) {
action_synced_wait(op, &data);
sigchld_cleanup(&data);
} else {
crm_trace("Waiting async for '%s'[%d]", op->opaque->exec, op->pid);
mainloop_child_add_with_flags(op->pid,
op->timeout,
op->id,
op,
(op->flags & SVC_ACTION_LEAVE_GROUP) ? mainloop_leave_pid_group : 0,
operation_finished);
op->opaque->stdout_gsource = mainloop_add_fd(op->id,
G_PRIORITY_LOW,
op->opaque->stdout_fd, op, &stdout_callbacks);
op->opaque->stderr_gsource = mainloop_add_fd(op->id,
G_PRIORITY_LOW,
op->opaque->stderr_fd, op, &stderr_callbacks);
services_add_inflight_op(op);
}
return TRUE;
}
GList *
services_os_get_directory_list(const char *root, gboolean files, gboolean executable)
{
GList *list = NULL;
struct dirent **namelist;
int entries = 0, lpc = 0;
char buffer[PATH_MAX];
entries = scandir(root, &namelist, NULL, alphasort);
if (entries <= 0) {
return list;
}
for (lpc = 0; lpc < entries; lpc++) {
struct stat sb;
if ('.' == namelist[lpc]->d_name[0]) {
free(namelist[lpc]);
continue;
}
snprintf(buffer, sizeof(buffer), "%s/%s", root, namelist[lpc]->d_name);
if (stat(buffer, &sb)) {
continue;
}
if (S_ISDIR(sb.st_mode)) {
if (files) {
free(namelist[lpc]);
continue;
}
} else if (S_ISREG(sb.st_mode)) {
if (files == FALSE) {
free(namelist[lpc]);
continue;
} else if (executable
&& (sb.st_mode & S_IXUSR) == 0
&& (sb.st_mode & S_IXGRP) == 0 && (sb.st_mode & S_IXOTH) == 0) {
free(namelist[lpc]);
continue;
}
}
list = g_list_append(list, strdup(namelist[lpc]->d_name));
free(namelist[lpc]);
}
free(namelist);
return list;
}
GList *
resources_os_list_ocf_providers(void)
{
return get_directory_list(OCF_ROOT_DIR "/resource.d", FALSE, TRUE);
}
GList *
resources_os_list_ocf_agents(const char *provider)
{
GList *gIter = NULL;
GList *result = NULL;
GList *providers = NULL;
if (provider) {
char buffer[500];
snprintf(buffer, sizeof(buffer), "%s/resource.d/%s", OCF_ROOT_DIR, provider);
return get_directory_list(buffer, TRUE, TRUE);
}
providers = resources_os_list_ocf_providers();
for (gIter = providers; gIter != NULL; gIter = gIter->next) {
GList *tmp1 = result;
GList *tmp2 = resources_os_list_ocf_agents(gIter->data);
if (tmp2) {
result = g_list_concat(tmp1, tmp2);
}
}
g_list_free_full(providers, free);
return result;
}
gboolean
services__ocf_agent_exists(const char *provider, const char *agent)
{
char *buf = NULL;
gboolean rc = FALSE;
struct stat st;
if (provider == NULL || agent == NULL) {
return rc;
}
buf = crm_strdup_printf(OCF_ROOT_DIR "/resource.d/%s/%s", provider, agent);
if (stat(buf, &st) == 0) {
rc = TRUE;
}
free(buf);
return rc;
}
diff --git a/tools/crm_resource_runtime.c b/tools/crm_resource_runtime.c
index 7855766694..3cf7b9ca29 100644
--- a/tools/crm_resource_runtime.c
+++ b/tools/crm_resource_runtime.c
@@ -1,2089 +1,2090 @@
/*
* Copyright 2004-2020 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_resource.h>
int resource_verbose = 0;
bool do_force = FALSE;
const char *attr_set_type = XML_TAG_ATTR_SETS;
static int
do_find_resource(const char *rsc, pe_resource_t * the_rsc, pe_working_set_t * data_set)
{
int found = 0;
GListPtr lpc = NULL;
for (lpc = the_rsc->running_on; lpc != NULL; lpc = lpc->next) {
pe_node_t *node = (pe_node_t *) lpc->data;
if (BE_QUIET) {
fprintf(stdout, "%s\n", node->details->uname);
} else {
const char *state = "";
if (!pe_rsc_is_clone(the_rsc) && the_rsc->fns->state(the_rsc, TRUE) == RSC_ROLE_MASTER) {
state = "Master";
}
fprintf(stdout, "resource %s is running on: %s %s\n", rsc, node->details->uname, state);
}
found++;
}
if (BE_QUIET == FALSE && found == 0) {
fprintf(stderr, "resource %s is NOT running\n", rsc);
}
return found;
}
int
cli_resource_search(pe_resource_t *rsc, const char *requested_name,
pe_working_set_t *data_set)
{
int found = 0;
pe_resource_t *parent = uber_parent(rsc);
if (pe_rsc_is_clone(rsc)) {
for (GListPtr iter = rsc->children; iter != NULL; iter = iter->next) {
found += do_find_resource(requested_name, iter->data, data_set);
}
/* The anonymous clone children's common ID is supplied */
} else if (pe_rsc_is_clone(parent)
&& is_not_set(rsc->flags, pe_rsc_unique)
&& rsc->clone_name
&& safe_str_eq(requested_name, rsc->clone_name)
&& safe_str_neq(requested_name, rsc->id)) {
for (GListPtr iter = parent->children; iter; iter = iter->next) {
found += do_find_resource(requested_name, iter->data, data_set);
}
} else {
found += do_find_resource(requested_name, rsc, data_set);
}
return found;
}
#define XPATH_MAX 1024
static int
find_resource_attr(cib_t * the_cib, const char *attr, const char *rsc, const char *set_type,
const char *set_name, const char *attr_id, const char *attr_name, char **value)
{
int offset = 0;
int rc = pcmk_ok;
xmlNode *xml_search = NULL;
char *xpath_string = NULL;
if(value) {
*value = NULL;
}
if(the_cib == NULL) {
return -ENOTCONN;
}
xpath_string = calloc(1, XPATH_MAX);
offset +=
snprintf(xpath_string + offset, XPATH_MAX - offset, "%s", get_object_path("resources"));
offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "//*[@id=\"%s\"]", rsc);
if (set_type) {
offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "/%s", set_type);
if (set_name) {
offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "[@id=\"%s\"]", set_name);
}
}
offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "//nvpair[");
if (attr_id) {
offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "@id=\"%s\"", attr_id);
}
if (attr_name) {
if (attr_id) {
offset += snprintf(xpath_string + offset, XPATH_MAX - offset, " and ");
}
offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "@name=\"%s\"", attr_name);
}
offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "]");
CRM_LOG_ASSERT(offset > 0);
rc = the_cib->cmds->query(the_cib, xpath_string, &xml_search,
cib_sync_call | cib_scope_local | cib_xpath);
if (rc != pcmk_ok) {
goto bail;
}
crm_log_xml_debug(xml_search, "Match");
if (xml_has_children(xml_search)) {
xmlNode *child = NULL;
rc = -EINVAL;
printf("Multiple attributes match name=%s\n", attr_name);
for (child = __xml_first_child(xml_search); child != NULL; child = __xml_next(child)) {
printf(" Value: %s \t(id=%s)\n",
crm_element_value(child, XML_NVPAIR_ATTR_VALUE), ID(child));
}
} else if(value) {
const char *tmp = crm_element_value(xml_search, attr);
if (tmp) {
*value = strdup(tmp);
}
}
bail:
free(xpath_string);
free_xml(xml_search);
return rc;
}
/* PRIVATE. Use the find_matching_attr_resources instead. */
static void
find_matching_attr_resources_recursive(GList/* <pe_resource_t*> */ ** result, pe_resource_t * rsc, const char * rsc_id,
const char * attr_set, const char * attr_id,
const char * attr_name, cib_t * cib, const char * cmd, int depth)
{
int rc = pcmk_ok;
char *lookup_id = clone_strip(rsc->id);
char *local_attr_id = NULL;
/* visit the children */
for(GList *gIter = rsc->children; gIter; gIter = gIter->next) {
find_matching_attr_resources_recursive(result, (pe_resource_t*)gIter->data, rsc_id, attr_set, attr_id, attr_name, cib, cmd, depth+1);
/* do it only once for clones */
if(pe_clone == rsc->variant) {
break;
}
}
rc = find_resource_attr(cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id);
/* Post-order traversal.
* The root is always on the list and it is the last item. */
if((0 == depth) || (pcmk_ok == rc)) {
/* push the head */
*result = g_list_append(*result, rsc);
}
free(local_attr_id);
free(lookup_id);
}
/* The result is a linearized pre-ordered tree of resources. */
static GList/*<pe_resource_t*>*/ *
find_matching_attr_resources(pe_resource_t * rsc, const char * rsc_id, const char * attr_set, const char * attr_id,
const char * attr_name, cib_t * cib, const char * cmd)
{
int rc = pcmk_ok;
char *lookup_id = NULL;
char *local_attr_id = NULL;
GList * result = NULL;
/* If --force is used, update only the requested resource (clone or primitive).
* Otherwise, if the primitive has the attribute, use that.
* Otherwise use the clone. */
if(do_force == TRUE) {
return g_list_append(result, rsc);
}
if(rsc->parent && pe_clone == rsc->parent->variant) {
int rc = pcmk_ok;
char *local_attr_id = NULL;
rc = find_resource_attr(cib, XML_ATTR_ID, rsc_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id);
free(local_attr_id);
if(rc != pcmk_ok) {
rsc = rsc->parent;
if (BE_QUIET == FALSE) {
printf("Performing %s of '%s' on '%s', the parent of '%s'\n", cmd, attr_name, rsc->id, rsc_id);
}
}
return g_list_append(result, rsc);
} else if(rsc->parent == NULL && rsc->children && pe_clone == rsc->variant) {
pe_resource_t *child = rsc->children->data;
if(child->variant == pe_native) {
lookup_id = clone_strip(child->id); /* Could be a cloned group! */
rc = find_resource_attr(cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id);
if(rc == pcmk_ok) {
rsc = child;
if (BE_QUIET == FALSE) {
printf("A value for '%s' already exists in child '%s', performing %s on that instead of '%s'\n", attr_name, lookup_id, cmd, rsc_id);
}
}
free(local_attr_id);
free(lookup_id);
}
return g_list_append(result, rsc);
}
/* If the resource is a group ==> children inherit the attribute if defined. */
find_matching_attr_resources_recursive(&result, rsc, rsc_id, attr_set, attr_id, attr_name, cib, cmd, 0);
return result;
}
int
cli_resource_update_attribute(pe_resource_t *rsc, const char *requested_name,
const char *attr_set, const char *attr_id,
const char *attr_name, const char *attr_value,
bool recursive, cib_t *cib,
pe_working_set_t *data_set)
{
int rc = pcmk_ok;
static bool need_init = TRUE;
char *local_attr_id = NULL;
char *local_attr_set = NULL;
GList/*<pe_resource_t*>*/ *resources = NULL;
const char *common_attr_id = attr_id;
if(attr_id == NULL
&& do_force == FALSE
&& find_resource_attr(
cib, XML_ATTR_ID, uber_parent(rsc)->id, NULL, NULL, NULL, attr_name, NULL) == -EINVAL) {
printf("\n");
}
if (safe_str_eq(attr_set_type, XML_TAG_ATTR_SETS)) {
if (do_force == FALSE) {
rc = find_resource_attr(cib, XML_ATTR_ID, uber_parent(rsc)->id,
XML_TAG_META_SETS, attr_set, attr_id,
attr_name, &local_attr_id);
if (rc == pcmk_ok && BE_QUIET == FALSE) {
printf("WARNING: There is already a meta attribute for '%s' called '%s' (id=%s)\n",
uber_parent(rsc)->id, attr_name, local_attr_id);
printf(" Delete '%s' first or use the force option to override\n",
local_attr_id);
}
free(local_attr_id);
if (rc == pcmk_ok) {
return -ENOTUNIQ;
}
}
resources = g_list_append(resources, rsc);
} else {
resources = find_matching_attr_resources(rsc, requested_name, attr_set,
attr_id, attr_name, cib, "update");
}
/* If either attr_set or attr_id is specified,
* one clearly intends to modify a single resource.
* It is the last item on the resource list.*/
for(GList *gIter = (attr_set||attr_id) ? g_list_last(resources) : resources
; gIter; gIter = gIter->next) {
char *lookup_id = NULL;
xmlNode *xml_top = NULL;
xmlNode *xml_obj = NULL;
local_attr_id = NULL;
local_attr_set = NULL;
rsc = (pe_resource_t*)gIter->data;
attr_id = common_attr_id;
lookup_id = clone_strip(rsc->id); /* Could be a cloned group! */
rc = find_resource_attr(cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name,
&local_attr_id);
if (rc == pcmk_ok) {
crm_debug("Found a match for name=%s: id=%s", attr_name, local_attr_id);
attr_id = local_attr_id;
} else if (rc != -ENXIO) {
free(lookup_id);
free(local_attr_id);
g_list_free(resources);
return rc;
} else {
const char *tag = crm_element_name(rsc->xml);
if (attr_set == NULL) {
local_attr_set = crm_strdup_printf("%s-%s", lookup_id,
attr_set_type);
attr_set = local_attr_set;
}
if (attr_id == NULL) {
local_attr_id = crm_strdup_printf("%s-%s", attr_set, attr_name);
attr_id = local_attr_id;
}
xml_top = create_xml_node(NULL, tag);
crm_xml_add(xml_top, XML_ATTR_ID, lookup_id);
xml_obj = create_xml_node(xml_top, attr_set_type);
crm_xml_add(xml_obj, XML_ATTR_ID, attr_set);
}
xml_obj = crm_create_nvpair_xml(xml_obj, attr_id, attr_name, attr_value);
if (xml_top == NULL) {
xml_top = xml_obj;
}
crm_log_xml_debug(xml_top, "Update");
rc = cib->cmds->modify(cib, XML_CIB_TAG_RESOURCES, xml_top, cib_options);
if (rc == pcmk_ok && BE_QUIET == FALSE) {
printf("Set '%s' option: id=%s%s%s%s%s value=%s\n", lookup_id, local_attr_id,
attr_set ? " set=" : "", attr_set ? attr_set : "",
attr_name ? " name=" : "", attr_name ? attr_name : "", attr_value);
}
free_xml(xml_top);
free(lookup_id);
free(local_attr_id);
free(local_attr_set);
if(recursive && safe_str_eq(attr_set_type, XML_TAG_META_SETS)) {
GListPtr lpc = NULL;
if(need_init) {
xmlNode *cib_constraints = get_object_root(XML_CIB_TAG_CONSTRAINTS, data_set->input);
need_init = FALSE;
unpack_constraints(cib_constraints, data_set);
for (lpc = data_set->resources; lpc != NULL; lpc = lpc->next) {
pe_resource_t *r = (pe_resource_t *) lpc->data;
clear_bit(r->flags, pe_rsc_allocating);
}
}
crm_debug("Looking for dependencies %p", rsc->rsc_cons_lhs);
set_bit(rsc->flags, pe_rsc_allocating);
for (lpc = rsc->rsc_cons_lhs; lpc != NULL; lpc = lpc->next) {
rsc_colocation_t *cons = (rsc_colocation_t *) lpc->data;
pe_resource_t *peer = cons->rsc_lh;
crm_debug("Checking %s %d", cons->id, cons->score);
if (cons->score > 0 && is_not_set(peer->flags, pe_rsc_allocating)) {
/* Don't get into colocation loops */
crm_debug("Setting %s=%s for dependent resource %s", attr_name, attr_value, peer->id);
cli_resource_update_attribute(peer, peer->id, NULL, NULL,
attr_name, attr_value, recursive,
cib, data_set);
}
}
}
}
g_list_free(resources);
return rc;
}
int
cli_resource_delete_attribute(pe_resource_t *rsc, const char *requested_name,
const char *attr_set, const char *attr_id,
const char *attr_name, cib_t *cib,
pe_working_set_t *data_set)
{
int rc = pcmk_ok;
GList/*<pe_resource_t*>*/ *resources = NULL;
if(attr_id == NULL
&& do_force == FALSE
&& find_resource_attr(
cib, XML_ATTR_ID, uber_parent(rsc)->id, NULL, NULL, NULL, attr_name, NULL) == -EINVAL) {
printf("\n");
}
if(safe_str_eq(attr_set_type, XML_TAG_META_SETS)) {
resources = find_matching_attr_resources(rsc, requested_name, attr_set,
attr_id, attr_name, cib, "delete");
} else {
resources = g_list_append(resources, rsc);
}
for(GList *gIter = resources; gIter; gIter = gIter->next) {
char *lookup_id = NULL;
xmlNode *xml_obj = NULL;
char *local_attr_id = NULL;
rsc = (pe_resource_t*)gIter->data;
lookup_id = clone_strip(rsc->id);
rc = find_resource_attr(cib, XML_ATTR_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name,
&local_attr_id);
if (rc == -ENXIO) {
free(lookup_id);
rc = pcmk_ok;
continue;
} else if (rc != pcmk_ok) {
free(lookup_id);
g_list_free(resources);
return rc;
}
if (attr_id == NULL) {
attr_id = local_attr_id;
}
xml_obj = crm_create_nvpair_xml(NULL, attr_id, attr_name, NULL);
crm_log_xml_debug(xml_obj, "Delete");
CRM_ASSERT(cib);
rc = cib->cmds->remove(cib, XML_CIB_TAG_RESOURCES, xml_obj, cib_options);
if (rc == pcmk_ok && BE_QUIET == FALSE) {
printf("Deleted '%s' option: id=%s%s%s%s%s\n", lookup_id, local_attr_id,
attr_set ? " set=" : "", attr_set ? attr_set : "",
attr_name ? " name=" : "", attr_name ? attr_name : "");
}
free(lookup_id);
free_xml(xml_obj);
free(local_attr_id);
}
g_list_free(resources);
return rc;
}
// \return Standard Pacemaker return code
static int
send_lrm_rsc_op(pcmk_controld_api_t *controld_api, bool do_fail_resource,
const char *host_uname, const char *rsc_id,
pe_working_set_t *data_set)
{
const char *router_node = host_uname;
const char *rsc_api_id = NULL;
const char *rsc_long_id = NULL;
const char *rsc_class = NULL;
const char *rsc_provider = NULL;
const char *rsc_type = NULL;
bool cib_only = false;
pe_resource_t *rsc = pe_find_resource(data_set->resources, rsc_id);
if (rsc == NULL) {
CMD_ERR("Resource %s not found", rsc_id);
return ENXIO;
} else if (rsc->variant != pe_native) {
CMD_ERR("We can only process primitive resources, not %s", rsc_id);
return EINVAL;
}
rsc_class = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS);
rsc_provider = crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER),
rsc_type = crm_element_value(rsc->xml, XML_ATTR_TYPE);
if ((rsc_class == NULL) || (rsc_type == NULL)) {
CMD_ERR("Resource %s does not have a class and type", rsc_id);
return EINVAL;
}
if (host_uname == NULL) {
CMD_ERR("Please specify a node name");
return EINVAL;
} else {
pe_node_t *node = pe_find_node(data_set->nodes, host_uname);
if (node == NULL) {
CMD_ERR("Node %s not found", host_uname);
return pcmk_rc_node_unknown;
}
if (!(node->details->online)) {
if (do_fail_resource) {
CMD_ERR("Node %s is not online", host_uname);
return ENOTCONN;
} else {
cib_only = true;
}
}
if (!cib_only && pe__is_guest_or_remote_node(node)) {
node = pe__current_node(node->details->remote_rsc);
if (node == NULL) {
CMD_ERR("No cluster connection to Pacemaker Remote node %s detected",
host_uname);
return ENOTCONN;
}
router_node = node->details->uname;
}
}
if (rsc->clone_name) {
rsc_api_id = rsc->clone_name;
rsc_long_id = rsc->id;
} else {
rsc_api_id = rsc->id;
}
if (do_fail_resource) {
return controld_api->fail_resource(controld_api, host_uname,
router_node, rsc_api_id, rsc_long_id,
rsc_class, rsc_provider, rsc_type);
} else {
return controld_api->refresh_resource(controld_api, host_uname,
router_node, rsc_api_id,
rsc_long_id, rsc_class,
rsc_provider, rsc_type, cib_only);
}
}
/*!
* \internal
* \brief Get resource name as used in failure-related node attributes
*
* \param[in] rsc Resource to check
*
* \return Newly allocated string containing resource's fail name
* \note The caller is responsible for freeing the result.
*/
static inline char *
rsc_fail_name(pe_resource_t *rsc)
{
const char *name = (rsc->clone_name? rsc->clone_name : rsc->id);
return is_set(rsc->flags, pe_rsc_unique)? strdup(name) : clone_strip(name);
}
// \return Standard Pacemaker return code
static int
clear_rsc_history(pcmk_controld_api_t *controld_api, const char *host_uname,
const char *rsc_id, pe_working_set_t *data_set)
{
int rc = pcmk_ok;
/* Erase the resource's entire LRM history in the CIB, even if we're only
* clearing a single operation's fail count. If we erased only entries for a
* single operation, we might wind up with a wrong idea of the current
* resource state, and we might not re-probe the resource.
*/
rc = send_lrm_rsc_op(controld_api, false, host_uname, rsc_id, data_set);
if (rc != pcmk_rc_ok) {
return rc;
}
crm_trace("Processing %d mainloop inputs",
controld_api->replies_expected(controld_api));
while (g_main_context_iteration(NULL, FALSE)) {
crm_trace("Processed mainloop input, %d still remaining",
controld_api->replies_expected(controld_api));
}
return rc;
}
static int
clear_rsc_failures(pcmk_controld_api_t *controld_api, const char *node_name,
const char *rsc_id, const char *operation,
const char *interval_spec, pe_working_set_t *data_set)
{
int rc = pcmk_ok;
const char *failed_value = NULL;
const char *failed_id = NULL;
const char *interval_ms_s = NULL;
GHashTable *rscs = NULL;
GHashTableIter iter;
/* Create a hash table to use as a set of resources to clean. This lets us
* clean each resource only once (per node) regardless of how many failed
* operations it has.
*/
rscs = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, NULL);
// Normalize interval to milliseconds for comparison to history entry
if (operation) {
interval_ms_s = crm_strdup_printf("%u",
crm_parse_interval_spec(interval_spec));
}
for (xmlNode *xml_op = __xml_first_child(data_set->failed); xml_op != NULL;
xml_op = __xml_next(xml_op)) {
failed_id = crm_element_value(xml_op, XML_LRM_ATTR_RSCID);
if (failed_id == NULL) {
// Malformed history entry, should never happen
continue;
}
// No resource specified means all resources match
if (rsc_id) {
pe_resource_t *fail_rsc = pe_find_resource_with_flags(data_set->resources,
failed_id,
pe_find_renamed|pe_find_anon);
if (!fail_rsc || safe_str_neq(rsc_id, fail_rsc->id)) {
continue;
}
}
// Host name should always have been provided by this point
failed_value = crm_element_value(xml_op, XML_ATTR_UNAME);
if (safe_str_neq(node_name, failed_value)) {
continue;
}
// No operation specified means all operations match
if (operation) {
failed_value = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
if (safe_str_neq(operation, failed_value)) {
continue;
}
// Interval (if operation was specified) defaults to 0 (not all)
failed_value = crm_element_value(xml_op, XML_LRM_ATTR_INTERVAL_MS);
if (safe_str_neq(interval_ms_s, failed_value)) {
continue;
}
}
/* not available until glib 2.32
g_hash_table_add(rscs, (gpointer) failed_id);
*/
g_hash_table_insert(rscs, (gpointer) failed_id, (gpointer) failed_id);
}
g_hash_table_iter_init(&iter, rscs);
while (g_hash_table_iter_next(&iter, (gpointer *) &failed_id, NULL)) {
crm_debug("Erasing failures of %s on %s", failed_id, node_name);
rc = clear_rsc_history(controld_api, node_name, failed_id, data_set);
if (rc != pcmk_rc_ok) {
return pcmk_rc2legacy(rc);
}
}
g_hash_table_destroy(rscs);
return rc;
}
static int
clear_rsc_fail_attrs(pe_resource_t *rsc, const char *operation,
const char *interval_spec, pe_node_t *node)
{
int rc = pcmk_ok;
int attr_options = pcmk__node_attr_none;
char *rsc_name = rsc_fail_name(rsc);
if (pe__is_guest_or_remote_node(node)) {
attr_options |= pcmk__node_attr_remote;
}
rc = pcmk__node_attr_request_clear(NULL, node->details->uname, rsc_name,
operation, interval_spec, NULL,
attr_options);
free(rsc_name);
return rc;
}
int
cli_resource_delete(pcmk_controld_api_t *controld_api, const char *host_uname,
pe_resource_t *rsc, const char *operation,
const char *interval_spec, bool just_failures,
pe_working_set_t *data_set)
{
int rc = pcmk_ok;
pe_node_t *node = NULL;
if (rsc == NULL) {
return -ENXIO;
} else if (rsc->children) {
GListPtr lpc = NULL;
for (lpc = rsc->children; lpc != NULL; lpc = lpc->next) {
pe_resource_t *child = (pe_resource_t *) lpc->data;
rc = cli_resource_delete(controld_api, host_uname, child, operation,
interval_spec, just_failures, data_set);
if (rc != pcmk_ok) {
return rc;
}
}
return pcmk_ok;
} else if (host_uname == NULL) {
GListPtr lpc = NULL;
GListPtr nodes = g_hash_table_get_values(rsc->known_on);
if(nodes == NULL && do_force) {
nodes = node_list_dup(data_set->nodes, FALSE, FALSE);
} else if(nodes == NULL && rsc->exclusive_discover) {
GHashTableIter iter;
pe_node_t *node = NULL;
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void**)&node)) {
if(node->weight >= 0) {
nodes = g_list_prepend(nodes, node);
}
}
} else if(nodes == NULL) {
nodes = g_hash_table_get_values(rsc->allowed_nodes);
}
for (lpc = nodes; lpc != NULL; lpc = lpc->next) {
node = (pe_node_t *) lpc->data;
if (node->details->online) {
rc = cli_resource_delete(controld_api, node->details->uname,
rsc, operation, interval_spec,
just_failures, data_set);
}
if (rc != pcmk_ok) {
g_list_free(nodes);
return rc;
}
}
g_list_free(nodes);
return pcmk_ok;
}
node = pe_find_node(data_set->nodes, host_uname);
if (node == NULL) {
printf("Unable to clean up %s because node %s not found\n",
rsc->id, host_uname);
return -ENODEV;
}
if (!node->details->rsc_discovery_enabled) {
printf("Unable to clean up %s because resource discovery disabled on %s\n",
rsc->id, host_uname);
return -EOPNOTSUPP;
}
if (controld_api == NULL) {
printf("Dry run: skipping clean-up of %s on %s due to CIB_file\n",
rsc->id, host_uname);
return pcmk_ok;
}
rc = clear_rsc_fail_attrs(rsc, operation, interval_spec, node);
if (rc != pcmk_rc_ok) {
printf("Unable to clean up %s failures on %s: %s\n",
rsc->id, host_uname, pcmk_rc_str(rc));
return pcmk_rc2legacy(rc);
}
if (just_failures) {
rc = clear_rsc_failures(controld_api, host_uname, rsc->id, operation,
interval_spec, data_set);
} else {
rc = clear_rsc_history(controld_api, host_uname, rsc->id, data_set);
rc = pcmk_rc2legacy(rc);
}
if (rc != pcmk_ok) {
printf("Cleaned %s failures on %s, but unable to clean history: %s\n",
rsc->id, host_uname, pcmk_strerror(rc));
} else {
printf("Cleaned up %s on %s\n", rsc->id, host_uname);
}
return rc;
}
int
cli_cleanup_all(pcmk_controld_api_t *controld_api, const char *node_name,
const char *operation, const char *interval_spec,
pe_working_set_t *data_set)
{
int rc = pcmk_ok;
int attr_options = pcmk__node_attr_none;
const char *display_name = node_name? node_name : "all nodes";
if (controld_api == NULL) {
printf("Dry run: skipping clean-up of %s due to CIB_file\n",
display_name);
return pcmk_ok;
}
if (node_name) {
pe_node_t *node = pe_find_node(data_set->nodes, node_name);
if (node == NULL) {
CMD_ERR("Unknown node: %s", node_name);
return -ENXIO;
}
if (pe__is_guest_or_remote_node(node)) {
attr_options |= pcmk__node_attr_remote;
}
}
rc = pcmk__node_attr_request_clear(NULL, node_name, NULL, operation,
interval_spec, NULL, attr_options);
if (rc != pcmk_rc_ok) {
printf("Unable to clean up all failures on %s: %s\n",
display_name, pcmk_rc_str(rc));
return pcmk_rc2legacy(rc);
}
if (node_name) {
rc = clear_rsc_failures(controld_api, node_name, NULL,
operation, interval_spec, data_set);
if (rc != pcmk_ok) {
printf("Cleaned all resource failures on %s, but unable to clean history: %s\n",
node_name, pcmk_strerror(rc));
return rc;
}
} else {
for (GList *iter = data_set->nodes; iter; iter = iter->next) {
pe_node_t *node = (pe_node_t *) iter->data;
rc = clear_rsc_failures(controld_api, node->details->uname, NULL,
operation, interval_spec, data_set);
if (rc != pcmk_ok) {
printf("Cleaned all resource failures on all nodes, but unable to clean history: %s\n",
pcmk_strerror(rc));
return rc;
}
}
}
printf("Cleaned up all resources on %s\n", display_name);
return pcmk_ok;
}
void
cli_resource_check(cib_t * cib_conn, pe_resource_t *rsc)
{
bool printed = false;
char *role_s = NULL;
char *managed = NULL;
pe_resource_t *parent = uber_parent(rsc);
find_resource_attr(cib_conn, XML_NVPAIR_ATTR_VALUE, parent->id,
NULL, NULL, NULL, XML_RSC_ATTR_MANAGED, &managed);
find_resource_attr(cib_conn, XML_NVPAIR_ATTR_VALUE, parent->id,
NULL, NULL, NULL, XML_RSC_ATTR_TARGET_ROLE, &role_s);
if(role_s) {
enum rsc_role_e role = text2role(role_s);
free(role_s);
if(role == RSC_ROLE_UNKNOWN) {
// Treated as if unset
} else if(role == RSC_ROLE_STOPPED) {
printf("\n * Configuration specifies '%s' should remain stopped\n",
parent->id);
printed = true;
} else if (is_set(parent->flags, pe_rsc_promotable)
&& (role == RSC_ROLE_SLAVE)) {
printf("\n * Configuration specifies '%s' should not be promoted\n",
parent->id);
printed = true;
}
}
if (managed && !crm_is_true(managed)) {
printf("%s * Configuration prevents cluster from stopping or starting unmanaged '%s'\n",
(printed? "" : "\n"), parent->id);
printed = true;
}
free(managed);
if (rsc->lock_node) {
printf("%s * '%s' is locked to node %s due to shutdown\n",
(printed? "" : "\n"), parent->id, rsc->lock_node->details->uname);
}
if (printed) {
printf("\n");
}
}
// \return Standard Pacemaker return code
int
cli_resource_fail(pcmk_controld_api_t *controld_api, const char *host_uname,
const char *rsc_id, pe_working_set_t *data_set)
{
crm_notice("Failing %s on %s", rsc_id, host_uname);
return send_lrm_rsc_op(controld_api, true, host_uname, rsc_id, data_set);
}
static GHashTable *
generate_resource_params(pe_resource_t * rsc, pe_working_set_t * data_set)
{
GHashTable *params = NULL;
GHashTable *meta = NULL;
GHashTable *combined = NULL;
GHashTableIter iter;
if (!rsc) {
crm_err("Resource does not exist in config");
return NULL;
}
params = crm_str_table_new();
meta = crm_str_table_new();
combined = crm_str_table_new();
get_rsc_attributes(params, rsc, NULL /* TODO: Pass in local node */ , data_set);
get_meta_attributes(meta, rsc, NULL /* TODO: Pass in local node */ , data_set);
if (params) {
char *key = NULL;
char *value = NULL;
g_hash_table_iter_init(&iter, params);
while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) {
g_hash_table_insert(combined, strdup(key), strdup(value));
}
g_hash_table_destroy(params);
}
if (meta) {
char *key = NULL;
char *value = NULL;
g_hash_table_iter_init(&iter, meta);
while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) {
char *crm_name = crm_meta_name(key);
g_hash_table_insert(combined, crm_name, strdup(value));
}
g_hash_table_destroy(meta);
}
return combined;
}
static bool resource_is_running_on(pe_resource_t *rsc, const char *host)
{
bool found = TRUE;
GListPtr hIter = NULL;
GListPtr hosts = NULL;
if(rsc == NULL) {
return FALSE;
}
rsc->fns->location(rsc, &hosts, TRUE);
for (hIter = hosts; host != NULL && hIter != NULL; hIter = hIter->next) {
pe_node_t *node = (pe_node_t *) hIter->data;
if(strcmp(host, node->details->uname) == 0) {
crm_trace("Resource %s is running on %s\n", rsc->id, host);
goto done;
} else if(strcmp(host, node->details->id) == 0) {
crm_trace("Resource %s is running on %s\n", rsc->id, host);
goto done;
}
}
if(host != NULL) {
crm_trace("Resource %s is not running on: %s\n", rsc->id, host);
found = FALSE;
} else if(host == NULL && hosts == NULL) {
crm_trace("Resource %s is not running\n", rsc->id);
found = FALSE;
}
done:
g_list_free(hosts);
return found;
}
/*!
* \internal
* \brief Create a list of all resources active on host from a given list
*
* \param[in] host Name of host to check whether resources are active
* \param[in] rsc_list List of resources to check
*
* \return New list of resources from list that are active on host
*/
static GList *
get_active_resources(const char *host, GList *rsc_list)
{
GList *rIter = NULL;
GList *active = NULL;
for (rIter = rsc_list; rIter != NULL; rIter = rIter->next) {
pe_resource_t *rsc = (pe_resource_t *) rIter->data;
/* Expand groups to their members, because if we're restarting a member
* other than the first, we can't otherwise tell which resources are
* stopping and starting.
*/
if (rsc->variant == pe_group) {
active = g_list_concat(active,
get_active_resources(host, rsc->children));
} else if (resource_is_running_on(rsc, host)) {
active = g_list_append(active, strdup(rsc->id));
}
}
return active;
}
GList*
subtract_lists(GList *from, GList *items, GCompareFunc cmp)
{
GList *item = NULL;
GList *result = g_list_copy(from);
for (item = items; item != NULL; item = item->next) {
GList *candidate = NULL;
for (candidate = from; candidate != NULL; candidate = candidate->next) {
crm_info("Comparing %s with %s", (const char *) candidate->data,
(const char *) item->data);
if(cmp(candidate->data, item->data) == 0) {
result = g_list_remove(result, candidate->data);
break;
}
}
}
return result;
}
static void dump_list(GList *items, const char *tag)
{
int lpc = 0;
GList *item = NULL;
for (item = items; item != NULL; item = item->next) {
crm_trace("%s[%d]: %s", tag, lpc, (char*)item->data);
lpc++;
}
}
static void display_list(GList *items, const char *tag)
{
GList *item = NULL;
for (item = items; item != NULL; item = item->next) {
fprintf(stdout, "%s%s\n", tag, (const char *)item->data);
}
}
/*!
* \internal
* \brief Upgrade XML to latest schema version and use it as working set input
*
* This also updates the working set timestamp to the current time.
*
* \param[in] data_set Working set instance to update
* \param[in] xml XML to use as input
*
* \return pcmk_ok on success, -ENOKEY if unable to upgrade XML
* \note On success, caller is responsible for freeing memory allocated for
* data_set->now.
* \todo This follows the example of other callers of cli_config_update()
* and returns -ENOKEY ("Required key not available") if that fails,
* but perhaps -pcmk_err_schema_validation would be better in that case.
*/
int
update_working_set_xml(pe_working_set_t *data_set, xmlNode **xml)
{
if (cli_config_update(xml, NULL, FALSE) == FALSE) {
return -ENOKEY;
}
data_set->input = *xml;
data_set->now = crm_time_new(NULL);
return pcmk_ok;
}
/*!
* \internal
* \brief Update a working set's XML input based on a CIB query
*
* \param[in] data_set Data set instance to initialize
* \param[in] cib Connection to the CIB manager
*
* \return pcmk_ok on success, -errno on failure
* \note On success, caller is responsible for freeing memory allocated for
* data_set->input and data_set->now.
*/
static int
update_working_set_from_cib(pe_working_set_t * data_set, cib_t *cib)
{
xmlNode *cib_xml_copy = NULL;
int rc;
rc = cib->cmds->query(cib, NULL, &cib_xml_copy, cib_scope_local | cib_sync_call);
if (rc != pcmk_ok) {
fprintf(stderr, "Could not obtain the current CIB: %s (%d)\n", pcmk_strerror(rc), rc);
return rc;
}
rc = update_working_set_xml(data_set, &cib_xml_copy);
if (rc != pcmk_ok) {
fprintf(stderr, "Could not upgrade the current CIB XML\n");
free_xml(cib_xml_copy);
return rc;
}
return pcmk_ok;
}
static int
update_dataset(cib_t *cib, pe_working_set_t * data_set, bool simulate)
{
char *pid = NULL;
char *shadow_file = NULL;
cib_t *shadow_cib = NULL;
int rc;
pe_reset_working_set(data_set);
rc = update_working_set_from_cib(data_set, cib);
if (rc != pcmk_ok) {
return rc;
}
if(simulate) {
pid = pcmk__getpid_s();
shadow_cib = cib_shadow_new(pid);
shadow_file = get_shadow_file(pid);
if (shadow_cib == NULL) {
fprintf(stderr, "Could not create shadow cib: '%s'\n", pid);
rc = -ENXIO;
goto cleanup;
}
rc = write_xml_file(data_set->input, shadow_file, FALSE);
if (rc < 0) {
fprintf(stderr, "Could not populate shadow cib: %s (%d)\n", pcmk_strerror(rc), rc);
goto cleanup;
}
rc = shadow_cib->cmds->signon(shadow_cib, crm_system_name, cib_command);
if(rc != pcmk_ok) {
fprintf(stderr, "Could not connect to shadow cib: %s (%d)\n", pcmk_strerror(rc), rc);
goto cleanup;
}
pcmk__schedule_actions(data_set, data_set->input, NULL);
run_simulation(data_set, shadow_cib, NULL, TRUE);
rc = update_dataset(shadow_cib, data_set, FALSE);
} else {
cluster_status(data_set);
}
cleanup:
/* Do not free data_set->input here, we need rsc->xml to be valid later on */
cib_delete(shadow_cib);
free(pid);
if(shadow_file) {
unlink(shadow_file);
free(shadow_file);
}
return rc;
}
static int
max_delay_for_resource(pe_working_set_t * data_set, pe_resource_t *rsc)
{
int delay = 0;
int max_delay = 0;
if(rsc && rsc->children) {
GList *iter = NULL;
for(iter = rsc->children; iter; iter = iter->next) {
pe_resource_t *child = (pe_resource_t *)iter->data;
delay = max_delay_for_resource(data_set, child);
if(delay > max_delay) {
double seconds = delay / 1000.0;
crm_trace("Calculated new delay of %.1fs due to %s", seconds, child->id);
max_delay = delay;
}
}
} else if(rsc) {
char *key = crm_strdup_printf("%s_%s_0", rsc->id, RSC_STOP);
pe_action_t *stop = custom_action(rsc, key, RSC_STOP, NULL, TRUE, FALSE, data_set);
const char *value = g_hash_table_lookup(stop->meta, XML_ATTR_TIMEOUT);
max_delay = value? (int) crm_parse_ll(value, NULL) : -1;
pe_free_action(stop);
}
return max_delay;
}
static int
max_delay_in(pe_working_set_t * data_set, GList *resources)
{
int max_delay = 0;
GList *item = NULL;
for (item = resources; item != NULL; item = item->next) {
int delay = 0;
pe_resource_t *rsc = pe_find_resource(data_set->resources, (const char *)item->data);
delay = max_delay_for_resource(data_set, rsc);
if(delay > max_delay) {
double seconds = delay / 1000.0;
crm_trace("Calculated new delay of %.1fs due to %s", seconds, rsc->id);
max_delay = delay;
}
}
return 5 + (max_delay / 1000);
}
#define waiting_for_starts(d, r, h) ((d != NULL) || \
(resource_is_running_on((r), (h)) == FALSE))
/*!
* \internal
* \brief Restart a resource (on a particular host if requested).
*
* \param[in] rsc The resource to restart
* \param[in] host The host to restart the resource on (or NULL for all)
* \param[in] timeout_ms Consider failed if actions do not complete in this time
* (specified in milliseconds, but a two-second
* granularity is actually used; if 0, a timeout will be
* calculated based on the resource timeout)
* \param[in] cib Connection to the CIB manager
*
* \return pcmk_ok on success, -errno on failure (exits on certain failures)
*/
int
cli_resource_restart(pe_resource_t *rsc, const char *host, int timeout_ms,
cib_t *cib)
{
int rc = 0;
int lpc = 0;
int before = 0;
int step_timeout_s = 0;
int sleep_interval = 2;
int timeout = timeout_ms / 1000;
bool stop_via_ban = FALSE;
char *rsc_id = NULL;
char *orig_target_role = NULL;
GList *list_delta = NULL;
GList *target_active = NULL;
GList *current_active = NULL;
GList *restart_target_active = NULL;
pe_working_set_t *data_set = NULL;
if(resource_is_running_on(rsc, host) == FALSE) {
const char *id = rsc->clone_name?rsc->clone_name:rsc->id;
if(host) {
printf("%s is not running on %s and so cannot be restarted\n", id, host);
} else {
printf("%s is not running anywhere and so cannot be restarted\n", id);
}
return -ENXIO;
}
/* We might set the target-role meta-attribute */
attr_set_type = XML_TAG_META_SETS;
rsc_id = strdup(rsc->id);
if ((pe_rsc_is_clone(rsc) || pe_bundle_replicas(rsc)) && host) {
stop_via_ban = TRUE;
}
/*
grab full cib
determine originally active resources
disable or ban
poll cib and watch for affected resources to get stopped
without --timeout, calculate the stop timeout for each step and wait for that
if we hit --timeout or the service timeout, re-enable or un-ban, report failure and indicate which resources we couldn't take down
if everything stopped, re-enable or un-ban
poll cib and watch for affected resources to get started
without --timeout, calculate the start timeout for each step and wait for that
if we hit --timeout or the service timeout, report (different) failure and indicate which resources we couldn't bring back up
report success
Optimizations:
- use constraints to determine ordered list of affected resources
- Allow a --no-deps option (aka. --force-restart)
*/
data_set = pe_new_working_set();
if (data_set == NULL) {
crm_perror(LOG_ERR, "Could not allocate working set");
rc = -ENOMEM;
goto done;
}
set_bit(data_set->flags, pe_flag_no_counts);
set_bit(data_set->flags, pe_flag_no_compat);
rc = update_dataset(cib, data_set, FALSE);
if(rc != pcmk_ok) {
fprintf(stdout, "Could not get new resource list: %s (%d)\n", pcmk_strerror(rc), rc);
goto done;
}
restart_target_active = get_active_resources(host, data_set->resources);
current_active = get_active_resources(host, data_set->resources);
dump_list(current_active, "Origin");
if (stop_via_ban) {
/* Stop the clone or bundle instance by banning it from the host */
BE_QUIET = TRUE;
rc = cli_resource_ban(rsc_id, host, NULL, cib);
} else {
/* Stop the resource by setting target-role to Stopped.
* Remember any existing target-role so we can restore it later
* (though it only makes any difference if it's Slave).
*/
char *lookup_id = clone_strip(rsc->id);
find_resource_attr(cib, XML_NVPAIR_ATTR_VALUE, lookup_id, NULL, NULL,
NULL, XML_RSC_ATTR_TARGET_ROLE, &orig_target_role);
free(lookup_id);
rc = cli_resource_update_attribute(rsc, rsc_id, NULL, NULL,
XML_RSC_ATTR_TARGET_ROLE,
RSC_STOPPED, FALSE, cib, data_set);
}
if(rc != pcmk_ok) {
fprintf(stderr, "Could not set target-role for %s: %s (%d)\n", rsc_id, pcmk_strerror(rc), rc);
if (current_active) {
g_list_free_full(current_active, free);
}
if (restart_target_active) {
g_list_free_full(restart_target_active, free);
}
goto done;
}
rc = update_dataset(cib, data_set, TRUE);
if(rc != pcmk_ok) {
fprintf(stderr, "Could not determine which resources would be stopped\n");
goto failure;
}
target_active = get_active_resources(host, data_set->resources);
dump_list(target_active, "Target");
list_delta = subtract_lists(current_active, target_active, (GCompareFunc) strcmp);
fprintf(stdout, "Waiting for %d resources to stop:\n", g_list_length(list_delta));
display_list(list_delta, " * ");
step_timeout_s = timeout / sleep_interval;
while (list_delta != NULL) {
before = g_list_length(list_delta);
if(timeout_ms == 0) {
step_timeout_s = max_delay_in(data_set, list_delta) / sleep_interval;
}
/* We probably don't need the entire step timeout */
for(lpc = 0; (lpc < step_timeout_s) && (list_delta != NULL); lpc++) {
sleep(sleep_interval);
if(timeout) {
timeout -= sleep_interval;
crm_trace("%ds remaining", timeout);
}
rc = update_dataset(cib, data_set, FALSE);
if(rc != pcmk_ok) {
fprintf(stderr, "Could not determine which resources were stopped\n");
goto failure;
}
if (current_active) {
g_list_free_full(current_active, free);
}
current_active = get_active_resources(host, data_set->resources);
g_list_free(list_delta);
list_delta = subtract_lists(current_active, target_active, (GCompareFunc) strcmp);
dump_list(current_active, "Current");
dump_list(list_delta, "Delta");
}
crm_trace("%d (was %d) resources remaining", g_list_length(list_delta), before);
if(before == g_list_length(list_delta)) {
/* aborted during stop phase, print the contents of list_delta */
fprintf(stderr, "Could not complete shutdown of %s, %d resources remaining\n", rsc_id, g_list_length(list_delta));
display_list(list_delta, " * ");
rc = -ETIME;
goto failure;
}
}
if (stop_via_ban) {
rc = cli_resource_clear(rsc_id, host, NULL, cib, TRUE);
} else if (orig_target_role) {
rc = cli_resource_update_attribute(rsc, rsc_id, NULL, NULL,
XML_RSC_ATTR_TARGET_ROLE,
orig_target_role, FALSE, cib,
data_set);
free(orig_target_role);
orig_target_role = NULL;
} else {
rc = cli_resource_delete_attribute(rsc, rsc_id, NULL, NULL,
XML_RSC_ATTR_TARGET_ROLE, cib,
data_set);
}
if(rc != pcmk_ok) {
fprintf(stderr, "Could not unset target-role for %s: %s (%d)\n", rsc_id, pcmk_strerror(rc), rc);
goto done;
}
if (target_active) {
g_list_free_full(target_active, free);
}
target_active = restart_target_active;
list_delta = subtract_lists(target_active, current_active, (GCompareFunc) strcmp);
fprintf(stdout, "Waiting for %d resources to start again:\n", g_list_length(list_delta));
display_list(list_delta, " * ");
step_timeout_s = timeout / sleep_interval;
while (waiting_for_starts(list_delta, rsc, host)) {
before = g_list_length(list_delta);
if(timeout_ms == 0) {
step_timeout_s = max_delay_in(data_set, list_delta) / sleep_interval;
}
/* We probably don't need the entire step timeout */
for (lpc = 0; (lpc < step_timeout_s) && waiting_for_starts(list_delta, rsc, host); lpc++) {
sleep(sleep_interval);
if(timeout) {
timeout -= sleep_interval;
crm_trace("%ds remaining", timeout);
}
rc = update_dataset(cib, data_set, FALSE);
if(rc != pcmk_ok) {
fprintf(stderr, "Could not determine which resources were started\n");
goto failure;
}
if (current_active) {
g_list_free_full(current_active, free);
}
/* It's OK if dependent resources moved to a different node,
* so we check active resources on all nodes.
*/
current_active = get_active_resources(NULL, data_set->resources);
g_list_free(list_delta);
list_delta = subtract_lists(target_active, current_active, (GCompareFunc) strcmp);
dump_list(current_active, "Current");
dump_list(list_delta, "Delta");
}
if(before == g_list_length(list_delta)) {
/* aborted during start phase, print the contents of list_delta */
fprintf(stdout, "Could not complete restart of %s, %d resources remaining\n", rsc_id, g_list_length(list_delta));
display_list(list_delta, " * ");
rc = -ETIME;
goto failure;
}
}
rc = pcmk_ok;
goto done;
failure:
if (stop_via_ban) {
cli_resource_clear(rsc_id, host, NULL, cib, TRUE);
} else if (orig_target_role) {
cli_resource_update_attribute(rsc, rsc_id, NULL, NULL,
XML_RSC_ATTR_TARGET_ROLE,
orig_target_role, FALSE, cib, data_set);
free(orig_target_role);
} else {
cli_resource_delete_attribute(rsc, rsc_id, NULL, NULL,
XML_RSC_ATTR_TARGET_ROLE, cib, data_set);
}
done:
if (list_delta) {
g_list_free(list_delta);
}
if (current_active) {
g_list_free_full(current_active, free);
}
if (target_active && (target_active != restart_target_active)) {
g_list_free_full(target_active, free);
}
if (restart_target_active) {
g_list_free_full(restart_target_active, free);
}
free(rsc_id);
pe_free_working_set(data_set);
return rc;
}
static inline int action_is_pending(pe_action_t *action)
{
if(is_set(action->flags, pe_action_optional)) {
return FALSE;
} else if(is_set(action->flags, pe_action_runnable) == FALSE) {
return FALSE;
} else if(is_set(action->flags, pe_action_pseudo)) {
return FALSE;
} else if(safe_str_eq("notify", action->task)) {
return FALSE;
}
return TRUE;
}
/*!
* \internal
* \brief Return TRUE if any actions in a list are pending
*
* \param[in] actions List of actions to check
*
* \return TRUE if any actions in the list are pending, FALSE otherwise
*/
static bool
actions_are_pending(GListPtr actions)
{
GListPtr action;
for (action = actions; action != NULL; action = action->next) {
pe_action_t *a = (pe_action_t *)action->data;
if (action_is_pending(a)) {
crm_notice("Waiting for %s (flags=0x%.8x)", a->uuid, a->flags);
return TRUE;
}
}
return FALSE;
}
/*!
* \internal
* \brief Print pending actions to stderr
*
* \param[in] actions List of actions to check
*
* \return void
*/
static void
print_pending_actions(GListPtr actions)
{
GListPtr action;
fprintf(stderr, "Pending actions:\n");
for (action = actions; action != NULL; action = action->next) {
pe_action_t *a = (pe_action_t *) action->data;
if (action_is_pending(a)) {
fprintf(stderr, "\tAction %d: %s", a->id, a->uuid);
if (a->node) {
fprintf(stderr, "\ton %s", a->node->details->uname);
}
fprintf(stderr, "\n");
}
}
}
/* For --wait, timeout (in seconds) to use if caller doesn't specify one */
#define WAIT_DEFAULT_TIMEOUT_S (60 * 60)
/* For --wait, how long to sleep between cluster state checks */
#define WAIT_SLEEP_S (2)
/*!
* \internal
* \brief Wait until all pending cluster actions are complete
*
* This waits until either the CIB's transition graph is idle or a timeout is
* reached.
*
* \param[in] timeout_ms Consider failed if actions do not complete in this time
* (specified in milliseconds, but one-second granularity
* is actually used; if 0, a default will be used)
* \param[in] cib Connection to the CIB manager
*
* \return pcmk_ok on success, -errno on failure
*/
int
wait_till_stable(int timeout_ms, cib_t * cib)
{
pe_working_set_t *data_set = NULL;
int rc = -1;
int timeout_s = timeout_ms? ((timeout_ms + 999) / 1000) : WAIT_DEFAULT_TIMEOUT_S;
time_t expire_time = time(NULL) + timeout_s;
time_t time_diff;
bool printed_version_warning = BE_QUIET; // i.e. don't print if quiet
data_set = pe_new_working_set();
if (data_set == NULL) {
return -ENOMEM;
}
set_bit(data_set->flags, pe_flag_no_counts);
set_bit(data_set->flags, pe_flag_no_compat);
do {
/* Abort if timeout is reached */
time_diff = expire_time - time(NULL);
if (time_diff > 0) {
crm_info("Waiting up to %ld seconds for cluster actions to complete", time_diff);
} else {
print_pending_actions(data_set->actions);
pe_free_working_set(data_set);
return -ETIME;
}
if (rc == pcmk_ok) { /* this avoids sleep on first loop iteration */
sleep(WAIT_SLEEP_S);
}
/* Get latest transition graph */
pe_reset_working_set(data_set);
rc = update_working_set_from_cib(data_set, cib);
if (rc != pcmk_ok) {
pe_free_working_set(data_set);
return rc;
}
pcmk__schedule_actions(data_set, data_set->input, NULL);
if (!printed_version_warning) {
/* If the DC has a different version than the local node, the two
* could come to different conclusions about what actions need to be
* done. Warn the user in this case.
*
* @TODO A possible long-term solution would be to reimplement the
* wait as a new controller operation that would be forwarded to the
* DC. However, that would have potential problems of its own.
*/
const char *dc_version = g_hash_table_lookup(data_set->config_hash,
"dc-version");
if (safe_str_neq(dc_version, PACEMAKER_VERSION "-" BUILD_VERSION)) {
printf("warning: wait option may not work properly in "
"mixed-version cluster\n");
printed_version_warning = TRUE;
}
}
} while (actions_are_pending(data_set->actions));
pe_free_working_set(data_set);
return pcmk_ok;
}
int
cli_resource_execute_from_params(const char *rsc_name, const char *rsc_class,
const char *rsc_prov, const char *rsc_type,
const char *action, GHashTable *params,
GHashTable *override_hash, int timeout_ms)
{
GHashTable *params_copy = NULL;
int rc = pcmk_ok;
svc_action_t *op = NULL;
if (safe_str_eq(rsc_class, PCMK_RESOURCE_CLASS_STONITH)) {
CMD_ERR("Sorry, the %s option doesn't support %s resources yet",
action, rsc_class);
crm_exit(CRM_EX_UNIMPLEMENT_FEATURE);
}
/* If no timeout was provided, grab the default. */
if (timeout_ms == 0) {
timeout_ms = crm_get_msec(CRM_DEFAULT_OP_TIMEOUT_S);
}
/* add meta_timeout env needed by some resource agents */
g_hash_table_insert(params, strdup("CRM_meta_timeout"),
crm_strdup_printf("%d", timeout_ms));
/* add crm_feature_set env needed by some resource agents */
g_hash_table_insert(params, strdup(XML_ATTR_CRM_VERSION), strdup(CRM_FEATURE_SET));
/* resources_action_create frees the params hash table it's passed, but we
* may need to reuse it in a second call to resources_action_create. Thus
* we'll make a copy here so that gets freed and the original remains for
* reuse.
*/
params_copy = crm_str_table_dup(params);
op = resources_action_create(rsc_name, rsc_class, rsc_prov, rsc_type, action, 0,
timeout_ms, params_copy, 0);
if (op == NULL) {
/* Re-run with stderr enabled so we can display a sane error message */
crm_enable_stderr(TRUE);
params_copy = crm_str_table_dup(params);
op = resources_action_create(rsc_name, rsc_class, rsc_prov, rsc_type, action, 0,
timeout_ms, params_copy, 0);
/* Callers of cli_resource_execute expect that the params hash table will
* be freed. That function uses this one, so for that reason and for
* making the two act the same, we should free the hash table here too.
*/
g_hash_table_destroy(params);
/* We know op will be NULL, but this makes static analysis happy */
services_action_free(op);
crm_exit(CRM_EX_DATAERR);
+ return rc; // Never reached, but helps static analysis
}
setenv("HA_debug", resource_verbose > 0 ? "1" : "0", 1);
if(resource_verbose > 1) {
setenv("OCF_TRACE_RA", "1", 1);
}
/* A resource agent using the standard ocf-shellfuncs library will not print
* messages to stderr if it doesn't have a controlling terminal (e.g. if
* crm_resource is called via script or ssh). This forces it to do so.
*/
setenv("OCF_TRACE_FILE", "/dev/stderr", 0);
if (override_hash) {
GHashTableIter iter;
char *name = NULL;
char *value = NULL;
g_hash_table_iter_init(&iter, override_hash);
while (g_hash_table_iter_next(&iter, (gpointer *) & name, (gpointer *) & value)) {
printf("Overriding the cluster configuration for '%s' with '%s' = '%s'\n",
rsc_name, name, value);
g_hash_table_replace(op->params, strdup(name), strdup(value));
}
}
if (services_action_sync(op)) {
int more, lpc, last;
char *local_copy = NULL;
rc = op->rc;
if (op->status == PCMK_LRM_OP_DONE) {
printf("Operation %s for %s (%s:%s:%s) returned: '%s' (%d)\n",
action, rsc_name, rsc_class, rsc_prov ? rsc_prov : "", rsc_type,
services_ocf_exitcode_str(op->rc), op->rc);
} else {
printf("Operation %s for %s (%s:%s:%s) failed: '%s' (%d)\n",
action, rsc_name, rsc_class, rsc_prov ? rsc_prov : "", rsc_type,
services_lrm_status_str(op->status), op->status);
}
/* hide output for validate-all if not in verbose */
if (resource_verbose == 0 && safe_str_eq(action, "validate-all"))
goto done;
if (op->stdout_data) {
local_copy = strdup(op->stdout_data);
more = strlen(local_copy);
last = 0;
for (lpc = 0; lpc < more; lpc++) {
if (local_copy[lpc] == '\n' || local_copy[lpc] == 0) {
local_copy[lpc] = 0;
printf(" > stdout: %s\n", local_copy + last);
last = lpc + 1;
}
}
free(local_copy);
}
if (op->stderr_data) {
local_copy = strdup(op->stderr_data);
more = strlen(local_copy);
last = 0;
for (lpc = 0; lpc < more; lpc++) {
if (local_copy[lpc] == '\n' || local_copy[lpc] == 0) {
local_copy[lpc] = 0;
printf(" > stderr: %s\n", local_copy + last);
last = lpc + 1;
}
}
free(local_copy);
}
} else {
rc = op->rc == 0 ? pcmk_err_generic : op->rc;
}
done:
services_action_free(op);
/* See comment above about why we free params here. */
g_hash_table_destroy(params);
return rc;
}
int
cli_resource_execute(pe_resource_t *rsc, const char *requested_name,
const char *rsc_action, GHashTable *override_hash,
int timeout_ms, cib_t * cib, pe_working_set_t *data_set)
{
int rc = pcmk_ok;
const char *rid = NULL;
const char *rtype = NULL;
const char *rprov = NULL;
const char *rclass = NULL;
const char *action = NULL;
GHashTable *params = NULL;
if (safe_str_eq(rsc_action, "validate")) {
action = "validate-all";
} else if (safe_str_eq(rsc_action, "force-check")) {
action = "monitor";
} else if (safe_str_eq(rsc_action, "force-stop")) {
action = rsc_action+6;
} else if (safe_str_eq(rsc_action, "force-start")
|| safe_str_eq(rsc_action, "force-demote")
|| safe_str_eq(rsc_action, "force-promote")) {
action = rsc_action+6;
if(pe_rsc_is_clone(rsc)) {
rc = cli_resource_search(rsc, requested_name, data_set);
if(rc > 0 && do_force == FALSE) {
CMD_ERR("It is not safe to %s %s here: the cluster claims it is already active",
action, rsc->id);
CMD_ERR("Try setting target-role=Stopped first or specifying "
"the force option");
crm_exit(CRM_EX_UNSAFE);
}
}
} else {
action = rsc_action;
}
if(pe_rsc_is_clone(rsc)) {
/* Grab the first child resource in the hope it's not a group */
rsc = rsc->children->data;
}
if(rsc->variant == pe_group) {
CMD_ERR("Sorry, the %s option doesn't support group resources",
rsc_action);
crm_exit(CRM_EX_UNIMPLEMENT_FEATURE);
}
rclass = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS);
rprov = crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER);
rtype = crm_element_value(rsc->xml, XML_ATTR_TYPE);
params = generate_resource_params(rsc, data_set);
if (timeout_ms == 0) {
timeout_ms = pe_get_configured_timeout(rsc, action, data_set);
}
rid = pe_rsc_is_anon_clone(rsc->parent)? requested_name : rsc->id;
rc = cli_resource_execute_from_params(rid, rclass, rprov, rtype, action,
params, override_hash, timeout_ms);
return rc;
}
int
cli_resource_move(pe_resource_t *rsc, const char *rsc_id, const char *host_name,
cib_t *cib, pe_working_set_t *data_set)
{
int rc = pcmk_ok;
unsigned int count = 0;
pe_node_t *current = NULL;
pe_node_t *dest = pe_find_node(data_set->nodes, host_name);
bool cur_is_dest = FALSE;
if (dest == NULL) {
return -pcmk_err_node_unknown;
}
if (scope_master && is_not_set(rsc->flags, pe_rsc_promotable)) {
pe_resource_t *p = uber_parent(rsc);
if (is_set(p->flags, pe_rsc_promotable)) {
CMD_ERR("Using parent '%s' for move instead of '%s'.", rsc->id, rsc_id);
rsc_id = p->id;
rsc = p;
} else {
CMD_ERR("Ignoring master option: %s is not promotable", rsc_id);
scope_master = FALSE;
}
}
current = pe__find_active_requires(rsc, &count);
if (is_set(rsc->flags, pe_rsc_promotable)) {
GListPtr iter = NULL;
unsigned int master_count = 0;
pe_node_t *master_node = NULL;
for(iter = rsc->children; iter; iter = iter->next) {
pe_resource_t *child = (pe_resource_t *)iter->data;
enum rsc_role_e child_role = child->fns->state(child, TRUE);
if(child_role == RSC_ROLE_MASTER) {
rsc = child;
master_node = pe__current_node(child);
master_count++;
}
}
if (scope_master || master_count) {
count = master_count;
current = master_node;
}
}
if (count > 1) {
if (pe_rsc_is_clone(rsc)) {
current = NULL;
} else {
return -pcmk_err_multiple;
}
}
if (current && (current->details == dest->details)) {
cur_is_dest = TRUE;
if (do_force) {
crm_info("%s is already %s on %s, reinforcing placement with location constraint.",
rsc_id, scope_master?"promoted":"active", dest->details->uname);
} else {
return -pcmk_err_already;
}
}
/* Clear any previous prefer constraints across all nodes. */
cli_resource_clear(rsc_id, NULL, data_set->nodes, cib, FALSE);
/* Clear any previous ban constraints on 'dest'. */
cli_resource_clear(rsc_id, dest->details->uname, data_set->nodes, cib, TRUE);
/* Record an explicit preference for 'dest' */
rc = cli_resource_prefer(rsc_id, dest->details->uname, cib);
crm_trace("%s%s now prefers node %s%s",
rsc->id, scope_master?" (master)":"", dest->details->uname, do_force?"(forced)":"");
/* only ban the previous location if current location != destination location.
* it is possible to use -M to enforce a location without regard of where the
* resource is currently located */
if(do_force && (cur_is_dest == FALSE)) {
/* Ban the original location if possible */
if(current) {
(void)cli_resource_ban(rsc_id, current->details->uname, NULL, cib);
} else if(count > 1) {
CMD_ERR("Resource '%s' is currently %s in %d locations. "
"One may now move to %s",
rsc_id, (scope_master? "promoted" : "active"),
count, dest->details->uname);
CMD_ERR("To prevent '%s' from being %s at a specific location, "
"specify a node.",
rsc_id, (scope_master? "promoted" : "active"));
} else {
crm_trace("Not banning %s from its current location: not active", rsc_id);
}
}
return rc;
}
static void
cli_resource_why_without_rsc_and_host(cib_t *cib_conn,GListPtr resources)
{
GListPtr lpc = NULL;
GListPtr hosts = NULL;
for (lpc = resources; lpc != NULL; lpc = lpc->next) {
pe_resource_t *rsc = (pe_resource_t *) lpc->data;
rsc->fns->location(rsc, &hosts, TRUE);
if (hosts == NULL) {
printf("Resource %s is not running\n", rsc->id);
} else {
printf("Resource %s is running\n", rsc->id);
}
cli_resource_check(cib_conn, rsc);
g_list_free(hosts);
hosts = NULL;
}
}
static void
cli_resource_why_with_rsc_and_host(cib_t *cib_conn, GListPtr resources,
pe_resource_t *rsc, const char *host_uname)
{
if (resource_is_running_on(rsc, host_uname)) {
printf("Resource %s is running on host %s\n",rsc->id,host_uname);
} else {
printf("Resource %s is not running on host %s\n", rsc->id, host_uname);
}
cli_resource_check(cib_conn, rsc);
}
static void
cli_resource_why_without_rsc_with_host(cib_t *cib_conn,GListPtr resources,pe_node_t *node)
{
const char* host_uname = node->details->uname;
GListPtr allResources = node->details->allocated_rsc;
GListPtr activeResources = node->details->running_rsc;
GListPtr unactiveResources = subtract_lists(allResources,activeResources,(GCompareFunc) strcmp);
GListPtr lpc = NULL;
for (lpc = activeResources; lpc != NULL; lpc = lpc->next) {
pe_resource_t *rsc = (pe_resource_t *) lpc->data;
printf("Resource %s is running on host %s\n",rsc->id,host_uname);
cli_resource_check(cib_conn,rsc);
}
for(lpc = unactiveResources; lpc != NULL; lpc = lpc->next) {
pe_resource_t *rsc = (pe_resource_t *) lpc->data;
printf("Resource %s is assigned to host %s but not running\n",
rsc->id, host_uname);
cli_resource_check(cib_conn,rsc);
}
g_list_free(allResources);
g_list_free(activeResources);
g_list_free(unactiveResources);
}
static void
cli_resource_why_with_rsc_without_host(cib_t *cib_conn, GListPtr resources,
pe_resource_t *rsc)
{
GListPtr hosts = NULL;
rsc->fns->location(rsc, &hosts, TRUE);
printf("Resource %s is %srunning\n", rsc->id, (hosts? "" : "not "));
cli_resource_check(cib_conn, rsc);
g_list_free(hosts);
}
void cli_resource_why(cib_t *cib_conn, GListPtr resources, pe_resource_t *rsc,
pe_node_t *node)
{
const char *host_uname = (node == NULL)? NULL : node->details->uname;
if ((rsc == NULL) && (host_uname == NULL)) {
cli_resource_why_without_rsc_and_host(cib_conn, resources);
} else if ((rsc != NULL) && (host_uname != NULL)) {
cli_resource_why_with_rsc_and_host(cib_conn, resources, rsc,
host_uname);
} else if ((rsc == NULL) && (host_uname != NULL)) {
cli_resource_why_without_rsc_with_host(cib_conn, resources, node);
} else if ((rsc != NULL) && (host_uname == NULL)) {
cli_resource_why_with_rsc_without_host(cib_conn, resources, rsc);
}
}

File Metadata

Mime Type
text/x-diff
Expires
Tue, Jul 8, 6:32 PM (11 h, 48 m)
Storage Engine
blob
Storage Format
Raw Data
Storage Handle
2002679
Default Alt Text
(1006 KB)

Event Timeline