diff --git a/.travis.yml b/.travis.yml index 61621baf81..3198a18263 100644 --- a/.travis.yml +++ b/.travis.yml @@ -1,143 +1,143 @@ # Copyright 2012-2019 the Pacemaker project contributors # # The version control history for this file may have further details. # Control file for the Travis autobuilder # https://docs.travis-ci.com/user/customizing-the-build/ language: c matrix: include: - compiler: gcc env: MAINT_EXTRA=0 - compiler: clang env: MAINT_EXTRA=0 #- compiler: cov-build # env: MAINT_EXTRA=0 - compiler: gcc env: MAINT_EXTRA=1 env: global: # -- BEGIN Coverity Scan ENV # Used by https://scan.coverity.com/scripts/travisci_build_coverity_scan.sh # The build command with all of the arguments that you would apply to a manual `cov-build` - COVERITY_SCAN_BUILD_COMMAND="make" # Email address for notifications related to this build - - OWNER_EMAIL="andrew@beekhof.net" + - OWNER_EMAIL="admin@clusterlabs.org" # Regular expression selects on which branches to run analysis # Be aware of quotas. Do not run on every branch/commit - COVERITY_SCAN_BRANCH_PATTERN="1.1" # COVERITY_SCAN_TOKEN via "travis encrypt" using the repo's public key - secure: "qnrF7L8RejLUY7URdNe7XP4Hu4R55C0tvAuMRg4EjVtelOpw+nIgA7BLiX19q/70VjFuKcGnMhW28TdYl0uwMMdWKKxmwTim04Sy3UfOE2BPeuQOBphr+8s9gd0U1MO8j2dZ84A40t5Mkk946wWZwT0okpjOr/PfBOZkU3o87FM=" # -- END Coverity Scan ENV cache: directories: - xml/.relaxng.org # sudo add-apt-repository ppa:hotot-team before_install: - sudo add-apt-repository "deb http://archive.ubuntu.com/ubuntu/ trusty main" - sudo apt-get update -qq # To switch to Travis-CI's containerized (non-sudo) architecture, # all our dependencies need to be on Travis's whitelist: # https://github.com/travis-ci/apt-package-whitelist # # The only ones that aren't already are: # - cluster-glue-dev: see open issue: # https://github.com/travis-ci/apt-package-whitelist/issues/2936 # - resource-agents: see open issue: # https://github.com/travis-ci/apt-package-whitelist/issues/4261 # - libdbus-1-dev: see multiple open issues: # https://github.com/travis-ci/apt-package-whitelist/issues?utf8=%E2%9C%93&q=is%3Aissue+libdbus+-1-dev # (a workaround is to install libdbus-glib-1-dev, which depends on it and is whitelisted) install: - test $MAINT_EXTRA -ne 0 || sudo apt-get install -qq automake autoconf libtool python python-dev libbz2-dev libdbus-1-dev libglib2.0-dev libgnutls-dev libltdl-dev libncurses5-dev libpam0g-dev libxml2-dev libxslt1-dev uuid-dev libqb-dev libcfg-dev libcmap-dev libcorosync-common-dev libcpg-dev libquorum-dev libsam-dev libtotem-pg-dev libvotequorum-dev cluster-glue-dev resource-agents - test $MAINT_EXTRA -eq 0 || sudo apt-get install -qq make libxml2-utils xsltproc before_script: # Save and restore CC so that ./configure can pass; # some tests (e.g. cts-exec-helper) require actual system-wide credentials - test $MAINT_EXTRA -ne 0 || { export CC_SAVED=$CC; export CC=$(echo ${CC} | sed s/cov-build/gcc/); sed -e 's|^\(CRM_DAEMON_USER=\).*$|\1nobody|' -e 's|^\(CRM_DAEMON_GROUP=\).*$|\1nobody|' -i -- configure.ac; ./autogen.sh; ./configure; export CC=$CC_SAVED; } - test $MAINT_EXTRA -eq 0 || ( cd xml; { cat Makefile.am; printf 'hack_rng\x3a %s' '${RNG_generated}'; } | make -f- top_srcdir=$(pwd)/.. top_builddir=$(pwd)/.. hack_rng ) script: # XXX Is the following item needed? - test $MAINT_EXTRA -ne 0 || test ${CC} = cov-build || sudo make install-exec-local || true - test $MAINT_EXTRA -ne 0 || test ${CC} = cov-build || { make && ./cts/cts-regression -V cli scheduler exec; } - test $MAINT_EXTRA -ne 0 || test ${CC} != cov-build || export CC=gcc; bash ./maint/travisci_build_coverity_scan.sh - test $MAINT_EXTRA -eq 0 || { { echo 'looking for presence of control characters...'; { git ls-files | xargs grep -Ensv "^([^[:cntrl:]]*|$(printf '\t'))*$"||:; } 2>/dev/null | { ! grep -Ev '^Binary file' && echo 'ALL OK'; }; } && ( cd xml; ./regression.sh && ./regression.sh -B && ./regression.sh -S && { schemas=; for schema in *.rng; do case ${schema} in *cibtr*);; *)schemas="${schemas} ${schema}";; esac; done; test -s .relaxng.org/relaxng.rng 2>/dev/null || curl --create-dirs -SsLo .relaxng.org/relaxng.rng 'https://raw.githubusercontent.com/relaxng/relaxng.org/master/relaxng.rng'; xmllint --noout --relaxng .relaxng.org/relaxng.rng ${schemas}; } ); } #after_script: #after_success: after_failure: - lsb_release -a - sudo cat /etc/apt/sources.list - whoami - env | sort - cat include/config.h notifications: irc: "irc.freenode.org#pcmk" # email: # recipients: # - developers@clusterlabs.org # whitelist branches: only: - master - "1.1" - "2.0" diff --git a/ChangeLog b/ChangeLog index 78b4fed778..934e4b39eb 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,2749 +1,2764 @@ +* Thu May 30 2019 Ken Gaillot Pacemaker-2.0.2-rc3 +- Changesets: 8 +- Diff: 5 files changed, 36 insertions(+), 11 deletions(-) + +- Features added since Pacemaker-2.0.2-rc2 + +- Changes since Pacemaker-2.0.2-rc2 + + libstonithd: avoid use-after-free when retrying a failed action + + libcrmcommon: export logfile environment variable if using default + +* Mon May 20 2019 Ken Gaillot Pacemaker-2.0.2-rc2 +- Changesets: 19 +- Diff: 11 files changed, 63 insertions(+), 37 deletions(-) + +- Changes since Pacemaker-2.0.2-rc1 + + controller: avoid memory leak when duplicate monitor is scheduled + + fencing: avoid use-after-free on early failure return + + libcrmcommon: avoid segfault when beginning formatted text list + * Wed Apr 24 2019 Ken Gaillot Pacemaker-2.0.2-rc1 - Changesets: 258 - Diff: 225 files changed, 28421 insertions(+), 24443 deletions(-) - Features added since Pacemaker-2.0.1 + tools: crm_resource --validate can get resource parameters from command line + tools: crm_resource --clear prints out any cleared constraints + tools: new crm_rule tool for checking rule expiration (experimental) + tools: stonith_admin supports XML output for machine parsing (experimental) + resources: new HealthIOWait resource agent for node health tracking - Changes since Pacemaker-2.0.1 + Important security fixes for CVE-2018-16878, CVE-2018-16877, CVE-2019-3885 + build: crm_report bug report URL is now configurable at build time + build: private libpengine/libtransitioner libraries combined as libpacemaker - + attrd: prevent race when writer leaves where new writer is not elected - + controller: prevent race when DC leaves where new DC is not elected + scheduler: respect order constraints when resources are being probed - + scheduler: don't disable waiting for DC fencing - + scheduler: don't order non-DC shutdowns before DC fencing + scheduler: one group stop shouldn't make another required + libcrmcommon: handle out-of-range integers in configuration better - + libcrmservice: cancel in-flight systemd/upstart DBus calls immediately + libcrmservice: fix use-after-free memory error in alert handling + + libstonithd: handle more than 64KB output from fence agents * Mon Mar 4 2019 Ken Gaillot Pacemaker-2.0.1-1 - Changesets: 592 173 files changed, 9268 insertions(+), 5344 deletions(-) - Features added since Pacemaker-2.0.0 + Pacemaker bundles support podman for container management + fencing: SBD may be used in a cluster that has guest nodes or bundles + fencing: fencing history is synchronized among all nodes + fencing: stonith_admin has option to clear fence history + tools: crm_mon can show fencing action failures and history + tools: crm_resource --clear supports new --expired option + Pacemaker Remote: new options to restrict TLS Diffie-Hellman prime length - Changes since Pacemaker-2.0.0 + scheduler: clone notifications could be scheduled for a stopped Pacemaker Remote node and block all further cluster actions (regression since 2.0.0) + libcrmcommon: correct behavior for completing interrupted live migrations (regression since 2.0.0) + tools: crm_resource -C could fail to clean up all failures in one run (regression since 2.0.0) + Pacemaker Remote: avoid unnecessary downtime when moving resource to Pacemaker Remote node that fails to come up (regression since 1.1.18) + tools: restore stonith_admin ability to confirm unseen nodes are down (regression since 1.1.12) + build: minor logging fixes to allow compatibility with GCC 9 -Werror + build: spec file now puts XML schemas in new pacemaker-schemas package + build: spec file now provides virtual pcmk-cluster-manager package + pacemaker-attrd: wait a short time before re-attempting failed writes + pacemaker-attrd: ignore attribute delays when writing after node (re-)join + pacemaker-attrd: start new election immediately if writer is lost + pacemaker-attrd: clear election dampening when the writer leaves + pacemaker-attrd: detect alert configuration changes when CIB is replaced + CIB: inform originator of CIB upgrade failure + controller: support resource agents that require node name even for meta-data + controller: don't record pending clone notifications in CIB + controller: DC detects completion of another node's shutdown more accurately + controller: shut down DC if unable to update node attributes + controller: handle corosync peer/join notifications for new node in any order + controller: clear election dampening when DC is lost + executor: cancel recurring monitors if fence device registration is lost + fencing: check for fence device update when resource defaults change + fencing: avoid pacemaker-fenced crash possible with stonith_admin misuse + fencing: limit fencing history to 500 entries + fencing: stonith_admin now complains if no action option is specified + pacemakerd: do not modify kernel.sysrq on Linux + scheduler: regression test compatibility with glib 2.59.0 + scheduler: avoid unnecessary recovery of cleaned guest nodes and bundles + scheduler: ensure failures causing fencing not expired until fencing done + scheduler: start unique clone instances in numerical order + scheduler: convert unique clones to anonymous clones when not supported + scheduler: associate pending tasks with correct clone instance + scheduler: ensure bundle clone notifications are directed to correct host + scheduler: avoid improper bundle monitor rescheduling or fail count clearing + scheduler: honor asymmetric orderings even when restarting + scheduler: don't order non-DC shutdowns before DC fencing + ACLs: assume unprivileged ACL user if can't get user info + Pacemaker Remote: get Diffie-Hellman prime bit length from GnuTLS API + libcrmservice: cancel DBus call when cancelling systemd/upstart actions + libcrmservice: order systemd resources relative to pacemaker_remote + libpe_status: add public API constructor/destructor for pe_working_set_t + tools: fix crm_resource --clear when lifetime was used with ban/move + tools: fix crm_resource --move when lifetime was used with previous move + tools: make crm_mon CIB connection errors non-fatal if previously successful + tools: improve crm_mon messages when generating HTML output + tools: crm_mon cluster connection failure is now "critical" in nagios mode + tools: crm_mon listing of standby nodes shows if they have active resources + tools: crm_diff now ignores attribute ordering when comparing in CIB mode + tools: improve crm_report detection of logs, CIB directory, and processes + tools: crm_verify returns reliable exit codes + tools: crm_simulate resource history uses same name as live cluster would * Fri Jul 6 2018 Ken Gaillot Pacemaker-2.0.0-1 - Changesets: 885 - Diff: 549 files changed, 89865 insertions(+), 95100 deletions(-) - Deprecated features removed since Pacemaker-1.1.18 + All of these have newer forms, and the cluster will automatically convert most older syntax usage in saved configurations to newer syntax as needed + Drop support for heartbeat and corosync 1 (whether using CMAN or plugin) + Drop support for rolling upgrades from Pacemaker versions older than 1.1.11 + Drop support for built-in SMTP and SNMP in crm_mon + Drop support for legacy option aliases including default-action-timeout, default-resource-stickiness, resource-failure-stickiness, default-resource-failure-stickiness, is-managed-default, and all names using underbar instead of dash + Drop support for "requires" operation meta-attribute + Drop support for the pcmk_*_cmd, pcmk_arg_map, and pcmk_poweroff_action fence resource parameters + Drop support for deprecated command-line options to crmadmin, crm_attribute, crm_resource, crm_verify, crm_mon, and stonith_admin + Drop support for operation meta-attributes in instance_attributes + Drop support for PCMK_legacy and LRMD_MAX_CHILDREN environment variables + Drop support for undocumented resource isolation feature + Drop support for processing very old saved CIB files (including pre-0.6.0 start failure entries, pre-0.6.5 operation history entries, pre-0.7 transition keys, pre-1.1.4 migration history entries, pre-1.0 XML configuration schemas, pre-1.1.6 ticket state entries, and pre-1.1.7 failed recurring operation history entries) - Features added since Pacemaker-1.1.18 + The pacemaker daemons have been renamed to make logs more intuitive and easier to search + The default location of the Pacemaker detail log is now /var/log/pacemaker/pacemaker.log (instead of being directly in /var/log), and Pacemaker will no longer use Corosync's logging preferences; configure script options are available to change default log locations + The detail log's message format has been improved + The master XML tag is deprecated in favor of using a standard clone tag with a new "promotable" meta-attribute set to true, and the "master-max" and "master-node-max" master meta-attributes are deprecated in favor of new "promoted-max" and "promoted-node-max" clone meta-attributes; documentation now refers to these as promotable clones rather than master/slave, stateful, or multistate clones, and refers to promotion scores instead of master scores + Administration-related documentation has been moved from the "Pacemaker Explained" document to a new "Pacemaker Administration" document + record-pending now defaults to TRUE (pending actions are shown in status) + All Python code in Pacemaker now supports both Python 2.7 and Python 3 + The command-line tools now return consistent, well-defined exit codes; crm_error has an --exit option to list these + Pacemaker's systemd unit files now remove systemd's spawned process limit + mount, path, and timer systemd unit types are now supported as resources + A negative stonith-watchdog-timeout now tells the cluster to automatically calculate the value based on SBD_WATCHDOG_TIMEOUT (which was the behavior of 0 before 1.1.15; 0 retains its post-1.1.15 behavior of disabling use of the watchdog as a fencing device) + The undocumented restart-type and role_after_failure resource meta-attributes are now deprecated + Regression testing code has been consolidated and overhauled (the most obvious change is new command names) + build: create /etc/pacemaker directory when installing + build: improved portability to BSD-based platforms + tools: crm_resource --cleanup now cleans only failed operation history; crm_resource --reprobe retains the previous behavior of cleaning all operation history + tools: add stonith_admin --validate option to check device configuration + tools: crm_node is now in the pacemaker-cli package (instead of pacemaker) + alerts: add epoch and usec alert variables for improved SNMP alerts + controller: deprecate "crmd-*" cluster options in favor of new names + scheduler: deprecate stonith-action value "poweroff" (use "off" instead) + scheduler: deprecate require-all in rsc_order + libcrmcluster: prefer corosync name over ring0_addr + xml: allow local "kind" in resource_set within rsc_order - Changes since Pacemaker-1.1.18 + Restore systemd unit dependency on DBus (regression in 1.1.17) + CIB: handle mixed-case node names when modifying attributes (regression in 1.1.17) + scheduler: avoid crash when logging ignored failure timeout (regression in 1.1.17) + attrd: ensure node name is broadcast at start-up (regression in 1.1.18) + scheduler: unfence before probing or starting fence devices (regression in 1.1.18) + tools: treat INFINITY correctly in crm_failcount (regression in 1.1.17) + tools: show master scores with crm_simulate -sL (regression in 1.1.18) + tools: crm_master did not work without explicit --lifetime (regression in 1.1.18) + Numerous changes to public C API of libraries + Choose current node correctly when a resource is multiply active + controller,executor,tools: avoid minor memory leaks + CIB: don't use empty CIB if real CIB has bad permissions + controller: avoid double free after ACL rejection of resource deletion + controller: don't record pending clone notifications in CIB + controller: always write faked failures to CIB whenever possible + controller: quorum gain without a node join should cause new transition + executor: handle systemd actions correctly when used with "service:" + executor: find absolute LSB paths when used with "service:" + scheduler: handle "requires" of "quorum" or "nothing" properly + scheduler: ensure orphaned recurring monitors have interval set + scheduler: handle pending migrations correctly when record-pending is true + scheduler: don't time out failures that cause fencing until fencing completes + scheduler: handle globally-unique bundle children correctly + scheduler: use correct default timeout for monitors + scheduler: "symmetrical" defaults to "false" for serialize orders + scheduler: avoid potential use-of-NULL when unpacking ordering constraint + scheduler: properly cancel recurring monitors + scheduler: do not schedule notifications for unrunnable actions + scheduler: ensure stops occur after stopped remote connections come back up + scheduler: consider only allowed nodes when ordering start after all recovery + scheduler: avoid graph loop from ordering bundle child stops/demotes after container fencing + scheduler: remote connection resources are safe to require only quorum + scheduler: correctly observe colocation with bundles in Master role + scheduler: restart resource after failed demote when appropriate + Pacemaker Remote: always use most recent remote proxy + tools: crm_node now gets correct node name and ID on Pacemaker Remote nodes + tools: correctly check crm_resource --move for master role + tools: cibsecret --help/--version doesn't require cluster to be running + tools: ignore attribute placement when crm_diff compares in cib mode + tools: prevent notify actions from causing crm_resource --wait to hang + resources: drop broken configdir parameter from ocf:pacemaker:controld - For further details, see: https://wiki.clusterlabs.org/wiki/Pacemaker_2.0_Changes * Tue Nov 14 2017 Ken Gaillot Pacemaker-1.1.18-1 - Update source tarball to revision: a9fbd15 - Changesets: 644 - Diff: 167 files changed, 9753 insertions(+), 5596 deletions(-) - Features added since Pacemaker-1.1.17 + warnings are now logged when using legacy syntax to be removed in 2.0 + agents: ifspeed agent is now installed when building + agents: ifspeed agent can optionally detect interface name from IP address + alerts: support alert filters + alerts: experimental support for alerts for node attribute changes + crmd,pengine: support unfencing of remote nodes + pengine: bundles now support all constraint types + pengine: bundles now support rkt containers + pengine: bundles support new container-attribute-target parameter + pengine,tools: logs and crm_mon show why resources changed state + stonith-ng: support new fencing resource parameter pcmk_delay_base + tools: new crm_resource option --why explains why resources are stopped - Changes since Pacemaker-1.1.17 + many documentation improvements + agents: ifspeed properly calculates speed of hfi1 interfaces + agents: ClusterMon now interprets "update" less than 1000 as seconds + attrd: don't lose attributes set between attrd start-up and cluster join + attrd: fix multiple minor memory leaks + crmd: correctly record that unfencing is complete + crmd: error more quickly if remote start fails due to missing key + lrmd: remote resource operations return immediate error if key setup fails + lrmd: allow pre-1.1.15 cluster nodes to connect to current Pacemaker Remote + pengine: guest nodes are now probed like other nodes + pengine: probe remote nodes for guest node resources + pengine: do not probe guest/bundle connections until guest/bundle is active + pengine: allow resources to stop prior to probes completing + pengine: bundles wait only for other containers on same node to be probed + pengine: have bundles log to stderr so 'docker logs'/'journalctl -M' works + pengine: only pass requests for promote/demote flags onto the bundle's child + pengine: do not map ports into Docker container when net=host is specified + pengine: allow resources inside bundles to receive clone notifications + pengine: default to non-interleaved bundle ordering for safety + pengine: ensure bundle nodes and child resources are correctly cleaned up + pengine: prevent graph loops when fencing the host underneath a bundle + pengine: fix multiple memory issues (use-after-free, use-of-NULL) with bundles + pengine: resources in bundles respect failcounts + pengine: ensure nested container connections run on the same host + pengine: ensure unrecoverable remote nodes are fenced even with no resources + pengine: handle resource migrating behind a migrating remote connection + pengine: don't prefer to keep unique instances on same node + pengine: exclude exclusive resources and nodes from symmetric default score + pengine: if ignoring failure, also ignore migration-threshold + pengine: restore the ability to send the transition graph via the disk if it gets too big + pengine: validate no-quorum-policy=suicide correctly + pengine: avoid crash when alerts section has comments + pengine: detect permanent master scores at start-up + pengine: do not re-add a node's default score for each location constraint + pengine: make sure calculated resource scores are consistent on different architectures + pengine: retrigger unfencing for changed device parameters only when necessary + pengine: don't schedule reload and restart in same transition (CLBZ#5309, regression introduced in 1.1.15) + stonith-ng: make fencing-device reappear properly after reenabling + stonith-ng: include pcmk_on_action in meta-data so 'on' can be overridden + tools: allow crm_report to work with no log files specified + tools: fix use-after-free in crm_diff introduced in 1.1.17 + tools: allow crm_resource to operate on anonymous clones in unknown states + tools: crm_resource --cleanup on appropriate nodes if we don't know state of resource + tools: prevent disconnection from crmd during crm_resource --cleanup + tools: improve messages for crm_resource --force-* options + tools: crm_mon: avoid infinite process spawning if -E script can't be run + tools: crm_mon: don't show previous exit-reason for failed action with none + libcrmservice: list systemd unit files, not only active units (CLBZ#5299) + libcrmservice: parse long description correctly for LSB meta-data * Thu Jul 06 2017 Ken Gaillot Pacemaker-1.1.17-1 - Update source tarball to revision: 301bc44 - Changesets: 539 - Diff: 177 files changed, 11525 insertions(+), 5036 deletions(-) - Features added since Pacemaker-1.1.16 + New "bundle" resource type for Docker container use cases (experimental) + New "PCMK_node_start_state" environment variable to start node in standby + New "value-source" rule expression attribute in location constraints to compare a node attribute against a resource parameter + New "stonith-max-attempts" cluster option to specify how many times fencing can fail for a target before the cluster will no longer immediately re-attempt it (previously hard-coded at 10) + New "cluster-ipc-limit" cluster option to avoid IPC client eviction in large clusters + Failures are now tracked per operation type, as well as per node and resource (the "fail-count" and "last-failure" node attribute names now end in "#OPERATION_INTERVAL") + attrd: Pacemaker Remote node attributes and regular expressions are now supported on legacy cluster stacks (heartbeat, CMAN, and corosync plugin) + tools: New "crm_resource --validate" option + tools: New "stonith_admin --list-targets" option + tools: New "crm_attribute --pattern" option to match a regular expression + tools: "crm_resource --cleanup" and "crm_failcount" can now take --operation and --interval options to operate on a single operation type - Changes since Pacemaker-1.1.16 + Fix multiple memory issues (leaks, use-after-free) in libraries + pengine: unmanaging a guest node resource puts guest in maintenance mode + cib: broadcasts of cib changes should always pass ACL checks + crmd,libcrmcommon: update throttling when CPUs are hot-plugged + crmd: abort transition whenever we lose quorum + crmd: avoid attribute write-out on join when atomic attrd is used + crmd: check for too many stonith failures only when aborting for that reason + crmd: correctly clear failure counts only for a specified node + crmd: don't fence old DC if it's shutting down as soon-to-be DC joins + crmd: forget stonith failures when forgetting node + crmd: all nodes should track stonith failure counts in case they become DC + crmd: update cache status for guest node whose host is fenced + dbus: prevent lrmd from hanging on dbus calls + fencing: detect newly added constraints for stonith devices + pengine: order remote actions after connection recovery (regression introduced in 1.1.15) + pengine: quicker recovery from failed demote + libcib: determine remote nodes correctly from node status entries + libcrmcommon: avoid evicting IPC client if messages spike briefly + libcrmcommon: better XML comment handling prevents infinite election loop + libcrmcommon: set month correctly in date/time string sent to alert agents + libfencing,fencing: intelligently remap "action" wrongly specified in config + libservices: ensure completed ops aren't on blocked ops list + libservices: properly detect and cancel in-flight systemd/upstart ops + libservices: properly watch writable DBus handles + libservices: systemd service that is reloading doesn't cause monitor failure + pacemaker_remoted: allow graceful shutdown while unmanaged + pengine,libpe_status: don't clear same fail-count twice + pengine: consider guest node unclean if its host is unclean + pengine: do not re-add a node's default score for each location constraint + pengine: avoid restarting services when recovering remote connection + pengine: better guest node recovery when host fails + pengine: guest node fencing doesn't require stonith enabled + pengine: allow probes of guest node connection resources + pengine: properly handle allow-migrate explicitly set for remote connection + pengine: fence failed remote nodes even if no resources can run on them + tools: resource agents will now get the correct node name on Pacemaker Remote nodes when using crm_node and crm_attribute + tools: avoid grep crashes in crm_report when looking for system logs + tools: crm_resource -C now clears last-failure as well as fail-count + tools: implement crm_failcount command-line options correctly + tools: properly ignore version with crm_diff --no-version * Wed Nov 30 2016 Ken Gaillot Pacemaker-1.1.16-1 - Update source tarball to revision: 76876b3 - Changesets: 382 - Diff: 145 files changed, 7200 insertions(+), 5621 deletions(-) - Features added since Pacemaker-1.1.15 + Location constraints may use rsc-pattern, with submatches expanded + node-health-base available with node-health-strategy=progressive + new Pacemaker Development document for working on pacemaker code base + new PCMK_panic_action variable allows crash instead of reboot on panic + resources: add resource agent for managing a node attribute + systemd: include socket units when listing all systemd agents - Changes since Pacemaker-1.1.15 + Important security fix for CVE-2016-7035 + Logging is now synchronous when blackboxes are enabled + All python code except CTS is now compatible with python 2.6+ and 3.2+ + build: take advantage of compiler features for security and performance + build: update SuSE spec modifications for recent spec changes + build: avoid watchdog reboot when upgrading pacemaker_remote with sbd + build: numerous other improvements in environment detection, etc. + cib: fix infinite loop when no schema validates + crmd: cl#5185 - record pending operations in CIB before they are performed + crmd: don't abort transitions for CIB comment changes + crmd: resend shutdown request if DC loses original request + documentation: install improved README in doc instead of now-removed AUTHORS + documentation: clarify licensing and provide copy of all licenses + documentation: document various features and upgrades better + fence_legacy: use "list" action when searching cluster-glue agents + libcib: don't stop sending alerts after releasing DC role + libcrmcommon: properly handle XML comments when comparing v2 patchset diffs + libcrmcommon: report errors consistently when waiting for data on connection + libpengine: avoid potential use-of-NULL + libservices: use DBusError API properly + pacemaker_remote: init script stop should always return 0 + pacemaker_remote: allow remote clients to timeout/reconnect + pacemaker_remote: correctly calculate remaining timeout when receiving messages + pengine: avoid transition loop for start-then-stop + unfencing + pengine: correctly update dependent actions of un-runnable clones + pengine: do not fence a node in maintenance mode if it shuts down cleanly + pengine: set OCF_RESKEY_CRM_meta_notify_active_* for multistate resources + resources: ping - avoid temporary files for fping check, support FreeBSD + resources: SysInfo - better support for FreeBSD + resources: variable name typo in docker-wrapper + systemd: order pacemaker after time-sync target + tools: correct attrd_updater help and error messages when using CMAN + tools: crm_standby --version/--help should work without cluster running + tools: make crm_report sanitize CIB before generating readable version + tools: display pending resource state by default when available + tools: avoid matching other process with same PID in ClusterMon * Tue Jun 21 2016 Ken Gaillot Pacemaker-1.1.15-1 - Update source tarball to revision: 32fa6a5 - Changesets: 533 - Diff: 219 files changed, 6659 insertions(+), 3989 deletions(-) - Features added since Pacemaker-1.1.14 + Event-driven alerts allow scripts to be called after significant events + build: Some files moved from pacemaker package to pacemaker-cli for cleaner pacemaker-remote dependencies + build: ./configure --with-configdir argument for /etc/sysconfig, /etc/default, etc. + fencing: Simplify watchdog integration + fencing: Support concurrent fencing actions via new pcmk_action_limit option + remote: pacemaker_remote may be stopped without disabling resource first + remote: Report integration status of Pacemaker Remote nodes in CIB node_state + tools: crm_mon now reports why resources are not starting + tools: crm_report now obscures passwords in logfiles + tools: attrd_updater --update-both/--update-delay options allow changing dampening value + tools: allow stonith_admin -H '*' to show history for all nodes - Changes since Pacemaker-1.1.14 + Fix multiple memory issues (leaks, use-after-free) in daemons, libraries and tools + Make various log messages more user-friendly + Improve FreeBSD and Hurd support + attrd: Prevent possible segfault on exit + cib: Fix regression to restore support for compressed CIB larger than 1MB + common: fix regression in 1.1.14 that made have-watchdog always true + controld: handle DLM "wait fencing" state better + crmd: Fix regression so that fenced unseen nodes do not remain unclean + crmd: Take start-delay into account when calculation action timeouts + crmd: Avoid timeout on older peers when cancelling a resource operation + fencing: Allow fencing by node ID (e.g. by DLM) even if node left cluster + lrmd: Fix potential issues when cluster is stopped via systemd shutdown + pacemakerd: Properly respawn stonithd if it fails + pengine: Fix regression with multiple monitor levels that could ignore failure + pengine: Correctly set OCF_RESKEY_CRM_meta_timeout when start-delay is configured + pengine: Properly order actions for master/slave resources in anti-colocations + pengine: Respect asymmetrical ordering when trying to move resources + pengine: Properly order stop actions on guest node relative to host stonith + pengine: Correctly block actions dependent on unrunnable clones + remote: Allow remote nodes to have node attributes even with legacy attrd + remote: Recover from remote node fencing more quickly + remote: Place resources on newly rejoined remote nodes more quickly + resources: ping agent can now use fping6 for IPv6 hosts + resources: SysInfo now resets #health_disk to green when there's sufficient free disk + tools: crm_report is now more efficient and handles Pacemaker Remote nodes better + tools: Prevent crm_resource segfault when --resource is not supplied with --restart + tools: crm_shadow --display option now works + tools: crm_resource --restart handles groups, target-roles and moving resources better * Thu Jan 14 2016 Ken Gaillot Pacemaker-1.1.14-1 - Update source tarball to revision: f0b585a - Changesets: 724 - Diff: 179 files changed, 13142 insertions(+), 7695 deletions(-) - Features added since Pacemaker-1.1.13 + crm_resource: Indicate common reasons why a resource may not start after a cleanup + crm_resource: New --force-promote and --force-demote options for debugging + fencing: Support targeting fencing topologies by node name pattern or node attribute + fencing: Remap sequential topology reboots to all-off-then-all-on + pengine: Allow resources to start and stop as soon as their state is known on all nodes + pengine: Include a list of all and available nodes with clone notifications + pengine: Addition of the clone resource clone-min metadata option + pengine: Support of multiple-active=block for resource groups + remote: Resources that create guest nodes can be included in a group resource + remote: reconnect_interval option for remote nodes to delay reconnect after fence - Changes since Pacemaker-1.1.13 + improve support for building on FreeBSD and Debian + fix multiple memory issues (leaks, use-after-free, double free, use-of-NULL) in components and tools + cib: Do not terminate due to badly behaving clients + cman: handle corosync-invented node names of the form Node{id} for peers not in its node list + controld: replace bashism + crm_node: Display node state with -l and quorum status with -q, if available + crmd: resources would sometimes be restarted when only non-unique parameters changed + crmd: fence remote node after connection failure only once + crmd: handle resources named the same as cluster nodes + crmd: Pre-emptively fail in-flight actions when lrmd connections fail + crmd: Record actions in the CIB as failed if we cannot execute them + crm_report: Enable password sanitizing by default + crm_report: Allow log file discovery to be disabled + crm_resource: Allow the resource configuration to be modified for --force-{check,start,..} calls + crm_resource: Compensate for -C and -p being called with the child resource for clones + crm_resource: Correctly clean up all children for anonymous cloned groups + crm_resource: Correctly clean up failcounts for inactive anonymous clones + crm_resource: Correctly observe --force when deleting and updating attributes + crm_shadow: Fix "crm_shadow --diff" + crm_simulate: Prevent segfault on arches with 64bit time_t + fencing: ensure "required"/"automatic" only apply to "on" actions + fencing: Return a provider for the internal fencing agent "#watchdog" instead of logging an error + fencing: ignore stderr output of fence agents (often used for debug messages) + fencing: fix issue where deleting a fence device attribute can delete the device + libcib: potential user input overflow + libcluster: overhaul peer cache management + log: make syslog less noisy + log: fix various misspellings in log messages + lrmd: cancel currently pending STONITH op if stonithd connection is lost + lrmd: Finalize all pending and recurring operations when cleaning up a resource + pengine: Bug cl#5247 - Imply resources running on a container are stopped when the container is stopped + pengine: cl#5235 - Prevent graph loops that can be introduced by "load_stopped -> migrate_to" ordering + pengine: Correctly bypass fencing for resources that do not require it + pengine: do not timeout remote node recurring monitor op failure until after fencing + pengine: Ensure recurring monitor operations are cancelled when clone instances are de-allocated + pengine: fixes segfault in pengine when fencing remote node + pengine: properly handle blocked clone actions + pengine: ensure failed actions that occurred in node shutdown are displayed + remote: Correctly display the usage of the ocf:pacemaker:remote resource agent + remote: do not fail operations because of a migration + remote: enable reloads for select remote connection options + resources: allow for top output with or without percent sign in HealthCPU + resources: Prevent an error message on stopping "Dummy" resource + systemd: Prevent segfault when logging failed operations + systemd: Reconnect to System DBus if the connection is closed + systemd: set systemd resources' timeout values higher than systemd's own default + tools: Do not send command lines to syslog + tools: update SNMP MIB + upstart: Ensure pending structs are correctly unreferenced * Wed Jun 24 2015 Andrew Beekhof Pacemaker-1.1.13-1 - Update source tarball to revision: 2a1847e - Changesets: 750 - Diff: 156 files changed, 11323 insertions(+), 3725 deletions(-) - Features added since Pacemaker-1.1.12 + Allow fail-counts to be removed en-mass when the new attrd is in operation + attrd supports private attributes (not written to CIB) + crmd: Ensure a watchdog device is in use if stonith-watchdog-timeout is configured + crmd: If configured, trigger the watchdog immediately if we lose quorum and no-quorum-policy=suicide + crm_diff: Support generating a difference without versions details if --no-version/-u is supplied + crm_resource: Implement an intelligent restart capability + Fencing: Advertise the watchdog device for fencing operations + Fencing: Allow the cluster to recover resources if the watchdog is in use + fencing: cl#5134 - Support random fencing delay to avoid double fencing + mcp: Allow orphan children to initiate node panic via SIGQUIT + mcp: Turn on sbd integration if pacemakerd finds it running + mcp: Two new error codes that result in machine reset or power off + Officially support the resource-discovery attribute for location constraints + PE: Allow natural ordering of colocation sets + PE: Support non-actionable degraded mode for OCF + pengine: cl#5207 - Display "UNCLEAN" for resources running on unclean offline nodes + remote: pcmk remote client tool for use with container wrapper script + Support machine panics for some kinds of errors (via sbd if available) + tools: add crm_resource --wait option + tools: attrd_updater supports --query and --all options + tools: attrd_updater: Allow attributes to be set for other nodes - Changes since Pacemaker-1.1.12 + pengine: exclusive discovery implies rsc is only allowed on exclusive subset of nodes + acl: Correctly implement the 'reference' acl directive + acl: Do not delay evaluation of added nodes in some situations + attrd: b22b1fe did uuid test too early + attrd: Clean out the node cache when requested by the admin + attrd: fixes double free in attrd legacy + attrd: properly write attributes for peers once uuid is discovered + attrd: refresh should force an immediate write-out of all attributes + attrd: Simplify how node deletions happen + Bug rhbz#1067544 - Tools: Correctly handle --ban, --move and --locate for master/slave groups + Bug rhbz#1181824 - Ensure the DC can be reliably fenced + cib: Ability to upgrade cib validation schema in legacy mode + cib: Always generate digests for cib diffs in legacy mode + cib: assignment where comparison intended + cib: Avoid nodeid conflicts we don't care about + cib: Correctly add "update-origin", "update-client" and "update-user" attributes for cib + cib: Correctly set up signal handlers + cib: Correctly track node state + cib: Do not update on disk backups if we're just querying them + cib: Enable cib legacy mode for plugin-based clusters + cib: Ensure file-based backends treat '-o section' consistently with the native backend + cib: Ensure upgrade operations from a non-DC get an acknowledgement + cib: No need to enforce cib digests for v2 diffs in legacy mode + cib: Revert d153b86 to instantly get cib synchronized in legacy mode + cib: tls sock cleanup for remote cib connections + cli: Ensure subsequent unknown long options are correctly detected + cluster: Invoke crm_remove_conflicting_peer() only when the new node's uname is being assigned in the node cache + common: Increment current and age for lib common as a result of APIs being added + corosync: Bug cl#5232 - Somewhat gracefully handle nodes with invalid UUIDs + corosync: Avoid unnecessary repeated CMAP API calls + crmd/pengine: handle on-fail=ignore properly + crmd: Add "on_node" attribute for *_last_failure_0 lrm resource operations + crmd: All peers need to track node shutdown requests + crmd: Cached copies of transient attributes cease to be valid once a node leaves the membership + crmd: Correctly add the local option that validates against schema for pengine to calculate + crmd: Disable debug logging that results in significant overhead + crmd: do not remove connection resources during re-probe + crmd: don't update fail count twice for same failure + crmd: Ensure remote connection resources timeout properly during 'migrate_from' action + crmd: Ensure throttle_mode() does something on Linux + crmd: Fixes crash when remote connection migration fails + crmd: gracefully handle remote node disconnects during op execution + crmd: Handle remote connection failures while executing ops on remote connection + crmd: include remote nodes when forcing cluster wide resource reprobe + crmd: never stop recurring monitor ops for pcmk remote during incomplete migration + crmd: Prevent the old version of DC from being fenced when it shuts down for rolling-upgrade + crmd: Prevent use-of-NULL during reprobe + crmd: properly update job limit for baremetal remote-nodes + crmd: Remote-node throttle jobs count towards cluster-node hosting conneciton rsc + crmd: Reset stonith failcount to recover transitioner when the node rejoins + crmd: resolves memory leak in crmd. + crmd: respect start-failure-is-fatal even for artifically injected events + crmd: Wait for all pending operations to complete before poking the policy engine + crmd: When container's host is fenced, cancel in-flight operations + crm_attribute: Correctly update config options when -o crm_config is specified + crm_failcount: Better error reporting when no resource is specified + crm_mon: add exit reason to resource failure output + crm_mon: Fill CRM_notify_node in traps with node's uname rather than node's id if possible + crm_mon: Repair notification delivery when the v2 patch format is in use + crm_node: Correctly remove nodes from the CIB by nodeid + crm_report: More patterns for finding logs on non-DC nodes + crm_resource: Allow resource restart operations to be node specific + crm_resource: avoid deletion of lrm cache on node with resource discovery disabled. + crm_resource: Calculate how long to wait for a restart based on the resource timeouts + crm_resource: Clean up memory in --restart error paths + crm_resource: Display the locations of all anonymous clone children when supplying the children's common ID + crm_resource: Ensure --restart sets/clears meta attributes + crm_resource: Ensure fail-counts are purged when we redetect the state of all resources + crm_resource: Implement --timeout for resource restart operations + crm_resource: Include group members when calculating the next timeout + crm_resource: Memory leak in error paths + crm_resource: Prevent use-after-free + crm_resource: Repair regression test outputs + crm_resource: Use-after-free when restarting a resource + dbus: ref count leaks + dbus: Ensure both the read and write queues get dispatched + dbus: Fail gracefully if malloc fails + dbus: handle dispatch queue when multiple replies need to be processed + dbus: Notice when dbus connections get disabled + dbus: Remove double-free introduced while trying to make coverity shut up + ensure if B is colocated with A, B can never run without A + fence_legacy: Avoid passing 'port' to cluster-glue agents + fencing: Allow nodes to be purged from the member cache + fencing: Correctly make args for fencing agents + fencing: Correctly wait for self-fencing to occur when the watchdog is in use + fencing: Ensure the hostlist parameter is set for watchdog agents + fencing: Force 'stonith-ng' as the system name + fencing: Gracefully handle invalid metadata from agents + fencing: If configured, wait stonith-watchdog-timer seconds for self-fencing to complete + fencing: Reject actions for devices that haven't been explicitly registered yet + ipc: properly allocate server enforced buffer size on client + ipc: use server enforced buffer during ipc client send + lrmd, services: interpret LSB status codes properly + lrmd: add back support for class heartbeat agents + lrmd: cancel pending async connection during disconnect + lrmd: enable ipc proxy for docker-wrapper privileged mode + lrmd: fix rescheduling of systemd monitor op during start + lrmd: Handle systemd reporting 'done' before a resource is actually stopped + lrmd: Hint to child processes that using sd_notify is not required + lrmd: Log with the correct personality + lrmd: Prevent glib assert triggered by timers being removed from mainloop more than once + lrmd: report original timeout when systemd operation completes + lrmd: store failed operation exit reason in cib + mainloop: resolves race condition mainloop poll involving modification of ipc connections + make targetted reprobe for remote node work, crm_resource -C -N + mcp: Allow a configurable delay when debugging shutdown issues + mcp: Avoid requiring 'export' for SYS-V sysconfig options + Membership: Detect and resolve nodes that change their ID + pacemakerd: resolves memory leak of xml structure in pacemakerd + pengine: ability to launch resources in isolated containers + pengine: add #kind=remote for baremetal remote-nodes + pengine: allow baremetal remote-nodes to recover without requiring fencing when cluster-node fails + pengine: allow remote-nodes to be placed in maintenance mode + pengine: Avoid trailing whitespaces when printing resource state + pengine: cl#5130 - Choose nodes capable of running all the colocated utilization resources + pengine: cl#5130 - Only check the capacities of the nodes that are allowed to run the resource + pengine: Correctly compare feature set to determine how to unpack meta attributes + pengine: disable migrations for resources with isolation containers + pengine: disable reloading of resources within isolated container wrappers + pengine: Do not aggregate children in a pending state into the started/stopped/etc lists + pengine: Do not record duplicate copies of the failed actions + pengine: Do not reschedule monitors that are no longer needed while resource definitions have changed + pengine: Fence baremetal remote when recurring monitor op fails + pengine: Fix colocation with unmanaged resources + pengine: Fix the behaviors of multi-state resources with asymmetrical ordering + pengine: fixes pengine crash with orphaned remote node connection resource + pengine: fixes segfault caused by malformed log warning + pengine: handle cloned isolated resources in a sane way + pengine: handle isolated resource scenario, cloned group of isolated resources + pengine: Handle ordering between stateful and migratable resources + pengine: imply stop in container node resources when host node is fenced + pengine: only fence baremetal remote when connection can fails or can not be recovered + pengine: only kill process group on timeout when on-fail does not equal block. + pengine: per-node control over resource discovery + pengine: prefer migration target for remote node connections + pengine: prevent disabling rsc discovery per node in certain situations + pengine: Prevent use-after-free in sort_rsc_process_order() + pengine: properly handle ordering during remote connection partial migration + pengine: properly recover remote-nodes when cluster-node proxy goes offline + pengine: remove unnecessary whitespace from notify environment variables + pengine: require-all feature for ordered clones + pengine: Resolve memory leaks + pengine: resource discovery mode for location constraints + pengine: restart master instances on instance attribute changes + pengine: Turn off legacy unpacking of resource options into the meta hashtable + pengine: Watchdog integration is sufficient for fencing + Perform systemd reloads asynchronously + ping: Correctly advertise multiplier default + Prefer to inherit the watchdog timeout from SBD + properly record stop args after reload + provide fake meta data for ra class heartbeat + remote: report timestamps for remote connection resource operations + remote: Treat recv msg timeout as a disconnect + service: Prevent potential use-of-NULL in metadata lookups + solaris: Allow compilation when dirent.d_type is not available + solaris: Correctly replace the linux swab functions + solaris: Disable throttling since /proc doesn't exist + stonith-ng: Correctly observe the watchdog completion timeout + stonith-ng: Correctly track node state + stonith-ng: Reset mainloop source IDs after removing them + systemd: Correctly handle long running stop actions + systemd: Ensure failed monitor operations always return + systemd: Ensure we don't call dbus_message_unref() with NULL + systemd: fix crash caused when canceling in-flight operation + systemd: Kindly ask dbus NOT to kill the process if the dbus connection fails + systemd: Perform actions asynchronously + systemd: Perform monitor operations without blocking + systemd: Tell systemd not to take DBus down from underneath us + systemd: Trick systemd into not stopping our services before us during shutdown + tools: Improve crm_mon output with certain option combinations + upstart: Monitor actions always return 'ok' or 'not running' + upstart: Perform more parts of monitor operations without blocking + xml: add 'require-all' to xml schema for constraints + xml: cl#5231 - Unset the deleted attributes in the resulting diffs + xml: Clone the latest constraint schema in preparation for changes" + xml: Correctly create v1 patchsets when deleting attributes + xml: Do not change the ordering of properties when applying v1 cib diffs + xml: Do not dump deleted attributes + xml: Do not prune leaves from v1 cib diffs that are being created with digests + xml: Ensure ACLs are reapplied before calculating what a replace operation changed + xml: Fix upgrade-1.3.xsl to correctly transform ACL rules with "attribute" + xml: Prevent assert errors in crm_element_value() on applying a patch without version information + xml: Prevent potential use-of-NULL * Tue Jul 22 2014 Andrew Beekhof Pacemaker-1.1.12-1 - Update source tarball to revision: 93a037d - Changesets: 795 - Diff: 195 files changed, 13772 insertions(+), 6176 deletions(-) - Features added since Pacemaker-1.1.11 + Changes to the ACL schema to support nodes and unix groups + cib: Check ACLs prior to making the update instead of parsing the diff afterwards + cib: Default ACL support to on + cib: Enable the more efficient xml patchset format + cib: Implement zero-copy status update + cib: Send all r/w operations via the cluster connection and have all nodes process them + crmd: Set "cluster-name" property to corosync's "cluster_name" by default for corosync-2 + crm_mon: Display brief output if "-b/--brief" is supplied or 'b' is toggled + crm_report: Allow ssh alternatives to be used + crm_ticket: Support multiple modifications for a ticket in an atomic operation + extra: Add logrotate configuration file for /var/log/pacemaker.log + Fencing: Add the ability to call stonith_api_time() from stonith_admin + logging: daemons always get a log file, unless explicitly set to configured 'none' + logging: allows the user to specify a log level that is output to syslog + PE: Automatically re-unfence a node if the fencing device definition changes + pengine: cl#5174 - Allow resource sets and templates for location constraints + pengine: Support cib object tags + pengine: Support cluster-specific instance attributes based on rules + pengine: Support id-ref in nvpair with optional "name" + pengine: Support per-resource maintenance mode + pengine: Support site-specific instance attributes based on rules + tools: Allow crm_shadow to create older configuration versions + tools: Display pending state in crm_mon/crm_resource/crm_simulate if --pending/-j is supplied (cl#5178) + xml: Add the ability to have lightweight schema revisions + xml: Enable resource sets in location constraints for 1.2 schema + xml: Support resources that require unfencing - Changes since Pacemaker-1.1.11 + acl: Authenticate pacemaker-remote requests with the node name as the client + acl: Read access must be explicitly granted + attrd: Ensure attribute dampening is always observed + attrd: Remove offline nodes from node cache for "peer-remove" requests + Bug cl#5055 - Improved migration support. + Bug cl#5184 - Ensure pending probes that ultimately fail are correctly updated + Bug cl#5196 - pengine: Check values after expanding templates + Bug cl#5212 - Do not promote instances when quorum is lots and no-quorum-policy=freeze + Bug cl#5213 - Ensure role colocation with -INFINITY is enforced + Bug cl#5213 - Limit the scope of the previous commit to the masters role + Bug cl#5219 - pengine: Allow unrelated resources with a common colocation target to remain promoted + Bug cl#5222 - cib: Repair rolling update capability + Bug cl#5222 - Enable legacy mode whenever a broadcast update is detected + Bug rhbz#1036631 - Stop members of cloned groups when dependencies are stopped + Bug rhbz#1054307 - cname pattern match should be more restrictive in init script + Bug rhbz#1057697 - Use native DBus library for systemd/upstart support to avoid problematic use of threads + Bug rhbz#1097457 - Limit the scope of the previous fix and include a helpful comment + Bug rhbz#1097457 - Prevent invalid transition when resource are ordered to start after the container they're started in + cib: allow setting permanent remote-node attributes + cib: Auto-detect which patchset format to use + cib: Determine the best value of validate-with if one is not supplied + cib: Do not disable cib disk writes if on-disk cib is corrupt + cib: Ensure 'cibadmin -R/--replace' commands get replies + cib: Erasing the cib is an admin action, bump the admin_epoch instead + cib: Fix remote cib based on TLS + cib: Ignore patch failures if we already have their contents + cib: Validate that everyone still sees the same configuration once all updates have completed + cibadmin: Allow priviliged clients to perform tasks as unpriviliged users + cibadmin: Remove dangerous commands that exposed unnecessary implementation internal details + cluster: Fix segfault on removing a node + cluster: Prevent search of unames from attempting to create node entries for unknown nodes + cluster: Remove unknown offline nodes with conflicting unames from node cache + controld: Do not consider the dlm up until the address list is present + controld: handling startup fencing within the controld agent, not the dlm + controld: Return OCF_ERR_INSTALLED instead of OCF_NOT_INSTALLED + crmd: Ack pending operations that were cancelled due to rsc deletion + crmd: Actions can only be executed if their pre-requisits completed successfully + crmd: avoid double free caused by nested hash table removal + crmd: Avoid spamming the cib by triggering a transition only once per non-status change + crmd: Correctly react to successful unfencing operations + crmd: Correctly recognise operation cancellations we initiated + crmd: Do not erase the status section for unfenced nodes + crmd: Do not overwrite existing node state when fencing completes + crmd: Do not start timers for already completed operations + crmd: Ensure crm_config options are re-read on updates + crmd: Fenced nodes that return prior to an election do not need to have their status section reset + crmd: make lrm_state hash table not case sensitive + crmd: make node_state erase correctly + crmd: Only write fence_averride if open() returns a positive file descriptor + crmd: Prevent manual fencing confirmations from attempting to create node entries for unknown nodes + crmd: Prevent SIGPIPE when notifying CMAN about fencing operations + crmd: Remove state of unknown nodes with conflicting unames from CIB + crmd: Remove unknown nodes with conflicting unames from CIB + crmd: Report unsuccessful unfencing operations + crm_diff: Allow the generation of xml patchsets without digests + crm_mon: Allow the file created by --as-html to be world readable + crm_mon: Ensure resource attributes have been unpacked before displaying connectivity data + crm_node: Only remove the named resource from the cib + crm_report: Gracefully handle rediculously large logfiles + crm_report: Only gather dlm data if dlm_controld is running + crm_resource: Gracefully handle -EACCESS when querying the cib + crm_verify: Perform a full set of calculations whenever the status section is present + fencing: Advertise support for reboot/on/off in the metadata for legacy agents + fencing: Automatically switch from 'list' to 'status' to 'static-list' if those actions are not advertised in the metadata + fencing: Cache metadata lookups to avoid repeated blocking during device registration + fencing: Correctly record which peer performed the fencing operation + fencing: default to 'off' when agent does not advertise 'reboot' in metadata + fencing: Do not unregister/register all stonith devices on every resource agent change + fencing: Execute all required fencing devices regardless of what topology level they are at + fencing: Fence using all required devices + fencing: Pass the correct options when looking up the history by node name + fencing: Update stonith device list only if stonith is enabled + get_cluster_type: failing concurrent tool invocations on heartbeat + ignore SIGPIPE when gnutls is in use + iso8601: Different logic is needed when logging and calculating durations + iso8601: Fix memory leak in duration calculation + Logging: Bootstrap daemon logging before processing arguments but configure it afterwards + lrmd: Cancel recurring operations before stop action is executed + lrmd: Expose logging variables expected by OCF agents + lrmd: Handle systemd reporting 'done' before a resource is actually stopped/started + lrmd: Merge duplicate recurring monitor operations + lrmd: Prevent OCF agents from logging to random files due to "value" of setenv() being NULL + lrmd: Provide stderr output from agents if available, otherwise fall back to stdout + mainloop: Better handle the killing of processes in the act of exiting + mainloop: Canceling in-flight operations should not fail if child process has already exited. + mainloop: Fixes use after free in process monitor code + mcp: Tell systemd not to respawn us if we exit with rc=100 + membership: Avoid duplicate peer entries in the peer cache + pengine: Allow container nodes to migrate with connection resource + pengine: avoid assert by searching for stop action on correct node during LogActions + pengine: Block restart of resources if any dependent resource in a group is unmanaged + pengine: cl#5186 - Avoid running rsc on two nodes when node is fenced during migration + pengine: cl#5187 - Prevent resources in an anti-colocation from even temporarily running on a same node + pengine: cl#5200 - Before migrating utilization-using resources to a node, take off the load that will no longer run there if it's not introducing transition loop + pengine: Correctly handle origin offsets in the future + pengine: Correctly observe requires=nothing + pengine: Default sequential to TRUE for resource sets for consistency with colocation sets + pengine: Delay unfencing until after we know the state of all resources that require unfencing + pengine: Do not initiate fencing for unclean nodes when fencing is disabled + pengine: Ensure instance numbers are preserved for cloned templates + pengine: Ensure unfencing only happens once, even if the transition is interrupted + pengine: Fencing devices default to only requiring quorum in order to start + pengine: fixes invalid transition caused by clones with more than 10 instances + pengine: Force record pending for migrate_to actions + pengine: handles edge case where container order constraints are not honored during migration + pengine: Ignore failure-timeout only if the failed operation has on-fail="block" + pengine: Mark unrunnable stop actions as "blocked" and show the correct current locations + pengine: Memory leaks + pengine: properly handle fencing of container remote-nodes when the container is orphaned + pengine: properly place resource within a container when container is a remote-node. + pengine: Unfencing is based on device probes, there is no need to unfence when normal resources are found active + pengine: Use "#cluster-name" in rules for setting cluster-specific instance attributes + pengine: Use "#site-name" in rules for setting site-specific instance attributes + remote: Allow baremetal remote-node connection resources to migrate + remote: clear remote-node status correctly + remote: Enable migration support for baremetal connection resources by default + remote: Handle request/response ipc proxy correctly + services: Correctly reset the nice value for lrmd's children + services: Do not allow duplicate recurring op entries + services: Do not block synced service executions + services: Fixes segfault associated with cancelling in-flight recurring operations. + services: Remove cancelled recurring ops from internal lists as early as possible + services: Remove file descriptors from mainloop as soon as we have drained them + services: Reset the scheduling policy and priority for lrmd's children without replying on SCHED_RESET_ON_FORK + services_action_cancel: Interpret return code from mainloop_child_kill() correctly + stonith_admin: Ensure pointers passed to sscanf() are properly initialized + stonith_api_time_helper now returns when the most recent fencing operation completed + systemd: Prevent use-of-NULL when determining if an agent exists + systemd: Try to handle dbus actions that complete prior to configuring a callback + Tools: Non-daemons shouldn't abort just because xml parsing failed + Upstart: Allow comilation with glib versions older than 2.28 + Upstart: Do not attempt upstart jobs if we cannot connect to dbus + When data was old, it fixed so that the newest cib might not be acquired. + xml: Check all available schemas when doing upgrades + xml: Correctly determine the lowest allowed schema version + xml: Correctly enforce ACLs after a replace operation + xml: Correctly infer attribute changes after a replace operation + xml: Create the correct diff when only part of a document is changed + xml: Detect attribute ordering changes + xml: Detect content that is added and removed in the same update + xml: Do not prune meaningful leaves from v1 patchsets + xml: Empty patchsets are considered to have applied cleanly + xml: Ensure patches always have version details set + xml: Find the minimal set of changes when part of a document is replaced + xml: If validate-with is missing, we find the most recent schema that accepts it and go from there + xml: Introduce a 'move' primitive for v2 patch sets + xml: Preserve the attribute order in the patch for subsequent digest validation + xml: Resolve memory leak when logging xml blobs + xml: Update xml validation to allow '' * Thu Feb 13 2014 David Vossel Pacemaker-1.1.11-1 - Update source tarball to revision: 33f9d09 - Changesets: 462 - Diff: 147 files changed, 6810 insertions(+), 4057 deletions(-) - Features added since Pacemaker-1.1.10 + attrd: A truly atomic version of attrd for use where CPG is used for cluster communication + cib: Allow values to be added/updated and removed in a single update + cib: Support XML comments in diffs + Core: Allow blackbox logging to be disabled with SIGUSR2 + crmd: Do not block on proxied calls from pacemaker_remoted + crmd: Enable cluster-wide throttling when the cib heavily exceeds its target load + crmd: Make the per-node action limit directly configurable in the CIB + crmd: Slow down recovery on nodes with IO load + crmd: Track CPU usage on cluster nodes and slow down recovery on nodes with high CPU/IO load + crm_mon: add --hide-headers option to hide all headers + crm_node: Display partition output in sorted order + crm_report: Collect logs directly from journald if available + Fencing: On timeout, clean up the agent's entire process group + Fencing: Support agents that need the host to be unfenced at startup + ipc: Raise the default buffer size to 128k + PE: Add a special attribute for distinguishing between real nodes and containers in constraint rules + PE: Allow location constraints to take a regex pattern to match against resource IDs + pengine: Distinguish between the agent being missing and something the agent needs being missing + remote: Properly version the remote connection protocol - Changes since Pacemaker-1.1.10 + Bug rhbz#1011618 - Consistently use 'Slave' as the role for unpromoted master/slave resources + Bug rhbz#1057697 - Use native DBus library for systemd and upstart support to avoid problematic use of threads + attrd: Any variable called 'cluster' makes the daemon crash before reaching main() + attrd: Avoid infinite write loop for unknown peers + attrd: Drop all attributes for peers that left the cluster + attrd: Give remote-nodes ability to set attributes with attrd + attrd: Prevent inflation of attribute dampen intervals + attrd: Support SI units for attribute dampening + Bug cl#5171 - pengine: Don't prevent clones from running due to dependent resources + Bug cl#5179 - Corosync: Attempt to retrieve a peer's node name if it is not already known + Bug cl#5181 - corosync: Ensure node IDs are written to the CIB as unsigned integers + Bug rhbz#902407 - crm_resource: Handle --ban for master/slave resources as advertised + cib: Correctly check for archived configuration files + cib: Correctly log short-form xml diffs + cib: Fix remote cib based on TLS + cibadmin: Report errors during sign-off + cli: Do not enabled blackbox for cli tools + cluster: Fix segfault on removing a node + cman: Do not start pacemaker if cman startup fails + cman: Start clvmd and friends from the init script if enabled + Command-line tools should stop after an assertion failure + controld: Use the correct variant of dlm_controld for corosync-2 clusters + cpg: Correctly set the group name length + cpg: Ensure the CPG group is always null-terminated + cpg: Only process one message at a time to allow other priority jobs to be performed + crmd: Correctly observe the configured batch-limit + crmd: Correctly update expected state when the previous DC shuts down + crmd: Correcty update the history cache when recurring ops change their return code + crmd: Don't add node_state to cib, if we have not seen or fenced this node yet + crmd: don't segfault on shutdown when using heartbeat + crmd: Prevent recurring monitors being cancelled due to notify operations + crmd: Reliably detect and act on reprobe operations from the policy engine + crmd: When a peer expectedly shuts down, record the new join and expected states into the cib + crmd: When the DC gracefully shuts down, record the new expected state into the cib + crm_attribute: Do not swallow hostname lookup failures + crm_mon: Do not display duplicates of failed actions + crm_mon: Reduce flickering in interactive mode + crm_resource: Observe --master modifier for --move + crm_resource: Provide a meaningful error if --master is used for primitives and groups + fencing: Allow fencing for node after topology entries are deleted + fencing: Apply correct score to the resource of group + fencing: Ignore changes to non-fencing resources + fencing: Observe pcmk_host_list during automatic unfencing + fencing: Put all fencing agent processes into their own process group + fencing: Wait until all possible replies are recieved before continuing with unverified devices + ipc: Compress msgs based on client's actual max send size + ipc: Have the ipc server enforce a minimum buffer size all clients must use. + iso8601: Prevent dates from jumping backwards a day in some timezones + lrmd: Correctly calculate metadata for the 'service' class + lrmd: Correctly cancel monitor actions for lsb/systemd/service resources on cleaning up + mcp: Remove LSB hints that instruct chkconfig to start pacemaker at boot time + mcp: Some distros complain when LSB scripts do not include Default-Start/Stop directives + pengine: Allow fencing of baremetal remote nodes + pengine: cl#5186 - Avoid running rsc on two nodes when node is fenced during migration + pengine: Correctly account for the location preferences of things colocated with a group + pengine: Correctly handle demotion of grouped masters that are partially demoted + pengine: Disable container node probes due to constraint conflicts + pengine: Do not allow colocation with blocked clone instances + pengine: Do not re-allocate clone instances that are blocked in the Stopped state + pengine: Do not restart resources that depend on unmanaged resources + pengine: Force record pending for migrate_to actions + pengine: Location constraints with role=Started should prevent masters from running at all + pengine: Order demote/promote of resources on remote nodes to happen only once the connection is up + pengine: Properly handle orphaned multistate resources living on remote-nodes + pengine: Properly shutdown orphaned remote connection resources + pengine: Recover unexpectedly running container nodes. + remote: Add support for ipv6 into pacemaker_remote daemon + remote: Handle endian changes between client and server and improve forward compatibility + services: Fixes segfault associated with cancelling in-flight recurring operations. + services: Reset the scheduling policy and priority for lrmd's children without replying on SCHED_RESET_ON_FORK * Fri Jul 26 2013 Andrew Beekhof Pacemaker-1.1.10-1 - Update source tarball to revision: ab2e209 - Changesets: 602 - Diff: 143 files changed, 8162 insertions(+), 5159 deletions(-) - Features added since Pacemaker-1.1.9 + Core: Convert all exit codes to positive errno values + crm_error: Add the ability to list and print error symbols + crm_resource: Allow individual resources to be reprobed + crm_resource: Allow options to be set recursively + crm_resource: Implement --ban for moving resources away from nodes and --clear (replaces --unmove) + crm_resource: Support OCF tracing when using --force-(check|start|stop) + PE: Allow active nodes in our current membership to be fenced without quorum + PE: Suppress meaningless IDs when displaying anonymous clone status + Turn off auto-respawning of systemd services when the cluster starts them + Bug cl#5128 - pengine: Support maintenance mode for a single node - Changes since Pacemaker-1.1.9 + crmd: cib: stonithd: Memory leaks resolved and improved use of glib reference counting + attrd: Fixes deleted attributes during dc election + Bug cf#5153 - Correctly display clone failcounts in crm_mon + Bug cl#5133 - pengine: Correctly observe on-fail=block for failed demote operation + Bug cl#5148 - legacy: Correctly remove a node that used to have a different nodeid + Bug cl#5151 - Ensure node names are consistently compared without case + Bug cl#5152 - crmd: Correctly clean up fenced nodes during membership changes + Bug cl#5154 - Do not expire failures when on-fail=block is present + Bug cl#5155 - pengine: Block the stop of resources if any depending resource is unmanaged + Bug cl#5157 - Allow migration in the absence of some colocation constraints + Bug cl#5161 - crmd: Prevent memory leak in operation cache + Bug cl#5164 - crmd: Fixes crash when using pacemaker-remote + Bug cl#5164 - pengine: Fixes segfault when calculating transition with remote-nodes. + Bug cl#5167 - crm_mon: Only print "stopped" node list for incomplete clone sets + Bug cl#5168 - Prevent clones from being bounced around the cluster due to location constraints + Bug cl#5170 - Correctly support on-fail=block for clones + cib: Correctly read back archived configurations if the primary is corrupted + cib: The result is not valid when diffs fail to apply cleanly for CLI tools + cib: Restore the ability to embed comments in the configuration + cluster: Detect and warn about node names with capitals + cman: Do not pretend we know the state of nodes we've never seen + cman: Do not unconditionally start cman if it is already running + cman: Support non-blocking CPG calls + Core: Ensure the blackbox is saved on abnormal program termination + corosync: Detect the loss of members for which we only know the nodeid + corosync: Do not pretend we know the state of nodes we've never seen + corosync: Ensure removed peers are erased from all caches + corosync: Nodes that can persist in sending CPG messages must be alive afterall + crmd: Do not get stuck in S_POLICY_ENGINE if a node we couldn't fence returns + crmd: Do not update fail-count and last-failure for old failures + crmd: Ensure all membership operations can complete while trying to cancel a transition + crmd: Ensure operations for cleaned up resources don't block recovery + crmd: Ensure we return to a stable state if there have been too many fencing failures + crmd: Initiate node shutdown if another node claims to have successfully fenced us + crmd: Prevent messages for remote crmd clients from being relayed to wrong daemons + crmd: Properly handle recurring monitor operations for remote-node agent + crmd: Store last-run and last-rc-change for all operations + crm_mon: Ensure stale pid files are updated when a new process is started + crm_report: Correctly collect logs when 'uname -n' reports fully qualified names + fencing: Fail the operation once all peers have been exhausted + fencing: Restore the ability to manually confirm that fencing completed + ipc: Allow unpriviliged clients to clean up after server failures + ipc: Restore the ability for members of the haclient group to connect to the cluster + legacy: Support "crm_node --remove" with a node name for corosync plugin (bnc#805278) + lrmd: Default to the upstream location for resource agent scratch directory + lrmd: Pass errors from lsb metadata generation back to the caller + pengine: Correctly handle resources that recover before we operate on them + pengine: Delete the old resource state on every node whenever the resource type is changed + pengine: Detect constraints with inappropriate actions (ie. promote for a clone) + pengine: Ensure per-node resource parameters are used during probes + pengine: If fencing is unavailable or disabled, block further recovery for resources that fail to stop + pengine: Implement the rest of get_timet_now() and rename to get_effective_time + pengine: Re-initiate _active_ recurring monitors that previously failed but have timed out + remote: Workaround for inconsistent tls handshake behavior between gnutls versions + systemd: Ensure we get shut down correctly by systemd + systemd: Reload systemd after adding/removing override files for cluster services + xml: Check for and replace non-printing characters with their octal equivalent while exporting xml text + xml: Prevent lockups by setting a more reliable buffer allocation strategy * Fri Mar 08 2013 Andrew Beekhof Pacemaker-1.1.9-1 - Update source tarball to revision: 7e42d77 - Statistics: Changesets: 731 Diff: 1301 files changed, 92909 insertions(+), 57455 deletions(-) - Features added in Pacemaker-1.1.9 + corosync: Allow cman and corosync 2.0 nodes to use a name other than uname() + corosync: Use queues to avoid blocking when sending CPG messages + ipc: Compress messages that exceed the configured IPC message limit + ipc: Use queues to prevent slow clients from blocking the server + ipc: Use shared memory by default + lrmd: Support nagios remote monitoring + lrmd: Pacemaker Remote Daemon for extending pacemaker functionality outside corosync cluster. + pengine: Check for master/slave resources that are not OCF agents + pengine: Support a 'requires' resource meta-attribute for controlling whether it needs quorum, fencing or nothing + pengine: Support for resource container + pengine: Support resources that require unfencing before start - Changes since Pacemaker-1.1.8 + attrd: Correctly handle deletion of non-existant attributes + Bug cl#5135 - Improved detection of the active cluster type + Bug rhbz#913093 - Use crm_node instead of uname + cib: Avoid use-after-free by correctly support cib_no_children for non-xpath queries + cib: Correctly process XML diff's involving element removal + cib: Performance improvements for non-DC nodes + cib: Prevent error message by correctly handling peer replies + cib: Prevent ordering changes when applying xml diffs + cib: Remove text nodes from cib replace operations + cluster: Detect node name collisions in corosync + cluster: Preserve corosync membership state when matching node name/id entries + cman: Force fenced to terminate on shutdown + cman: Ignore qdisk 'nodes' + core: Drop per-user core directories + corosync: Avoid errors when closing failed connections + corosync: Ensure peer state is preserved when matching names to nodeids + corosync: Clean up CMAP connections after querying node name + corosync: Correctly detect corosync 2.0 clusters even if we don't have permission to access it + crmd: Bug cl#5144 - Do not updated the expected status of failed nodes + crmd: Correctly determin if cluster disconnection was abnormal + crmd: Correctly relay messages for remote clients (bnc#805626, bnc#804704) + crmd: Correctly stall the FSA when waiting for additional inputs + crmd: Detect and recover when we are evicted from CPG + crmd: Differentiate between a node that is up and coming up in peer_update_callback() + crmd: Have cib operation timeouts scale with node count + crmd: Improved continue/wait logic in do_dc_join_finalize() + crmd: Prevent election storms caused by getrusage() values being too close + crmd: Prevent timeouts when performing pacemaker level membership negotiation + crmd: Prevent use-after-free of fsa_message_queue during exit + crmd: Store all current actions when stalling the FSA + crm_mon: Do not try to render a blank cib and indicate the previous output is now stale + crm_mon: Fixes crm_mon crash when using snmp traps. + crm_mon: Look for the correct error codes when applying configuration updates + crm_report: Ensure policy engine logs are found + crm_report: Fix node list detection + crm_resource: Have crm_resource generate a valid transition key when sending resource commands to the crmd + date/time: Bug cl#5118 - Correctly convert seconds-since-epoch to the current time + fencing: Attempt to provide more information that just 'generic error' for failed actions + fencing: Correctly record completed but previously unknown fencing operations + fencing: Correctly terminate when all device options have been exhausted + fencing: cov#739453 - String not null terminated + fencing: Do not merge new fencing requests with stale ones from dead nodes + fencing: Do not start fencing until entire device topology is found or query results timeout. + fencing: Do not wait for the query timeout if all replies have arrived + fencing: Fix passing of parameters from CMAN containing '=' + fencing: Fix non-comparison when sorting devices by priority + fencing: On failure, only try a topology device once from the remote level. + fencing: Only try peers for non-topology based operations once + fencing: Retry stonith device for duration of action's timeout period. + heartbeat: Remove incorrect assert during cluster connect + ipc: Bug cl#5110 - Prevent 100% CPU usage when looking for synchronous replies + ipc: Use 50k as the default compression threshold + legacy: Prevent assertion failure on routing ais messages (bnc#805626) + legacy: Re-enable logging from the pacemaker plugin + legacy: Relax the 'active' check for plugin based clusters to avoid false negatives + legacy: Skip peer process check if the process list is empty in crm_is_corosync_peer_active() + mcp: Only define HA_DEBUGLOG to avoid agent calls to ocf_log printing everything twice + mcp: Re-attach to existing pacemaker components when mcp fails + pengine: Any location constraint for the slave role applies to all roles + pengine: Avoid leaking memory when cleaning up failcounts and using containers + pengine: Bug cl#5101 - Ensure stop order is preserved for partially active groups + pengine: Bug cl#5140 - Allow set members to be stopped when the subseqent set has require-all=false + pengine: Bug cl#5143 - Prevent shuffling of anonymous master/slave instances + pengine: Bug rhbz#880249 - Ensure orphan masters are demoted before being stopped + pengine: Bug rhbz#880249 - Teach the PE how to recover masters into primitives + pengine: cl#5025 - Automatically clear failcount for start/monitor failures after resource parameters change + pengine: cl#5099 - Probe operation uses the timeout value from the minimum interval monitor by default (#bnc776386) + pengine: cl#5111 - When clone/master child rsc has on-fail=stop, insure all children stop on failure. + pengine: cl#5142 - Do not delete orphaned children of an anonymous clone + pengine: Correctly unpack active anonymous clones + pengine: Ensure previous migrations are closed out before attempting another one + pengine: Introducing the whitebox container resources feature + pengine: Prevent double-free for cloned primitive from template + pengine: Process rsc_ticket dependencies earlier for correctly allocating resources (bnc#802307) + pengine: Remove special cases for fencing resources + pengine: rhbz#902459 - Remove rsc node status for orphan resources + systemd: Gracefully handle unexpected DBus return types + Replace the use of the insecure mktemp(3) with mkstemp(3) * Thu Sep 20 2012 Andrew Beekhof Pacemaker-1.1.8-1 - Update source tarball to revision: 1a5341f - Statistics: Changesets: 1019 Diff: 2107 files changed, 117258 insertions(+), 73606 deletions(-) - All APIs have been cleaned up and reduced to essentials - Pacemaker now includes a replacement lrmd that supports systemd and upstart agents - Config and state files (cib.xml, PE inputs and core files) have moved to new locations - The crm shell has become a separate project and no longer included with Pacemaker - All daemons/tools now have a unified set of error codes based on errno.h (see crm_error) - Changes since Pacemaker-1.1.7 + Core: Bug cl#5032 - Rewrite the iso8601 date handling code + Core: Correctly extract the version details from a diff + Core: Log blackbox contents, if enabled, when an error occurs + Core: Only LOG_NOTICE and higher are sent to syslog + Core: Replace use of IPC from clplumbing with IPC from libqb + Core: SIGUSR1 now enables blackbox logging, SIGTRAP to write out + Core: Support a blackbox for additional logging detail after crashes/errors + Promote support for advanced fencing logic to the stable schema + Promote support for node starting scores to the stable schema + Promote support for service and systemd to the stable schema + attrd: Differentiate between updating all our attributes and everybody updating all theirs too + attrd: Have single-shot clients wait for an ack before disconnecting + cib: cl#5026 - Synced cib updates should not return until the cpg broadcast is complete. + corosync: Detect when the first corosync has not yet formed and handle it gracefully + corosync: Obtain a full list of configured nodes, including their names, when we connect to the quorum API + corosync: Obtain a node name from DNS if one was not already known + corosync: Populate the cib nodelist from corosync if available + corosync: Use the CFG API and DNS to determine node names if not configured in corosync.conf + crmd: Block after 10 failed fencing attempts for a node + crmd: cl#5051 - Fixes file leak in PE ipc connection initialization. + crmd: cl#5053 - Fixes fail-count not being updated properly. + crmd: cl#5057 - Restart sub-systems correctly (bnc#755671) + crmd: cl#5068 - Fixes crm_node -R option so it works with corosync 2.0 + crmd: Correctly re-establish failed attrd connections + crmd: Detect when the quorum API isn't configured for corosync 2.0 + crmd: Do not overwrite any configured node type (eg. quorum node) + crmd: Enable use of new lrmd daemon and client library in crmd. + crmd: Overhaul the way node state is recorded and updated in the CIB + fencing: Bug rhbz#853537 - Prevent use-of-NULL when the cib libraries are not available + fencing: cl#5073 - Add 'off' as an valid value for stonith-action option. + fencing: cl#5092 - Always timeout stonith operations if timeout period expires. + fencing: cl#5093 - Stonith per device timeout option + fencing: Clean up if we detect a failed connection + fencing: Delegate complex self fencing requests - we wont be around to see it to completion + fencing: Ensure all peers are notified of complex fencing op completion + fencing: Fix passing of fence_legacy parameters containing '=' + fencing: Gracefully handle metadata requests for unknown agents + fencing: Return cached dynamic target list for busy devices. + fencing: rhbz#801355 - Abort transition on DC when external fencing operation is detected + fencing: rhbz#801355 - Merge fence requests for identical operations already in progress. + fencing: rhbz#801355 - Report fencing operations external of pacemaker to cib + fencing: Specify the action to perform using action= instead of the older option= + fencing: Stop building fake metadata for broken agents + fencing: Tolerate agents that report empty metadata in the admin tool + mcp: Correctly retry the connection to corosync on failure + mcp: Do not shut down IPC until the last client exits + mcp: Prevent use-after-free when running against corosync 1.x + pengine: Bug cl#5059 - Use the correct action's status when calculating required actions for interleaved clones + pengine: Bypass online/offline checking resource detection for ping/quorum nodes + pengine: cl#5044 - migrate_to no longer requires load_stopped for avoiding possible transition loop + pengine: cl#5069 - Honor 'on-fail=ignore' even when operation is disabled. + pengine: cl#5070 - Allow influence of promotion score when multistate rsc is left hand of colocation + pengine: cl#5072 - Fixes monitor op stopping after rsc promotion. + pengine: cl#5072 - Fixes pengine regression test failures + pengine: Correctly set the status for nodes not intended to run Pacemaker + pengine: Do not append instance numbers to anonymous clones + pengine: Fix failcount expiration + pengine: Fix memory leaks found by valgrind + pengine: Fix use-after-free and use-of-NULL errors detected by coverity + pengine: Fixes use of colocation scores other than +/- INFINITY + pengine: Improve detection of rejoining nodes + pengine: Prevent use-of-NULL when tracing is enabled + pengine: Stonith resources are allowed to start even if their probes haven't completed on partially active nodes + services: New class called 'service' which expands to the correct (LSB/systemd/upstart) standard + services: Support Asynchronous systemd/upstart actions + Tools: crm_shadow - Bug cl#5062 - Correctly set argv[0] when forking a shell process + Tools: crm_report: Always include system logs (if we can find them) * Wed Mar 28 2012 Andrew Beekhof Pacemaker-1.1.7-1 - Update source tarball to revision: bc7ff2c - Statistics: Changesets: 513 Diff: 1171 files changed, 90472 insertions(+), 19368 deletions(-) - Changes since Pacemaker-1.1.6.1 + ais: Prepare for corosync versions using IPC from libqb + cib: Correctly shutdown in the presence of peers without relying on timers + cib: Don't halt disk writes if the previous digest is missing + cib: Determine when there are no peers to respond to our shutdown request and exit + cib: Ensure no additional messages are processed after we begin terminating + Cluster: Hook up the callbacks to the corosync quorum notifications + Core: basename() may modify its input, do not pass in a constant + Core: Bug cl#5016 - Prevent failures in recurring ops from being lost + Core: Bug rhbz#800054 - Correctly retrieve heartbeat uuids + Core: Correctly determine when an XML file should be decompressed + Core: Correctly track the length of a string without reading from uninitialzied memory (valgrind) + Core: Ensure signals are handled eventually in the absense of timer sources or IPC messages + Core: Prevent use-of-NULL in crm_update_peer() + Core: Strip text nodes from on disk xml files + Core: Support libqb for logging + corosync: Consistently set the correct uuid with get_node_uuid() + Corosync: Correctly disconnect from corosync variants + Corosync: Correctly extract the node id from membership udpates + corosync: Correctly infer lost members from the quorum API + Corosync: Default to using the nodeid as the node's uuid (instead of uname) + corosync: Ensure we catch nodes that leave the membership, even if the ringid doesn't change + corosync: Hook up CPG membership + corosync: Relax a development assert and gracefully handle the error condition + corosync: Remove deprecated member of the CFG API + corosync: Treat CS_ERR_QUEUE_FULL the same as CS_ERR_TRY_AGAIN + corosync: Unset the process list when nodes dissappear on us + crmd: Also purge fencing results when we enter S_NOT_DC + crmd: Bug cl#5015 - Remove the failed operation as well as the resulting fail-count and last-failure attributes + crmd: Correctly determine when a node can suicide with fencing + crmd: Election - perform the age comparison only once + crmd: Fast-track shutdown if we couldn't request it via attrd + crmd: Leave it up to the PE to decide which ops can/cannot be reload + crmd: Prevent use-after-free when calling delete_resource due to CRM_OP_REPROBE + crmd: Supply format arguments in the correct order + fencing: Add missing format parameter + fencing: Add the fencing topology section to the 1.1 configuration schema + fencing: fence_legacy - Drop spurilous host argument from status query + fencing: fence_legacy - Ensure port is available as an environment variable when calling monitor + fencing: fence_pcmk - don't block if nothing is specified on stdin + fencing: Fix log format error + fencing: Fix segfault caused by passing garbage to dlsym() + fencing: Fix use-of-NULL in process_remote_stonith_query() + fencing: Fix use-of-NULL when listing installed devices + fencing: Implement support for advanced fencing topologies: eg. kdump || (network && disk) || power + fencing: More gracefully handle failed 'list' operations for devices that only support a single connection + fencing: Prevent duplicate free when listing devices + fencing: Prevent uninitialized pointers being passed to free + fencing: Prevent use-after-free, we may need the query result for subsequent operations + fencing: Provide enough data to construct an entry in the node's fencing history + fencing: Standardize on /one/ method for clients to request members be fenced + fencing: Supress errors when listing all registered devices + mcp: corosync_cfg_state_track was removed from the corosync API, luckily we didnt use it for anything + mcp: Do not specify a WorkingDirectory in the systemd unit file - startup fails if its not available + mcp: Set the HA_quorum_type env variable consistently with our corosync plugin + mcp: Shut down if one of our child processes can/should not be respawned + pengine: Bug cl#5000 - Ensure ordering is preserved when depending on partial sets + pengine: Bug cl#5028 - Unmanaged services should block shutdown unless in maintenance mode + pengine: Bug cl#5038 - Prevent restart of anonymous clones when clone-max decreases + pengine: Bug cl#5007 - Fixes use of colocation constraints with multi-state resources + pengine: Bug cl#5014 - Prevent asymmetrical order constraints from causing resource stops + pengine: Bug cl#5000 - Implements ability to create rsc_order constraint sets such that A can start after B or C has started. + pengine: Correctly migrate a resource that has just migrated + pengine: Correct return from error path + pengine: Detect reloads of previously migrated resources + pengine: Ensure post-migration stop actions occur before node shutdown + pengine: Log as loudly as possible when we cannot shut down a cluster node + pengine: Reload of a resource no longer causes a restart of dependent resources + pengine: Support limiting the number of concurrent live migrations + pengine: Support referencing templates in constraints + pengine: Support of referencing resource templates in resource sets + pengine: Support to make tickets standby for relinquishing tickets gracefully + stonith: A "start" operation of a stonith resource does a "monitor" on the device beyond registering it + stonith: Bug rhbz#745526 - Ensure stonith_admin actually gets called by fence_pcmk + Stonith: Ensure all nodes receive and deliver notifications of the manual override + stonith: Fix the stonith timeout issue (cl#5009, bnc#727498) + Stonith: Implement a manual override for when nodes are known to be safely off + Tools: Bug cl#5003 - Prevent use-after-free in crm_simlate + Tools: crm_mon - Support to display tickets (based on Yuusuke Iida's work) + Tools: crm_simulate - Support to grant/revoke/standby/activate tickets from the new ticket state section + Tools: Implement crm_node functionality for native corosync + Fix a number of potential problems reported by coverity * Wed Aug 31 2011 Andrew Beekhof 1.1.6-1 - Update source tarball to revision: 676e5f25aa46 tip - Statistics: Changesets: 376 Diff: 1761 files changed, 36259 insertions(+), 140578 deletions(-) - Changes since Pacemaker-1.1.5 + ais: check for retryable errors when dispatching AIS messages + ais: Correctly disconnect from Corosync and Cman based clusters + ais: Followup to previous patch - Ensure we drain the corosync queue of messages when Glib tells us there is input + ais: Handle IPC error before checking for NULL data (bnc#702907) + cib: Check the validation version before adding the originator details of a CIB change + cib: Remove disconnected remote connections from mainloop + cman: Correctly override existing fenced operations + cman: Dequeue all the cman emitted events and not only the first one leaving the others in the event's queue. + cman: Don't call fenced_join and fenced_leave when notifying cman of a fencing event. + cman: We need to run the crmd as root for CMAN so that we can ACK fencing operations + Core: Cancelled and pending operations do not count as failed + Core: Ensure there is sufficient space for EOS when building short-form option strings + Core: Fix variable expansion in pkg-config files + Core: Partial revert of accidental commit in previous patch + Core: Use dlopen to load heartbeat libraries on-demand + crmd: Bug lf#2509 - Watch for config option changes from the CIB even if we're not the DC + crmd: Bug lf#2528 - Introduce a slight delay when creating a transition to allow attrd time to perform its updates + crmd: Bug lf#2559 - Fail actions that were scheduled for a failed/fenced node + crmd: Bug lf#2584 - Allow nodes to fence themselves if they're the last one standing + crmd: Bug lf#2632 - Correctly handle nodes that return faster than stonith + crmd: Cancel timers for actions that were pending on dead nodes + crmd: Catch fence operations that claim to succeed but did not really + crmd: Do not wait for actions that were pending on dead nodes + crmd: Ensure we do not attempt to perform action on failed nodes + crmd: Prevent use-of-NULL by g_hash_table_iter_next() + crmd: Recurring actions shouldn't cause the last non-recurring action to be forgotten + crmd: Store only the last and last failed operation in the CIB + mcp: dirname() modifies the input path - pass in a copy of the logfile path + mcp: Enable stack detection logic instead of forcing 'corosync' + mcp: Fix spelling mistake in systemd service script that prevents shutdown + mcp: Shut down if corosync becomes unavailable + mcp: systemd control file is now functional + pengine: Before migrating an utilization-using resource to a node, take off the load which will no longer run there (lf#2599, bnc#695440) + pengine: Before migrating an utilization-using resource to a node, take off the load which will no longer run there (regression tests) (lf#2599, bnc#695440) + pengine: Bug lf#2574 - Prevent shuffling by choosing the correct clone instance to stop + pengine: Bug lf#2575 - Use uname for migration variables, id is a UUID on heartbeat + pengine: Bug lf#2581 - Avoid group restart when clone (re)starts on an unrelated node + pengine: Bug lf#2613, lf#2619 - Group migration after failures and non-default utilization policies + pengine: Bug suse#707150 - Prevent services being active if dependencies on clones are not satisfied + pengine: Correctly recognise which recurring operations are currently active + pengine: Demote from Master does not clear previous errors + pengine: Ensure restarts due to definition changes cause the start action to be re-issued not probes + pengine: Ensure role is preserved for unmanaged resources + pengine: Ensure unmanaged resources have the correct role set so the correct monitor operation is chosen + pengine: Fix memory leak for re-allocated resources reported by valgrind + pengine: Implement cluster ticket and deadman + pengine: Implement resource template + pengine: Correctly determine the state of multi-state resources with a partial operation history + pengine: Only allocate master/slave resources once + pengine: Partial revert of 'Minor code cleanup CS: cf6bca32376c On: 2011-08-15' + pengine: Resolve memory leak reported by valgrind + pengine: Restore the ability to save inputs to disk + Shell: implement -w,--wait option to wait for the transition to finish + Shell: repair template list command + Shell: set of commands to examine logs, reports, etc + Stonith: Consolidate pcmk_host_map into run_stonith_agent so that it is applied consistently + Stonith: Deprecate pcmk_arg_map for the saner pcmk_host_argument + Stonith: Fix use-of-NULL by g_hash_table_lookup + Stonith: Improved pcmk_host_map parsing + Stonith: Prevent use-of-NULL by g_hash_table_lookup + Stonith: Prevent use-of-NULL when no Linux-HA stonith agents are present + stonith: Add missing entries to stonith_error2string() + Stonith: Correctly finish sending agent options if the initial write is interrupted + stonith: Correctly handle synchronous calls + stonith: Coverity - Correctly construct result list for the query API call + stonith: Coverity - Remove badly constructed memory allocation from the query API call + stonith: Ensure completed operations are recorded as such in the history + Stonith: Ensure device parameters are passed to the daemon during registration + stonith: Fix use-of-NULL in stonith_api_device_list() + stonith: stonith_admin - Prevent use of uninitialized pointer by --history command + Tools: Bug lf#2528 - Make progress when attrd_updater is called repeatedly within the dampen interval but with the same value + Tools: crm_report - Correctly extract data from the local node + Tools: crm_report - Remove newlines when detecting the node list + Tools: crm_report - Repair the ability to extract data from the local machine + Tools: crm_report - Report on all detected backtraces * Fri Feb 11 2011 Andrew Beekhof 1.1.5-1 - Update source tarball to revision: baad6636a053 - Statistics: Changesets: 184 Diff: 605 files changed, 46103 insertions(+), 26417 deletions(-) - Changes since Pacemaker-1.1.4 + Add the ability to delegate sub-sections of the cluster to non-root users via ACLs Needs to be enabled at compile time, not enabled by default. + ais: Bug lf#2550 - Report failed processes immediately + Core: Prevent recently introduced use-after-free in replace_xml_child() + Core: Reinstate the logic that skips past non-XML_ELEMENT_NODE children + Core: Remove extra calls to xmlCleanupParser resulting in use-after-free + Core: Repair reference to child-of-child after removal of xml_child_iter_filter from get_message_xml() + crmd: Bug lf#2545 - Ensure notify variables are accurate for stop operations + crmd: Cancel recurring operations while we're still connected to the lrmd + crmd: Reschedule the PE_START action if its not already running when we try to use it + crmd: Update failcount for failed promote and demote operations + pengine: Bug lf#2445 - Avoid relying on stickness for stable clone placement + pengine: Bug lf#2445 - Do not override configured clone stickiness values + pengine: Bug lf#2493 - Don't imply colocation requirements when applying ordering constraints with clones + pengine: Bug lf#2495 - Prevent segfault by validating the contents of ordering sets + pengine: Bug lf#2508 - Correctly reconstruct the status of anonymous cloned groups + pengine: Bug lf#2518 - Avoid spamming the logs with errors for orphan resources + pengine: Bug lf#2544 - Prevent unstable clone placement by factoring in the current node's score before all others + pengine: Bug lf#2554 - target-role alone is not sufficient to promote resources + pengine: Correct target_rc for probes of inactive resources (fix regression introduced by cs:ac3f03006e95) + pengine: Ensure that fencing has completed for stop actions on stonith-dependent resources (lf#2551) + pengine: Only update the node's promotion score if the resource is active there + pengine: Only use the promotion score from the current clone instance + pengine: Prevent use-of-NULL resulting from variable shadowing spotted by Coverity + pengine: Prevent use-of-NULL when there is status for an undefined node + pengine: Prevet use-after-free resulting from unintended recursion when chosing a node to promote master/slave resources + Shell: don't create empty optional sections (bnc#665131) + Stonith: Teach stonith_admin to automagically obtain the current node attributes for the target from the CIB + tools: Bug lf#2527 - Prevent use-of-NULL in crm_simulate + Tools: Prevent crm_resource commands from being lost due to the use of cib_scope_local * Wed Oct 20 2010 Andrew Beekhof 1.1.4-1 - Update source tarball to revision: 75406c3eb2c1 tip - Statistics: Changesets: 169 Diff: 772 files changed, 56172 insertions(+), 39309 deletions(-) - Changes since Pacemaker-1.1.3 + Italian translation of Clusters from Scratch + Significant performance enhancements to the Policy Engine and CIB + cib: Bug lf#2506 - Don't remove client's when notifications fail, they might just be too big + cib: Drop invalid/failed connections from the client hashtable + cib: Ensure all diffs sent to peers have sufficient ordering information + cib: Ensure non-change diffs can preserve the ordering on the other side + cib: Fix the feature set check + cib: Include version information on our synthesised diffs when nothing changed + cib: Optimize the way we detect group/set ordering changes - 15% speedup + cib: Prevent false detection of config updates with the new diff format + cib: Reduce unnecessary copying when comparing xml objects + cib: Repair the processing of updates sent from peer nodes + cib: Revert part of a recent commit that purged still valid connections + cib: The feature set version check is only valid if the current value is non-NULL + Core: Actually removing diff markers is necessary + Core: Bug lf#2506 - Drop the compression limit because Heartbeat's IPC code sucks + Core: Cache Relax-NG schemas - profiling indicates many cycles are wasted needlessly re-parsing them + Core: Correctly compare against crm_log_level in the logging macros + Core: Correctly extract the version details from a diff + Core: Correctly hook up the RNG schema cache + Core: Correctly use lazy_xml_sort() for v2 digests + Core: Don't compress large payload elements unless we're approaching message limits + Core: Don't insert empty ID tags when applying diffs + Core: Enable the improve v2 digests + Core: Ensure ordering is preserved when applying diffs + Core: Fix the CRM_CHECK macro + Core: Modify the v2 digest algorithm so that some fields are sorted + Core: Prevent use-after-free when creating a CIB update for a timed out action + Core: Prevent use-of-NULL when cleaning up RelaxNG data structures + Core: Provide significant performance improvements by implementing versioned diffs and digests + crmd: All pending operations should be recorded, even recurring ones with high start delays + crmd: Don't abort transitions when probes are completed on a node + crmd: Don't hide stop events that time out - allowing faster recovery in the presence of overloaded hosts + crmd: Ensure the CIB is always writable on the DC by removing a timing hole + crmd: Include the correct transition details for timed out operations + crmd: Prevent use of NULL by making copies of the operation's hash table + crmd: There's no need to check the cib version from the 'added' part of diff updates + crmd: Use the supplied timeout for stop actions + mcp: Ensure valgrind is able to log its output somewhere + mcp: Use 99/01 for the start/stop sequence to avoid problems with services (such as libvirtd) started by init - Patch from Vladislav Bogdanov + pengine: Ensure fencing of the DC preceeds the STONITH_DONE operation + pengine: Fix memory leak introduced as part of the conversion to GHashTables + pengine: Fix memory leak when processing completed migration actions + pengine: Fix typo leading to use-of-NULL in the new ordering code + pengine: Free memory in recently introduced helper function + pengine: lf#2478 - Implement improved handling and recovery of atomic resource migrations + pengine: Obtain massive speedup by prepending to the list of ordering constraints (which can grow quite large) + pengine: Optimize the logic for deciding which non-grouped anonymous clone instances to probe for + pengine: Prevent clones from being stopped because resources colocated with them cannot be active + pengine: Try to ensure atomic migration ops occur within a single transition + pengine: Use hashtables instead of linked lists for performance sensitive datastructures + pengine: Use the original digest algorithm for parameter lists + stonith: cleanup children on timeout in fence_legacy + Stonith: Fix two memory leaks + Tools: crm_shadow - Avoid replacing the entire configuration (including status) * Tue Sep 21 2010 Andrew Beekhof 1.1.3-1 - Update source tarball to revision: e3bb31c56244 tip - Statistics: Changesets: 352 Diff: 481 files changed, 14130 insertions(+), 11156 deletions(-) - Changes since Pacemaker-1.1.2.1 + ais: Bug lf#2401 - Improved processing when the peer crmd processes join/leave + ais: Correct the logic for conecting to plugin based clusters + ais: Do not supply a process list in mcp-mode + ais: Drop support for whitetank in the 1.1 release series + ais: Get an initial dump of the node membership when connecting to quorum-based clusters + ais: Guard against saturated cpg connections + ais: Handle CS_ERR_TRY_AGAIN in more cases + ais: Move the code for finding uid before the fork so that the child does no logging + ais: Never allow quorum plugins to affect connection to the pacemaker plugin + ais: Sign everyone up for peer process updates, not just the crmd + ais: The cluster type needs to be set before initializing classic openais connections + cib: Also free query result for xpath operations that return more than one hit + cib: Attempt to resolve memory corruption when forking a child to write the cib to disk + cib: Correctly free memory when writing out the cib to disk + cib: Fix the application of unversioned diffs + cib: Remove old developmental error logging + cib: Restructure the 'valid peer' check for deciding which instructions to ignore + cman: Correctly process membership/quorum changes from the pcmk plugin. Allow other message types through untouched + cman: Filter directed messages not intended for us + cman: Grab the initial membership when we connect + cman: Keep the list of peer processes up-to-date + cman: Make sure our common hooks are called after a cman membership update + cman: Make sure we can compile without cman present + cman: Populate sender details for cpg messages + cman: Update the ringid for cman based clusters + Core: Correctly unpack HA_Messages containing multiple entries with the same name + Core: crm_count_member() should only track nodes that have the full stack up + Core: New developmental logging system inspired by the kernel and a PoC from Lars Ellenberg + crmd: All nodes should see status updates, not just he DC + crmd: Allow non-DC nodes to clear failcounts too + crmd: Base DC election on process relative uptime + crmd: Bug lf#2439 - cancel_op() can also return HA_RSCBUSY + crmd: Bug lf#2439 - Handle asynchronous notification of resource deletion events + crmd: Bug lf#2458 - Ensure stop actions always have the relevant resource attributes + crmd: Disable age as a criteria for cman based clusters, its not reliable enough + crmd: Ensure we activate the DC timer if we detect an alternate DC + crmd: Factor the nanosecond component of process uptime in elections + crmd: Fix assertion failure when performing async resource failures + crmd: Fix handling of async resource deletion results + crmd: Include the action for crm graph operations + crmd: Make sure the membership cache is accurate after a sucessful fencing operation + crmd: Make sure we always poke the FSA after a transition to clear any TE_HALT actions + crmd: Offer crm-level membership once the peer starts the crmd process + crmd: Only need to request quorum update for plugin based clusters + crmd: Prevent assertion failure for stop actions resulting from cs: 3c0bc17c6daf + crmd: Prevent everyone from loosing DC elections by correctly initializing all relevant variables + crmd: Prevent segmentation fault + crmd: several fixes for async resource delete (thanks to beekhof) + crmd: Use the correct define/size for lrm resource IDs + Introduce two new cluster types 'cman' and 'corosync', replaces 'quorum_provider' concept + mcp: Add missing headers when built without heartbeat support + mcp: Correctly initialize the string containing the list of active daemons + mcp: Fix macro expansion in init script + mcp: Fix the expansion of the pid file in the init script + mcp: Handle CS_ERR_TRY_AGAIN when connecting to libcfg + mcp: Make sure we can compile the mcp without cman present + mcp: New master control process for (re)spawning pacemaker daemons + mcp: Read config early so we can re-initialize logging asap if daemonizing + mcp: Rename the mcp binary to pacemakerd and create a 'pacemaker' init script + mcp: Resend our process list after every CPG change + mcp: Tell chkconfig we need to shut down early on + pengine: Avoid creating invalid ordering constraints for probes that are not needed + pengine: Bug lf#1959 - Fail unmanaged resources should not prevent other services from shutting down + pengine: Bug lf#2422 - Ordering dependencies on partially active groups not observed properly + pengine: Bug lf#2424 - Use notify oepration definition if it exists in the configuration + pengine: Bug lf#2433 - No services should be stopped until probes finish + pengine: Bug lf#2453 - Enforce clone ordering in the absense of colocation constraints + pengine: Bug lf#2476 - Repair on-fail=block for groups and primitive resources + pengine: Correctly detect when there is a real failcount that expired and needs to be cleared + pengine: Correctly handle pseudo action creation + pengine: Correctly order clone startup after group/clone start + pengine: Correct use-after-free introduced in the prior patch + pengine: Do not demote resources because something that requires it can not run + pengine: Fix colocation for interleaved clones + pengine: Fix colocation with partially active groups + pengine: Fix potential use-after-free defect from coverity + pengine: Fix previous merge + pengine: Fix use-after-free in order_actions() reported by valgrind + pengine: Make the current data set a global variable so it does not need to be passed around everywhere + pengine: Prevent endless loop when looking for operation definitions in the configuration + pengine: Prevent segfault by ensuring the arguments to do_calculations() are initialized + pengine: Rewrite the ordering constraint logic to be simplicity, clarity and maintainability + pengine: Wait until stonith is available, do not fall back to shutdown for nodes requesting termination + Resolve coverity RESOURCE_LEAK defects + Shell: Complete the transition to using crm_attribute instead of crm_failcount and crm_standby + stonith: Advertise stonith-ng options in the metadata + stonith: Bug lf#2461 - Prevent segfault by not looking up operations if the hashtable has not been initialized yet + stonith: Bug lf#2473 - Add the timeout at the top level where the daemon is looking for it + Stonith: Bug lf#2473 - Ensure stonith operations complete within the timeout and are terminated if they run too long + stonith: Bug lf#2473 - Ensure timeouts are included for fencing operations + stonith: Bug lf#2473 - Gracefully handle remote operations that arrive late (after we have done notifications) + stonith: Correctly parse pcmk_host_list parameters that appear on a single line + stonith: Map poweron/poweroff back to on/off expected by the stonith tool from cluster-glue + stonith: pass the configuration to the stonith program via environment variables (bnc#620781) + Stonith: Use the timeout specified by the user + Support starting plugin-based Pacemaker clusters with the MCP as well + Tools: Bug lf#2456 - Fix assertion failure in crm_resource + tools: crm_node - Repair the ability to connect to openais based clusters + tools: crm_node - Use the correct short option for --cman + tools: crm_report - corosync.conf wont necessarily contain the text 'pacemaker' anymore + Tools: crm_simulate - Fix use-after-free in when terminating + tools: crm_simulate - Resolve coverity USE_AFTER_FREE defect + Tools: Drop the 'pingd' daemon and resource agent in favor of ocf:pacemaker:ping + Tools: Fix recently introduced use-of-NULL + Tools: Fix use-after-free defects from coverity * Wed May 12 2010 Andrew Beekhof 1.1.2-1 - Update source tarball to revision: c25c972a25cc tip - Statistics: Changesets: 339 Diff: 708 files changed, 37918 insertions(+), 10584 deletions(-) - Changes since Pacemaker-1.1.1 + ais: Do not count votes from offline nodes and calculate current votes before sending quorum data + ais: Ensure the list of active processes sent to clients is always up-to-date + ais: Look for the correct conf variable for turning on file logging + ais: Need to find a better and thread-safe way to set core_uses_pid. Disable for now. + ais: Use the threadsafe version of getpwnam + Core: Bump the feature set due to the new failcount expiry feature + Core: fix memory leaks exposed by valgrind + Core: Bug lf#2414 - Prevent use-after-free reported by valgrind when doing xpath based deletions + crmd: Bug lf#2414 - Prevent use-after-free of the PE connection after it dies + crmd: Bug lf#2414 - Prevent use-after-free of the stonith-ng connection + crmd: Bug lf#2401 - Improved detection of partially active peers + crmd: Bug lf#2379 - Ensure the cluster terminates when the PE is not available + crmd: Do not allow the target_rc to be misused by resource agents + crmd: Do not ignore action timeouts based on FSA state + crmd: Ensure we don't get stuck in S_PENDING if we lose an election to someone that never talks to us again + crmd: Fix memory leaks exposed by valgrind + crmd: Remove race condition that could lead to multiple instances of a clone being active on a machine + crmd: Send erase_status_tag() calls to the local CIB when the DC is fenced, since there is no DC to accept them + crmd: Use global fencing notifications to prevent secondary fencing operations of the DC + pengine: Bug lf#2317 - Avoid needless restart of primitive depending on a clone + pengine: Bug lf#2361 - Ensure clones observe mandatory ordering constraints if the LHS is unrunnable + pengine: Bug lf#2383 - Combine failcounts for all instances of an anonymous clone on a host + pengine: Bug lf#2384 - Fix intra-set colocation and ordering + pengine: Bug lf#2403 - Enforce mandatory promotion (colocation) constraints + pengine: Bug lf#2412 - Correctly find clone instances by their prefix + pengine: Do not be so quick to pull the trigger on nodes that are coming up + pengine: Fix memory leaks exposed by valgrind + pengine: Rewrite native_merge_weights() to avoid Fix use-after-free + Shell: Bug bnc#590035 - always reload status if working with the cluster + Shell: Bug bnc#592762 - Default to using the status section from the live CIB + Shell: Bug lf#2315 - edit multiple meta_attributes sets in resource management + Shell: Bug lf#2221 - enable comments + Shell: Bug bnc#580492 - implement new cibstatus interface and commands + Shell: Bug bnc#585471 - new cibstatus import command + Shell: check timeouts also against the default-action-timeout property + Shell: new configure filter command + Tools: crm_mon - fix memory leaks exposed by valgrind * Tue Feb 16 2010 Andrew Beekhof - 1.1.1-1 - First public release of Pacemaker 1.1 - Package reference documentation in a doc subpackage - Move cts into a subpackage so that it can be easily consumed by others - Update source tarball to revision: 17d9cd4ee29f + New stonith daemon that supports global notifications + Service placement influenced by the physical resources + A new tool for simulating failures and the cluster’s reaction to them + Ability to serialize an otherwise unrelated a set of resource actions (eg. Xen migrations) * Mon Jan 18 2010 Andrew Beekhof - 1.0.7-1 - Update source tarball to revision: 2eed906f43e9 (stable-1.0) tip - Statistics: Changesets: 193 Diff: 220 files changed, 15933 insertions(+), 8782 deletions(-) - Changes since 1.0.5-4 + pengine: Bug 2213 - Ensure groups process location constraints so that clone-node-max works for cloned groups + pengine: Bug lf#2153 - non-clones should not restart when clones stop/start on other nodes + pengine: Bug lf#2209 - Clone ordering should be able to prevent startup of dependent clones + pengine: Bug lf#2216 - Correctly identify the state of anonymous clones when deciding when to probe + pengine: Bug lf#2225 - Operations that require fencing should wait for 'stonith_complete' not 'all_stopped'. + pengine: Bug lf#2225 - Prevent clone peers from stopping while another is instance is (potentially) being fenced + pengine: Correctly anti-colocate with a group + pengine: Correctly unpack ordering constraints for resource sets to avoid graph loops + Tools: crm: load help from crm_cli.txt + Tools: crm: resource sets (bnc#550923) + Tools: crm: support for comments (LF 2221) + Tools: crm: support for description attribute in resources/operations (bnc#548690) + Tools: hb2openais: add EVMS2 CSM processing (and other changes) (bnc#548093) + Tools: hb2openais: do not allow empty rules, clones, or groups (LF 2215) + Tools: hb2openais: refuse to convert pure EVMS volumes + cib: Ensure the loop for login message terminates + cib: Finally fix reliability of receiving large messages over remote plaintext connections + cib: Fix remote notifications + cib: For remote connections, default to CRM_DAEMON_USER since thats the only one that the cib can validate the password for using PAM + cib: Remote plaintext - Retry sending parts of the message that did not fit the first time + crmd: Ensure batch-limit is correctly enforced + crmd: Ensure we have the latest status after a transition abort + (bnc#547579,547582): Tools: crm: status section editing support + shell: Add allow-migrate as allowed meta-attribute (bnc#539968) + Medium: Build: Do not automatically add -L/lib, it could cause 64-bit arches to break + Medium: pengine: Bug lf#2206 - rsc_order constraints always use score at the top level + Medium: pengine: Only complain about target-role=master for non m/s resources + Medium: pengine: Prevent non-multistate resources from being promoted through target-role + Medium: pengine: Provide a default action for resource-set ordering + Medium: pengine: Silently fix requires=fencing for stonith resources so that it can be set in op_defaults + Medium: Tools: Bug lf#2286 - Allow the shell to accept template parameters on the command line + Medium: Tools: Bug lf#2307 - Provide a way to determin the nodeid of past cluster members + Medium: Tools: crm: add update method to template apply (LF 2289) + Medium: Tools: crm: direct RA interface for ocf class resource agents (LF 2270) + Medium: Tools: crm: direct RA interface for stonith class resource agents (LF 2270) + Medium: Tools: crm: do not add score which does not exist + Medium: Tools: crm: do not consider warnings as errors (LF 2274) + Medium: Tools: crm: do not remove sets which contain id-ref attribute (LF 2304) + Medium: Tools: crm: drop empty attributes elements + Medium: Tools: crm: exclude locations when testing for pathological constraints (LF 2300) + Medium: Tools: crm: fix exit code on single shot commands + Medium: Tools: crm: fix node delete (LF 2305) + Medium: Tools: crm: implement -F (--force) option + Medium: Tools: crm: rename status to cibstatus (LF 2236) + Medium: Tools: crm: revisit configure commit + Medium: Tools: crm: stay in crm if user specified level only (LF 2286) + Medium: Tools: crm: verify changes on exit from the configure level + Medium: ais: Some clients such as gfs_controld want a cluster name, allow one to be specified in corosync.conf + Medium: cib: Clean up logic for receiving remote messages + Medium: cib: Create valid notification control messages + Medium: cib: Indicate where the remote connection came from + Medium: cib: Send password prompt to stderr so that stdout can be redirected + Medium: cts: Fix rsh handling when stdout is not required + Medium: doc: Fill in the section on removing a node from an AIS-based cluster + Medium: doc: Update the docs to reflect the 0.6/1.0 rolling upgrade problem + Medium: doc: Use Publican for docbook based documentation + Medium: fencing: stonithd: add metadata for stonithd instance attributes (and support in the shell) + Medium: fencing: stonithd: ignore case when comparing host names (LF 2292) + Medium: tools: Make crm_mon functional with remote connections + Medium: xml: Add stopped as a supported role for operations + Medium: xml: Bug bnc#552713 - Treat node unames as text fields not IDs + Medium: xml: Bug lf#2215 - Create an always-true expression for empty rules when upgrading from 0.6 * Thu Oct 29 2009 Andrew Beekhof - 1.0.5-4 - Include the fixes from CoroSync integration testing - Move the resource templates - they are not documentation - Ensure documentation is placed in a standard location - Exclude documentation that is included elsewhere in the package - Update the tarball from upstream to version ee19d8e83c2a + cib: Correctly clean up when both plaintext and tls remote ports are requested + pengine: Bug bnc#515172 - Provide better defaults for lt(e) and gt(e) comparisions + pengine: Bug lf#2197 - Allow master instances placemaker to be influenced by colocation constraints + pengine: Make sure promote/demote pseudo actions are created correctly + pengine: Prevent target-role from promoting more than master-max instances + ais: Bug lf#2199 - Prevent expected-quorum-votes from being populated with garbage + ais: Prevent deadlock - don't try to release IPC message if the connection failed + cib: For validation errors, send back the full CIB so the client can display the errors + cib: Prevent use-after-free for remote plaintext connections + crmd: Bug lf#2201 - Prevent use-of-NULL when running heartbeat * Wed Oct 13 2009 Andrew Beekhof - 1.0.5-3 - Update the tarball from upstream to version 38cd629e5c3c + Core: Bug lf#2169 - Allow dtd/schema validation to be disabled + pengine: Bug lf#2106 - Not all anonymous clone children are restarted after configuration change + pengine: Bug lf#2170 - stop-all-resources option had no effect + pengine: Bug lf#2171 - Prevent groups from starting if they depend on a complex resource which can not + pengine: Disable resource management if stonith-enabled=true and no stonith resources are defined + pengine: do not include master score if it would prevent allocation + ais: Avoid excessive load by checking for dead children every 1s (instead of 100ms) + ais: Bug rh#525589 - Prevent shutdown deadlocks when running on CoroSync + ais: Gracefully handle changes to the AIS nodeid + crmd: Bug bnc#527530 - Wait for the transition to complete before leaving S_TRANSITION_ENGINE + crmd: Prevent use-after-free with LOG_DEBUG_3 + Medium: xml: Mask the "symmetrical" attribute on rsc_colocation constraints (bnc#540672) + Medium (bnc#520707): Tools: crm: new templates ocfs2 and clvm + Medium: Build: Invert the disable ais/heartbeat logic so that --without (ais|heartbeat) is available to rpmbuild + Medium: pengine: Bug lf#2178 - Indicate unmanaged clones + Medium: pengine: Bug lf#2180 - Include node information for all failed ops + Medium: pengine: Bug lf#2189 - Incorrect error message when unpacking simple ordering constraint + Medium: pengine: Correctly log resources that would like to start but can not + Medium: pengine: Stop ptest from logging to syslog + Medium: ais: Include version details in plugin name + Medium: crmd: Requery the resource metadata after every start operation * Fri Aug 21 2009 Tomas Mraz - 1.0.5-2.1 - rebuilt with new openssl * Wed Aug 19 2009 Andrew Beekhof - 1.0.5-2 - Add versioned perl dependency as specified by https://fedoraproject.org/wiki/Packaging/Perl#Packages_that_link_to_libperl - No longer remove RPATH data, it prevents us finding libperl.so and no other libraries were being hardcoded - Compile in support for heartbeat - Conditionally add heartbeat-devel and corosynclib-devel to the -devel requirements depending on which stacks are supported * Mon Aug 17 2009 Andrew Beekhof - 1.0.5-1 - Add dependency on resource-agents - Use the version of the configure macro that supplies --prefix, --libdir, etc - Update the tarball from upstream to version 462f1569a437 (Pacemaker 1.0.5 final) + Tools: crm_resource - Advertise --move instead of --migrate + Medium: Extra: New node connectivity RA that uses system ping and attrd_updater + Medium: crmd: Note that dc-deadtime can be used to mask the brokeness of some switches * Tue Aug 11 2009 Ville Skyttä - 1.0.5-0.7.c9120a53a6ae.hg - Use bzipped upstream tarball. * Wed Jul 29 2009 Andrew Beekhof - 1.0.5-0.6.c9120a53a6ae.hg - Add back missing build auto* dependencies - Minor cleanups to the install directive * Tue Jul 28 2009 Andrew Beekhof - 1.0.5-0.5.c9120a53a6ae.hg - Add a leading zero to the revision when alphatag is used * Tue Jul 28 2009 Andrew Beekhof - 1.0.5-0.4.c9120a53a6ae.hg - Incorporate the feedback from the cluster-glue review - Realistically, the version is a 1.0.5 pre-release - Use the global directive instead of define for variables - Use the haclient/hacluster group/user instead of daemon - Use the _configure macro - Fix install dependencies * Fri Jul 24 2009 Andrew Beekhof - 1.0.4-3 - Initial Fedora checkin - Include an AUTHORS and license file in each package - Change the library package name to pacemaker-libs to be more Fedora compliant - Remove execute permissions from xml related files - Reference the new cluster-glue devel package name - Update the tarball from upstream to version c9120a53a6ae + pengine: Only prevent migration if the clone dependency is stopping/starting on the target node + pengine: Bug 2160 - Don't shuffle clones due to colocation + pengine: New implementation of the resource migration (not stop/start) logic + Medium: Tools: crm_resource - Prevent use-of-NULL by requiring a resource name for the -A and -a options + Medium: pengine: Prevent use-of-NULL in find_first_action() * Tue Jul 14 2009 Andrew Beekhof - 1.0.4-2 - Reference authors from the project AUTHORS file instead of listing in description - Change Source0 to reference the Mercurial repo - Cleaned up the summaries and descriptions - Incorporate the results of Fedora package self-review * Thu Jun 04 2009 Andrew Beekhof - 1.0.4-1 - Update source tarball to revision: 1d87d3e0fc7f (stable-1.0) - Statistics: Changesets: 209 Diff: 266 files changed, 12010 insertions(+), 8276 deletions(-) - Changes since Pacemaker-1.0.3 + (bnc#488291): ais: do not rely on byte endianness on ptr cast + (bnc#507255): Tools: crm: delete rsc/op_defaults (these meta_attributes are killing me) + (bnc#507255): Tools: crm: import properly rsc/op_defaults + (LF 2114): Tools: crm: add support for operation instance attributes + ais: Bug lf#2126 - Messages replies cannot be routed to transient clients + ais: Fix compilation for the latest Corosync API (v1719) + attrd: Do not perform all updates as complete refreshes + cib: Fix huge memory leak affecting heartbeat-based clusters + Core: Allow xpath queries to match attributes + Core: Generate the help text directly from a tool options struct + Core: Handle differences in 0.6 messaging format + crmd: Bug lf#2120 - All transient node attribute updates need to go via attrd + crmd: Correctly calculate how long an FSA action took to avoid spamming the logs with errors + crmd: Fix another large memory leak affecting Heartbeat based clusters + lha: Restore compatibility with older versions + pengine: Bug bnc#495687 - Filesystem is not notified of successful STONITH under some conditions + pengine: Make running a cluster with STONITH enabled but no STONITH resources an error and provide details on resolutions + pengine: Prevent use-ofNULL when using resource ordering sets + pengine: Provide inter-notification ordering guarantees + pengine: Rewrite the notification code to be understanable and extendable + Tools: attrd - Prevent race condition resulting in the cluster forgetting the node wishes to shut down + Tools: crm: regression tests + Tools: crm_mon - Fix smtp notifications + Tools: crm_resource - Repair the ability to query meta attributes + Low Build: Bug lf#2105 - Debian package should contain pacemaker doc and crm templates + Medium (bnc#507255): Tools: crm: handle empty rsc/op_defaults properly + Medium (bnc#507255): Tools: crm: use the right obj_type when creating objects from xml nodes + Medium (LF 2107): Tools: crm: revisit exit codes in configure + Medium: cib: Do not bother validating updates that only affect the status section + Medium: Core: Include supported stacks in version information + Medium: crmd: Record in the CIB, the cluster infrastructure being used + Medium: cts: Do not combine crm_standby arguments - the wrapper can not process them + Medium: cts: Fix the CIBAusdit class + Medium: Extra: Refresh showscores script from Dominik + Medium: pengine: Build a statically linked version of ptest + Medium: pengine: Correctly log the actions for resources that are being recovered + Medium: pengine: Correctly log the occurrence of promotion events + Medium: pengine: Implememt node health based on a patch from Mark Hamzy + Medium: Tools: Add examples to help text outputs + Medium: Tools: crm: catch syntax errors for configure load + Medium: Tools: crm: implement erasing nodes in configure erase + Medium: Tools: crm: work with parents only when managing xml objects + Medium: Tools: crm_mon - Add option to run custom notification program on resource operations (Patch by Dominik Klein) + Medium: Tools: crm_resource - Allow --cleanup to function on complex resources and cluster-wide + Medium: Tools: haresource2cib.py - Patch from horms to fix conversion error + Medium: Tools: Include stack information in crm_mon output + Medium: Tools: Two new options (--stack,--constraints) to crm_resource for querying how a resource is configured * Wed Apr 08 2009 Andrew Beekhof - 1.0.3-1 - Update source tarball to revision: b133b3f19797 (stable-1.0) tip - Statistics: Changesets: 383 Diff: 329 files changed, 15471 insertions(+), 15119 deletions(-) - Changes since Pacemaker-1.0.2 + Added tag SLE11-HAE-GMC for changeset 9196be9830c2 + ais plugin: Fix quorum calculation (bnc#487003) + ais: Another memory fix leak in error path + ais: Bug bnc#482847, bnc#482905 - Force a clean exit of OpenAIS once Pacemaker has finished unloading + ais: Bug bnc#486858 - Fix update_member() to prevent spamming clients with membership events containing no changes + ais: Centralize all quorum calculations in the ais plugin and allow expected votes to be configured int he cib + ais: Correctly handle a return value of zero from openais_dispatch_recv() + ais: Disable logging to a file + ais: Fix memory leak in error path + ais: IPC messages are only in scope until a response is sent + All signal handlers used with CL_SIGNAL() need to be as minimal as possible + cib: Bug bnc#482885 - Simplify CIB disk-writes to prevent data loss. Required a change to the backup filename format + cib: crmd: Revert part of 9782ab035003. Complex shutdown routines need G_main_add_SignalHandler to avoid race coditions + crm: Avoid infinite loop during crm configure edit (bnc#480327) + crmd: Avoid a race condition by waiting for the attrd update to trigger a transition automatically + crmd: Bug bnc#480977 - Prevent extra, partial, shutdown when a node restarts too quickly + crmd: Bug bnc#480977 - Prevent extra, partial, shutdown when a node restarts too quickly (verified) + crmd: Bug bnc#489063 - Ensure the DC is always unset after we 'lose' an election + crmd: Bug BSC#479543 - Correctly find the migration source for timed out migrate_from actions + crmd: Call crm_peer_init() before we start the FSA - prevents a race condition when used with Heartbeat + crmd: Erasing the status section should not be forced to the local node + crmd: Fix memory leak in cib notication processing code + crmd: Fix memory leak in transition graph processing + crmd: Fix memory leaks found by valgrind + crmd: More memory leaks fixes found by valgrind + fencing: stonithd: is_heartbeat_cluster is a no-no if there is no heartbeat support + pengine: Bug bnc#466788 - Exclude nodes that can not run resources + pengine: Bug bnc#466788 - Make colocation based on node attributes work + pengine: Bug BNC#478687 - Do not crash when clone-max is 0 + pengine: Bug bnc#488721 - Fix id-ref expansion for clones, the doc-root for clone children is not the cib root + pengine: Bug bnc#490418 - Correctly determine node state for nodes wishing to be terminated + pengine: Bug LF#2087 - Correctly parse the state of anonymous clones that have multiple instances on a given node + pengine: Bug lf#2089 - Meta attributes are not inherited by clone children + pengine: Bug lf#2091 - Correctly restart modified resources that were found active by a probe + pengine: Bug lf#2094 - Fix probe ordering for cloned groups + pengine: Bug LF:2075 - Fix large pingd memory leaks + pengine: Correctly attach orphaned clone children to their parent + pengine: Correctly handle terminate node attributes that are set to the output from time() + pengine: Ensure orphaned clone members are hooked up to the parent when clone-max=0 + pengine: Fix memory leak in LogActions + pengine: Fix the determination of whether a group is active + pengine: Look up the correct promotion preference for anonymous masters + pengine: Simplify handling of start failures by changing the default migration-threshold to INFINITY + pengine: The ordered option for clones no longer causes extra start/stop operations + RA: Bug bnc#490641 - Shut down dlm_controld with -TERM instead of -KILL + RA: pingd: Set default ping interval to 1 instead of 0 seconds + Resources: pingd - Correctly tell the ping daemon to shut down + Tools: Bug bnc#483365 - Ensure the command from cluster_test includes a value for --log-facility + Tools: cli: fix and improve delete command + Tools: crm: add and implement templates + Tools: crm: add support for command aliases and some common commands (i.e. cd,exit) + Tools: crm: create top configuration nodes if they are missing + Tools: crm: fix parsing attributes for rules (broken by the previous changeset) + Tools: crm: new ra set of commands + Tools: crm: resource agents information management + Tools: crm: rsc/op_defaults + Tools: crm: support for no value attribute in nvpairs + Tools: crm: the new configure monitor command + Tools: crm: the new configure node command + Tools: crm_mon - Prevent use-of-NULL when summarizing an orphan + Tools: hb2openais: create clvmd clone for respawn evmsd in ha.cf + Tools: hb2openais: fix a serious recursion bug in xml node processing + Tools: hb2openais: fix ocfs2 processing + Tools: pingd - prevent double free of getaddrinfo() output in error path + Tools: The default re-ping interval for pingd should be 1s not 1ms + Medium (bnc#479049): Tools: crm: add validation of resource type for the configure primitive command + Medium (bnc#479050): Tools: crm: add help for RA parameters in tab completion + Medium (bnc#479050): Tools: crm: add tab completion for primitive params/meta/op + Medium (bnc#479050): Tools: crm: reimplement cluster properties completion + Medium (bnc#486968): Tools: crm: listnodes function requires no parameters (do not mix completion with other stuff) + Medium: ais: Remove the ugly hack for dampening AIS membership changes + Medium: cib: Fix memory leaks by using mainloop_add_signal + Medium: cib: Move more logging to the debug level (was info) + Medium: cib: Overhaul the processing of synchronous replies + Medium: Core: Add library functions for instructing the cluster to terminate nodes + Medium: crmd: Add new expected-quorum-votes option + Medium: crmd: Allow up to 5 retires when an attrd update fails + Medium: crmd: Automatically detect and use new values for crm_config options + Medium: crmd: Bug bnc#490426 - Escalated shutdowns stall when there are pending resource operations + Medium: crmd: Clean up and optimize the DC election algorithm + Medium: crmd: Fix memory leak in shutdown + Medium: crmd: Fix memory leaks spotted by Valgrind + Medium: crmd: Ignore join messages from hosts other than our DC + Medium: crmd: Limit the scope of resource updates to the status section + Medium: crmd: Prevent the crmd from being respawned if its told to shut down when it did not ask to be + Medium: crmd: Re-check the election status after membership events + Medium: crmd: Send resource updates via the local CIB during elections + Medium: pengine: Bug bnc#491441 - crm_mon does not display operations returning 'uninstalled' correctly + Medium: pengine: Bug lf#2101 - For location constraints, role=Slave is equivalent to role=Started + Medium: pengine: Clean up the API - removed ->children() and renamed ->find_child() to fine_rsc() + Medium: pengine: Compress the display of healthy anonymous clones + Medium: pengine: Correctly log the actions for resources that are being recovered + Medium: pengine: Determin a promotion score for complex resources + Medium: pengine: Ensure clones always have a value for globally-unique + Medium: pengine: Prevent orphan clones from being allocated + Medium: RA: controld: Return proper exit code for stop op. + Medium: Tools: Bug bnc#482558 - Fix logging test in cluster_test + Medium: Tools: Bug bnc#482828 - Fix quoting in cluster_test logging setup + Medium: Tools: Bug bnc#482840 - Include directory path to CTSlab.py + Medium: Tools: crm: add more user input checks + Medium: Tools: crm: do not check resource status of we are working with a shadow + Medium: Tools: crm: fix id-refs and allow reference to top objects (i.e. primitive) + Medium: Tools: crm: ignore comments in the CIB + Medium: Tools: crm: multiple column output would not work with small lists + Medium: Tools: crm: refuse to delete running resources + Medium: Tools: crm: rudimentary if-else for templates + Medium: Tools: crm: Start/stop clones via target-role. + Medium: Tools: crm_mon - Compress the node status for healthy and offline nodes + Medium: Tools: crm_shadow - Return 0/cib_ok when --create-empty succeeds + Medium: Tools: crm_shadow - Support -e, the short form of --create-empty + Medium: Tools: Make attrd quieter + Medium: Tools: pingd - Avoid using various clplumbing functions as they seem to leak + Medium: Tools: Reduce pingd logging * Mon Feb 16 2009 Andrew Beekhof - 1.0.2-1 - Update source tarball to revision: d232d19daeb9 (stable-1.0) tip - Statistics: Changesets: 441 Diff: 639 files changed, 20871 insertions(+), 21594 deletions(-) - Changes since Pacemaker-1.0.1 + (bnc#450815): Tools: crm cli: do not generate id for the operations tag + ais: Add support for the new AIS IPC layer + ais: Always set header.error to the correct default: SA_AIS_OK + ais: Bug BNC#456243 - Ensure the membership cache always contains an entry for the local node + ais: Bug BNC:456208 - Prevent deadlocks by not logging in the child process before exec() + ais: By default, disable supprt for the WIP openais IPC patch + ais: Detect and handle situations where ais and the crm disagree on the node name + ais: Ensure crm_peer_seq is updated after a membership update + ais: Make sure all IPC header fields are set to sane defaults + ais: Repair and streamline service load now that whitetank startup functions correctly + build: create and install doc files + cib: Allow clients without mainloop to connect to the cib + cib: CID:18 - Fix use-of-NULL in cib_perform_op + cib: CID:18 - Repair errors introduced in b5a18704477b - Fix use-of-NULL in cib_perform_op + cib: Ensure diffs contain the correct values of admin_epoch + cib: Fix four moderately sized memory leaks detected by Valgrind + Core: CID:10 - Prevent indexing into an array of schemas with a negative value + Core: CID:13 - Fix memory leak in log_data_element + Core: CID:15 - Fix memory leak in crm_get_peer + Core: CID:6 - Fix use-of-NULL in copy_ha_msg_input + Core: Fix crash in the membership code preventing node shutdown + Core: Fix more memory leaks foudn by valgrind + Core: Prevent unterminated strings after decompression + crmd: Bug BNC:467995 - Delay marking STONITH operations complete until STONITH tells us so + crmd: Bug LF:1962 - Do not NACK peers because they are not (yet) in our membership. Just ignore them. + crmd: Bug LF:2010 - Ensure fencing cib updates create the node_state entry if needed to preent re-fencing during cluster startup + crmd: Correctly handle reconnections to attrd + crmd: Ensure updates for lost migrate operations indicate which node it tried to migrating to + crmd: If there are no nodes to finalize, start an election. + crmd: If there are no nodes to welcome, start an election. + crmd: Prevent node attribute loss by detecting attrd disconnections immediately + crmd: Prevent node re-probe loops by ensuring mandatory actions always complete + pengine: Bug 2005 - Fix startup ordering of cloned stonith groups + pengine: Bug 2006 - Correctly reprobe cloned groups + pengine: Bug BNC:465484 - Fix the no-quorum-policy=suicide option + pengine: Bug LF:1996 - Correctly process disabled monitor operations + pengine: CID:19 - Fix use-of-NULL in determine_online_status + pengine: Clones now default to globally-unique=false + pengine: Correctly calculate the number of available nodes for the clone to use + pengine: Only shoot online nodes with no-quorum-policy=suicide + pengine: Prevent on-fail settings being ignored after a resource is successfully stopped + pengine: Prevent use-of-NULL for failed migrate actions in process_rsc_state() + pengine: Remove an optimization for the terminate node attribute that caused the cluster to block indefinitly + pengine: Repar the ability to colocate based on node attributes other than uname + pengine: Start the correct monitor operation for unmanaged masters + stonith: CID:3 - Fix another case of exceptionally poor error handling by the original stonith developers + stonith: CID:5 - Checking for NULL and then dereferencing it anyway is an interesting approach to error handling + stonithd: Sending IPC to the cluster is a privileged operation + stonithd: wrong checks for shmid (0 is a valid id) + Tools: attrd - Correctly determine when an attribute has stopped changing and should be committed to the CIB + Tools: Bug 2003 - pingd does not correctly detect failures when the interface is down + Tools: Bug 2003 - pingd does not correctly handle node-down events on multi-NIC systems + Tools: Bug 2021 - pingd does not detect sequence wrapping correctly, incorrectly reports nodes offline + Tools: Bug BNC:468066 - Do not use the result of uname() when its no longer in scope + Tools: Bug BNC:473265 - crm_resource -L dumps core + Tools: Bug LF:2001 - Transient node attributes should be set via attrd + Tools: Bug LF:2036 - crm_resource cannot set/get parameters for cloned resources + Tools: Bug LF:2046 - Node attribute updates are lost because attrd can take too long to start + Tools: Cause the correct clone instance to be failed with crm_resource -F + Tools: cluster_test - Allow the user to select a stack and fix CTS invocation + Tools: crm cli: allow rename only if the resource is stopped + Tools: crm cli: catch system errors on file operations + Tools: crm cli: completion for ids in configure + Tools: crm cli: drop '-rsc' from attributes for order constraint + Tools: crm cli: exit with an appropriate exit code + Tools: crm cli: fix wrong order of action and resource in order constraint + Tools: crm cli: fox wrong exit code + Tools: crm cli: improve handling of cib attributes + Tools: crm cli: new command: configure rename + Tools: crm cli: new command: configure upgrade + Tools: crm cli: new command: node delete + Tools: crm cli: prevent key errors on missing cib attributes + Tools: crm cli: print long help for help topics + Tools: crm cli: return on syntax error when parsing score + Tools: crm cli: rsc_location can be without nvpairs + Tools: crm cli: short node preference location constraint + Tools: crm cli: sometimes, on errors, level would change on single shot use + Tools: crm cli: syntax: drop a bunch of commas (remains of help tables conversion) + Tools: crm cli: verify user input for sanity + Tools: crm: find expressions within rules (do not always skip xml nodes due to used id) + Tools: crm_master should not define a set id now that attrd is used. Defining one can break lookups + Tools: crm_mon Use the OID assigned to the project by IANA for SNMP traps + Medium (bnc#445622): Tools: crm cli: improve the node show command and drop node status + Medium (LF 2009): stonithd: improve timeouts for remote fencing + Medium: ais: Allow dead peers to be removed from membership calculations + Medium: ais: Pass node deletion events on to clients + Medium: ais: Sanitize ipc usage + Medium: ais: Supply the node uname in addtion to the id + Medium: Build: Clean up configure to ensure NON_FATAL_CFLAGS is consistent with CFLAGS (ie. includes -g) + Medium: Build: Install cluster_test + Medium: Build: Use more restrictive CFLAGS and fix the resulting errors + Medium: cib: CID:20 - Fix potential use-after-free in cib_native_signon + Medium: Core: Bug BNC:474727 - Set a maximum time to wait for IPC messages + Medium: Core: CID:12 - Fix memory leak in decode_transition_magic error path + Medium: Core: CID:14 - Fix memory leak in calculate_xml_digest error path + Medium: Core: CID:16 - Fix memory leak in date_to_string error path + Medium: Core: Try to track down the cause of XML parsing errors + Medium: crmd: Bug BNC:472473 - Do not wait excessive amounts of time for lost actions + Medium: crmd: Bug BNC:472473 - Reduce the transition timeout to action_timeout+network_delay + Medium: crmd: Do not fast-track the processing of LRM refreshes when there are pending actions. + Medium: crmd: do_dc_join_filter_offer - Check the 'join' message is for the current instance before deciding to NACK peers + Medium: crmd: Find option values without having to do a config upgrade + Medium: crmd: Implement shutdown using a transient node attribute + Medium: crmd: Update the crmd options to use dashes instead of underscores + Medium: cts: Add 'cluster reattach' to the suite of automated regression tests + Medium: cts: cluster_test - Make some usability enhancements + Medium: CTS: cluster_test - suggest a valid port number + Medium: CTS: Fix python import order + Medium: cts: Implement an automated SplitBrain test + Medium: CTS: Remove references to deleted classes + Medium: Extra: Resources - Use HA_VARRUN instead of HA_RSCTMP for state files as Heartbeat removes HA_RSCTMP at startup + Medium: HB: Bug 1933 - Fake crmd_client_status_callback() calls because HB does not provide them for already running processes + Medium: pengine: CID:17 - Fix memory leak in find_actions_by_task error path + Medium: pengine: CID:7,8 - Prevent hypothetical use-of-NULL in LogActions + Medium: pengine: Defer logging the actions performed on a resource until we have processed ordering constraints + Medium: pengine: Remove the symmetrical attribute of colocation constraints + Medium: Resources: pingd - fix the meta defaults + Medium: Resources: Stateful - Add missing meta defaults + Medium: stonithd: exit if we the pid file cannot be locked + Medium: Tools: Allow attrd clients to specify the ID the attribute should be created with + Medium: Tools: attrd - Allow attribute updates to be performed from a hosts peer + Medium: Tools: Bug LF:1994 - Clean up crm_verify return codes + Medium: Tools: Change the pingd defaults to ping hosts once every second (instead of 5 times every 10 seconds) + Medium: Tools: cibmin - Detect resource operations with a view to providing email/snmp/cim notification + Medium: Tools: crm cli: add back symmetrical for order constraints + Medium: Tools: crm cli: generate role in location when converting from xml + Medium: Tools: crm cli: handle shlex exceptions + Medium: Tools: crm cli: keep order of help topics + Medium: Tools: crm cli: refine completion for ids in configure + Medium: Tools: crm cli: replace inf with INFINITY + Medium: Tools: crm cli: streamline cib load and parsing + Medium: Tools: crm cli: supply provider only for ocf class primitives + Medium: Tools: crm_mon - Add support for sending mail notifications of resource events + Medium: Tools: crm_mon - Include the DC version in status summary + Medium: Tools: crm_mon - Sanitize startup and option processing + Medium: Tools: crm_mon - switch to event-driven updates and add support for sending snmp traps + Medium: Tools: crm_shadow - Replace the --locate option with the saner --edit + Medium: Tools: hb2openais: do not remove Evmsd resources, but replace them with clvmd + Medium: Tools: hb2openais: replace crmadmin with crm_mon + Medium: Tools: hb2openais: replace the lsb class with ocf for o2cb + Medium: Tools: hb2openais: reuse code + Medium: Tools: LF:2029 - Display an error if crm_resource is used to reset the operation history of non-primitive resources + Medium: Tools: Make pingd resilient to attrd failures + Medium: Tools: pingd - fix the command line switches + Medium: Tools: Rename ccm_tool to crm_node * Tue Nov 18 2008 Andrew Beekhof - 1.0.1-1 - Update source tarball to revision: 6fc5ce8302ab (stable-1.0) tip - Statistics: Changesets: 170 Diff: 816 files changed, 7633 insertions(+), 6286 deletions(-) - Changes since Pacemaker-1.0.1 + ais: Allow the crmd to get callbacks whenever a node state changes + ais: Create an option for starting the mgmtd daemon automatically + ais: Ensure HA_RSCTMP exists for use by resource agents + ais: Hook up the openais.conf config logging options + ais: Zero out the PID of disconnecting clients + cib: Ensure global updates cause a disk write when appropriate + Core: Add an extra snaity check to getXpathResults() to prevent segfaults + Core: Do not redefine __FUNCTION__ unnecessarily + Core: Repair the ability to have comments in the configuration + crmd: Bug:1975 - crmd should wait indefinitely for stonith operations to complete + crmd: Ensure PE processing does not occur for all error cases in do_pe_invoke_callback + crmd: Requests to the CIB should cause any prior PE calculations to be ignored + heartbeat: Wait for membership 'up' events before removing stale node status data + pengine: Bug LF:1988 - Ensure recurring operations always have the correct target-rc set + pengine: Bug LF:1988 - For unmanaged resources we need to skip the usual can_run_resources() checks + pengine: Ensure the terminate node attribute is handled correctly + pengine: Fix optional colocation + pengine: Improve up the detection of 'new' nodes joining the cluster + pengine: Prevent assert failures in master_color() by ensuring unmanaged masters are always reallocated to their current location + Tools: crm cli: parser: return False on syntax error and None for comments + Tools: crm cli: unify template and edit commands + Tools: crm_shadow - Show more line number information after validation failures + Tools: hb2openais: add option to upgrade the CIB to v3.0 + Tools: hb2openais: add U option to getopts and update usage + Tools: hb2openais: backup improved and multiple fixes + Tools: hb2openais: fix class/provider reversal + Tools: hb2openais: fix testing + Tools: hb2openais: move the CIB update to the end + Tools: hb2openais: update logging and set logfile appropriately + Tools: LF:1969 - Attrd never sets any properties in the cib + Tools: Make attrd functional on OpenAIS + Medium: ais: Hook up the options for specifying the expected number of nodes and total quorum votes + Medium: ais: Look for pacemaker options inside the service block with 'name: pacemaker' instead of creating an addtional configuration block + Medium: ais: Provide better feedback when nodes change nodeids (in openais.conf) + Medium: cib: Always store cib contents on disk with num_updates=0 + Medium: cib: Ensure remote access ports are cleaned up on shutdown + Medium: crmd: Detect deleted resource operations automatically + Medium: crmd: Erase a nodes resource operations and transient attributes after a successful STONITH + Medium: crmd: Find a more appropriate place to update quorum and refresh attrd attributes + Medium: crmd: Fix the handling of unexpected PE exits to ensure the current CIB is stored + Medium: crmd: Fix the recording of pending operations in the CIB + Medium: crmd: Initiate an attrd refresh _after_ the status section has been fully repopulated + Medium: crmd: Only the DC should update quorum in an openais cluster + Medium: Ensure meta attributes are used consistantly + Medium: pengine: Allow group and clone level resource attributes + Medium: pengine: Bug N:437719 - Ensure scores from colocated resources count when allocating groups + Medium: pengine: Prevent lsb scripts from being used in globally unique clones + Medium: pengine: Make a best-effort guess at a migration threshold for people with 0.6 configs + Medium: Resources: controld - ensure we are part of a clone with globally_unique=false + Medium: Tools: attrd - Automatically refresh all attributes after a CIB replace operation + Medium: Tools: Bug LF:1985 - crm_mon - Correctly process failed cib queries to allow reconnection after cluster restarts + Medium: Tools: Bug LF:1987 - crm_verify incorrectly warns of configuration upgrades for the most recent version + Medium: Tools: crm (bnc#441028): check for key error in attributes management + Medium: Tools: crm_mon - display the meaning of the operation rc code instead of the status + Medium: Tools: crm_mon - Fix the display of timing data + Medium: Tools: crm_verify - check that we are being asked to validate a complete config + Medium: xml: Relax the restriction on the contents of rsc_locaiton.node * Thu Oct 16 2008 Andrew Beekhof - 1.0.0-1 - Update source tarball to revision: 388654dfef8f tip - Statistics: Changesets: 261 Diff: 3021 files changed, 244985 insertions(+), 111596 deletions(-) - Changes since f805e1b30103 + add the crm cli program + ais: Move the service id definition to a common location and make sure it is always used + build: rename hb2openais.sh to .in and replace paths with vars + cib: Implement --create for crm_shadow + cib: Remove dead files + Core: Allow the expected number of quorum votes to be configrable + Core: cl_malloc and friends were removed from Heartbeat + Core: Only call xmlCleanupParser() if we parsed anything. Doing so unconditionally seems to cause a segfault + hb2openais.sh: improve pingd handling; several bugs fixed + hb2openais: fix clone creation; replace EVMS strings + new hb2openais.sh conversion script + pengine: Bug LF:1950 - Ensure the current values for all notification variables are always set (even if empty) + pengine: Bug LF:1955 - Ensure unmanaged masters are unconditionally repromoted to ensure they are monitored correctly. + pengine: Bug LF:1955 - Fix another case of filtering causing unmanaged master failures + pengine: Bug LF:1955 - Umanaged mode prevents master resources from being allocated correctly + pengine: Bug N:420538 - Anit-colocation caused a positive node preference + pengine: Correctly handle unmanaged resources to prevent them from being started elsewhere + pengine: crm_resource - Fix the --migrate command + pengine: MAke stonith-enabled default to true and warn if no STONITH resources are found + pengine: Make sure orphaned clone children are created correctly + pengine: Monitors for unmanaged resources do not need to wait for start/promote/demote actions to complete + stonithd (LF 1951): fix remote stonith operations + stonithd: fix handling of timeouts + stonithd: fix logic for stonith resource priorities + stonithd: implement the fence-timeout instance attribute + stonithd: initialize value before reading fence-timeout + stonithd: set timeouts for fencing ops to the timeout of the start op + stonithd: stonith rsc priorities (new feature) + Tools: Add hb2openais - a tool for upgrading a Heartbeat cluster to use OpenAIS instead + Tools: crm_verify - clean up the upgrade logic to prevent crash on invalid configurations + Tools: Make pingd functional on Linux + Update version numbers for 1.0 candidates + Medium: ais: Add support for a synchronous call to retrieve the nodes nodeid + Medium: ais: Use the agreed service number + Medium: Build: Reliably detect heartbeat libraries during configure + Medium: Build: Supply prototypes for libreplace functions when needed + Medium: Build: Teach configure how to find corosync + Medium: Core: Provide better feedback if Pacemaker is started by a stack it does not support + Medium: crmd: Avoid calling GHashTable functions with NULL + Medium: crmd: Delay raising I_ERROR when the PE exits until we have had a chance to save the current CIB + Medium: crmd: Hook up the stonith-timeout option to stonithd + Medium: crmd: Prevent potential use-of-NULL in global_timer_callback + Medium: crmd: Rationalize the logging of graph aborts + Medium: pengine: Add a stonith_timeout option and remove new options that are better set in rsc_defaults + Medium: pengine: Allow external entities to ask for a node to be shot by creating a terminate=true transient node attribute + Medium: pengine: Bug LF:1950 - Notifications do not contain all documented resource state fields + Medium: pengine: Bug N:417585 - Do not restart group children whos individual score drops below zero + Medium: pengine: Detect clients that disconnect before receiving their reply + Medium: pengine: Implement a true maintenance mode + Medium: pengine: Implement on-fail=standby for NTT. Derived from a patch by Satomi TANIGUCHI + Medium: pengine: Print the correct message when stonith is disabled + Medium: pengine: ptest - check the input is valid before proceeding + Medium: pengine: Revert group stickiness to the 'old way' + Medium: pengine: Use the correct attribute for action 'requires' (was prereq) + Medium: stonithd: Fix compilation without full heartbeat install + Medium: stonithd: exit with better code on empty host list + Medium: tools: Add a new regression test for CLI tools + Medium: tools: crm_resource - return with non-zero when a resource migration command is invalid + Medium: tools: crm_shadow - Allow the admin to start with an empty CIB (and no cluster connection) + Medium: xml: pacemaker-0.7 is now an alias for the 1.0 schema * Mon Sep 22 2008 Andrew Beekhof - 0.7.3-1 - Update source tarball to revision: 33e677ab7764+ tip - Statistics: Changesets: 133 Diff: 89 files changed, 7492 insertions(+), 1125 deletions(-) - Changes since f805e1b30103 + Tools: add the crm cli program + Core: cl_malloc and friends were removed from Heartbeat + Core: Only call xmlCleanupParser() if we parsed anything. Doing so unconditionally seems to cause a segfault + new hb2openais.sh conversion script + pengine: Bug LF:1950 - Ensure the current values for all notification variables are always set (even if empty) + pengine: Bug LF:1955 - Ensure unmanaged masters are unconditionally repromoted to ensure they are monitored correctly. + pengine: Bug LF:1955 - Fix another case of filtering causing unmanaged master failures + pengine: Bug LF:1955 - Umanaged mode prevents master resources from being allocated correctly + pengine: Bug N:420538 - Anit-colocation caused a positive node preference + pengine: Correctly handle unmanaged resources to prevent them from being started elsewhere + pengine: crm_resource - Fix the --migrate command + pengine: MAke stonith-enabled default to true and warn if no STONITH resources are found + pengine: Make sure orphaned clone children are created correctly + pengine: Monitors for unmanaged resources do not need to wait for start/promote/demote actions to complete + stonithd (LF 1951): fix remote stonith operations + Tools: crm_verify - clean up the upgrade logic to prevent crash on invalid configurations + Medium: ais: Add support for a synchronous call to retrieve the nodes nodeid + Medium: ais: Use the agreed service number + Medium: pengine: Allow external entities to ask for a node to be shot by creating a terminate=true transient node attribute + Medium: pengine: Bug LF:1950 - Notifications do not contain all documented resource state fields + Medium: pengine: Bug N:417585 - Do not restart group children whos individual score drops below zero + Medium: pengine: Implement a true maintenance mode + Medium: pengine: Print the correct message when stonith is disabled + Medium: stonithd: exit with better code on empty host list + Medium: xml: pacemaker-0.7 is now an alias for the 1.0 schema * Wed Aug 20 2008 Andrew Beekhof - 0.7.1-1 - Update source tarball to revision: f805e1b30103+ tip - Statistics: Changesets: 184 Diff: 513 files changed, 43408 insertions(+), 43783 deletions(-) - Changes since 0.7.0-19 + Fix compilation when GNUTLS isn't found + admin: Fix use-after-free in crm_mon + Build: Remove testing code that prevented heartbeat-only builds + cib: Use single quotes so that the xpath queries for nvpairs will succeed + crmd: Always connect to stonithd when the TE starts and ensure we notice if it dies + crmd: Correctly handle a dead PE process + crmd: Make sure async-failures cause the failcount to be incremented + pengine: Bug LF:1941 - Handle failed clone instance probes when clone-max < #nodes + pengine: Parse resource ordering sets correctly + pengine: Prevent use-of-NULL - order->rsc_rh will not always be non-NULL + pengine: Unpack colocation sets correctly + Tools: crm_mon - Prevent use-of-NULL for orphaned resources + Medium: ais: Add support for a synchronous call to retrieve the nodes nodeid + Medium: ais: Allow transient clients to receive membership updates + Medium: ais: Avoid double-free in error path + Medium: ais: Include in the mebership nodes for which we have not determined their hostname + Medium: ais: Spawn the PE from the ais plugin instead of the crmd + Medium: cib: By default, new configurations use the latest schema + Medium: cib: Clean up the CIB if it was already disconnected + Medium: cib: Only increment num_updates if something actually changed + Medium: cib: Prevent use-after-free in client after abnormal termination of the CIB + Medium: Core: Fix memory leak in xpath searches + Medium: Core: Get more details regarding parser errors + Medium: Core: Repair expand_plus_plus - do not call char2score on unexpanded values + Medium: Core: Switch to the libxml2 parser - its significantly faster + Medium: Core: Use a libxml2 library function for xml -> text conversion + Medium: crmd: Asynchronous failure actions have no parameters + Medium: crmd: Avoid calling glib functions with NULL + Medium: crmd: Do not allow an election to promote a node from S_STARTING + Medium: crmd: Do not vote if we have not completed the local startup + Medium: crmd: Fix te_update_diff() now that get_object_root() functions differently + Medium: crmd: Fix the lrmd xpath expressions to not contain quotes + Medium: crmd: If we get a join offer during an election, better restart the election + Medium: crmd: No further processing is needed when using the LRMs API call for failing resources + Medium: crmd: Only update have-quorum if the value changed + Medium: crmd: Repair the input validation logic in do_te_invoke + Medium: cts: CIBs can no longer contain comments + Medium: cts: Enable a bunch of tests that were incorrectly disabled + Medium: cts: The libxml2 parser wont allow v1 resources to use integers as parameter names + Medium: Do not use the cluster UID and GID directly. Look them up based on the configured value of HA_CCMUSER + Medium: Fix compilation when heartbeat is not supported + Medium: pengine: Allow groups to be involved in optional ordering constraints + Medium: pengine: Allow sets of operations to be reused by multiple resources + Medium: pengine: Bug LF:1941 - Mark extra clone instances as orphans and do not show inactive ones + Medium: pengine: Determin the correct migration-threshold during resource expansion + Medium: pengine: Implement no-quorum-policy=suicide (FATE #303619) + Medium: pengine: Clean up resources after stopping old copies of the PE + Medium: pengine: Teach the PE how to stop old copies of itself + Medium: Tools: Backport hb_report updates + Medium: Tools: cib_shadow - On create, spawn a new shell with CIB_shadow and PS1 set accordingly + Medium: Tools: Rename cib_shadow to crm_shadow * Fri Jul 18 2008 Andrew Beekhof - 0.7.0-19 - Update source tarball to revision: 007c3a1c50f5 (unstable) tip - Statistics: Changesets: 108 Diff: 216 files changed, 4632 insertions(+), 4173 deletions(-) - Changes added since unstable-0.7 + admin: Fix use-after-free in crm_mon + ais: Change the tag for the ais plugin to "pacemaker" (used in openais.conf) + ais: Log terminated processes as an error + cib: Performance - Reorganize things to avoid calculating the XML diff twice + pengine: Bug LF:1941 - Handle failed clone instance probes when clone-max < #nodes + pengine: Fix memory leak in action2xml + pengine: Make OCF_ERR_ARGS a node-level error rather than a cluster-level one + pengine: Properly handle clones that are not installed on all nodes + Medium: admin: cibadmin - Show any validation errors if the upgrade failed + Medium: admin: cib_shadow - Implement --locate to display the underlying filename + Medium: admin: cib_shadow - Implement a --diff option + Medium: admin: cib_shadow - Implement a --switch option + Medium: admin: crm_resource - create more compact constraints that do not use lifetime (which is deprecated) + Medium: ais: Approximate born_on for OpenAIS based clusters + Medium: cib: Remove do_id_check, it is a poor substitute for ID validation by a schema + Medium: cib: Skip construction of pre-notify messages if no-one wants one + Medium: Core: Attempt to streamline some key functions to increase performance + Medium: Core: Clean up XML parser after validation + Medium: crmd: Detect and optimize the CRMs behavior when processing diffs of an LRM refresh + Medium: Fix memory leaks when resetting the name of an XML object + Medium: pengine: Prefer the current location if it is one of a group of nodes with the same (highest) score * Wed Jun 25 2008 Andrew Beekhof - 0.7.0-1 - Update source tarball to revision: bde0c7db74fb tip - Statistics: Changesets: 439 Diff: 676 files changed, 41310 insertions(+), 52071 deletions(-) - Changes added since stable-0.6 + A new tool for setting up and invoking CTS + Admin: All tools now use --node (-N) for specifying node unames + Admin: All tools now use --xml-file (-x) and --xml-text (-X) for specifying where to find XML blobs + cib: Cleanup the API - remove redundant input fields + cib: Implement CIB_shadow - a facility for making and testing changes before uploading them to the cluster + cib: Make registering per-op callbacks an API call and renamed (for clarity) the API call for requesting notifications + Core: Add a facility for automatically upgrading old configurations + Core: Adopt libxml2 as the XML processing library - all external clients need to be recompiled + Core: Allow sending TLS messages larger than the MTU + Core: Fix parsing of time-only ISO dates + Core: Smarter handling of XML values containing quotes + Core: XML memory corruption - catch, and handle, cases where we are overwriting an attribute value with itself + Core: The xml ID type does not allow UUIDs that start with a number + Core: Implement XPath based versions of query/delete/replace/modify + Core: Remove some HA2.0.(3,4) compatibility code + crmd: Overhaul the detection of nodes that are starting vs. failed + pengine: Bug LF:1459 - Allow failures to expire + pengine: Have the PE do non-persistent configuration upgrades before performing calculations + pengine: Replace failure-stickiness with a simple 'migration-threshold' + tengine: Simplify the design by folding the tengine process into the crmd + Medium: Admin: Bug LF:1438 - Allow the list of all/active resource operations to be queried by crm_resource + Medium: Admin: Bug LF:1708 - crm_resource should print a warning if an attribute is already set as a meta attribute + Medium: Admin: Bug LF:1883 - crm_mon should display fail-count and operation history + Medium: Admin: Bug LF:1883 - crm_mon should display operation timing data + Medium: Admin: Bug N:371785 - crm_resource -C does not also clean up fail-count attributes + Medium: Admin: crm_mon - include timing data for failed actions + Medium: ais: Read options from the environment since objdb is not completely usable yet + Medium: cib: Add sections for op_defaults and rsc_defaults + Medium: cib: Better matching notification callbacks (for detecting duplicates and removal) + Medium: cib: Bug LF:1348 - Allow rules and attribute sets to be referenced for use in other objects + Medium: cib: BUG LF:1918 - By default, all cib calls now timeout after 30s + Medium: cib: Detect updates that decrease the version tuple + Medium: cib: Implement a client-side operation timeout - Requires LHA update + Medium: cib: Implement callbacks and async notifications for remote connections + Medium: cib: Make cib->cmds->update() an alias for modify at the API level (also implemented in cibadmin) + Medium: cib: Mark the CIB as disconnected if the IPC connection is terminated + Medium: cib: New call option 'cib_can_create' which can be passed to modify actions - allows the object to be created if it does not exist yet + Medium: cib: Reimplement get|set|delete attributes using XPath + Medium: cib: Remove some useless parts of the API + Medium: cib: Remove the 'attributes' scaffolding from the new format + Medium: cib: Implement the ability for clients to connect to remote servers + Medium: Core: Add support for validating xml against RelaxNG schemas + Medium: Core: Allow more than one item to be modified/deleted in XPath based operations + Medium: Core: Fix the sort_pairs function for creating sorted xml objects + Medium: Core: iso8601 - Implement subtract_duration and fix subtract_time + Medium: Core: Reduce the amount of xml copying + Medium: Core: Support value='value+=N' XML updates (in addtion to value='value++') + Medium: crmd: Add support for lrm_ops->fail_rsc if its available + Medium: crmd: HB - watch link status for node leaving events + Medium: crmd: Bug LF:1924 - Improved handling of lrmd disconnects and shutdowns + Medium: crmd: Do not wait for actions with a start_delay over 5 minutes. Confirm them immediately + Medium: pengine: Bug LF:1328 - Do not fencing nodes in clusters without managed resources + Medium: pengine: Bug LF:1461 - Give transient node attributes (in ) preference over persistent ones (in ) + Medium: pengine: Bug LF:1884, Bug LF:1885 - Implement N:M ordering and colocation constraints + Medium: pengine: Bug LF:1886 - Create a resource and operation 'defaults' config section + Medium: pengine: Bug LF:1892 - Allow recurring actions to be triggered at known times + Medium: pengine: Bug LF:1926 - Probes should complete before stop actions are invoked + Medium: pengine: Fix the standby when its set as a transient attribute + Medium: pengine: Implement a global 'stop-all-resources' option + Medium: pengine: Implement cibpipe, a tool for performing/simulating config changes "offline" + Medium: pengine: We do not allow colocation with specific clone instances + Medium: Tools: pingd - Implement a stack-independent version of pingd + Medium: xml: Ship an xslt for upgrading from 0.6 to 0.7 * Thu Jun 19 2008 Andrew Beekhof - 0.6.5-1 - Update source tarball to revision: b9fe723d1ac5 tip - Statistics: Changesets: 48 Diff: 37 files changed, 1204 insertions(+), 234 deletions(-) - Changes since Pacemaker-0.6.4 + Admin: Repair the ability to delete failcounts + ais: Audit IPC handling between the AIS plugin and CRM processes + ais: Have the plugin create needed /var/lib directories + ais: Make sure the sync and async connections are assigned correctly (not swapped) + cib: Correctly detect configuration changes - num_updates does not count + pengine: Apply stickiness values to the whole group, not the individual resources + pengine: Bug N:385265 - Ensure groups are migrated instead of remaining partially active on the current node + pengine: Bug N:396293 - Enforce mandatory group restarts due to ordering constraints + pengine: Correctly recover master instances found active on more than one node + pengine: Fix memory leaks reported by Valgrind + Medium: Admin: crm_mon - Misc improvements from Satomi Taniguchi + Medium: Bug LF:1900 - Resource stickiness should not allow placement in asynchronous clusters + Medium: crmd: Ensure joins are completed promptly when a node taking part dies + Medium: pengine: Avoid clone instance shuffling in more cases + Medium: pengine: Bug LF:1906 - Remove an optimization in native_merge_weights() causing group scores to behave eratically + Medium: pengine: Make use of target_rc data to correctly process resource operations + Medium: pengine: Prevent a possible use of NULL in sort_clone_instance() + Medium: tengine: Include target rc in the transition key - used to correctly determin operation failure * Thu May 22 2008 Andrew Beekhof - 0.6.4-1 - Update source tarball to revision: 226d8e356924 tip - Statistics: Changesets: 55 Diff: 199 files changed, 7103 insertions(+), 12378 deletions(-) - Changes since Pacemaker-0.6.3 + crmd: Bug LF:1881 LF:1882 - Overhaul the logic for operation cancelation and deletion + crmd: Bug LF:1894 - Make sure cancelled recurring operations are cleaned out from the CIB + pengine: Bug N:387749 - Colocation with clones causes unnecessary clone instance shuffling + pengine: Ensure 'master' monitor actions are cancelled _before_ we demote the resource + pengine: Fix assert failure leading to core dump - make sure variable is properly initialized + pengine: Make sure 'slave' monitoring happens after the resource has been demoted + pengine: Prevent failure stickiness underflows (where too many failures become a _positive_ preference) + Medium: Admin: crm_mon - Only complain if the output file could not be opened + Medium: Common: filter_action_parameters - enable legacy handling only for older versions + Medium: pengine: Bug N:385265 - The failure stickiness of group children is ignored until it reaches -INFINITY + Medium: pengine: Implement master and clone colocation by exlcuding nodes rather than setting ones score to INFINITY (similar to cs: 756afc42dc51) + Medium: tengine: Bug LF:1875 - Correctly find actions to cancel when their node leaves the cluster * Wed Apr 23 2008 Andrew Beekhof - 0.6.3-1 - Update source tarball to revision: fd8904c9bc67 tip - Statistics: Changesets: 117 Diff: 354 files changed, 19094 insertions(+), 11338 deletions(-) - Changes since Pacemaker-0.6.2 + Admin: Bug LF:1848 - crm_resource - Pass set name and id to delete_resource_attr() in the correct order + Build: SNMP has been moved to the management/pygui project + crmd: Bug LF1837 - Unmanaged resources prevent crmd from shutting down + crmd: Prevent use-after-free in lrm interface code (Patch based on work by Keisuke MORI) + pengine: Allow the cluster to make progress by not retrying failed demote actions + pengine: Anti-colocation with slave should not prevent master colocation + pengine: Bug LF 1768 - Wait more often for STONITH ops to complete before starting resources + pengine: Bug LF1836 - Allow is-managed-default=false to be overridden by individual resources + pengine: Bug LF185 - Prevent pointless master/slave instance shuffling by ignoring the master-pref of stopped instances + pengine: Bug N-191176 - Implement interleaved ordering for clone-to-clone scenarios + pengine: Bug N-347004 - Ensure clone notifications are always sent when an instance is stopped/started + pengine: Bug N-347004 - Include notification ordering is correct for interleaved clones + pengine: Bug PM-11 - Directly link probe_complete to starting clone instances + pengine: Bug PM1 - Fix setting failcounts when applied to complex resources + pengine: Bug PM12, LF1648 - Extensive revision of group ordering + pengine: Bug PM7 - Ensure masters are always demoted before they are stopped + pengine: Create probes after allocation to allow smarter handling of anonymous clones + pengine: Do not prioritize clone instances that must be moved + pengine: Fix error in previous commit that allowed more than the required number of masters to be promoted + pengine: Group start ordering fixes + pengine: Implement promote/demote ordering for cloned groups + tengine: Repair failcount updates + tengine: Use the correct offset when updating failcount + Medium: Admin: Add a summary output that can be easily parsed by CTS for audit purposes + Medium: Build: Make configure fail if bz2 or libxml2 are not present + Medium: Build: Re-instate a better default for LCRSODIR + Medium: CIB: Bug LF-1861 - Filter irrelvant error status from synchronous CIB clients + Medium: Core: Bug 1849 - Invalid conversion of ordinal leap year to gregorian date + Medium: Core: Drop compatibility code for 2.0.4 and 2.0.5 clusters + Medium: crmd: Bug LF-1860 - Automatically cancel recurring ops before demote and promote operations (not only stops) + Medium: crmd: Save the current CIB contents if we detect the PE crashed + Medium: pengine: Bug LF:1866 - Fix version check when applying compatibility handling for failed start operations + Medium: pengine: Bug LF:1866 - Restore the ability to have start failures not be fatal + Medium: pengine: Bug PM1 - Failcount applies to all instances of non-unique clone + Medium: pengine: Correctly set the state of partially active master/slave groups + Medium: pengine: Do not claim to be stopping an already stopped orphan + Medium: pengine: Ensure implies_left ordering constraints are always effective + Medium: pengine: Indicate each resources 'promotion' score + Medium: pengine: Prevent a possible use-of-NULL + Medium: pengine: Reprocess the current action if it changed (so that any prior dependencies are updated) + Medium: tengine: Bug LF-1859 - Wait for fail-count updates to complete before terminating the transition + Medium: tengine: Bug LF:1859 - Do not abort graphs due to our own failcount updates + Medium: tengine: Bug LF:1859 - Prevent the TE from interupting itself * Thu Feb 14 2008 Andrew Beekhof - 0.6.2-1 - Update source tarball to revision: 28b1a8c1868b tip - Statistics: Changesets: 11 Diff: 7 files changed, 58 insertions(+), 18 deletions(-) - Changes since Pacemaker-0.6.1 + haresources2cib.py: set default-action-timeout to the default (20s) + haresources2cib.py: update ra parameters lists + Medium: SNMP: Allow the snmp subagent to be built (patch from MATSUDA, Daiki) + Medium: Tools: Make sure the autoconf variables in haresources2cib are expanded * Tue Feb 12 2008 Andrew Beekhof - 0.6.1-1 - Update source tarball to revision: e7152d1be933 tip - Statistics: Changesets: 25 Diff: 37 files changed, 1323 insertions(+), 227 deletions(-) - Changes since Pacemaker-0.6.0 + CIB: Ensure changes to top-level attributes (like admin_epoch) cause a disk write + CIB: Ensure the archived file hits the disk before returning + CIB: Repair the ability to do 'atomic increment' updates (value="value++") + crmd: Bug #7 - Connecting to the crmd immediately after startup causes use-of-NULL + Medium: CIB: Mask cib_diff_resync results from the caller - they do not need to know + Medium: crmd: Delay starting the IPC server until we are fully functional + Medium: CTS: Fix the startup patterns + Medium: pengine: Bug 1820 - Allow the first resource in a group to be migrated + Medium: pengine: Bug 1820 - Check the colocation dependencies of resources to be migrated * Mon Jan 14 2008 Andrew Beekhof - 0.6.0-1 - This is the first release of the Pacemaker Cluster Resource Manager formerly part of Heartbeat. - For those looking for the GUI, mgmtd, CIM or TSA components, they are now found in the new pacemaker-pygui project. Build dependencies prevent them from being included in Heartbeat (since the built-in CRM is no longer supported) and, being non-core components, are not included with Pacemaker. - Update source tarball to revision: c94b92d550cf - Statistics: Changesets: 347 Diff: 2272 files changed, 132508 insertions(+), 305991 deletions(-) - Test hardware: + 6-node vmware cluster (sles10-sp1/256MB/vmware stonith) on a single host (opensuse10.3/2GB/2.66GHz Quad Core2) + 7-node EMC Centera cluster (sles10/512MB/2GHz Xeon/ssh stonith) - Notes: Heartbeat Stack + All testing was performed with STONITH enabled + The CRM was enabled using the "crm respawn" directive - Notes: OpenAIS Stack + This release contains a preview of support for the OpenAIS cluster stack + The current release of the OpenAIS project is missing two important patches that we require. OpenAIS packages containing these patches are available for most major distributions at: http://download.opensuse.org/repositories/server:/ha-clustering + The OpenAIS stack is not currently recommended for use in clusters that have shared data as STONITH support is not yet implimented + pingd is not yet available for use with the OpenAIS stack + 3 significant OpenAIS issues were found during testing of 4 and 6 node clusters. We are activly working together with the OpenAIS project to get these resolved. - Pending bugs encountered during testing: + OpenAIS #1736 - Openais membership took 20s to stabilize + Heartbeat #1750 - ipc_bufpool_update: magic number in head does not match + OpenAIS #1793 - Assertion failure in memb_state_gather_enter() + OpenAIS #1796 - Cluster message corruption - Changes since Heartbeat-2.1.2-24 + Add OpenAIS support + Admin: crm_uuid - Look in the right place for Heartbeat UUID files + admin: Exit and indicate a problem if the crmd exits while crmadmin is performing a query + cib: Fix CIB_OP_UPDATE calls that modify the whole CIB + cib: Fix compilation when supporting the heartbeat stack + cib: Fix memory leaks caused by the switch to get_message_xml() + cib: HA_VALGRIND_ENABLED needs to be set _and_ set to 1|yes|true + cib: Use get_message_xml() in preference to cl_get_struct() + cib: Use the return value from call to write() in cib_send_plaintext() + Core: ccm nodes can legitimately have a node id of 0 + Core: Fix peer-process tracking for the Heartbeat stack + Core: Heartbeat does not send status notifications for nodes that were already part of the cluster. Fake them instead + CRM: Add children to HA_Messages such that the field name matches F_XML_TAGNAME + crm: Adopt a more flexible appraoch to enabling Valgrind + crm: Fix compilation when bzip2 is not installed + CRM: Future-proof get_message_xml() + crmd: Filter election responses based on time not FSA state + crmd: Handle all possible peer states in crmd_ha_status_callback() + crmd: Make sure the current date/time is set - prevents use-of-NULL when evaluating rules + crmd: Relax an assertion regrading ccm membership instances + crmd: Use (node->processes&crm_proc_ais) to accurately update the CIB after replace operations + crmd: Heartbeat: Accurately record peer client status + pengine: Bug 1777 - Allow colocation with a resource in the Stopped state + pengine: Bug 1822 - Prevent use-of-NULL in PromoteRsc() + pengine: Implement three recovery policies based on op_status and op_rc + pengine: Parse fail-count correctly (it may be set to ININFITY) + pengine: Prevent graph-loop when stonith agents need to be moved around before a STONITH op + pengine: Prevent graph-loops when two operations have the same name+interval + tengine: Cancel active timers when destroying graphs + tengine: Ensure failcount is set correctly for failed stops/starts + tengine: Update failcount for oeprations that time out + Medium: admin: Prevent hang in crm_mon -1 when there is no cib connection - Patch from Junko IKEDA + Medium: cib: Require --force|-f when performing potentially dangerous commands with cibadmin + Medium: cib: Tweak the shutdown code + Medium: Common: Only count peer processes of active nodes + Medium: Core: Create generic cluster sign-in method + Medium: core: Fix compilation when Heartbeat support is disabled + Medium: Core: General cleanup for supporting two stacks + Medium: Core: iso6601 - Support parsing of time-only strings + Medium: core: Isolate more code that is only needed when SUPPORT_HEARTBEAT is enabled + Medium: crm: Improved logging of errors in the XML parser + Medium: crmd: Fix potential use-of-NULL in string comparison + Medium: crmd: Reimpliment syncronizing of CIB queries and updates when invoking the PE + Medium: crm_mon: Indicate when a node is both in standby mode and offline + Medium: pengine: Bug 1822 - Do not try an promote groups if not all of it is active + Medium: pengine: on_fail=nothing is an alias for 'ignore' not 'restart' + Medium: pengine: Prevent a potential use-of-NULL in cron_range_satisfied() + snmp subagent: fix a problem on displaying an unmanaged group + snmp subagent: use the syslog setting + snmp: v2 support (thanks to Keisuke MORI) + snmp_subagent - made it not complain about some things if shutting down diff --git a/cts/cts-fencing.in b/cts/cts-fencing.in index af5915ad1a..1771882ef5 100644 --- a/cts/cts-fencing.in +++ b/cts/cts-fencing.in @@ -1,1441 +1,1469 @@ #!@PYTHON@ """ Regression tests for Pacemaker's fencer """ # Pacemaker targets compatibility with Python 2.7 and 3.2+ from __future__ import print_function, unicode_literals, absolute_import, division __copyright__ = "Copyright 2012-2018 the Pacemaker project contributors" __license__ = "GNU General Public License version 2 or later (GPLv2+) WITHOUT ANY WARRANTY" import io import os import sys import subprocess import shlex import time import tempfile +import signal # Where to find test binaries # Prefer the source tree if available BUILD_DIR = "@abs_top_builddir@" TEST_DIR = sys.path[0] AUTOGEN_COROSYNC_TEMPLATE = """ totem { version: 2 cluster_name: cts-fencing crypto_cipher: none crypto_hash: none transport: udp } nodelist { node { nodeid: 1 name: %s ring0_addr: 127.0.0.1 } } logging { debug: off to_syslog: no to_stderr: no to_logfile: yes logfile: %s } """ # These values must be kept in sync with include/crm/crm.h class CrmExit(object): OK = 0 ERROR = 1 INVALID_PARAM = 2 UNIMPLEMENT_FEATURE = 3 INSUFFICIENT_PRIV = 4 NOT_INSTALLED = 5 NOT_CONFIGURED = 6 NOT_RUNNING = 7 USAGE = 64 DATAERR = 65 NOINPUT = 66 NOUSER = 67 NOHOST = 68 UNAVAILABLE = 69 SOFTWARE = 70 OSERR = 71 OSFILE = 72 CANTCREAT = 73 IOERR = 74 TEMPFAIL = 75 PROTOCOL = 76 NOPERM = 77 CONFIG = 78 FATAL = 100 PANIC = 101 DISCONNECT = 102 SOLO = 103 DIGEST = 104 NOSUCH = 105 QUORUM = 106 UNSAFE = 107 EXISTS = 108 MULTIPLE = 109 OLD = 110 TIMEOUT = 124 MAX = 255 def update_path(): """ Set the PATH environment variable appropriately for the tests """ new_path = os.environ['PATH'] if os.path.exists("%s/cts-fencing.in" % TEST_DIR): print("Running tests from the source tree: %s (%s)" % (BUILD_DIR, TEST_DIR)) # For pacemaker-fenced and cts-fence-helper new_path = "%s/daemons/fenced:%s" % (BUILD_DIR, new_path) new_path = "%s/tools:%s" % (BUILD_DIR, new_path) # For stonith_admin new_path = "%s/cts:%s" % (BUILD_DIR, new_path) # For cts-support else: print("Running tests from the install tree: @CRM_DAEMON_DIR@ (not %s)" % TEST_DIR) # For pacemaker-fenced, cts-fence-helper, and cts-support new_path = "@CRM_DAEMON_DIR@:%s" % (new_path) print('Using PATH="{}"'.format(new_path)) os.environ['PATH'] = new_path def pipe_output(pipes, stdout=True, stderr=False): """ Wrapper to get text output from pipes regardless of Python version """ output = "" pipe_outputs = pipes.communicate() if sys.version_info < (3,): if stdout: output = output + pipe_outputs[0] if stderr: output = output + pipe_outputs[1] else: if stdout: output = output + pipe_outputs[0].decode(sys.stdout.encoding) if stderr: output = output + pipe_outputs[1].decode(sys.stderr.encoding) return output def output_from_command(command): """ Execute command and return its standard output """ test = subprocess.Popen(shlex.split(command), stdout=subprocess.PIPE) test.wait() return pipe_output(test).split("\n") def localname(): """ Return the uname of the local host """ our_uname = output_from_command("uname -n") if our_uname: our_uname = our_uname[0] else: our_uname = "localhost" return our_uname def killall(process): """ Kill all instances of a process """ cmd = shlex.split("killall -9 -q %s" % process) test = subprocess.Popen(cmd, stdout=subprocess.PIPE) test.wait() class TestError(Exception): """ Base class for exceptions in this module """ pass class ExitCodeError(TestError): """ Exception raised when command exit status is unexpected """ def __init__(self, exit_code): self.exit_code = exit_code def __str__(self): return repr(self.exit_code) class OutputNotFoundError(TestError): """ Exception raised when command output does not contain wanted string """ def __init__(self, output): self.output = output def __str__(self): return repr(self.output) class OutputFoundError(TestError): """ Exception raised when command output contains unwanted string """ def __init__(self, exit_code): self.output = output def __str__(self): return repr(self.output) class Test(object): """ Executor for a single test """ def __init__(self, name, description, verbose=0, with_cpg=0): self.name = name self.description = description self.cmds = [] self.verbose = verbose self.result_txt = "" self.cmd_tool_output = "" self.result_exitcode = CrmExit.OK if with_cpg: self.stonith_options = "-c" self.enable_corosync = 1 else: self.stonith_options = "-s" self.enable_corosync = 0 self.stonith_process = None self.stonith_output = "" self.stonith_patterns = [] self.negative_stonith_patterns = [] self.executed = 0 def __new_cmd(self, cmd, args, exitcode, stdout_match="", no_wait=0, stdout_negative_match="", kill=None): """ Add a command to be executed as part of this test """ self.cmds.append( { "cmd" : cmd, "kill" : kill, "args" : args, "expected_exitcode" : exitcode, "stdout_match" : stdout_match, "stdout_negative_match" : stdout_negative_match, "no_wait" : no_wait, } ) def start_environment(self): """ Prepare the host for executing a test """ # Make sure we are in full control killall("pacemakerd") killall("pacemaker-fenced") if self.verbose: self.stonith_options = self.stonith_options + " -V" print("Starting pacemaker-fenced with %s" % self.stonith_options) if os.path.exists("/tmp/stonith-regression.log"): os.remove('/tmp/stonith-regression.log') cmd = "pacemaker-fenced %s -l /tmp/stonith-regression.log" % self.stonith_options self.stonith_process = subprocess.Popen(shlex.split(cmd)) time.sleep(1) def clean_environment(self): """ Clean up the host after executing a test """ if self.stonith_process: - self.stonith_process.terminate() - self.stonith_process.wait() + if self.stonith_process.poll() == None: + self.stonith_process.terminate() + self.stonith_process.wait() + else: + return_code = { + getattr(signal, _signame): _signame + for _signame in dir(signal) + if _signame.startswith('SIG') and not _signame.startswith("SIG_") + }.get(-self.stonith_process.returncode, "RET=%d" % (self.stonith_process.returncode)) + msg = "FAILURE - '%s' failed. pacemaker-fenced abnormally exited during test (%s)." + self.result_txt = msg % (self.name, return_code) + self.result_exitcode = CrmExit.ERROR self.stonith_output = "" self.stonith_process = None - logfile = io.open('/tmp/stonith-regression.log', 'rt') + # the default for utf-8 encoding would error out if e.g. memory corruption + # makes fenced output any kind of 8 bit value - while still interesting + # for debugging and we'd still like the regression-test to go over the + # full set of test-cases + logfile = io.open('/tmp/stonith-regression.log', 'rt', encoding = "ISO-8859-1") for line in logfile.readlines(): self.stonith_output = self.stonith_output + line if self.verbose: print("Daemon Output Start") print(self.stonith_output) print("Daemon Output End") os.remove('/tmp/stonith-regression.log') def add_stonith_log_pattern(self, pattern): """ Add a log pattern to expect from this test """ self.stonith_patterns.append(pattern) def add_stonith_neg_log_pattern(self, pattern): """ Add a log pattern that should not occur with this test """ self.negative_stonith_patterns.append(pattern) def add_cmd(self, cmd, args): """ Add a simple command to be executed as part of this test """ self.__new_cmd(cmd, args, CrmExit.OK, "") def add_cmd_no_wait(self, cmd, args): """ Add a simple command to be executed (without waiting) as part of this test """ self.__new_cmd(cmd, args, CrmExit.OK, "", 1) def add_cmd_check_stdout(self, cmd, args, match, no_match=""): """ Add a simple command with expected output to be executed as part of this test """ self.__new_cmd(cmd, args, CrmExit.OK, match, 0, no_match) def add_expected_fail_cmd(self, cmd, args, exitcode=CrmExit.ERROR): """ Add a command to be executed as part of this test and expected to fail """ self.__new_cmd(cmd, args, exitcode, "") def get_exitcode(self): """ Return the exit status of the last test execution """ return self.result_exitcode def print_result(self, filler): """ Print the result of the last test execution """ print("%s%s" % (filler, self.result_txt)) def run_cmd(self, args): """ Execute a command as part of this test """ cmd = shlex.split(args['args']) cmd.insert(0, args['cmd']) if self.verbose: print("\n\nRunning: "+" ".join(cmd)) test = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE) if args['kill']: if self.verbose: print("Also running: "+args['kill']) subprocess.Popen(shlex.split(args['kill'])) if args['no_wait'] == 0: test.wait() else: return CrmExit.OK output = pipe_output(test, stderr=True) if self.verbose: print(output) if test.returncode != args['expected_exitcode']: raise ExitCodeError(test.returncode) if args['stdout_match'] != "" and output.count(args['stdout_match']) == 0: raise OutputNotFoundError(output) if args['stdout_negative_match'] != "" and output.count(args['stdout_negative_match']) != 0: raise OutputFoundError(output) def count_negative_matches(self, outline): """ Return 1 if a line matches patterns that shouldn't have occurred """ count = 0 for line in self.negative_stonith_patterns: if outline.count(line): count = 1 if self.verbose: print("This pattern should not have matched = '%s" % (line)) return count def match_stonith_patterns(self): """ Check test output for expected patterns """ negative_matches = 0 cur = 0 pats = self.stonith_patterns total_patterns = len(self.stonith_patterns) if len(self.stonith_patterns) == 0 and len(self.negative_stonith_patterns) == 0: return for line in self.stonith_output.split("\n"): negative_matches = negative_matches + self.count_negative_matches(line) if len(pats) == 0: continue cur = -1 for pat in pats: cur = cur + 1 if line.count(pats[cur]): del pats[cur] break if len(pats) > 0 or negative_matches: if self.verbose: for pat in pats: print("Pattern Not Matched = '%s'" % pat) msg = "FAILURE - '%s' failed. %d patterns out of %d not matched. %d negative matches." self.result_txt = msg % (self.name, len(pats), total_patterns, negative_matches) self.result_exitcode = CrmExit.ERROR def set_error(self, step, cmd): """ Record failure of this test """ msg = "FAILURE - '%s' failed at step %d. Command: %s %s" self.result_txt = msg % (self.name, step, cmd['cmd'], cmd['args']) self.result_exitcode = CrmExit.ERROR def run(self): """ Execute this test. """ res = 0 i = 1 self.start_environment() if self.verbose: print("\n--- START TEST - %s" % self.name) self.result_txt = "SUCCESS - '%s'" % (self.name) self.result_exitcode = CrmExit.OK for cmd in self.cmds: try: self.run_cmd(cmd) except ExitCodeError as e: print("Step %d FAILED - command returned %s, expected %d" % (i, e, cmd['expected_exitcode'])) self.set_error(i, cmd); break except OutputNotFoundError as e: print("Step %d FAILED - '%s' was not found in command output: %s" % (i, cmd['stdout_match'], e)) self.set_error(i, cmd); break except OutputFoundError as e: print("Step %d FAILED - '%s' was found in command output: %s" % (i, cmd['stdout_negative_match'], e)) self.set_error(i, cmd); break if self.verbose: print("Step %d SUCCESS" % (i)) i = i + 1 self.clean_environment() if self.result_exitcode == CrmExit.OK: self.match_stonith_patterns() print(self.result_txt) if self.verbose: print("--- END TEST - %s\n" % self.name) self.executed = 1 return res class Tests(object): """ Collection of all fencing regression tests """ def __init__(self, verbose=0): self.tests = [] self.verbose = verbose self.autogen_corosync_cfg = not os.path.exists("/etc/corosync/corosync.conf") def new_test(self, name, description, with_cpg=0): """ Create a named test """ test = Test(name, description, self.verbose, with_cpg) self.tests.append(test) return test def print_list(self): """ List all registered tests """ print("\n==== %d TESTS FOUND ====" % (len(self.tests))) print("%35s - %s" % ("TEST NAME", "TEST DESCRIPTION")) print("%35s - %s" % ("--------------------", "--------------------")) for test in self.tests: print("%35s - %s" % (test.name, test.description)) print("==== END OF LIST ====\n") def start_corosync(self): """ Start the corosync process """ if self.verbose: print("Starting corosync") test = subprocess.Popen("corosync", stdout=subprocess.PIPE) test.wait() time.sleep(10) def run_single(self, name): """ Run a single named test """ for test in self.tests: if test.name == name: test.run() break def run_tests_matching(self, pattern): """ Run all tests whose name matches a pattern """ for test in self.tests: if test.name.count(pattern) != 0: test.run() def run_cpg_only(self): """ Run all corosync-enabled tests """ for test in self.tests: if test.enable_corosync: test.run() def run_no_cpg(self): """ Run all standalone tests """ for test in self.tests: if not test.enable_corosync: test.run() def run_tests(self): """ Run all tests """ for test in self.tests: test.run() def exit(self): """ Exit (with error status code if any test failed) """ for test in self.tests: if test.executed == 0: continue if test.get_exitcode() != CrmExit.OK: sys.exit(CrmExit.ERROR) sys.exit(CrmExit.OK) def print_results(self): """ Print summary of results of executed tests """ failures = 0 success = 0 print("\n\n======= FINAL RESULTS ==========") print("\n--- FAILURE RESULTS:") for test in self.tests: if test.executed == 0: continue if test.get_exitcode() != CrmExit.OK: failures = failures + 1 test.print_result(" ") else: success = success + 1 if failures == 0: print(" None") print("\n--- TOTALS\n Pass:%d\n Fail:%d\n" % (success, failures)) def build_api_sanity_tests(self): """ Register tests to verify basic API usage """ verbose_arg = "" if self.verbose: verbose_arg = "-V" test = self.new_test("standalone_low_level_api_test", "Sanity test client api in standalone mode.") test.add_cmd("cts-fence-helper", "-t %s" % (verbose_arg)) test = self.new_test("cpg_low_level_api_test", "Sanity test client api using mainloop and cpg.", 1) test.add_cmd("cts-fence-helper", "-m %s" % (verbose_arg)) def build_custom_timeout_tests(self): """ Register tests to verify custom timeout usage """ # custom timeout without topology test = self.new_test("cpg_custom_timeout_1", "Verify per device timeouts work as expected without using topology.", 1) test.add_cmd('stonith_admin', '-R false1 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=node1 node2 node3"') test.add_cmd('stonith_admin', '-R true1 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node3" -o "pcmk_off_timeout=1"') test.add_cmd('stonith_admin', '-R false2 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=node3" -o "pcmk_off_timeout=4"') - test.add_cmd("stonith_admin", "-F node3 -t 2") - # timeout is 2+1+4 = 7 - test.add_stonith_log_pattern("Total timeout set to 7") + test.add_cmd("stonith_admin", "-F node3 -t 5") + # timeout is 5+1+4 = 10 + test.add_stonith_log_pattern("Total timeout set to 10") # custom timeout _WITH_ topology test = self.new_test("cpg_custom_timeout_2", "Verify per device timeouts work as expected _WITH_ topology.", 1) test.add_cmd('stonith_admin', '-R false1 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=node1 node2 node3"') test.add_cmd('stonith_admin', '-R true1 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node3" -o "pcmk_off_timeout=1"') test.add_cmd('stonith_admin', '-R false2 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=node3" -o "pcmk_off_timeout=4000"') test.add_cmd("stonith_admin", "-r node3 -i 1 -v false1") test.add_cmd("stonith_admin", "-r node3 -i 2 -v true1") test.add_cmd("stonith_admin", "-r node3 -i 3 -v false2") - test.add_cmd("stonith_admin", "-F node3 -t 2") - # timeout is 2+1+4000 = 4003 - test.add_stonith_log_pattern("Total timeout set to 4003") + test.add_cmd("stonith_admin", "-F node3 -t 5") + # timeout is 5+1+4000 = 4006 + test.add_stonith_log_pattern("Total timeout set to 4006") def build_fence_merge_tests(self): """ Register tests to verify when fence operations should be merged """ ### Simple test that overlapping fencing operations get merged test = self.new_test("cpg_custom_merge_single", "Verify overlapping identical fencing operations are merged, no fencing levels used.", 1) test.add_cmd("stonith_admin", "-R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3\"") test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node3\" ") test.add_cmd("stonith_admin", "-R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3\"") test.add_cmd_no_wait("stonith_admin", "-F node3 -t 10") test.add_cmd("stonith_admin", "-F node3 -t 10") ### one merger will happen test.add_stonith_log_pattern("Merging stonith action off for node node3 originating from client") ### the pattern below signifies that both the original and duplicate operation completed test.add_stonith_log_pattern("Operation off of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") ### Test that multiple mergers occur test = self.new_test("cpg_custom_merge_multiple", "Verify multiple overlapping identical fencing operations are merged", 1) test.add_cmd("stonith_admin", "-R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3\"") test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"delay=2\" -o \"pcmk_host_list=node3\" ") test.add_cmd("stonith_admin", "-R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3\"") test.add_cmd_no_wait("stonith_admin", "-F node3 -t 10") test.add_cmd_no_wait("stonith_admin", "-F node3 -t 10") test.add_cmd_no_wait("stonith_admin", "-F node3 -t 10") test.add_cmd_no_wait("stonith_admin", "-F node3 -t 10") test.add_cmd("stonith_admin", "-F node3 -t 10") ### 4 mergers should occur test.add_stonith_log_pattern("Merging stonith action off for node node3 originating from client") test.add_stonith_log_pattern("Merging stonith action off for node node3 originating from client") test.add_stonith_log_pattern("Merging stonith action off for node node3 originating from client") test.add_stonith_log_pattern("Merging stonith action off for node node3 originating from client") ### the pattern below signifies that both the original and duplicate operation completed test.add_stonith_log_pattern("Operation off of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") ### Test that multiple mergers occur with topologies used test = self.new_test("cpg_custom_merge_with_topology", "Verify multiple overlapping identical fencing operations are merged with fencing levels.", 1) test.add_cmd("stonith_admin", "-R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3\"") test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node3\" ") test.add_cmd("stonith_admin", "-R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3\"") test.add_cmd("stonith_admin", "-r node3 -i 1 -v false1") test.add_cmd("stonith_admin", "-r node3 -i 1 -v false2") test.add_cmd("stonith_admin", "-r node3 -i 2 -v true1") test.add_cmd_no_wait("stonith_admin", "-F node3 -t 10") test.add_cmd_no_wait("stonith_admin", "-F node3 -t 10") test.add_cmd_no_wait("stonith_admin", "-F node3 -t 10") test.add_cmd_no_wait("stonith_admin", "-F node3 -t 10") test.add_cmd("stonith_admin", "-F node3 -t 10") ### 4 mergers should occur test.add_stonith_log_pattern("Merging stonith action off for node node3 originating from client") test.add_stonith_log_pattern("Merging stonith action off for node node3 originating from client") test.add_stonith_log_pattern("Merging stonith action off for node node3 originating from client") test.add_stonith_log_pattern("Merging stonith action off for node node3 originating from client") ### the pattern below signifies that both the original and duplicate operation completed test.add_stonith_log_pattern("Operation off of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") def build_fence_no_merge_tests(self): """ Register tests to verify when fence operations should not be merged """ test = self.new_test("cpg_custom_no_merge", "Verify differing fencing operations are not merged", 1) test.add_cmd("stonith_admin", "-R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3 node2\"") test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node3 node2\" ") test.add_cmd("stonith_admin", "-R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3 node2\"") test.add_cmd("stonith_admin", "-r node3 -i 1 -v false1") test.add_cmd("stonith_admin", "-r node3 -i 1 -v false2") test.add_cmd("stonith_admin", "-r node3 -i 2 -v true1") test.add_cmd_no_wait("stonith_admin", "-F node2 -t 10") test.add_cmd("stonith_admin", "-F node3 -t 10") test.add_stonith_neg_log_pattern("Merging stonith action off for node node3 originating from client") def build_standalone_tests(self): """ Register a grab bag of tests that can be executed in standalone or corosync mode """ test_types = [ { "prefix" : "standalone", "use_cpg" : 0, }, { "prefix" : "cpg", "use_cpg" : 1, }, ] # test what happens when all devices timeout for test_type in test_types: test = self.new_test("%s_fence_multi_device_failure" % test_type["prefix"], "Verify that all devices timeout, a fencing failure is returned.", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R false3 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"") if test_type["use_cpg"] == 1: test.add_expected_fail_cmd("stonith_admin", "-F node3 -t 2", CrmExit.TIMEOUT) test.add_stonith_log_pattern("Total timeout set to 6") else: test.add_expected_fail_cmd("stonith_admin", "-F node3 -t 2", CrmExit.ERROR) test.add_stonith_log_pattern("for host 'node3' with device 'false1' returned: ") test.add_stonith_log_pattern("for host 'node3' with device 'false2' returned: ") test.add_stonith_log_pattern("for host 'node3' with device 'false3' returned: ") # test what happens when multiple devices can fence a node, but the first device fails. for test_type in test_types: test = self.new_test("%s_fence_device_failure_rollover" % test_type["prefix"], "Verify that when one fence device fails for a node, the others are tried.", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"") - test.add_cmd("stonith_admin", "-F node3 -t 2") + test.add_cmd("stonith_admin", "-F node3 -t 5") if test_type["use_cpg"] == 1: - test.add_stonith_log_pattern("Total timeout set to 6") + test.add_stonith_log_pattern("Total timeout set to 15") + + # test what happens when we try to use a missing fence-agent. + for test_type in test_types: + test = self.new_test("%s_fence_missing_agent" % test_type["prefix"], + "Verify proper error-handling when using a non-existent fence-agent.", + test_type["use_cpg"]) + test.add_cmd("stonith_admin", + "-R true1 -a fence_missing -o \"mode=pass\" -o \"pcmk_host_list=node3\"") + test.add_cmd("stonith_admin", + "-R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node2\"") + + test.add_expected_fail_cmd("stonith_admin", "-F node3 -t 5", CrmExit.ERROR) + test.add_cmd("stonith_admin", "-F node2 -t 5") # simple topology test for one device for test_type in test_types: if test_type["use_cpg"] == 0: continue test = self.new_test("%s_topology_simple" % test_type["prefix"], "Verify all fencing devices at a level are used.", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R true -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-r node3 -i 1 -v true") - test.add_cmd("stonith_admin", "-F node3 -t 2") + test.add_cmd("stonith_admin", "-F node3 -t 5") - test.add_stonith_log_pattern("Total timeout set to 2") + test.add_stonith_log_pattern("Total timeout set to 5") test.add_stonith_log_pattern("for host 'node3' with device 'true' returned: 0") # add topology, delete topology, verify fencing still works for test_type in test_types: if test_type["use_cpg"] == 0: continue test = self.new_test("%s_topology_add_remove" % test_type["prefix"], "Verify fencing occurrs after all topology levels are removed", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R true -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-r node3 -i 1 -v true") test.add_cmd("stonith_admin", "-d node3 -i 1") - test.add_cmd("stonith_admin", "-F node3 -t 2") + test.add_cmd("stonith_admin", "-F node3 -t 5") - test.add_stonith_log_pattern("Total timeout set to 2") + test.add_stonith_log_pattern("Total timeout set to 5") test.add_stonith_log_pattern("for host 'node3' with device 'true' returned: 0") # test what happens when the first fencing level has multiple devices. for test_type in test_types: if test_type["use_cpg"] == 0: continue test = self.new_test("%s_topology_device_fails" % test_type["prefix"], "Verify if one device in a level fails, the other is tried.", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R false -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R true -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-r node3 -i 1 -v false") test.add_cmd("stonith_admin", "-r node3 -i 2 -v true") test.add_cmd("stonith_admin", "-F node3 -t 20") test.add_stonith_log_pattern("Total timeout set to 40") test.add_stonith_log_pattern("for host 'node3' with device 'false' returned: -201") test.add_stonith_log_pattern("for host 'node3' with device 'true' returned: 0") # test what happens when the first fencing level fails. for test_type in test_types: if test_type["use_cpg"] == 0: continue test = self.new_test("%s_topology_multi_level_fails" % test_type["prefix"], "Verify if one level fails, the next leve is tried.", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R true3 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R true4 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-r node3 -i 1 -v false1") test.add_cmd("stonith_admin", "-r node3 -i 1 -v true1") test.add_cmd("stonith_admin", "-r node3 -i 2 -v true2") test.add_cmd("stonith_admin", "-r node3 -i 2 -v false2") test.add_cmd("stonith_admin", "-r node3 -i 3 -v true3") test.add_cmd("stonith_admin", "-r node3 -i 3 -v true4") test.add_cmd("stonith_admin", "-F node3 -t 3") test.add_stonith_log_pattern("Total timeout set to 18") test.add_stonith_log_pattern("for host 'node3' with device 'false1' returned: -201") test.add_stonith_log_pattern("for host 'node3' with device 'false2' returned: -201") test.add_stonith_log_pattern("for host 'node3' with device 'true3' returned: 0") test.add_stonith_log_pattern("for host 'node3' with device 'true4' returned: 0") # test what happens when the first fencing level had devices that no one has registered for test_type in test_types: if test_type["use_cpg"] == 0: continue test = self.new_test("%s_topology_missing_devices" % test_type["prefix"], "Verify topology can continue with missing devices.", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R true3 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R true4 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-r node3 -i 1 -v false1") test.add_cmd("stonith_admin", "-r node3 -i 1 -v true1") test.add_cmd("stonith_admin", "-r node3 -i 2 -v true2") test.add_cmd("stonith_admin", "-r node3 -i 2 -v false2") test.add_cmd("stonith_admin", "-r node3 -i 3 -v true3") test.add_cmd("stonith_admin", "-r node3 -i 3 -v true4") - test.add_cmd("stonith_admin", "-F node3 -t 2") + test.add_cmd("stonith_admin", "-F node3 -t 5") # Test what happens if multiple fencing levels are defined, and then the first one is removed. for test_type in test_types: if test_type["use_cpg"] == 0: continue test = self.new_test("%s_topology_level_removal" % test_type["prefix"], "Verify level removal works.", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R true3 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R true4 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-r node3 -i 1 -v false1") test.add_cmd("stonith_admin", "-r node3 -i 1 -v true1") test.add_cmd("stonith_admin", "-r node3 -i 2 -v true2") test.add_cmd("stonith_admin", "-r node3 -i 2 -v false2") test.add_cmd("stonith_admin", "-r node3 -i 3 -v true3") test.add_cmd("stonith_admin", "-r node3 -i 3 -v true4") # Now remove level 2, verify none of the devices in level two are hit. test.add_cmd("stonith_admin", "-d node3 -i 2") test.add_cmd("stonith_admin", "-F node3 -t 20") test.add_stonith_log_pattern("Total timeout set to 8") test.add_stonith_log_pattern("for host 'node3' with device 'false1' returned: -201") test.add_stonith_neg_log_pattern("for host 'node3' with device 'false2' returned: ") test.add_stonith_log_pattern("for host 'node3' with device 'true3' returned: 0") test.add_stonith_log_pattern("for host 'node3' with device 'true4' returned: 0") # Test targeting a topology level by node name pattern. for test_type in test_types: if test_type["use_cpg"] == 0: continue test = self.new_test("%s_topology_level_pattern" % test_type["prefix"], "Verify targeting topology by node name pattern works.", test_type["use_cpg"]) test.add_cmd("stonith_admin", """-R true -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node1 node2 node3" """) test.add_cmd("stonith_admin", """-r '@node.*' -i 1 -v true""") - test.add_cmd("stonith_admin", "-F node3 -t 2") + test.add_cmd("stonith_admin", "-F node3 -t 5") test.add_stonith_log_pattern("for host 'node3' with device 'true' returned: 0") # test allowing commas and semicolons as delimiters in pcmk_host_list for test_type in test_types: test = self.new_test("%s_host_list_delimiters" % test_type["prefix"], "Verify commas and semicolons can be used as pcmk_host_list delimiters", test_type["use_cpg"]) test.add_cmd("stonith_admin", """-R true1 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node1,node2,node3" """) test.add_cmd("stonith_admin", """-R true2 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=pcmk1;pcmk2;pcmk3" """) - test.add_cmd("stonith_admin", "stonith_admin -F node2 -t 2") - test.add_cmd("stonith_admin", "stonith_admin -F pcmk3 -t 2") + test.add_cmd("stonith_admin", "stonith_admin -F node2 -t 5") + test.add_cmd("stonith_admin", "stonith_admin -F pcmk3 -t 5") test.add_stonith_log_pattern("for host 'node2' with device 'true1' returned: 0") test.add_stonith_log_pattern("for host 'pcmk3' with device 'true2' returned: 0") # test the stonith builds the correct list of devices that can fence a node. for test_type in test_types: test = self.new_test("%s_list_devices" % test_type["prefix"], "Verify list of devices that can fence a node is correct", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node3\"") test.add_cmd("stonith_admin", "-R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-R true3 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd_check_stdout("stonith_admin", "-l node1 -V", "true2", "true1") test.add_cmd_check_stdout("stonith_admin", "-l node1 -V", "true3", "true1") # simple test of device monitor for test_type in test_types: test = self.new_test("%s_monitor" % test_type["prefix"], "Verify device is reachable", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node3\"") test.add_cmd("stonith_admin", "-R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3\"") test.add_cmd("stonith_admin", "-Q true1") test.add_cmd("stonith_admin", "-Q false1") test.add_expected_fail_cmd("stonith_admin", "-Q true2", CrmExit.ERROR) # Verify monitor occurs for duration of timeout period on failure for test_type in test_types: test = self.new_test("%s_monitor_timeout" % test_type["prefix"], "Verify monitor uses duration of timeout period given.", test_type["use_cpg"]) test.add_cmd("stonith_admin", '-R true1 -a fence_dummy -o "mode=fail" -o "monitor_mode=fail" -o "pcmk_host_list=node3"') test.add_expected_fail_cmd("stonith_admin", "-Q true1 -t 5", CrmExit.ERROR) test.add_stonith_log_pattern("Attempt 2 to execute") # Verify monitor occurs for duration of timeout period on failure, but stops at max retries for test_type in test_types: test = self.new_test("%s_monitor_timeout_max_retries" % test_type["prefix"], "Verify monitor retries until max retry value or timeout is hit.", test_type["use_cpg"]) test.add_cmd("stonith_admin", '-R true1 -a fence_dummy -o "mode=fail" -o "monitor_mode=fail" -o "pcmk_host_list=node3"') test.add_expected_fail_cmd("stonith_admin", "-Q true1 -t 15", CrmExit.ERROR) test.add_stonith_log_pattern("Attempted to execute agent fence_dummy (list) the maximum number of times") # simple register test for test_type in test_types: test = self.new_test("%s_register" % test_type["prefix"], "Verify devices can be registered and un-registered", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node3\"") test.add_cmd("stonith_admin", "-Q true1") test.add_cmd("stonith_admin", "-D true1") test.add_expected_fail_cmd("stonith_admin", "-Q true1", CrmExit.ERROR) # simple reboot test for test_type in test_types: test = self.new_test("%s_reboot" % test_type["prefix"], "Verify devices can be rebooted", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node3\"") - test.add_cmd("stonith_admin", "-B node3 -t 2") + test.add_cmd("stonith_admin", "-B node3 -t 5") test.add_cmd("stonith_admin", "-D true1") test.add_expected_fail_cmd("stonith_admin", "-Q true1", CrmExit.ERROR) # test fencing history. for test_type in test_types: if test_type["use_cpg"] == 0: continue test = self.new_test("%s_fence_history" % test_type["prefix"], "Verify last fencing operation is returned.", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node3\"") - test.add_cmd("stonith_admin", "-F node3 -t 2 -V") + test.add_cmd("stonith_admin", "-F node3 -t 5 -V") test.add_cmd_check_stdout("stonith_admin", "-H node3", "succeeded turning off node node3", "") # simple test of dynamic list query for test_type in test_types: test = self.new_test("%s_dynamic_list_query" % test_type["prefix"], "Verify dynamic list of fencing devices can be retrieved.", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o mode=pass -o mock_dynamic_hosts=fake_port_1") test.add_cmd("stonith_admin", "-R true2 -a fence_dummy -o mode=pass -o mock_dynamic_hosts=fake_port_1") test.add_cmd("stonith_admin", "-R true3 -a fence_dummy -o mode=pass -o mock_dynamic_hosts=fake_port_1") test.add_cmd_check_stdout("stonith_admin", "-l fake_port_1", "3 fence devices found") # fence using dynamic list query for test_type in test_types: test = self.new_test("%s_fence_dynamic_list_query" % test_type["prefix"], "Verify dynamic list of fencing devices can be retrieved.", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o mode=pass -o mock_dynamic_hosts=fake_port_1") test.add_cmd("stonith_admin", "-R true2 -a fence_dummy -o mode=pass -o mock_dynamic_hosts=fake_port_1") test.add_cmd("stonith_admin", "-R true3 -a fence_dummy -o mode=pass -o mock_dynamic_hosts=fake_port_1") test.add_cmd("stonith_admin", "-F fake_port_1 -t 5 -V") # simple test of query using status action for test_type in test_types: test = self.new_test("%s_status_query" % test_type["prefix"], "Verify dynamic list of fencing devices can be retrieved.", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_check=status\"") test.add_cmd("stonith_admin", "-R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_check=status\"") test.add_cmd("stonith_admin", "-R true3 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_check=status\"") test.add_cmd_check_stdout("stonith_admin", "-l fake_port_1", "3 fence devices found") # test what happens when no reboot action is advertised for test_type in test_types: test = self.new_test("%s_no_reboot_support" % test_type["prefix"], "Verify reboot action defaults to off when no reboot action is advertised by agent.", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy_no_reboot -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-B node1 -t 5 -V") test.add_stonith_log_pattern("does not advertise support for 'reboot', performing 'off'") test.add_stonith_log_pattern("with device 'true1' returned: 0 (OK)") # make sure reboot is used when reboot action is advertised for test_type in test_types: test = self.new_test("%s_with_reboot_support" % test_type["prefix"], "Verify reboot action can be used when metadata advertises it.", test_type["use_cpg"]) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "-B node1 -t 5 -V") test.add_stonith_neg_log_pattern("does not advertise support for 'reboot', performing 'off'") test.add_stonith_log_pattern("with device 'true1' returned: 0 (OK)") def build_nodeid_tests(self): """ Register tests that use a corosync node id """ our_uname = localname() ### verify nodeid is supplied when nodeid is in the metadata parameters test = self.new_test("cpg_supply_nodeid", "Verify nodeid is given when fence agent has nodeid as parameter", 1) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=%s\"" % (our_uname)) test.add_cmd("stonith_admin", "-F %s -t 3" % (our_uname)) test.add_stonith_log_pattern("For stonith action (off) for victim %s, adding nodeid" % (our_uname)) ### verify nodeid is _NOT_ supplied when nodeid is not in the metadata parameters test = self.new_test("cpg_do_not_supply_nodeid", "Verify nodeid is _NOT_ given when fence agent does not have nodeid as parameter", 1) # use a host name that won't be in corosync.conf test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=regr-test\"") test.add_cmd("stonith_admin", "-F regr-test -t 3") test.add_stonith_neg_log_pattern("For stonith action (off) for victim regr-test, adding nodeid") ### verify nodeid use doesn't explode standalone mode test = self.new_test("standalone_do_not_supply_nodeid", "Verify nodeid in metadata parameter list doesn't kill standalone mode", 0) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=%s\"" % (our_uname)) test.add_cmd("stonith_admin", "-F %s -t 3" % (our_uname)) test.add_stonith_neg_log_pattern("For stonith action (off) for victim %s, adding nodeid" % (our_uname)) def build_unfence_tests(self): """ Register tests that verify unfencing """ our_uname = localname() ### verify unfencing using automatic unfencing test = self.new_test("cpg_unfence_required_1", "Verify require unfencing on all devices when automatic=true in agent's metadata", 1) test.add_cmd('stonith_admin', '-R true1 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s"' % (our_uname)) test.add_cmd('stonith_admin', '-R true2 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s"' % (our_uname)) test.add_cmd("stonith_admin", "-U %s -t 3" % (our_uname)) # both devices should be executed test.add_stonith_log_pattern("with device 'true1' returned: 0 (OK)") test.add_stonith_log_pattern("with device 'true2' returned: 0 (OK)") ### verify unfencing using automatic unfencing fails if any of the required agents fail test = self.new_test("cpg_unfence_required_2", "Verify require unfencing on all devices when automatic=true in agent's metadata", 1) test.add_cmd('stonith_admin', '-R true1 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s"' % (our_uname)) test.add_cmd('stonith_admin', '-R true2 -a fence_dummy_auto_unfence -o "mode=fail" -o "pcmk_host_list=%s"' % (our_uname)) test.add_expected_fail_cmd("stonith_admin", "-U %s -t 6" % (our_uname), CrmExit.ERROR) ### verify unfencing using automatic devices with topology test = self.new_test("cpg_unfence_required_3", "Verify require unfencing on all devices even when at different topology levels", 1) test.add_cmd('stonith_admin', '-R true1 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s node3"' % (our_uname)) test.add_cmd('stonith_admin', '-R true2 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s node3"' % (our_uname)) test.add_cmd("stonith_admin", "-r %s -i 1 -v true1" % (our_uname)) test.add_cmd("stonith_admin", "-r %s -i 2 -v true2" % (our_uname)) test.add_cmd("stonith_admin", "-U %s -t 3" % (our_uname)) test.add_stonith_log_pattern("with device 'true1' returned: 0 (OK)") test.add_stonith_log_pattern("with device 'true2' returned: 0 (OK)") ### verify unfencing using automatic devices with topology test = self.new_test("cpg_unfence_required_4", "Verify all required devices are executed even with topology levels fail.", 1) test.add_cmd('stonith_admin', '-R true1 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s node3"' % (our_uname)) test.add_cmd('stonith_admin', '-R true2 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s node3"' % (our_uname)) test.add_cmd('stonith_admin', '-R true3 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s node3"' % (our_uname)) test.add_cmd('stonith_admin', '-R true4 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s node3"' % (our_uname)) test.add_cmd('stonith_admin', '-R false1 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=%s node3"' % (our_uname)) test.add_cmd('stonith_admin', '-R false2 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=%s node3"' % (our_uname)) test.add_cmd('stonith_admin', '-R false3 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=%s node3"' % (our_uname)) test.add_cmd('stonith_admin', '-R false4 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=%s node3"' % (our_uname)) test.add_cmd("stonith_admin", "-r %s -i 1 -v true1" % (our_uname)) test.add_cmd("stonith_admin", "-r %s -i 1 -v false1" % (our_uname)) test.add_cmd("stonith_admin", "-r %s -i 2 -v false2" % (our_uname)) test.add_cmd("stonith_admin", "-r %s -i 2 -v true2" % (our_uname)) test.add_cmd("stonith_admin", "-r %s -i 2 -v false3" % (our_uname)) test.add_cmd("stonith_admin", "-r %s -i 2 -v true3" % (our_uname)) test.add_cmd("stonith_admin", "-r %s -i 3 -v false4" % (our_uname)) test.add_cmd("stonith_admin", "-r %s -i 4 -v true4" % (our_uname)) test.add_cmd("stonith_admin", "-U %s -t 3" % (our_uname)) test.add_stonith_log_pattern("with device 'true1' returned: 0 (OK)") test.add_stonith_log_pattern("with device 'true2' returned: 0 (OK)") test.add_stonith_log_pattern("with device 'true3' returned: 0 (OK)") test.add_stonith_log_pattern("with device 'true4' returned: 0 (OK)") def build_unfence_on_target_tests(self): """ Register tests that verify unfencing that runs on the target """ our_uname = localname() ### verify unfencing using on_target device test = self.new_test("cpg_unfence_on_target_1", "Verify unfencing with on_target = true", 1) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=%s\"" % (our_uname)) test.add_cmd("stonith_admin", "-U %s -t 3" % (our_uname)) test.add_stonith_log_pattern("(on) to be executed on the target node") ### verify failure of unfencing using on_target device test = self.new_test("cpg_unfence_on_target_2", "Verify failure unfencing with on_target = true", 1) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=%s node_fake_1234\"" % (our_uname)) test.add_expected_fail_cmd("stonith_admin", "-U node_fake_1234 -t 3", CrmExit.ERROR) test.add_stonith_log_pattern("(on) to be executed on the target node") ### verify unfencing using on_target device with topology test = self.new_test("cpg_unfence_on_target_3", "Verify unfencing with on_target = true using topology", 1) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=%s node3\"" % (our_uname)) test.add_cmd("stonith_admin", "-R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=%s node3\"" % (our_uname)) test.add_cmd("stonith_admin", "-r %s -i 1 -v true1" % (our_uname)) test.add_cmd("stonith_admin", "-r %s -i 2 -v true2" % (our_uname)) test.add_cmd("stonith_admin", "-U %s -t 3" % (our_uname)) test.add_stonith_log_pattern("(on) to be executed on the target node") ### verify unfencing using on_target device with topology fails when victim node doesn't exist test = self.new_test("cpg_unfence_on_target_4", "Verify unfencing failure with on_target = true using topology", 1) test.add_cmd("stonith_admin", "-R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=%s node_fake\"" % (our_uname)) test.add_cmd("stonith_admin", "-R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=%s node_fake\"" % (our_uname)) test.add_cmd("stonith_admin", "-r node_fake -i 1 -v true1") test.add_cmd("stonith_admin", "-r node_fake -i 2 -v true2") test.add_expected_fail_cmd("stonith_admin", "-U node_fake -t 3", CrmExit.ERROR) test.add_stonith_log_pattern("(on) to be executed on the target node") def build_remap_tests(self): """ Register tests that verify remapping of reboots to off-on """ test = self.new_test("cpg_remap_simple", "Verify sequential topology reboot is remapped to all-off-then-all-on", 1) test.add_cmd("stonith_admin", """-R true1 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """ """-o "pcmk_off_timeout=1" -o "pcmk_reboot_timeout=10" """) test.add_cmd("stonith_admin", """-R true2 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """ """-o "pcmk_off_timeout=2" -o "pcmk_reboot_timeout=20" """) test.add_cmd("stonith_admin", "-r node_fake -i 1 -v true1 -v true2") test.add_cmd("stonith_admin", "-B node_fake -t 5") test.add_stonith_log_pattern("Remapping multiple-device reboot of node_fake") # timeout should be sum of off timeouts (1+2=3), not reboot timeouts (10+20=30) test.add_stonith_log_pattern("Total timeout set to 3 for peer's fencing of node_fake") test.add_stonith_log_pattern("perform op 'node_fake off' with 'true1'") test.add_stonith_log_pattern("perform op 'node_fake off' with 'true2'") test.add_stonith_log_pattern("Remapped off of node_fake complete, remapping to on") # fence_dummy sets "on" as an on_target action test.add_stonith_log_pattern("Ignoring true1 'on' failure (no capable peers) for node_fake") test.add_stonith_log_pattern("Ignoring true2 'on' failure (no capable peers) for node_fake") test.add_stonith_log_pattern("Undoing remap of reboot of node_fake") test = self.new_test("cpg_remap_automatic", "Verify remapped topology reboot skips automatic 'on'", 1) test.add_cmd("stonith_admin", """-R true1 -a fence_dummy_auto_unfence """ """-o "mode=pass" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", """-R true2 -a fence_dummy_auto_unfence """ """-o "mode=pass" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", "-r node_fake -i 1 -v true1 -v true2") test.add_cmd("stonith_admin", "-B node_fake -t 5") test.add_stonith_log_pattern("Remapping multiple-device reboot of node_fake") test.add_stonith_log_pattern("perform op 'node_fake off' with 'true1'") test.add_stonith_log_pattern("perform op 'node_fake off' with 'true2'") test.add_stonith_log_pattern("Remapped off of node_fake complete, remapping to on") test.add_stonith_log_pattern("Undoing remap of reboot of node_fake") test.add_stonith_neg_log_pattern("perform op 'node_fake on' with") test.add_stonith_neg_log_pattern("'on' failure") test = self.new_test("cpg_remap_complex_1", "Verify remapped topology reboot in second level works if non-remapped first level fails", 1) test.add_cmd("stonith_admin", """-R false1 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", """-R true1 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", """-R true2 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", "-r node_fake -i 1 -v false1") test.add_cmd("stonith_admin", "-r node_fake -i 2 -v true1 -v true2") test.add_cmd("stonith_admin", "-B node_fake -t 5") test.add_stonith_log_pattern("perform op 'node_fake reboot' with 'false1'") test.add_stonith_log_pattern("Remapping multiple-device reboot of node_fake") test.add_stonith_log_pattern("perform op 'node_fake off' with 'true1'") test.add_stonith_log_pattern("perform op 'node_fake off' with 'true2'") test.add_stonith_log_pattern("Remapped off of node_fake complete, remapping to on") test.add_stonith_log_pattern("Ignoring true1 'on' failure (no capable peers) for node_fake") test.add_stonith_log_pattern("Ignoring true2 'on' failure (no capable peers) for node_fake") test.add_stonith_log_pattern("Undoing remap of reboot of node_fake") test = self.new_test("cpg_remap_complex_2", "Verify remapped topology reboot failure in second level proceeds to third level", 1) test.add_cmd("stonith_admin", """-R false1 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", """-R false2 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", """-R true1 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", """-R true2 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", """-R true3 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", "-r node_fake -i 1 -v false1") test.add_cmd("stonith_admin", "-r node_fake -i 2 -v true1 -v false2 -v true3") test.add_cmd("stonith_admin", "-r node_fake -i 3 -v true2") test.add_cmd("stonith_admin", "-B node_fake -t 5") test.add_stonith_log_pattern("perform op 'node_fake reboot' with 'false1'") test.add_stonith_log_pattern("Remapping multiple-device reboot of node_fake") test.add_stonith_log_pattern("perform op 'node_fake off' with 'true1'") test.add_stonith_log_pattern("perform op 'node_fake off' with 'false2'") test.add_stonith_log_pattern("Attempted to execute agent fence_dummy (off) the maximum number of times") test.add_stonith_log_pattern("Undoing remap of reboot of node_fake") test.add_stonith_log_pattern("perform op 'node_fake reboot' with 'true2'") test.add_stonith_neg_log_pattern("node_fake with true3") def setup_environment(self, use_corosync): """ Prepare the host before executing any tests """ if use_corosync: if self.autogen_corosync_cfg: (handle, self.autogen_corosync_log) = tempfile.mkstemp(prefix="cts-fencing-", suffix=".corosync.log") os.close(handle) corosync_cfg = io.open("/etc/corosync/corosync.conf", "w") corosync_cfg.write(AUTOGEN_COROSYNC_TEMPLATE % (localname(), self.autogen_corosync_log)) corosync_cfg.close() ### make sure we are in control ### killall("corosync") self.start_corosync() subprocess.call(["cts-support", "install"]) def cleanup_environment(self, use_corosync): """ Clean up the host after executing desired tests """ if use_corosync: killall("corosync") if self.autogen_corosync_cfg: if self.verbose: print("Corosync output") logfile = io.open(self.autogen_corosync_log, 'rt') for line in logfile.readlines(): print(line.strip()) logfile.close() os.remove(self.autogen_corosync_log) os.remove("/etc/corosync/corosync.conf") subprocess.call(["cts-support", "uninstall"]) class TestOptions(object): """ Option handler """ def __init__(self): self.options = {} self.options['list-tests'] = 0 self.options['run-all'] = 1 self.options['run-only'] = "" self.options['run-only-pattern'] = "" self.options['verbose'] = 0 self.options['invalid-arg'] = "" self.options['cpg-only'] = 0 self.options['no-cpg'] = 0 self.options['show-usage'] = 0 def build_options(self, argv): """ Set options based on command-line arguments """ args = argv[1:] skip = 0 for i in range(0, len(args)): if skip: skip = 0 continue elif args[i] == "-h" or args[i] == "--help": self.options['show-usage'] = 1 elif args[i] == "-l" or args[i] == "--list-tests": self.options['list-tests'] = 1 elif args[i] == "-V" or args[i] == "--verbose": self.options['verbose'] = 1 elif args[i] == "-n" or args[i] == "--no-cpg": self.options['no-cpg'] = 1 elif args[i] == "-c" or args[i] == "--cpg-only": self.options['cpg-only'] = 1 elif args[i] == "-r" or args[i] == "--run-only": self.options['run-only'] = args[i+1] skip = 1 elif args[i] == "-p" or args[i] == "--run-only-pattern": self.options['run-only-pattern'] = args[i+1] skip = 1 def show_usage(self): """ Show command usage """ print("usage: " + sys.argv[0] + " [options]") print("If no options are provided, all tests will run") print("Options:") print("\t [--help | -h] Show usage") print("\t [--list-tests | -l] Print out all registered tests.") print("\t [--cpg-only | -c] Only run tests that require corosync.") print("\t [--no-cpg | -n] Only run tests that do not require corosync") print("\t [--run-only | -r 'testname'] Run a specific test") print("\t [--verbose | -V] Verbose output") print("\t [--run-only-pattern | -p 'string'] Run only tests containing the string value") print("\n\tExample: Run only the test 'start_stop'") print("\t\t " + sys.argv[0] + " --run-only start_stop") print("\n\tExample: Run only the tests with the string 'systemd' present in them") print("\t\t " + sys.argv[0] + " --run-only-pattern systemd") def main(argv): """ Run fencing regression tests as specified by arguments """ update_path() opts = TestOptions() opts.build_options(argv) use_corosync = 1 tests = Tests(opts.options['verbose']) tests.build_standalone_tests() tests.build_custom_timeout_tests() tests.build_api_sanity_tests() tests.build_fence_merge_tests() tests.build_fence_no_merge_tests() tests.build_unfence_tests() tests.build_unfence_on_target_tests() tests.build_nodeid_tests() tests.build_remap_tests() if opts.options['list-tests']: tests.print_list() sys.exit(CrmExit.OK) elif opts.options['show-usage']: opts.show_usage() sys.exit(CrmExit.OK) print("Starting ...") if opts.options['no-cpg']: use_corosync = 0 tests.setup_environment(use_corosync) if opts.options['run-only-pattern'] != "": tests.run_tests_matching(opts.options['run-only-pattern']) tests.print_results() elif opts.options['run-only'] != "": tests.run_single(opts.options['run-only']) tests.print_results() elif opts.options['no-cpg']: tests.run_no_cpg() tests.print_results() elif opts.options['cpg-only']: tests.run_cpg_only() tests.print_results() else: tests.run_tests() tests.print_results() tests.cleanup_environment(use_corosync) tests.exit() if __name__ == "__main__": main(sys.argv) diff --git a/daemons/controld/controld_remote_ra.c b/daemons/controld/controld_remote_ra.c index 66a3788076..097e93a46d 100644 --- a/daemons/controld/controld_remote_ra.c +++ b/daemons/controld/controld_remote_ra.c @@ -1,1293 +1,1296 @@ /* - * Copyright 2013-2018 David Vossel + * Copyright 2013-2019 the Pacemaker project contributors + * + * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #define REMOTE_LRMD_RA "remote" /* The max start timeout before cmd retry */ #define MAX_START_TIMEOUT_MS 10000 typedef struct remote_ra_cmd_s { /*! the local node the cmd is issued from */ char *owner; /*! the remote node the cmd is executed on */ char *rsc_id; /*! the action to execute */ char *action; /*! some string the client wants us to give it back */ char *userdata; char *exit_reason; // descriptive text on error /*! start delay in ms */ int start_delay; /*! timer id used for start delay. */ int delay_id; /*! timeout in ms for cmd */ int timeout; int remaining_timeout; /*! recurring interval in ms */ guint interval_ms; /*! interval timer id */ int interval_id; int reported_success; int monitor_timeout_id; int takeover_timeout_id; /*! action parameters */ lrmd_key_value_t *params; /*! executed rc */ int rc; int op_status; int call_id; time_t start_time; gboolean cancel; } remote_ra_cmd_t; enum remote_migration_status { expect_takeover = 1, takeover_complete, }; typedef struct remote_ra_data_s { crm_trigger_t *work; remote_ra_cmd_t *cur_cmd; GList *cmds; GList *recurring_cmds; enum remote_migration_status migrate_status; gboolean active; /* Maintenance mode is difficult to determine from the controller's context, * so we have it signalled back with the transition from the scheduler. */ gboolean is_maintenance; /* Similar for whether we are controlling a guest node or remote node. * Fortunately there is a meta-attribute in the transition already and * as the situation doesn't change over time we can use the * resource start for noting down the information for later use when * the attributes aren't at hand. */ gboolean controlling_guest; } remote_ra_data_t; static int handle_remote_ra_start(lrm_state_t * lrm_state, remote_ra_cmd_t * cmd, int timeout_ms); static void handle_remote_ra_stop(lrm_state_t * lrm_state, remote_ra_cmd_t * cmd); static GList *fail_all_monitor_cmds(GList * list); static void free_cmd(gpointer user_data) { remote_ra_cmd_t *cmd = user_data; if (!cmd) { return; } if (cmd->delay_id) { g_source_remove(cmd->delay_id); } if (cmd->interval_id) { g_source_remove(cmd->interval_id); } if (cmd->monitor_timeout_id) { g_source_remove(cmd->monitor_timeout_id); } if (cmd->takeover_timeout_id) { g_source_remove(cmd->takeover_timeout_id); } free(cmd->owner); free(cmd->rsc_id); free(cmd->action); free(cmd->userdata); free(cmd->exit_reason); lrmd_key_value_freeall(cmd->params); free(cmd); } static int generate_callid(void) { static int remote_ra_callid = 0; remote_ra_callid++; if (remote_ra_callid <= 0) { remote_ra_callid = 1; } return remote_ra_callid; } static gboolean recurring_helper(gpointer data) { remote_ra_cmd_t *cmd = data; lrm_state_t *connection_rsc = NULL; cmd->interval_id = 0; connection_rsc = lrm_state_find(cmd->rsc_id); if (connection_rsc && connection_rsc->remote_ra_data) { remote_ra_data_t *ra_data = connection_rsc->remote_ra_data; ra_data->recurring_cmds = g_list_remove(ra_data->recurring_cmds, cmd); ra_data->cmds = g_list_append(ra_data->cmds, cmd); mainloop_set_trigger(ra_data->work); } return FALSE; } static gboolean start_delay_helper(gpointer data) { remote_ra_cmd_t *cmd = data; lrm_state_t *connection_rsc = NULL; cmd->delay_id = 0; connection_rsc = lrm_state_find(cmd->rsc_id); if (connection_rsc && connection_rsc->remote_ra_data) { remote_ra_data_t *ra_data = connection_rsc->remote_ra_data; mainloop_set_trigger(ra_data->work); } return FALSE; } /*! * \internal * \brief Handle cluster communication related to pacemaker_remote node joining * * \param[in] node_name Name of newly integrated pacemaker_remote node */ static void remote_node_up(const char *node_name) { int call_opt, call_id = 0; xmlNode *update, *state; crm_node_t *node; CRM_CHECK(node_name != NULL, return); crm_info("Announcing pacemaker_remote node %s", node_name); /* Clear node's operation history. The node's transient attributes should * and normally will be cleared when the node leaves, but since remote node * state has a number of corner cases, clear them here as well, to be sure. */ call_opt = crmd_cib_smart_opt(); erase_status_tag(node_name, XML_CIB_TAG_LRM, call_opt); erase_status_tag(node_name, XML_TAG_TRANSIENT_NODEATTRS, call_opt); /* Clear node's probed attribute */ update_attrd(node_name, CRM_OP_PROBED, NULL, NULL, TRUE); /* Ensure node is in the remote peer cache with member status */ node = crm_remote_peer_get(node_name); CRM_CHECK(node != NULL, return); crm_update_peer_state(__FUNCTION__, node, CRM_NODE_MEMBER, 0); /* pacemaker_remote nodes don't participate in the membership layer, * so cluster nodes don't automatically get notified when they come and go. * We send a cluster message to the DC, and update the CIB node state entry, * so the DC will get it sooner (via message) or later (via CIB refresh), * and any other interested parties can query the CIB. */ send_remote_state_message(node_name, TRUE); update = create_xml_node(NULL, XML_CIB_TAG_STATUS); state = create_node_state_update(node, node_update_cluster, update, __FUNCTION__); /* Clear the XML_NODE_IS_FENCED flag in the node state. If the node ever * needs to be fenced, this flag will allow various actions to determine * whether the fencing has happened yet. */ crm_xml_add(state, XML_NODE_IS_FENCED, "0"); /* TODO: If the remote connection drops, and this (async) CIB update either * failed or has not yet completed, later actions could mistakenly think the * node has already been fenced (if the XML_NODE_IS_FENCED attribute was * previously set, because it won't have been cleared). This could prevent * actual fencing or allow recurring monitor failures to be cleared too * soon. Ideally, we wouldn't rely on the CIB for the fenced status. */ fsa_cib_update(XML_CIB_TAG_STATUS, update, call_opt, call_id, NULL); if (call_id < 0) { crm_perror(LOG_WARNING, "%s CIB node state setup", node_name); } free_xml(update); } enum down_opts { DOWN_KEEP_LRM, DOWN_ERASE_LRM }; /*! * \internal * \brief Handle cluster communication related to pacemaker_remote node leaving * * \param[in] node_name Name of lost node * \param[in] opts Whether to keep or erase LRM history */ static void remote_node_down(const char *node_name, const enum down_opts opts) { xmlNode *update; int call_id = 0; int call_opt = crmd_cib_smart_opt(); crm_node_t *node; /* Purge node from attrd's memory */ update_attrd_remote_node_removed(node_name, NULL); /* Purge node's transient attributes */ erase_status_tag(node_name, XML_TAG_TRANSIENT_NODEATTRS, call_opt); /* Normally, the LRM operation history should be kept until the node comes * back up. However, after a successful fence, we want to clear it, so we * don't think resources are still running on the node. */ if (opts == DOWN_ERASE_LRM) { erase_status_tag(node_name, XML_CIB_TAG_LRM, call_opt); } /* Ensure node is in the remote peer cache with lost state */ node = crm_remote_peer_get(node_name); CRM_CHECK(node != NULL, return); crm_update_peer_state(__FUNCTION__, node, CRM_NODE_LOST, 0); /* Notify DC */ send_remote_state_message(node_name, FALSE); /* Update CIB node state */ update = create_xml_node(NULL, XML_CIB_TAG_STATUS); create_node_state_update(node, node_update_cluster, update, __FUNCTION__); fsa_cib_update(XML_CIB_TAG_STATUS, update, call_opt, call_id, NULL); if (call_id < 0) { crm_perror(LOG_ERR, "%s CIB node state update", node_name); } free_xml(update); } /*! * \internal * \brief Handle effects of a remote RA command on node state * * \param[in] cmd Completed remote RA command */ static void check_remote_node_state(remote_ra_cmd_t *cmd) { /* Only successful actions can change node state */ if (cmd->rc != PCMK_OCF_OK) { return; } if (safe_str_eq(cmd->action, "start")) { remote_node_up(cmd->rsc_id); } else if (safe_str_eq(cmd->action, "migrate_from")) { /* After a successful migration, we don't need to do remote_node_up() * because the DC already knows the node is up, and we don't want to * clear LRM history etc. We do need to add the remote node to this * host's remote peer cache, because (unless it happens to be DC) * it hasn't been tracking the remote node, and other code relies on * the cache to distinguish remote nodes from unseen cluster nodes. */ crm_node_t *node = crm_remote_peer_get(cmd->rsc_id); CRM_CHECK(node != NULL, return); crm_update_peer_state(__FUNCTION__, node, CRM_NODE_MEMBER, 0); } else if (safe_str_eq(cmd->action, "stop")) { lrm_state_t *lrm_state = lrm_state_find(cmd->rsc_id); remote_ra_data_t *ra_data = lrm_state? lrm_state->remote_ra_data : NULL; if (ra_data) { if (ra_data->migrate_status != takeover_complete) { /* Stop means down if we didn't successfully migrate elsewhere */ remote_node_down(cmd->rsc_id, DOWN_KEEP_LRM); } else if (AM_I_DC == FALSE) { /* Only the connection host and DC track node state, * so if the connection migrated elsewhere and we aren't DC, * un-cache the node, so we don't have stale info */ crm_remote_peer_cache_remove(cmd->rsc_id); } } } /* We don't do anything for successful monitors, which is correct for * routine recurring monitors, and for monitors on nodes where the * connection isn't supposed to be (the cluster will stop the connection in * that case). However, if the initial probe finds the connection already * active on the node where we want it, we probably should do * remote_node_up(). Unfortunately, we can't distinguish that case here. * Given that connections have to be initiated by the cluster, the chance of * that should be close to zero. */ } static void report_remote_ra_result(remote_ra_cmd_t * cmd) { lrmd_event_data_t op = { 0, }; check_remote_node_state(cmd); op.type = lrmd_event_exec_complete; op.rsc_id = cmd->rsc_id; op.op_type = cmd->action; op.user_data = cmd->userdata; op.exit_reason = cmd->exit_reason; op.timeout = cmd->timeout; op.interval_ms = cmd->interval_ms; op.rc = cmd->rc; op.op_status = cmd->op_status; op.t_run = cmd->start_time; op.t_rcchange = cmd->start_time; if (cmd->reported_success && cmd->rc != PCMK_OCF_OK) { op.t_rcchange = time(NULL); /* This edge case will likely never ever occur, but if it does the * result is that a failure will not be processed correctly. This is only * remotely possible because we are able to detect a connection resource's tcp * connection has failed at any moment after start has completed. The actual * recurring operation is just a connectivity ping. * * basically, we are not guaranteed that the first successful monitor op and * a subsequent failed monitor op will not occur in the same timestamp. We have to * make it look like the operations occurred at separate times though. */ if (op.t_rcchange == op.t_run) { op.t_rcchange++; } } if (cmd->params) { lrmd_key_value_t *tmp; op.params = crm_str_table_new(); for (tmp = cmd->params; tmp; tmp = tmp->next) { g_hash_table_insert(op.params, strdup(tmp->key), strdup(tmp->value)); } } op.call_id = cmd->call_id; op.remote_nodename = cmd->owner; lrm_op_callback(&op); if (op.params) { g_hash_table_destroy(op.params); } } static void update_remaining_timeout(remote_ra_cmd_t * cmd) { cmd->remaining_timeout = ((cmd->timeout / 1000) - (time(NULL) - cmd->start_time)) * 1000; } static gboolean retry_start_cmd_cb(gpointer data) { lrm_state_t *lrm_state = data; remote_ra_data_t *ra_data = lrm_state->remote_ra_data; remote_ra_cmd_t *cmd = NULL; int rc = -1; if (!ra_data || !ra_data->cur_cmd) { return FALSE; } cmd = ra_data->cur_cmd; if (safe_str_neq(cmd->action, "start") && safe_str_neq(cmd->action, "migrate_from")) { return FALSE; } update_remaining_timeout(cmd); if (cmd->remaining_timeout > 0) { rc = handle_remote_ra_start(lrm_state, cmd, cmd->remaining_timeout); } if (rc != 0) { cmd->rc = PCMK_OCF_UNKNOWN_ERROR; cmd->op_status = PCMK_LRM_OP_ERROR; report_remote_ra_result(cmd); if (ra_data->cmds) { mainloop_set_trigger(ra_data->work); } ra_data->cur_cmd = NULL; free_cmd(cmd); } else { /* wait for connection event */ } return FALSE; } static gboolean connection_takeover_timeout_cb(gpointer data) { lrm_state_t *lrm_state = NULL; remote_ra_cmd_t *cmd = data; crm_info("takeover event timed out for node %s", cmd->rsc_id); cmd->takeover_timeout_id = 0; lrm_state = lrm_state_find(cmd->rsc_id); handle_remote_ra_stop(lrm_state, cmd); free_cmd(cmd); return FALSE; } static gboolean monitor_timeout_cb(gpointer data) { lrm_state_t *lrm_state = NULL; remote_ra_cmd_t *cmd = data; lrm_state = lrm_state_find(cmd->rsc_id); crm_info("Timed out waiting for remote poke response from %s%s", cmd->rsc_id, (lrm_state? "" : " (no LRM state)")); cmd->monitor_timeout_id = 0; cmd->op_status = PCMK_LRM_OP_TIMEOUT; cmd->rc = PCMK_OCF_UNKNOWN_ERROR; if (lrm_state && lrm_state->remote_ra_data) { remote_ra_data_t *ra_data = lrm_state->remote_ra_data; if (ra_data->cur_cmd == cmd) { ra_data->cur_cmd = NULL; } if (ra_data->cmds) { mainloop_set_trigger(ra_data->work); } } report_remote_ra_result(cmd); free_cmd(cmd); if(lrm_state) { lrm_state_disconnect(lrm_state); } return FALSE; } static void synthesize_lrmd_success(lrm_state_t *lrm_state, const char *rsc_id, const char *op_type) { lrmd_event_data_t op = { 0, }; if (lrm_state == NULL) { /* if lrm_state not given assume local */ lrm_state = lrm_state_find(fsa_our_uname); } CRM_ASSERT(lrm_state != NULL); op.type = lrmd_event_exec_complete; op.rsc_id = rsc_id; op.op_type = op_type; op.rc = PCMK_OCF_OK; op.op_status = PCMK_LRM_OP_DONE; op.t_run = time(NULL); op.t_rcchange = op.t_run; op.call_id = generate_callid(); process_lrm_event(lrm_state, &op, NULL, NULL); } void remote_lrm_op_callback(lrmd_event_data_t * op) { gboolean cmd_handled = FALSE; lrm_state_t *lrm_state = NULL; remote_ra_data_t *ra_data = NULL; remote_ra_cmd_t *cmd = NULL; crm_debug("Processing '%s%s%s' event on remote connection to %s: %s " "(%d) status=%s (%d)", (op->op_type? op->op_type : ""), (op->op_type? " " : ""), lrmd_event_type2str(op->type), op->remote_nodename, services_ocf_exitcode_str(op->rc), op->rc, services_lrm_status_str(op->op_status), op->op_status); lrm_state = lrm_state_find(op->remote_nodename); if (!lrm_state || !lrm_state->remote_ra_data) { crm_debug("No state information found for remote connection event"); return; } ra_data = lrm_state->remote_ra_data; if (op->type == lrmd_event_new_client) { // Another client has connected to the remote daemon if (ra_data->migrate_status == expect_takeover) { // Great, we knew this was coming ra_data->migrate_status = takeover_complete; } else { crm_err("Unexpected pacemaker_remote client takeover for %s. Disconnecting", op->remote_nodename); /* In this case, lrmd_tls_connection_destroy() will be called under the control of mainloop. */ /* Do not free lrm_state->conn yet. */ /* It'll be freed in the following stop action. */ lrm_state_disconnect_only(lrm_state); } return; } /* filter all EXEC events up */ if (op->type == lrmd_event_exec_complete) { if (ra_data->migrate_status == takeover_complete) { crm_debug("ignoring event, this connection is taken over by another node"); } else { lrm_op_callback(op); } return; } if ((op->type == lrmd_event_disconnect) && (ra_data->cur_cmd == NULL)) { if (ra_data->active == FALSE) { crm_debug("Disconnection from Pacemaker Remote node %s complete", lrm_state->node_name); } else if (!remote_ra_is_in_maintenance(lrm_state)) { crm_err("Lost connection to Pacemaker Remote node %s", lrm_state->node_name); ra_data->recurring_cmds = fail_all_monitor_cmds(ra_data->recurring_cmds); ra_data->cmds = fail_all_monitor_cmds(ra_data->cmds); } else { crm_notice("Unmanaged Pacemaker Remote node %s disconnected", lrm_state->node_name); /* Do roughly what a 'stop' on the remote-resource would do */ handle_remote_ra_stop(lrm_state, NULL); remote_node_down(lrm_state->node_name, DOWN_KEEP_LRM); /* now fake the reply of a successful 'stop' */ synthesize_lrmd_success(NULL, lrm_state->node_name, "stop"); } return; } if (!ra_data->cur_cmd) { crm_debug("no event to match"); return; } cmd = ra_data->cur_cmd; /* Start actions and migrate from actions complete after connection * comes back to us. */ if (op->type == lrmd_event_connect && (safe_str_eq(cmd->action, "start") || safe_str_eq(cmd->action, "migrate_from"))) { if (op->connection_rc < 0) { update_remaining_timeout(cmd); if (op->connection_rc == -ENOKEY) { // Hard error, don't retry cmd->op_status = PCMK_LRM_OP_ERROR; cmd->rc = PCMK_OCF_INVALID_PARAM; cmd->exit_reason = strdup("Authentication key not readable"); } else if (cmd->remaining_timeout > 3000) { crm_trace("rescheduling start, remaining timeout %d", cmd->remaining_timeout); g_timeout_add(1000, retry_start_cmd_cb, lrm_state); return; } else { crm_trace("can't reschedule start, remaining timeout too small %d", cmd->remaining_timeout); cmd->op_status = PCMK_LRM_OP_TIMEOUT; cmd->rc = PCMK_OCF_UNKNOWN_ERROR; } } else { lrm_state_reset_tables(lrm_state, TRUE); cmd->rc = PCMK_OCF_OK; cmd->op_status = PCMK_LRM_OP_DONE; ra_data->active = TRUE; } crm_debug("Remote connection event matched %s action", cmd->action); report_remote_ra_result(cmd); cmd_handled = TRUE; } else if (op->type == lrmd_event_poke && safe_str_eq(cmd->action, "monitor")) { if (cmd->monitor_timeout_id) { g_source_remove(cmd->monitor_timeout_id); cmd->monitor_timeout_id = 0; } /* Only report success the first time, after that only worry about failures. * For this function, if we get the poke pack, it is always a success. Pokes * only fail if the send fails, or the response times out. */ if (!cmd->reported_success) { cmd->rc = PCMK_OCF_OK; cmd->op_status = PCMK_LRM_OP_DONE; report_remote_ra_result(cmd); cmd->reported_success = 1; } crm_debug("Remote poke event matched %s action", cmd->action); /* success, keep rescheduling if interval is present. */ if (cmd->interval_ms && (cmd->cancel == FALSE)) { ra_data->recurring_cmds = g_list_append(ra_data->recurring_cmds, cmd); cmd->interval_id = g_timeout_add(cmd->interval_ms, recurring_helper, cmd); cmd = NULL; /* prevent free */ } cmd_handled = TRUE; } else if (op->type == lrmd_event_disconnect && safe_str_eq(cmd->action, "monitor")) { if (ra_data->active == TRUE && (cmd->cancel == FALSE)) { cmd->rc = PCMK_OCF_UNKNOWN_ERROR; cmd->op_status = PCMK_LRM_OP_ERROR; report_remote_ra_result(cmd); crm_err("Remote connection to %s unexpectedly dropped during monitor", lrm_state->node_name); } cmd_handled = TRUE; } else if (op->type == lrmd_event_new_client && safe_str_eq(cmd->action, "stop")) { handle_remote_ra_stop(lrm_state, cmd); cmd_handled = TRUE; } else { crm_debug("Event did not match %s action", ra_data->cur_cmd->action); } if (cmd_handled) { ra_data->cur_cmd = NULL; if (ra_data->cmds) { mainloop_set_trigger(ra_data->work); } free_cmd(cmd); } } static void handle_remote_ra_stop(lrm_state_t * lrm_state, remote_ra_cmd_t * cmd) { remote_ra_data_t *ra_data = NULL; CRM_ASSERT(lrm_state); ra_data = lrm_state->remote_ra_data; if (ra_data->migrate_status != takeover_complete) { /* delete pending ops when ever the remote connection is intentionally stopped */ g_hash_table_remove_all(lrm_state->pending_ops); } else { /* we no longer hold the history if this connection has been migrated, * however, we keep metadata cache for future use */ lrm_state_reset_tables(lrm_state, FALSE); } ra_data->active = FALSE; lrm_state_disconnect(lrm_state); if (ra_data->cmds) { g_list_free_full(ra_data->cmds, free_cmd); } if (ra_data->recurring_cmds) { g_list_free_full(ra_data->recurring_cmds, free_cmd); } ra_data->cmds = NULL; ra_data->recurring_cmds = NULL; ra_data->cur_cmd = NULL; if (cmd) { cmd->rc = PCMK_OCF_OK; cmd->op_status = PCMK_LRM_OP_DONE; report_remote_ra_result(cmd); } } static int handle_remote_ra_start(lrm_state_t * lrm_state, remote_ra_cmd_t * cmd, int timeout_ms) { const char *server = NULL; lrmd_key_value_t *tmp = NULL; int port = 0; remote_ra_data_t *ra_data = lrm_state->remote_ra_data; int timeout_used = timeout_ms > MAX_START_TIMEOUT_MS ? MAX_START_TIMEOUT_MS : timeout_ms; for (tmp = cmd->params; tmp; tmp = tmp->next) { if (safe_str_eq(tmp->key, XML_RSC_ATTR_REMOTE_RA_ADDR) || safe_str_eq(tmp->key, XML_RSC_ATTR_REMOTE_RA_SERVER)) { server = tmp->value; } else if (safe_str_eq(tmp->key, XML_RSC_ATTR_REMOTE_RA_PORT)) { port = atoi(tmp->value); } else if (safe_str_eq(tmp->key, CRM_META"_"XML_RSC_ATTR_CONTAINER)) { ra_data->controlling_guest = TRUE; } } return lrm_state_remote_connect_async(lrm_state, server, port, timeout_used); } static gboolean handle_remote_ra_exec(gpointer user_data) { int rc = 0; lrm_state_t *lrm_state = user_data; remote_ra_data_t *ra_data = lrm_state->remote_ra_data; remote_ra_cmd_t *cmd; GList *first = NULL; if (ra_data->cur_cmd) { /* still waiting on previous cmd */ return TRUE; } while (ra_data->cmds) { first = ra_data->cmds; cmd = first->data; if (cmd->delay_id) { /* still waiting for start delay timer to trip */ return TRUE; } ra_data->cmds = g_list_remove_link(ra_data->cmds, first); g_list_free_1(first); if (!strcmp(cmd->action, "start") || !strcmp(cmd->action, "migrate_from")) { ra_data->migrate_status = 0; rc = handle_remote_ra_start(lrm_state, cmd, cmd->timeout); if (rc == 0) { /* take care of this later when we get async connection result */ crm_debug("Remote connection started, waiting for connect event"); ra_data->cur_cmd = cmd; return TRUE; } else { crm_debug("connect failed, not expecting to match any connection event later"); cmd->rc = PCMK_OCF_UNKNOWN_ERROR; cmd->op_status = PCMK_LRM_OP_ERROR; } report_remote_ra_result(cmd); } else if (!strcmp(cmd->action, "monitor")) { if (lrm_state_is_connected(lrm_state) == TRUE) { rc = lrm_state_poke_connection(lrm_state); if (rc < 0) { cmd->rc = PCMK_OCF_UNKNOWN_ERROR; cmd->op_status = PCMK_LRM_OP_ERROR; } } else { rc = -1; cmd->op_status = PCMK_LRM_OP_DONE; cmd->rc = PCMK_OCF_NOT_RUNNING; } if (rc == 0) { crm_debug("Poked Pacemaker Remote at node %s, waiting for async response", cmd->rsc_id); ra_data->cur_cmd = cmd; cmd->monitor_timeout_id = g_timeout_add(cmd->timeout, monitor_timeout_cb, cmd); return TRUE; } report_remote_ra_result(cmd); } else if (!strcmp(cmd->action, "stop")) { if (ra_data->migrate_status == expect_takeover) { /* briefly wait on stop for the takeover event to occur. If the * takeover event does not occur during the wait period, that's fine. * It just means that the remote-node's lrm_status section is going to get * cleared which will require all the resources running in the remote-node * to be explicitly re-detected via probe actions. If the takeover does occur * successfully, then we can leave the status section intact. */ cmd->takeover_timeout_id = g_timeout_add((cmd->timeout/2), connection_takeover_timeout_cb, cmd); ra_data->cur_cmd = cmd; return TRUE; } handle_remote_ra_stop(lrm_state, cmd); } else if (!strcmp(cmd->action, "migrate_to")) { ra_data->migrate_status = expect_takeover; cmd->rc = PCMK_OCF_OK; cmd->op_status = PCMK_LRM_OP_DONE; report_remote_ra_result(cmd); } else if (!strcmp(cmd->action, "reload")) { /* reloads are a no-op right now, add logic here when they become important */ cmd->rc = PCMK_OCF_OK; cmd->op_status = PCMK_LRM_OP_DONE; report_remote_ra_result(cmd); } free_cmd(cmd); } return TRUE; } static void remote_ra_data_init(lrm_state_t * lrm_state) { remote_ra_data_t *ra_data = NULL; if (lrm_state->remote_ra_data) { return; } ra_data = calloc(1, sizeof(remote_ra_data_t)); ra_data->work = mainloop_add_trigger(G_PRIORITY_HIGH, handle_remote_ra_exec, lrm_state); lrm_state->remote_ra_data = ra_data; } void remote_ra_cleanup(lrm_state_t * lrm_state) { remote_ra_data_t *ra_data = lrm_state->remote_ra_data; if (!ra_data) { return; } if (ra_data->cmds) { g_list_free_full(ra_data->cmds, free_cmd); } if (ra_data->recurring_cmds) { g_list_free_full(ra_data->recurring_cmds, free_cmd); } mainloop_destroy_trigger(ra_data->work); free(ra_data); lrm_state->remote_ra_data = NULL; } gboolean is_remote_lrmd_ra(const char *agent, const char *provider, const char *id) { if (agent && provider && !strcmp(agent, REMOTE_LRMD_RA) && !strcmp(provider, "pacemaker")) { return TRUE; } if (id && lrm_state_find(id) && safe_str_neq(id, fsa_our_uname)) { return TRUE; } return FALSE; } lrmd_rsc_info_t * remote_ra_get_rsc_info(lrm_state_t * lrm_state, const char *rsc_id) { lrmd_rsc_info_t *info = NULL; if ((lrm_state_find(rsc_id))) { info = calloc(1, sizeof(lrmd_rsc_info_t)); info->id = strdup(rsc_id); info->type = strdup(REMOTE_LRMD_RA); info->standard = strdup(PCMK_RESOURCE_CLASS_OCF); info->provider = strdup("pacemaker"); } return info; } static gboolean is_remote_ra_supported_action(const char *action) { if (!action) { return FALSE; } else if (strcmp(action, "start") && strcmp(action, "stop") && strcmp(action, "reload") && strcmp(action, "migrate_to") && strcmp(action, "migrate_from") && strcmp(action, "monitor")) { return FALSE; } return TRUE; } static GList * fail_all_monitor_cmds(GList * list) { GList *rm_list = NULL; remote_ra_cmd_t *cmd = NULL; GListPtr gIter = NULL; for (gIter = list; gIter != NULL; gIter = gIter->next) { cmd = gIter->data; if ((cmd->interval_ms > 0) && safe_str_eq(cmd->action, "monitor")) { rm_list = g_list_append(rm_list, cmd); } } for (gIter = rm_list; gIter != NULL; gIter = gIter->next) { cmd = gIter->data; cmd->rc = PCMK_OCF_UNKNOWN_ERROR; cmd->op_status = PCMK_LRM_OP_ERROR; crm_trace("Pre-emptively failing %s %s (interval=%u, %s)", cmd->action, cmd->rsc_id, cmd->interval_ms, cmd->userdata); report_remote_ra_result(cmd); list = g_list_remove(list, cmd); free_cmd(cmd); } /* frees only the list data, not the cmds */ g_list_free(rm_list); return list; } static GList * remove_cmd(GList * list, const char *action, guint interval_ms) { remote_ra_cmd_t *cmd = NULL; GListPtr gIter = NULL; for (gIter = list; gIter != NULL; gIter = gIter->next) { cmd = gIter->data; if ((cmd->interval_ms == interval_ms) && safe_str_eq(cmd->action, action)) { break; } cmd = NULL; } if (cmd) { list = g_list_remove(list, cmd); free_cmd(cmd); } return list; } int remote_ra_cancel(lrm_state_t *lrm_state, const char *rsc_id, const char *action, guint interval_ms) { lrm_state_t *connection_rsc = NULL; remote_ra_data_t *ra_data = NULL; connection_rsc = lrm_state_find(rsc_id); if (!connection_rsc || !connection_rsc->remote_ra_data) { return -EINVAL; } ra_data = connection_rsc->remote_ra_data; ra_data->cmds = remove_cmd(ra_data->cmds, action, interval_ms); ra_data->recurring_cmds = remove_cmd(ra_data->recurring_cmds, action, interval_ms); if (ra_data->cur_cmd && (ra_data->cur_cmd->interval_ms == interval_ms) && (safe_str_eq(ra_data->cur_cmd->action, action))) { ra_data->cur_cmd->cancel = TRUE; } return 0; } static remote_ra_cmd_t * handle_dup_monitor(remote_ra_data_t *ra_data, guint interval_ms, const char *userdata) { GList *gIter = NULL; remote_ra_cmd_t *cmd = NULL; /* there are 3 places a potential duplicate monitor operation * could exist. * 1. recurring_cmds list. where the op is waiting for its next interval * 2. cmds list, where the op is queued to get executed immediately * 3. cur_cmd, which means the monitor op is in flight right now. */ if (interval_ms == 0) { return NULL; } if (ra_data->cur_cmd && ra_data->cur_cmd->cancel == FALSE && (ra_data->cur_cmd->interval_ms == interval_ms) && safe_str_eq(ra_data->cur_cmd->action, "monitor")) { cmd = ra_data->cur_cmd; goto handle_dup; } for (gIter = ra_data->recurring_cmds; gIter != NULL; gIter = gIter->next) { cmd = gIter->data; if ((cmd->interval_ms == interval_ms) && safe_str_eq(cmd->action, "monitor")) { goto handle_dup; } } for (gIter = ra_data->cmds; gIter != NULL; gIter = gIter->next) { cmd = gIter->data; if ((cmd->interval_ms == interval_ms) && safe_str_eq(cmd->action, "monitor")) { goto handle_dup; } } return NULL; handle_dup: crm_trace("merging duplicate monitor cmd " CRM_OP_FMT, cmd->rsc_id, "monitor", interval_ms); /* update the userdata */ if (userdata) { free(cmd->userdata); cmd->userdata = strdup(userdata); } /* if we've already reported success, generate a new call id */ if (cmd->reported_success) { cmd->start_time = time(NULL); cmd->call_id = generate_callid(); cmd->reported_success = 0; } /* if we have an interval_id set, that means we are in the process of * waiting for this cmd's next interval. instead of waiting, cancel * the timer and execute the action immediately */ if (cmd->interval_id) { g_source_remove(cmd->interval_id); cmd->interval_id = 0; recurring_helper(cmd); } return cmd; } int remote_ra_exec(lrm_state_t *lrm_state, const char *rsc_id, const char *action, const char *userdata, guint interval_ms, int timeout, /* ms */ int start_delay, /* ms */ lrmd_key_value_t * params) { int rc = 0; lrm_state_t *connection_rsc = NULL; remote_ra_cmd_t *cmd = NULL; remote_ra_data_t *ra_data = NULL; if (is_remote_ra_supported_action(action) == FALSE) { rc = -EINVAL; goto exec_done; } connection_rsc = lrm_state_find(rsc_id); if (!connection_rsc) { rc = -EINVAL; goto exec_done; } remote_ra_data_init(connection_rsc); ra_data = connection_rsc->remote_ra_data; cmd = handle_dup_monitor(ra_data, interval_ms, userdata); if (cmd) { - return cmd->call_id; + rc = cmd->call_id; + goto exec_done; } cmd = calloc(1, sizeof(remote_ra_cmd_t)); cmd->owner = strdup(lrm_state->node_name); cmd->rsc_id = strdup(rsc_id); cmd->action = strdup(action); cmd->userdata = strdup(userdata); cmd->interval_ms = interval_ms; cmd->timeout = timeout; cmd->start_delay = start_delay; cmd->params = params; cmd->start_time = time(NULL); cmd->call_id = generate_callid(); if (cmd->start_delay) { cmd->delay_id = g_timeout_add(cmd->start_delay, start_delay_helper, cmd); } ra_data->cmds = g_list_append(ra_data->cmds, cmd); mainloop_set_trigger(ra_data->work); return cmd->call_id; exec_done: lrmd_key_value_freeall(params); return rc; } /*! * \internal * \brief Immediately fail all monitors of a remote node, if proxied here * * \param[in] node_name Name of pacemaker_remote node */ void remote_ra_fail(const char *node_name) { lrm_state_t *lrm_state = lrm_state_find(node_name); if (lrm_state && lrm_state_is_connected(lrm_state)) { remote_ra_data_t *ra_data = lrm_state->remote_ra_data; crm_info("Failing monitors on pacemaker_remote node %s", node_name); ra_data->recurring_cmds = fail_all_monitor_cmds(ra_data->recurring_cmds); ra_data->cmds = fail_all_monitor_cmds(ra_data->cmds); } } /* A guest node fencing implied by host fencing looks like: * * * * * * * */ #define XPATH_PSEUDO_FENCE "//" XML_GRAPH_TAG_PSEUDO_EVENT \ "[@" XML_LRM_ATTR_TASK "='stonith']/" XML_GRAPH_TAG_DOWNED \ "/" XML_CIB_TAG_NODE /*! * \internal * \brief Check a pseudo-action for Pacemaker Remote node side effects * * \param[in] xml XML of pseudo-action to check */ void remote_ra_process_pseudo(xmlNode *xml) { xmlXPathObjectPtr search = xpath_search(xml, XPATH_PSEUDO_FENCE); if (numXpathResults(search) == 1) { xmlNode *result = getXpathResult(search, 0); /* Normally, we handle the necessary side effects of a guest node stop * action when reporting the remote agent's result. However, if the stop * is implied due to fencing, it will be a fencing pseudo-event, and * there won't be a result to report. Handle that case here. * * This will result in a duplicate call to remote_node_down() if the * guest stop was real instead of implied, but that shouldn't hurt. * * There is still one corner case that isn't handled: if a guest node * isn't running any resources when its host is fenced, it will appear * to be cleanly stopped, so there will be no pseudo-fence, and our * peer cache state will be incorrect unless and until the guest is * recovered. */ if (result) { const char *remote = ID(result); if (remote) { remote_node_down(remote, DOWN_ERASE_LRM); } } } freeXpathObject(search); } static void remote_ra_maintenance(lrm_state_t * lrm_state, gboolean maintenance) { remote_ra_data_t *ra_data = lrm_state->remote_ra_data; xmlNode *update, *state; int call_opt, call_id = 0; crm_node_t *node; call_opt = crmd_cib_smart_opt(); node = crm_remote_peer_get(lrm_state->node_name); CRM_CHECK(node != NULL, return); update = create_xml_node(NULL, XML_CIB_TAG_STATUS); state = create_node_state_update(node, node_update_none, update, __FUNCTION__); crm_xml_add(state, XML_NODE_IS_MAINTENANCE, maintenance?"1":"0"); fsa_cib_update(XML_CIB_TAG_STATUS, update, call_opt, call_id, NULL); if (call_id < 0) { crm_perror(LOG_WARNING, "%s CIB node state update failed", lrm_state->node_name); } else { /* TODO: still not 100% sure that async update will succeed ... */ ra_data->is_maintenance = maintenance; } free_xml(update); } #define XPATH_PSEUDO_MAINTENANCE "//" XML_GRAPH_TAG_PSEUDO_EVENT \ "[@" XML_LRM_ATTR_TASK "='" CRM_OP_MAINTENANCE_NODES "']/" \ XML_GRAPH_TAG_MAINTENANCE /*! * \internal * \brief Check a pseudo-action holding updates for maintenance state * * \param[in] xml XML of pseudo-action to check */ void remote_ra_process_maintenance_nodes(xmlNode *xml) { xmlXPathObjectPtr search = xpath_search(xml, XPATH_PSEUDO_MAINTENANCE); if (numXpathResults(search) == 1) { xmlNode *node; int cnt = 0, cnt_remote = 0; for (node = first_named_child(getXpathResult(search, 0), XML_CIB_TAG_NODE); node; node = __xml_next(node)) { lrm_state_t *lrm_state = lrm_state_find(ID(node)); cnt++; if (lrm_state && lrm_state->remote_ra_data && ((remote_ra_data_t *) lrm_state->remote_ra_data)->active) { cnt_remote++; remote_ra_maintenance(lrm_state, crm_atoi(crm_element_value(node, XML_NODE_IS_MAINTENANCE), "0")); } } crm_trace("Action holds %d nodes (%d remotes found) " "adjusting maintenance-mode", cnt, cnt_remote); } freeXpathObject(search); } gboolean remote_ra_is_in_maintenance(lrm_state_t * lrm_state) { remote_ra_data_t *ra_data = lrm_state->remote_ra_data; return ra_data->is_maintenance; } gboolean remote_ra_controlling_guest(lrm_state_t * lrm_state) { remote_ra_data_t *ra_data = lrm_state->remote_ra_data; return ra_data->controlling_guest; } diff --git a/daemons/fenced/fenced_commands.c b/daemons/fenced/fenced_commands.c index d25b84c6f7..eba83cc6d6 100644 --- a/daemons/fenced/fenced_commands.c +++ b/daemons/fenced/fenced_commands.c @@ -1,2754 +1,2771 @@ /* * Copyright 2009-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #if SUPPORT_CIBSECRETS # include #endif #include GHashTable *device_list = NULL; GHashTable *topology = NULL; GList *cmd_list = NULL; struct device_search_s { /* target of fence action */ char *host; /* requested fence action */ char *action; /* timeout to use if a device is queried dynamically for possible targets */ int per_device_timeout; /* number of registered fencing devices at time of request */ int replies_needed; /* number of device replies received so far */ int replies_received; /* whether the target is eligible to perform requested action (or off) */ bool allow_suicide; /* private data to pass to search callback function */ void *user_data; /* function to call when all replies have been received */ void (*callback) (GList * devices, void *user_data); /* devices capable of performing requested action (or off if remapping) */ GListPtr capable; }; static gboolean stonith_device_dispatch(gpointer user_data); static void st_child_done(GPid pid, int rc, const char *output, gpointer user_data); static void stonith_send_reply(xmlNode * reply, int call_options, const char *remote_peer, const char *client_id); static void search_devices_record_result(struct device_search_s *search, const char *device, gboolean can_fence); typedef struct async_command_s { int id; int pid; int fd_stdout; int options; int default_timeout; /* seconds */ int timeout; /* seconds */ int start_delay; /* milliseconds */ int delay_id; char *op; char *origin; char *client; char *client_name; char *remote_op_id; char *victim; uint32_t victim_nodeid; char *action; char *device; char *mode; GListPtr device_list; GListPtr device_next; void *internal_user_data; void (*done_cb) (GPid pid, int rc, const char *output, gpointer user_data); guint timer_sigterm; guint timer_sigkill; /*! If the operation timed out, this is the last signal * we sent to the process to get it to terminate */ int last_timeout_signo; stonith_device_t *active_on; + stonith_device_t *activating_on; } async_command_t; static xmlNode *stonith_construct_async_reply(async_command_t * cmd, const char *output, xmlNode * data, int rc); static gboolean is_action_required(const char *action, stonith_device_t *device) { return device && device->automatic_unfencing && safe_str_eq(action, "on"); } static int get_action_delay_max(stonith_device_t * device, const char * action) { const char *value = NULL; int delay_max_ms = 0; if (safe_str_neq(action, "off") && safe_str_neq(action, "reboot")) { return 0; } value = g_hash_table_lookup(device->params, STONITH_ATTR_DELAY_MAX); if (value) { delay_max_ms = crm_get_msec(value); } return delay_max_ms; } static int get_action_delay_base(stonith_device_t * device, const char * action) { const char *value = NULL; int delay_base_ms = 0; if (safe_str_neq(action, "off") && safe_str_neq(action, "reboot")) { return 0; } value = g_hash_table_lookup(device->params, STONITH_ATTR_DELAY_BASE); if (value) { delay_base_ms = crm_get_msec(value); } return delay_base_ms; } /*! * \internal * \brief Override STONITH timeout with pcmk_*_timeout if available * * \param[in] device STONITH device to use * \param[in] action STONITH action name * \param[in] default_timeout Timeout to use if device does not have * a pcmk_*_timeout parameter for action * * \return Value of pcmk_(action)_timeout if available, otherwise default_timeout * \note For consistency, it would be nice if reboot/off/on timeouts could be * set the same way as start/stop/monitor timeouts, i.e. with an * entry in the fencing resource configuration. However that * is insufficient because fencing devices may be registered directly via * the fencer's register_device() API instead of going through the CIB * (e.g. stonith_admin uses it for its -R option, and the executor uses it * to ensure a device is registered when a command is issued). As device * properties, pcmk_*_timeout parameters can be grabbed by the fencer when * the device is registered, whether by CIB change or API call. */ static int get_action_timeout(stonith_device_t * device, const char *action, int default_timeout) { if (action && device && device->params) { char buffer[64] = { 0, }; const char *value = NULL; /* If "reboot" was requested but the device does not support it, * we will remap to "off", so check timeout for "off" instead */ if (safe_str_eq(action, "reboot") && is_not_set(device->flags, st_device_supports_reboot)) { crm_trace("%s doesn't support reboot, using timeout for off instead", device->id); action = "off"; } /* If the device config specified an action-specific timeout, use it */ snprintf(buffer, sizeof(buffer), "pcmk_%s_timeout", action); value = g_hash_table_lookup(device->params, buffer); if (value) { return atoi(value); } } return default_timeout; } static void free_async_command(async_command_t * cmd) { if (!cmd) { return; } if (cmd->delay_id) { g_source_remove(cmd->delay_id); } cmd_list = g_list_remove(cmd_list, cmd); g_list_free_full(cmd->device_list, free); free(cmd->device); free(cmd->action); free(cmd->victim); free(cmd->remote_op_id); free(cmd->client); free(cmd->client_name); free(cmd->origin); free(cmd->mode); free(cmd->op); free(cmd); } static async_command_t * create_async_command(xmlNode * msg) { async_command_t *cmd = NULL; xmlNode *op = get_xpath_object("//@" F_STONITH_ACTION, msg, LOG_ERR); const char *action = crm_element_value(op, F_STONITH_ACTION); CRM_CHECK(action != NULL, crm_log_xml_warn(msg, "NoAction"); return NULL); crm_log_xml_trace(msg, "Command"); cmd = calloc(1, sizeof(async_command_t)); crm_element_value_int(msg, F_STONITH_CALLID, &(cmd->id)); crm_element_value_int(msg, F_STONITH_CALLOPTS, &(cmd->options)); crm_element_value_int(msg, F_STONITH_TIMEOUT, &(cmd->default_timeout)); cmd->timeout = cmd->default_timeout; cmd->origin = crm_element_value_copy(msg, F_ORIG); cmd->remote_op_id = crm_element_value_copy(msg, F_STONITH_REMOTE_OP_ID); cmd->client = crm_element_value_copy(msg, F_STONITH_CLIENTID); cmd->client_name = crm_element_value_copy(msg, F_STONITH_CLIENTNAME); cmd->op = crm_element_value_copy(msg, F_STONITH_OPERATION); cmd->action = strdup(action); cmd->victim = crm_element_value_copy(op, F_STONITH_TARGET); cmd->mode = crm_element_value_copy(op, F_STONITH_MODE); cmd->device = crm_element_value_copy(op, F_STONITH_DEVICE); CRM_CHECK(cmd->op != NULL, crm_log_xml_warn(msg, "NoOp"); free_async_command(cmd); return NULL); CRM_CHECK(cmd->client != NULL, crm_log_xml_warn(msg, "NoClient")); cmd->done_cb = st_child_done; cmd_list = g_list_append(cmd_list, cmd); return cmd; } static int get_action_limit(stonith_device_t * device) { const char *value = NULL; int action_limit = 1; value = g_hash_table_lookup(device->params, STONITH_ATTR_ACTION_LIMIT); if (value) { action_limit = crm_parse_int(value, "1"); if (action_limit == 0) { /* pcmk_action_limit should not be 0. Enforce it to be 1. */ action_limit = 1; } } return action_limit; } static int get_active_cmds(stonith_device_t * device) { int counter = 0; GListPtr gIter = NULL; GListPtr gIterNext = NULL; CRM_CHECK(device != NULL, return 0); for (gIter = cmd_list; gIter != NULL; gIter = gIterNext) { async_command_t *cmd = gIter->data; gIterNext = gIter->next; if (cmd->active_on == device) { counter++; } } return counter; } +static void +fork_cb(GPid pid, gpointer user_data) +{ + async_command_t *cmd = (async_command_t *) user_data; + stonith_device_t * device = + /* in case of a retry we've done the move from + activating_on to active_on already + */ + cmd->activating_on?cmd->activating_on:cmd->active_on; + + CRM_ASSERT(device); + crm_debug("Operation %s%s%s on %s now running with pid=%d, timeout=%ds", + cmd->action, cmd->victim ? " for node " : "", cmd->victim ? cmd->victim : "", + device->id, pid, cmd->timeout); + cmd->active_on = device; + cmd->activating_on = NULL; +} + static gboolean stonith_device_execute(stonith_device_t * device) { int exec_rc = 0; const char *action_str = NULL; async_command_t *cmd = NULL; stonith_action_t *action = NULL; int active_cmds = 0; int action_limit = 0; CRM_CHECK(device != NULL, return FALSE); active_cmds = get_active_cmds(device); action_limit = get_action_limit(device); if (action_limit > -1 && active_cmds >= action_limit) { crm_trace("%s is over its action limit of %d (%u active action%s)", device->id, action_limit, active_cmds, active_cmds > 1 ? "s" : ""); return TRUE; } if (device->pending_ops) { GList *first = device->pending_ops; cmd = first->data; if (cmd && cmd->delay_id) { crm_trace ("Operation %s%s%s on %s was asked to run too early, waiting for start_delay timeout of %dms", cmd->action, cmd->victim ? " for node " : "", cmd->victim ? cmd->victim : "", device->id, cmd->start_delay); return TRUE; } device->pending_ops = g_list_remove_link(device->pending_ops, first); g_list_free_1(first); } if (cmd == NULL) { crm_trace("Nothing further to do for %s", device->id); return TRUE; } if(safe_str_eq(device->agent, STONITH_WATCHDOG_AGENT)) { if(safe_str_eq(cmd->action, "reboot")) { pcmk_panic(__FUNCTION__); return TRUE; } else if(safe_str_eq(cmd->action, "off")) { pcmk_panic(__FUNCTION__); return TRUE; } else { crm_info("Faking success for %s watchdog operation", cmd->action); cmd->done_cb(0, 0, NULL, cmd); return TRUE; } } #if SUPPORT_CIBSECRETS if (replace_secret_params(device->id, device->params) < 0) { /* replacing secrets failed! */ if (safe_str_eq(cmd->action,"stop")) { /* don't fail on stop! */ crm_info("proceeding with the stop operation for %s", device->id); } else { crm_err("failed to get secrets for %s, " "considering resource not configured", device->id); exec_rc = PCMK_OCF_NOT_CONFIGURED; cmd->done_cb(0, exec_rc, NULL, cmd); return TRUE; } } #endif action_str = cmd->action; if (safe_str_eq(cmd->action, "reboot") && is_not_set(device->flags, st_device_supports_reboot)) { crm_warn("Agent '%s' does not advertise support for 'reboot', performing 'off' action instead", device->agent); action_str = "off"; } action = stonith_action_create(device->agent, action_str, cmd->victim, cmd->victim_nodeid, cmd->timeout, device->params, device->aliases); - /* for async exec, exec_rc is pid if positive and error code if negative/zero */ - exec_rc = stonith_action_execute_async(action, (void *)cmd, cmd->done_cb); + /* for async exec, exec_rc is negative for early error exit + otherwise handling of success/errors is done via callbacks */ + cmd->activating_on = device; + exec_rc = stonith_action_execute_async(action, (void *)cmd, + cmd->done_cb, fork_cb); - if (exec_rc > 0) { - crm_debug("Operation %s%s%s on %s now running with pid=%d, timeout=%ds", - cmd->action, cmd->victim ? " for node " : "", cmd->victim ? cmd->victim : "", - device->id, exec_rc, cmd->timeout); - cmd->active_on = device; - - } else { + if (exec_rc < 0) { crm_warn("Operation %s%s%s on %s failed: %s (%d)", cmd->action, cmd->victim ? " for node " : "", cmd->victim ? cmd->victim : "", device->id, pcmk_strerror(exec_rc), exec_rc); + cmd->activating_on = NULL; cmd->done_cb(0, exec_rc, NULL, cmd); } return TRUE; } static gboolean stonith_device_dispatch(gpointer user_data) { return stonith_device_execute(user_data); } static gboolean start_delay_helper(gpointer data) { async_command_t *cmd = data; stonith_device_t *device = NULL; cmd->delay_id = 0; device = cmd->device ? g_hash_table_lookup(device_list, cmd->device) : NULL; if (device) { mainloop_set_trigger(device->work); } return FALSE; } static void schedule_stonith_command(async_command_t * cmd, stonith_device_t * device) { int delay_max = 0; int delay_base = 0; CRM_CHECK(cmd != NULL, return); CRM_CHECK(device != NULL, return); if (cmd->device) { free(cmd->device); } if (device->include_nodeid && cmd->victim) { crm_node_t *node = crm_get_peer(0, cmd->victim); cmd->victim_nodeid = node->id; } cmd->device = strdup(device->id); cmd->timeout = get_action_timeout(device, cmd->action, cmd->default_timeout); if (cmd->remote_op_id) { crm_debug("Scheduling %s on %s for remote peer %s with op id (%s) (timeout=%ds)", cmd->action, device->id, cmd->origin, cmd->remote_op_id, cmd->timeout); } else { crm_debug("Scheduling %s on %s for %s (timeout=%ds)", cmd->action, device->id, cmd->client, cmd->timeout); } device->pending_ops = g_list_append(device->pending_ops, cmd); mainloop_set_trigger(device->work); delay_max = get_action_delay_max(device, cmd->action); delay_base = get_action_delay_base(device, cmd->action); if (delay_max == 0) { delay_max = delay_base; } if (delay_max < delay_base) { crm_warn("Base-delay (%dms) is larger than max-delay (%dms) " "for %s on %s - limiting to max-delay", delay_base, delay_max, cmd->action, device->id); delay_base = delay_max; } if (delay_max > 0) { // coverity[dont_call] We're not using rand() for security cmd->start_delay = ((delay_max != delay_base)?(rand() % (delay_max - delay_base)):0) + delay_base; crm_notice("Delaying %s on %s for %dms (timeout=%ds, base=%dms, " "max=%dms)", cmd->action, device->id, cmd->start_delay, cmd->timeout, delay_base, delay_max); cmd->delay_id = g_timeout_add(cmd->start_delay, start_delay_helper, cmd); } } static void free_device(gpointer data) { GListPtr gIter = NULL; stonith_device_t *device = data; g_hash_table_destroy(device->params); g_hash_table_destroy(device->aliases); for (gIter = device->pending_ops; gIter != NULL; gIter = gIter->next) { async_command_t *cmd = gIter->data; crm_warn("Removal of device '%s' purged operation %s", device->id, cmd->action); cmd->done_cb(0, -ENODEV, NULL, cmd); } g_list_free(device->pending_ops); g_list_free_full(device->targets, free); mainloop_destroy_trigger(device->work); free_xml(device->agent_metadata); free(device->namespace); free(device->on_target_actions); free(device->agent); free(device->id); free(device); } void free_device_list() { if (device_list != NULL) { g_hash_table_destroy(device_list); device_list = NULL; } } void init_device_list() { if (device_list == NULL) { device_list = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, free_device); } } static GHashTable * build_port_aliases(const char *hostmap, GListPtr * targets) { char *name = NULL; int last = 0, lpc = 0, max = 0, added = 0; GHashTable *aliases = crm_strcase_table_new(); if (hostmap == NULL) { return aliases; } max = strlen(hostmap); for (; lpc <= max; lpc++) { switch (hostmap[lpc]) { /* Assignment chars */ case '=': case ':': if (lpc > last) { free(name); name = calloc(1, 1 + lpc - last); memcpy(name, hostmap + last, lpc - last); } last = lpc + 1; break; /* Delimeter chars */ /* case ',': Potentially used to specify multiple ports */ case 0: case ';': case ' ': case '\t': if (name) { char *value = NULL; value = calloc(1, 1 + lpc - last); memcpy(value, hostmap + last, lpc - last); crm_debug("Adding alias '%s'='%s'", name, value); g_hash_table_replace(aliases, name, value); if (targets) { *targets = g_list_append(*targets, strdup(value)); } value = NULL; name = NULL; added++; } else if (lpc > last) { crm_debug("Parse error at offset %d near '%s'", lpc - last, hostmap + last); } last = lpc + 1; break; } if (hostmap[lpc] == 0) { break; } } if (added == 0) { crm_info("No host mappings detected in '%s'", hostmap); } free(name); return aliases; } static void parse_host_line(const char *line, int max, GListPtr * output) { int lpc = 0; int last = 0; if (max <= 0) { return; } /* Check for any complaints about additional parameters that the device doesn't understand */ if (strstr(line, "invalid") || strstr(line, "variable")) { crm_debug("Skipping: %s", line); return; } crm_trace("Processing %d bytes: [%s]", max, line); /* Skip initial whitespace */ for (lpc = 0; lpc <= max && isspace(line[lpc]); lpc++) { last = lpc + 1; } /* Now the actual content */ for (lpc = 0; lpc <= max; lpc++) { gboolean a_space = isspace(line[lpc]); if (a_space && lpc < max && isspace(line[lpc + 1])) { /* fast-forward to the end of the spaces */ } else if (a_space || line[lpc] == ',' || line[lpc] == ';' || line[lpc] == 0) { int rc = 1; char *entry = NULL; if (lpc != last) { entry = calloc(1, 1 + lpc - last); rc = sscanf(line + last, "%[a-zA-Z0-9_-.]", entry); } if (entry == NULL) { /* Skip */ } else if (rc != 1) { crm_warn("Could not parse (%d %d): %s", last, lpc, line + last); } else if (safe_str_neq(entry, "on") && safe_str_neq(entry, "off")) { crm_trace("Adding '%s'", entry); *output = g_list_append(*output, entry); entry = NULL; } free(entry); last = lpc + 1; } } } static GListPtr parse_host_list(const char *hosts) { int lpc = 0; int max = 0; int last = 0; GListPtr output = NULL; if (hosts == NULL) { return output; } max = strlen(hosts); for (lpc = 0; lpc <= max; lpc++) { if (hosts[lpc] == '\n' || hosts[lpc] == 0) { int len = lpc - last; if(len > 1) { char *line = strndup(hosts + last, len); line[len] = 0; /* Because it might be '\n' */ parse_host_line(line, len, &output); free(line); } last = lpc + 1; } } crm_trace("Parsed %d entries from '%s'", g_list_length(output), hosts); return output; } GHashTable *metadata_cache = NULL; void free_metadata_cache() { if (metadata_cache != NULL) { g_hash_table_destroy(metadata_cache); metadata_cache = NULL; } } static void init_metadata_cache() { if (metadata_cache == NULL) { metadata_cache = crm_str_table_new(); } } static xmlNode * get_agent_metadata(const char *agent) { xmlNode *xml = NULL; char *buffer = NULL; init_metadata_cache(); buffer = g_hash_table_lookup(metadata_cache, agent); if(safe_str_eq(agent, STONITH_WATCHDOG_AGENT)) { return NULL; } else if(buffer == NULL) { stonith_t *st = stonith_api_new(); int rc = st->cmds->metadata(st, st_opt_sync_call, agent, NULL, &buffer, 10); stonith_api_delete(st); if (rc || !buffer) { crm_err("Could not retrieve metadata for fencing agent %s", agent); return NULL; } g_hash_table_replace(metadata_cache, strdup(agent), buffer); } xml = string2xml(buffer); return xml; } static gboolean is_nodeid_required(xmlNode * xml) { xmlXPathObjectPtr xpath = NULL; if (stand_alone) { return FALSE; } if (!xml) { return FALSE; } xpath = xpath_search(xml, "//parameter[@name='nodeid']"); if (numXpathResults(xpath) <= 0) { freeXpathObject(xpath); return FALSE; } freeXpathObject(xpath); return TRUE; } #define MAX_ACTION_LEN 256 static char * add_action(char *actions, const char *action) { int offset = 0; if (actions == NULL) { actions = calloc(1, MAX_ACTION_LEN); } else { offset = strlen(actions); } if (offset > 0) { offset += snprintf(actions+offset, MAX_ACTION_LEN - offset, " "); } offset += snprintf(actions+offset, MAX_ACTION_LEN - offset, "%s", action); return actions; } static void read_action_metadata(stonith_device_t *device) { xmlXPathObjectPtr xpath = NULL; int max = 0; int lpc = 0; if (device->agent_metadata == NULL) { return; } xpath = xpath_search(device->agent_metadata, "//action"); max = numXpathResults(xpath); if (max <= 0) { freeXpathObject(xpath); return; } for (lpc = 0; lpc < max; lpc++) { const char *on_target = NULL; const char *action = NULL; xmlNode *match = getXpathResult(xpath, lpc); CRM_LOG_ASSERT(match != NULL); if(match == NULL) { continue; }; on_target = crm_element_value(match, "on_target"); action = crm_element_value(match, "name"); if(safe_str_eq(action, "list")) { set_bit(device->flags, st_device_supports_list); } else if(safe_str_eq(action, "status")) { set_bit(device->flags, st_device_supports_status); } else if(safe_str_eq(action, "reboot")) { set_bit(device->flags, st_device_supports_reboot); } else if (safe_str_eq(action, "on")) { /* "automatic" means the cluster will unfence node when it joins */ const char *automatic = crm_element_value(match, "automatic"); /* "required" is a deprecated synonym for "automatic" */ const char *required = crm_element_value(match, "required"); if (crm_is_true(automatic) || crm_is_true(required)) { device->automatic_unfencing = TRUE; } } if (action && crm_is_true(on_target)) { device->on_target_actions = add_action(device->on_target_actions, action); } } freeXpathObject(xpath); } /*! * \internal * \brief Set a pcmk_*_action parameter if not already set * * \param[in,out] params Device parameters * \param[in] action Name of action * \param[in] value Value to use if action is not already set */ static void map_action(GHashTable *params, const char *action, const char *value) { char *key = crm_strdup_printf("pcmk_%s_action", action); if (g_hash_table_lookup(params, key)) { crm_warn("Ignoring %s='%s', see %s instead", STONITH_ATTR_ACTION_OP, value, key); free(key); } else { crm_warn("Mapping %s='%s' to %s='%s'", STONITH_ATTR_ACTION_OP, value, key, value); g_hash_table_insert(params, key, strdup(value)); } } /*! * \internal * \brief Create device parameter table from XML * * \param[in] name Device name (used for logging only) * \param[in,out] params Device parameters */ static GHashTable * xml2device_params(const char *name, xmlNode *dev) { GHashTable *params = xml2list(dev); const char *value; /* Action should never be specified in the device configuration, * but we support it for users who are familiar with other software * that worked that way. */ value = g_hash_table_lookup(params, STONITH_ATTR_ACTION_OP); if (value != NULL) { crm_warn("%s has '%s' parameter, which should never be specified in configuration", name, STONITH_ATTR_ACTION_OP); if (*value == '\0') { crm_warn("Ignoring empty '%s' parameter", STONITH_ATTR_ACTION_OP); } else if (strcmp(value, "reboot") == 0) { crm_warn("Ignoring %s='reboot' (see stonith-action cluster property instead)", STONITH_ATTR_ACTION_OP); } else if (strcmp(value, "off") == 0) { map_action(params, "reboot", value); } else { map_action(params, "off", value); map_action(params, "reboot", value); } g_hash_table_remove(params, STONITH_ATTR_ACTION_OP); } return params; } static stonith_device_t * build_device_from_xml(xmlNode * msg) { const char *value; xmlNode *dev = get_xpath_object("//" F_STONITH_DEVICE, msg, LOG_ERR); stonith_device_t *device = NULL; char *agent = crm_element_value_copy(dev, "agent"); CRM_CHECK(agent != NULL, return device); device = calloc(1, sizeof(stonith_device_t)); CRM_CHECK(device != NULL, {free(agent); return device;}); device->id = crm_element_value_copy(dev, XML_ATTR_ID); device->agent = agent; device->namespace = crm_element_value_copy(dev, "namespace"); device->params = xml2device_params(device->id, dev); value = g_hash_table_lookup(device->params, STONITH_ATTR_HOSTLIST); if (value) { device->targets = parse_host_list(value); } value = g_hash_table_lookup(device->params, STONITH_ATTR_HOSTMAP); device->aliases = build_port_aliases(value, &(device->targets)); device->agent_metadata = get_agent_metadata(device->agent); read_action_metadata(device); value = g_hash_table_lookup(device->params, "nodeid"); if (!value) { device->include_nodeid = is_nodeid_required(device->agent_metadata); } value = crm_element_value(dev, "rsc_provides"); if (safe_str_eq(value, "unfencing")) { device->automatic_unfencing = TRUE; } if (is_action_required("on", device)) { crm_info("The fencing device '%s' requires unfencing", device->id); } if (device->on_target_actions) { crm_info("The fencing device '%s' requires actions (%s) to be executed on the target node", device->id, device->on_target_actions); } device->work = mainloop_add_trigger(G_PRIORITY_HIGH, stonith_device_dispatch, device); /* TODO: Hook up priority */ return device; } static const char * target_list_type(stonith_device_t * dev) { const char *check_type = NULL; check_type = g_hash_table_lookup(dev->params, STONITH_ATTR_HOSTCHECK); if (check_type == NULL) { if (g_hash_table_lookup(dev->params, STONITH_ATTR_HOSTLIST)) { check_type = "static-list"; } else if (g_hash_table_lookup(dev->params, STONITH_ATTR_HOSTMAP)) { check_type = "static-list"; } else if(is_set(dev->flags, st_device_supports_list)){ check_type = "dynamic-list"; } else if(is_set(dev->flags, st_device_supports_status)){ check_type = "status"; } else { check_type = "none"; } } return check_type; } void schedule_internal_command(const char *origin, stonith_device_t * device, const char *action, const char *victim, int timeout, void *internal_user_data, void (*done_cb) (GPid pid, int rc, const char *output, gpointer user_data)) { async_command_t *cmd = NULL; cmd = calloc(1, sizeof(async_command_t)); cmd->id = -1; cmd->default_timeout = timeout ? timeout : 60; cmd->timeout = cmd->default_timeout; cmd->action = strdup(action); cmd->victim = victim ? strdup(victim) : NULL; cmd->device = strdup(device->id); cmd->origin = strdup(origin); cmd->client = strdup(crm_system_name); cmd->client_name = strdup(crm_system_name); cmd->internal_user_data = internal_user_data; cmd->done_cb = done_cb; /* cmd, not internal_user_data, is passed to 'done_cb' as the userdata */ schedule_stonith_command(cmd, device); } gboolean string_in_list(GListPtr list, const char *item) { int lpc = 0; int max = g_list_length(list); for (lpc = 0; lpc < max; lpc++) { const char *value = g_list_nth_data(list, lpc); if (safe_str_eq(item, value)) { return TRUE; } else { crm_trace("%d: '%s' != '%s'", lpc, item, value); } } return FALSE; } static void status_search_cb(GPid pid, int rc, const char *output, gpointer user_data) { async_command_t *cmd = user_data; struct device_search_s *search = cmd->internal_user_data; stonith_device_t *dev = cmd->device ? g_hash_table_lookup(device_list, cmd->device) : NULL; gboolean can = FALSE; free_async_command(cmd); if (!dev) { search_devices_record_result(search, NULL, FALSE); return; } mainloop_set_trigger(dev->work); if (rc == 1 /* unknown */ ) { crm_trace("Host %s is not known by %s", search->host, dev->id); } else if (rc == 0 /* active */ || rc == 2 /* inactive */ ) { crm_trace("Host %s is known by %s", search->host, dev->id); can = TRUE; } else { crm_notice("Unknown result when testing if %s can fence %s: rc=%d", dev->id, search->host, rc); } search_devices_record_result(search, dev->id, can); } static void dynamic_list_search_cb(GPid pid, int rc, const char *output, gpointer user_data) { async_command_t *cmd = user_data; struct device_search_s *search = cmd->internal_user_data; stonith_device_t *dev = cmd->device ? g_hash_table_lookup(device_list, cmd->device) : NULL; gboolean can_fence = FALSE; free_async_command(cmd); /* Host/alias must be in the list output to be eligible to be fenced * * Will cause problems if down'd nodes aren't listed or (for virtual nodes) * if the guest is still listed despite being moved to another machine */ if (!dev) { search_devices_record_result(search, NULL, FALSE); return; } mainloop_set_trigger(dev->work); /* If we successfully got the targets earlier, don't disable. */ if (rc != 0 && !dev->targets) { crm_notice("Disabling port list queries for %s (%d): %s", dev->id, rc, output); /* Fall back to status */ g_hash_table_replace(dev->params, strdup(STONITH_ATTR_HOSTCHECK), strdup("status")); g_list_free_full(dev->targets, free); dev->targets = NULL; } else if (!rc) { crm_info("Refreshing port list for %s", dev->id); g_list_free_full(dev->targets, free); dev->targets = parse_host_list(output); dev->targets_age = time(NULL); } if (dev->targets) { const char *alias = g_hash_table_lookup(dev->aliases, search->host); if (!alias) { alias = search->host; } if (string_in_list(dev->targets, alias)) { can_fence = TRUE; } } search_devices_record_result(search, dev->id, can_fence); } /*! * \internal * \brief Returns true if any key in first is not in second or second has a different value for key */ static int device_params_diff(GHashTable *first, GHashTable *second) { char *key = NULL; char *value = NULL; GHashTableIter gIter; g_hash_table_iter_init(&gIter, first); while (g_hash_table_iter_next(&gIter, (void **)&key, (void **)&value)) { if(strstr(key, "CRM_meta") == key) { continue; } else if(strcmp(key, "crm_feature_set") == 0) { continue; } else { char *other_value = g_hash_table_lookup(second, key); if (!other_value || safe_str_neq(other_value, value)) { crm_trace("Different value for %s: %s != %s", key, other_value, value); return 1; } } } return 0; } /*! * \internal * \brief Checks to see if an identical device already exists in the device_list */ static stonith_device_t * device_has_duplicate(stonith_device_t * device) { stonith_device_t *dup = g_hash_table_lookup(device_list, device->id); if (!dup) { crm_trace("No match for %s", device->id); return NULL; } else if (safe_str_neq(dup->agent, device->agent)) { crm_trace("Different agent: %s != %s", dup->agent, device->agent); return NULL; } /* Use calculate_operation_digest() here? */ if (device_params_diff(device->params, dup->params) || device_params_diff(dup->params, device->params)) { return NULL; } crm_trace("Match"); return dup; } int stonith_device_register(xmlNode * msg, const char **desc, gboolean from_cib) { stonith_device_t *dup = NULL; stonith_device_t *device = build_device_from_xml(msg); CRM_CHECK(device != NULL, return -ENOMEM); dup = device_has_duplicate(device); if (dup) { crm_debug("Device '%s' already existed in device list (%d active devices)", device->id, g_hash_table_size(device_list)); free_device(device); device = dup; } else { stonith_device_t *old = g_hash_table_lookup(device_list, device->id); if (from_cib && old && old->api_registered) { /* If the cib is writing over an entry that is shared with a stonith client, * copy any pending ops that currently exist on the old entry to the new one. * Otherwise the pending ops will be reported as failures */ crm_info("Overwriting an existing entry for %s from the cib", device->id); device->pending_ops = old->pending_ops; device->api_registered = TRUE; old->pending_ops = NULL; if (device->pending_ops) { mainloop_set_trigger(device->work); } } g_hash_table_replace(device_list, device->id, device); crm_notice("Added '%s' to the device list (%d active devices)", device->id, g_hash_table_size(device_list)); } if (desc) { *desc = device->id; } if (from_cib) { device->cib_registered = TRUE; } else { device->api_registered = TRUE; } return pcmk_ok; } int stonith_device_remove(const char *id, gboolean from_cib) { stonith_device_t *device = g_hash_table_lookup(device_list, id); if (!device) { crm_info("Device '%s' not found (%d active devices)", id, g_hash_table_size(device_list)); return pcmk_ok; } if (from_cib) { device->cib_registered = FALSE; } else { device->verified = FALSE; device->api_registered = FALSE; } if (!device->cib_registered && !device->api_registered) { g_hash_table_remove(device_list, id); crm_info("Removed '%s' from the device list (%d active devices)", id, g_hash_table_size(device_list)); } else { crm_trace("Not removing '%s' from the device list (%d active devices) " "- still %s%s_registered", id, g_hash_table_size(device_list), device->cib_registered?"cib":"", device->api_registered?"api":""); } return pcmk_ok; } /*! * \internal * \brief Return the number of stonith levels registered for a node * * \param[in] tp Node's topology table entry * * \return Number of non-NULL levels in topology entry * \note This function is used only for log messages. */ static int count_active_levels(stonith_topology_t * tp) { int lpc = 0; int count = 0; for (lpc = 0; lpc < ST_LEVEL_MAX; lpc++) { if (tp->levels[lpc] != NULL) { count++; } } return count; } static void free_topology_entry(gpointer data) { stonith_topology_t *tp = data; int lpc = 0; for (lpc = 0; lpc < ST_LEVEL_MAX; lpc++) { if (tp->levels[lpc] != NULL) { g_list_free_full(tp->levels[lpc], free); } } free(tp->target); free(tp->target_value); free(tp->target_pattern); free(tp->target_attribute); free(tp); } void free_topology_list() { if (topology != NULL) { g_hash_table_destroy(topology); topology = NULL; } } void init_topology_list() { if (topology == NULL) { topology = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, free_topology_entry); } } char *stonith_level_key(xmlNode *level, int mode) { if(mode == -1) { mode = stonith_level_kind(level); } switch(mode) { case 0: return crm_element_value_copy(level, XML_ATTR_STONITH_TARGET); case 1: return crm_element_value_copy(level, XML_ATTR_STONITH_TARGET_PATTERN); case 2: { const char *name = crm_element_value(level, XML_ATTR_STONITH_TARGET_ATTRIBUTE); const char *value = crm_element_value(level, XML_ATTR_STONITH_TARGET_VALUE); if(name && value) { return crm_strdup_printf("%s=%s", name, value); } } default: return crm_strdup_printf("Unknown-%d-%s", mode, ID(level)); } } int stonith_level_kind(xmlNode * level) { int mode = 0; const char *target = crm_element_value(level, XML_ATTR_STONITH_TARGET); if(target == NULL) { mode++; target = crm_element_value(level, XML_ATTR_STONITH_TARGET_PATTERN); } if(stand_alone == FALSE && target == NULL) { mode++; if(crm_element_value(level, XML_ATTR_STONITH_TARGET_ATTRIBUTE) == NULL) { mode++; } else if(crm_element_value(level, XML_ATTR_STONITH_TARGET_VALUE) == NULL) { mode++; } } return mode; } static stonith_key_value_t * parse_device_list(const char *devices) { int lpc = 0; int max = 0; int last = 0; stonith_key_value_t *output = NULL; if (devices == NULL) { return output; } max = strlen(devices); for (lpc = 0; lpc <= max; lpc++) { if (devices[lpc] == ',' || devices[lpc] == 0) { char *line = strndup(devices + last, lpc - last); output = stonith_key_value_add(output, NULL, line); free(line); last = lpc + 1; } } return output; } /*! * \internal * \brief Register a STONITH level for a target * * Given an XML request specifying the target name, level index, and device IDs * for the level, this will create an entry for the target in the global topology * table if one does not already exist, then append the specified device IDs to * the entry's device list for the specified level. * * \param[in] msg XML request for STONITH level registration * \param[out] desc If not NULL, will be set to string representation ("TARGET[LEVEL]") * * \return pcmk_ok on success, -EINVAL if XML does not specify valid level index */ int stonith_level_register(xmlNode *msg, char **desc) { int id = 0; xmlNode *level; int mode; char *target; stonith_topology_t *tp; stonith_key_value_t *dIter = NULL; stonith_key_value_t *devices = NULL; /* Allow the XML here to point to the level tag directly, or wrapped in * another tag. If directly, don't search by xpath, because it might give * multiple hits (e.g. if the XML is the CIB). */ if (safe_str_eq(TYPE(msg), XML_TAG_FENCING_LEVEL)) { level = msg; } else { level = get_xpath_object("//" XML_TAG_FENCING_LEVEL, msg, LOG_ERR); } CRM_CHECK(level != NULL, return -EINVAL); mode = stonith_level_kind(level); target = stonith_level_key(level, mode); crm_element_value_int(level, XML_ATTR_STONITH_INDEX, &id); if (desc) { *desc = crm_strdup_printf("%s[%d]", target, id); } /* Sanity-check arguments */ if (mode >= 3 || (id <= 0) || (id >= ST_LEVEL_MAX)) { crm_trace("Could not add %s[%d] (%d) to the topology (%d active entries)", target, id, mode, g_hash_table_size(topology)); free(target); crm_log_xml_err(level, "Bad topology"); return -EINVAL; } /* Find or create topology table entry */ tp = g_hash_table_lookup(topology, target); if (tp == NULL) { tp = calloc(1, sizeof(stonith_topology_t)); tp->kind = mode; tp->target = target; tp->target_value = crm_element_value_copy(level, XML_ATTR_STONITH_TARGET_VALUE); tp->target_pattern = crm_element_value_copy(level, XML_ATTR_STONITH_TARGET_PATTERN); tp->target_attribute = crm_element_value_copy(level, XML_ATTR_STONITH_TARGET_ATTRIBUTE); g_hash_table_replace(topology, tp->target, tp); crm_trace("Added %s (%d) to the topology (%d active entries)", target, mode, g_hash_table_size(topology)); } else { free(target); } if (tp->levels[id] != NULL) { crm_info("Adding to the existing %s[%d] topology entry", tp->target, id); } devices = parse_device_list(crm_element_value(level, XML_ATTR_STONITH_DEVICES)); for (dIter = devices; dIter; dIter = dIter->next) { const char *device = dIter->value; crm_trace("Adding device '%s' for %s[%d]", device, tp->target, id); tp->levels[id] = g_list_append(tp->levels[id], strdup(device)); } stonith_key_value_freeall(devices, 1, 1); crm_info("Target %s has %d active fencing levels", tp->target, count_active_levels(tp)); return pcmk_ok; } int stonith_level_remove(xmlNode *msg, char **desc) { int id = 0; stonith_topology_t *tp; char *target; /* Unlike additions, removal requests should always have one level tag */ xmlNode *level = get_xpath_object("//" XML_TAG_FENCING_LEVEL, msg, LOG_ERR); CRM_CHECK(level != NULL, return -EINVAL); target = stonith_level_key(level, -1); crm_element_value_int(level, XML_ATTR_STONITH_INDEX, &id); if (desc) { *desc = crm_strdup_printf("%s[%d]", target, id); } /* Sanity-check arguments */ if (id >= ST_LEVEL_MAX) { free(target); return -EINVAL; } tp = g_hash_table_lookup(topology, target); if (tp == NULL) { crm_info("Topology for %s not found (%d active entries)", target, g_hash_table_size(topology)); } else if (id == 0 && g_hash_table_remove(topology, target)) { crm_info("Removed all %s related entries from the topology (%d active entries)", target, g_hash_table_size(topology)); } else if (id > 0 && tp->levels[id] != NULL) { g_list_free_full(tp->levels[id], free); tp->levels[id] = NULL; crm_info("Removed level '%d' from topology for %s (%d active levels remaining)", id, target, count_active_levels(tp)); } free(target); return pcmk_ok; } static int stonith_device_action(xmlNode * msg, char **output) { int rc = pcmk_ok; xmlNode *dev = get_xpath_object("//" F_STONITH_DEVICE, msg, LOG_ERR); const char *id = crm_element_value(dev, F_STONITH_DEVICE); async_command_t *cmd = NULL; stonith_device_t *device = NULL; if (id) { crm_trace("Looking for '%s'", id); device = g_hash_table_lookup(device_list, id); } if (device && device->api_registered == FALSE) { rc = -ENODEV; } else if (device) { cmd = create_async_command(msg); if (cmd == NULL) { return -EPROTO; } schedule_stonith_command(cmd, device); rc = -EINPROGRESS; } else { crm_info("Device %s not found", id ? id : ""); rc = -ENODEV; } return rc; } static void search_devices_record_result(struct device_search_s *search, const char *device, gboolean can_fence) { search->replies_received++; if (can_fence && device) { search->capable = g_list_append(search->capable, strdup(device)); } if (search->replies_needed == search->replies_received) { crm_debug("Finished Search. %d devices can perform action (%s) on node %s", g_list_length(search->capable), search->action ? search->action : "", search->host ? search->host : ""); search->callback(search->capable, search->user_data); free(search->host); free(search->action); free(search); } } /*! * \internal * \brief Check whether the local host is allowed to execute a fencing action * * \param[in] device Fence device to check * \param[in] action Fence action to check * \param[in] target Hostname of fence target * \param[in] allow_suicide Whether self-fencing is allowed for this operation * * \return TRUE if local host is allowed to execute action, FALSE otherwise */ static gboolean localhost_is_eligible(const stonith_device_t *device, const char *action, const char *target, gboolean allow_suicide) { gboolean localhost_is_target = safe_str_eq(target, stonith_our_uname); if (device && action && device->on_target_actions && strstr(device->on_target_actions, action)) { if (!localhost_is_target) { crm_trace("%s operation with %s can only be executed for localhost not %s", action, device->id, target); return FALSE; } } else if (localhost_is_target && !allow_suicide) { crm_trace("%s operation does not support self-fencing", action); return FALSE; } return TRUE; } static void can_fence_host_with_device(stonith_device_t * dev, struct device_search_s *search) { gboolean can = FALSE; const char *check_type = NULL; const char *host = search->host; const char *alias = NULL; CRM_LOG_ASSERT(dev != NULL); if (dev == NULL) { goto search_report_results; } else if (host == NULL) { can = TRUE; goto search_report_results; } /* Short-circuit query if this host is not allowed to perform the action */ if (safe_str_eq(search->action, "reboot")) { /* A "reboot" *might* get remapped to "off" then "on", so short-circuit * only if all three are disallowed. If only one or two are disallowed, * we'll report that with the results. We never allow suicide for * remapped "on" operations because the host is off at that point. */ if (!localhost_is_eligible(dev, "reboot", host, search->allow_suicide) && !localhost_is_eligible(dev, "off", host, search->allow_suicide) && !localhost_is_eligible(dev, "on", host, FALSE)) { goto search_report_results; } } else if (!localhost_is_eligible(dev, search->action, host, search->allow_suicide)) { goto search_report_results; } alias = g_hash_table_lookup(dev->aliases, host); if (alias == NULL) { alias = host; } check_type = target_list_type(dev); if (safe_str_eq(check_type, "none")) { can = TRUE; } else if (safe_str_eq(check_type, "static-list")) { /* Presence in the hostmap is sufficient * Only use if all hosts on which the device can be active can always fence all listed hosts */ if (string_in_list(dev->targets, host)) { can = TRUE; } else if (g_hash_table_lookup(dev->params, STONITH_ATTR_HOSTMAP) && g_hash_table_lookup(dev->aliases, host)) { can = TRUE; } } else if (safe_str_eq(check_type, "dynamic-list")) { time_t now = time(NULL); if (dev->targets == NULL || dev->targets_age + 60 < now) { crm_trace("Running '%s' to check whether %s is eligible to fence %s (%s)", check_type, dev->id, search->host, search->action); schedule_internal_command(__FUNCTION__, dev, "list", NULL, search->per_device_timeout, search, dynamic_list_search_cb); /* we'll respond to this search request async in the cb */ return; } if (string_in_list(dev->targets, alias)) { can = TRUE; } } else if (safe_str_eq(check_type, "status")) { crm_trace("Running '%s' to check whether %s is eligible to fence %s (%s)", check_type, dev->id, search->host, search->action); schedule_internal_command(__FUNCTION__, dev, "status", search->host, search->per_device_timeout, search, status_search_cb); /* we'll respond to this search request async in the cb */ return; } else { crm_err("Invalid value for " STONITH_ATTR_HOSTCHECK ": %s", check_type); check_type = "Invalid " STONITH_ATTR_HOSTCHECK; } if (safe_str_eq(host, alias)) { crm_notice("%s is%s eligible to fence (%s) %s: %s", dev->id, (can? "" : " not"), search->action, host, check_type); } else { crm_notice("%s is%s eligible to fence (%s) %s (aka. '%s'): %s", dev->id, (can? "" : " not"), search->action, host, alias, check_type); } search_report_results: search_devices_record_result(search, dev ? dev->id : NULL, can); } static void search_devices(gpointer key, gpointer value, gpointer user_data) { stonith_device_t *dev = value; struct device_search_s *search = user_data; can_fence_host_with_device(dev, search); } #define DEFAULT_QUERY_TIMEOUT 20 static void get_capable_devices(const char *host, const char *action, int timeout, bool suicide, void *user_data, void (*callback) (GList * devices, void *user_data)) { struct device_search_s *search; int per_device_timeout = DEFAULT_QUERY_TIMEOUT; int devices_needing_async_query = 0; char *key = NULL; const char *check_type = NULL; GHashTableIter gIter; stonith_device_t *device = NULL; if (!g_hash_table_size(device_list)) { callback(NULL, user_data); return; } search = calloc(1, sizeof(struct device_search_s)); if (!search) { callback(NULL, user_data); return; } g_hash_table_iter_init(&gIter, device_list); while (g_hash_table_iter_next(&gIter, (void **)&key, (void **)&device)) { check_type = target_list_type(device); if (safe_str_eq(check_type, "status") || safe_str_eq(check_type, "dynamic-list")) { devices_needing_async_query++; } } /* If we have devices that require an async event in order to know what * nodes they can fence, we have to give the events a timeout. The total * query timeout is divided among those events. */ if (devices_needing_async_query) { per_device_timeout = timeout / devices_needing_async_query; if (!per_device_timeout) { crm_err("STONITH timeout %ds is too low; using %ds, but consider raising to at least %ds", timeout, DEFAULT_QUERY_TIMEOUT, DEFAULT_QUERY_TIMEOUT * devices_needing_async_query); per_device_timeout = DEFAULT_QUERY_TIMEOUT; } else if (per_device_timeout < DEFAULT_QUERY_TIMEOUT) { crm_notice("STONITH timeout %ds is low for the current configuration;" " consider raising to at least %ds", timeout, DEFAULT_QUERY_TIMEOUT * devices_needing_async_query); } } search->host = host ? strdup(host) : NULL; search->action = action ? strdup(action) : NULL; search->per_device_timeout = per_device_timeout; /* We are guaranteed this many replies. Even if a device gets * unregistered some how during the async search, we will get * the correct number of replies. */ search->replies_needed = g_hash_table_size(device_list); search->allow_suicide = suicide; search->callback = callback; search->user_data = user_data; /* kick off the search */ crm_debug("Searching through %d devices to see what is capable of action (%s) for target %s", search->replies_needed, search->action ? search->action : "", search->host ? search->host : ""); g_hash_table_foreach(device_list, search_devices, search); } struct st_query_data { xmlNode *reply; char *remote_peer; char *client_id; char *target; char *action; int call_options; }; /*! * \internal * \brief Add action-specific attributes to query reply XML * * \param[in,out] xml XML to add attributes to * \param[in] action Fence action * \param[in] device Fence device */ static void add_action_specific_attributes(xmlNode *xml, const char *action, stonith_device_t *device) { int action_specific_timeout; int delay_max; int delay_base; CRM_CHECK(xml && action && device, return); if (is_action_required(action, device)) { crm_trace("Action %s is required on %s", action, device->id); crm_xml_add_int(xml, F_STONITH_DEVICE_REQUIRED, 1); } action_specific_timeout = get_action_timeout(device, action, 0); if (action_specific_timeout) { crm_trace("Action %s has timeout %dms on %s", action, action_specific_timeout, device->id); crm_xml_add_int(xml, F_STONITH_ACTION_TIMEOUT, action_specific_timeout); } delay_max = get_action_delay_max(device, action); if (delay_max > 0) { crm_trace("Action %s has maximum random delay %dms on %s", action, delay_max, device->id); crm_xml_add_int(xml, F_STONITH_DELAY_MAX, delay_max / 1000); } delay_base = get_action_delay_base(device, action); if (delay_base > 0) { crm_xml_add_int(xml, F_STONITH_DELAY_BASE, delay_base / 1000); } if ((delay_max > 0) && (delay_base == 0)) { crm_trace("Action %s has maximum random delay %dms on %s", action, delay_max, device->id); } else if ((delay_max == 0) && (delay_base > 0)) { crm_trace("Action %s has a static delay of %dms on %s", action, delay_base, device->id); } else if ((delay_max > 0) && (delay_base > 0)) { crm_trace("Action %s has a minimum delay of %dms and a randomly chosen " "maximum delay of %dms on %s", action, delay_base, delay_max, device->id); } } /*! * \internal * \brief Add "disallowed" attribute to query reply XML if appropriate * * \param[in,out] xml XML to add attribute to * \param[in] action Fence action * \param[in] device Fence device * \param[in] target Fence target * \param[in] allow_suicide Whether self-fencing is allowed */ static void add_disallowed(xmlNode *xml, const char *action, stonith_device_t *device, const char *target, gboolean allow_suicide) { if (!localhost_is_eligible(device, action, target, allow_suicide)) { crm_trace("Action %s on %s is disallowed for local host", action, device->id); crm_xml_add(xml, F_STONITH_ACTION_DISALLOWED, XML_BOOLEAN_TRUE); } } /*! * \internal * \brief Add child element with action-specific values to query reply XML * * \param[in,out] xml XML to add attribute to * \param[in] action Fence action * \param[in] device Fence device * \param[in] target Fence target * \param[in] allow_suicide Whether self-fencing is allowed */ static void add_action_reply(xmlNode *xml, const char *action, stonith_device_t *device, const char *target, gboolean allow_suicide) { xmlNode *child = create_xml_node(xml, F_STONITH_ACTION); crm_xml_add(child, XML_ATTR_ID, action); add_action_specific_attributes(child, action, device); add_disallowed(child, action, device, target, allow_suicide); } static void stonith_query_capable_device_cb(GList * devices, void *user_data) { struct st_query_data *query = user_data; int available_devices = 0; xmlNode *dev = NULL; xmlNode *list = NULL; GListPtr lpc = NULL; /* Pack the results into XML */ list = create_xml_node(NULL, __FUNCTION__); crm_xml_add(list, F_STONITH_TARGET, query->target); for (lpc = devices; lpc != NULL; lpc = lpc->next) { stonith_device_t *device = g_hash_table_lookup(device_list, lpc->data); const char *action = query->action; if (!device) { /* It is possible the device got unregistered while * determining who can fence the target */ continue; } available_devices++; dev = create_xml_node(list, F_STONITH_DEVICE); crm_xml_add(dev, XML_ATTR_ID, device->id); crm_xml_add(dev, "namespace", device->namespace); crm_xml_add(dev, "agent", device->agent); crm_xml_add_int(dev, F_STONITH_DEVICE_VERIFIED, device->verified); /* If the originating fencer wants to reboot the node, and we have a * capable device that doesn't support "reboot", remap to "off" instead. */ if (is_not_set(device->flags, st_device_supports_reboot) && safe_str_eq(query->action, "reboot")) { crm_trace("%s doesn't support reboot, using values for off instead", device->id); action = "off"; } /* Add action-specific values if available */ add_action_specific_attributes(dev, action, device); if (safe_str_eq(query->action, "reboot")) { /* A "reboot" *might* get remapped to "off" then "on", so after * sending the "reboot"-specific values in the main element, we add * sub-elements for "off" and "on" values. * * We short-circuited earlier if "reboot", "off" and "on" are all * disallowed for the local host. However if only one or two are * disallowed, we send back the results and mark which ones are * disallowed. If "reboot" is disallowed, this might cause problems * with older fencer versions, which won't check for it. Older * versions will ignore "off" and "on", so they are not a problem. */ add_disallowed(dev, action, device, query->target, is_set(query->call_options, st_opt_allow_suicide)); add_action_reply(dev, "off", device, query->target, is_set(query->call_options, st_opt_allow_suicide)); add_action_reply(dev, "on", device, query->target, FALSE); } /* A query without a target wants device parameters */ if (query->target == NULL) { xmlNode *attrs = create_xml_node(dev, XML_TAG_ATTRS); g_hash_table_foreach(device->params, hash2field, attrs); } } crm_xml_add_int(list, F_STONITH_AVAILABLE_DEVICES, available_devices); if (query->target) { crm_debug("Found %d matching devices for '%s'", available_devices, query->target); } else { crm_debug("%d devices installed", available_devices); } if (list != NULL) { crm_log_xml_trace(list, "Add query results"); add_message_xml(query->reply, F_STONITH_CALLDATA, list); } stonith_send_reply(query->reply, query->call_options, query->remote_peer, query->client_id); free_xml(query->reply); free(query->remote_peer); free(query->client_id); free(query->target); free(query->action); free(query); free_xml(list); g_list_free_full(devices, free); } static void stonith_query(xmlNode * msg, const char *remote_peer, const char *client_id, int call_options) { struct st_query_data *query = NULL; const char *action = NULL; const char *target = NULL; int timeout = 0; xmlNode *dev = get_xpath_object("//@" F_STONITH_ACTION, msg, LOG_TRACE); crm_element_value_int(msg, F_STONITH_TIMEOUT, &timeout); if (dev) { const char *device = crm_element_value(dev, F_STONITH_DEVICE); target = crm_element_value(dev, F_STONITH_TARGET); action = crm_element_value(dev, F_STONITH_ACTION); if (device && safe_str_eq(device, "manual_ack")) { /* No query or reply necessary */ return; } } crm_log_xml_debug(msg, "Query"); query = calloc(1, sizeof(struct st_query_data)); query->reply = stonith_construct_reply(msg, NULL, NULL, pcmk_ok); query->remote_peer = remote_peer ? strdup(remote_peer) : NULL; query->client_id = client_id ? strdup(client_id) : NULL; query->target = target ? strdup(target) : NULL; query->action = action ? strdup(action) : NULL; query->call_options = call_options; get_capable_devices(target, action, timeout, is_set(call_options, st_opt_allow_suicide), query, stonith_query_capable_device_cb); } #define ST_LOG_OUTPUT_MAX 512 static void log_operation(async_command_t * cmd, int rc, int pid, const char *next, const char *output) { if (rc == 0) { next = NULL; } if (cmd->victim != NULL) { do_crm_log(rc == 0 ? LOG_NOTICE : LOG_ERR, "Operation '%s' [%d] (call %d from %s) for host '%s' with device '%s' returned: %d (%s)%s%s", cmd->action, pid, cmd->id, cmd->client_name, cmd->victim, cmd->device, rc, pcmk_strerror(rc), (next? ", retrying with " : ""), (next ? next : "")); } else { do_crm_log_unlikely(rc == 0 ? LOG_DEBUG : LOG_NOTICE, "Operation '%s' [%d] for device '%s' returned: %d (%s)%s%s", cmd->action, pid, cmd->device, rc, pcmk_strerror(rc), (next? ", retrying with " : ""), (next ? next : "")); } if (output) { /* Logging the whole string confuses syslog when the string is xml */ char *prefix = crm_strdup_printf("%s:%d", cmd->device, pid); crm_log_output(rc == 0 ? LOG_DEBUG : LOG_WARNING, prefix, output); free(prefix); } } static void stonith_send_async_reply(async_command_t * cmd, const char *output, int rc, GPid pid) { xmlNode *reply = NULL; gboolean bcast = FALSE; reply = stonith_construct_async_reply(cmd, output, NULL, rc); if (safe_str_eq(cmd->action, "metadata")) { /* Too verbose to log */ crm_trace("Metadata query for %s", cmd->device); output = NULL; } else if (crm_str_eq(cmd->action, "monitor", TRUE) || crm_str_eq(cmd->action, "list", TRUE) || crm_str_eq(cmd->action, "status", TRUE)) { crm_trace("Never broadcast %s replies", cmd->action); } else if (!stand_alone && safe_str_eq(cmd->origin, cmd->victim) && safe_str_neq(cmd->action, "on")) { crm_trace("Broadcast %s reply for %s", cmd->action, cmd->victim); crm_xml_add(reply, F_SUBTYPE, "broadcast"); bcast = TRUE; } log_operation(cmd, rc, pid, NULL, output); crm_log_xml_trace(reply, "Reply"); if (bcast) { crm_xml_add(reply, F_STONITH_OPERATION, T_STONITH_NOTIFY); send_cluster_message(NULL, crm_msg_stonith_ng, reply, FALSE); } else if (cmd->origin) { crm_trace("Directed reply to %s", cmd->origin); send_cluster_message(crm_get_peer(0, cmd->origin), crm_msg_stonith_ng, reply, FALSE); } else { crm_trace("Directed local %ssync reply to %s", (cmd->options & st_opt_sync_call) ? "" : "a-", cmd->client_name); do_local_reply(reply, cmd->client, cmd->options & st_opt_sync_call, FALSE); } if (stand_alone) { /* Do notification with a clean data object */ xmlNode *notify_data = create_xml_node(NULL, T_STONITH_NOTIFY_FENCE); crm_xml_add_int(notify_data, F_STONITH_RC, rc); crm_xml_add(notify_data, F_STONITH_TARGET, cmd->victim); crm_xml_add(notify_data, F_STONITH_OPERATION, cmd->op); crm_xml_add(notify_data, F_STONITH_DELEGATE, "localhost"); crm_xml_add(notify_data, F_STONITH_DEVICE, cmd->device); crm_xml_add(notify_data, F_STONITH_REMOTE_OP_ID, cmd->remote_op_id); crm_xml_add(notify_data, F_STONITH_ORIGIN, cmd->client); do_stonith_notify(0, T_STONITH_NOTIFY_FENCE, rc, notify_data); do_stonith_notify(0, T_STONITH_NOTIFY_HISTORY, 0, NULL); } free_xml(reply); } static void cancel_stonith_command(async_command_t * cmd) { stonith_device_t *device; CRM_CHECK(cmd != NULL, return); if (!cmd->device) { return; } device = g_hash_table_lookup(device_list, cmd->device); if (device) { crm_trace("Cancel scheduled %s on %s", cmd->action, device->id); device->pending_ops = g_list_remove(device->pending_ops, cmd); } } static void st_child_done(GPid pid, int rc, const char *output, gpointer user_data) { stonith_device_t *device = NULL; stonith_device_t *next_device = NULL; async_command_t *cmd = user_data; GListPtr gIter = NULL; GListPtr gIterNext = NULL; CRM_CHECK(cmd != NULL, return); cmd->active_on = NULL; /* The device is ready to do something else now */ device = g_hash_table_lookup(device_list, cmd->device); if (device) { if (rc == pcmk_ok && (safe_str_eq(cmd->action, "list") || safe_str_eq(cmd->action, "monitor") || safe_str_eq(cmd->action, "status"))) { device->verified = TRUE; } mainloop_set_trigger(device->work); } crm_debug("Operation '%s' on '%s' completed with rc=%d (%d remaining)", cmd->action, cmd->device, rc, g_list_length(cmd->device_next)); if (rc == 0) { GListPtr iter; /* see if there are any required devices left to execute for this op */ for (iter = cmd->device_next; iter != NULL; iter = iter->next) { next_device = g_hash_table_lookup(device_list, iter->data); if (next_device != NULL && is_action_required(cmd->action, next_device)) { cmd->device_next = iter->next; break; } next_device = NULL; } } else if (rc != 0 && cmd->device_next && (is_action_required(cmd->action, device) == FALSE)) { /* if this device didn't work out, see if there are any others we can try. * if the failed device was 'required', we can't pick another device. */ next_device = g_hash_table_lookup(device_list, cmd->device_next->data); cmd->device_next = cmd->device_next->next; } /* this operation requires more fencing, hooray! */ if (next_device) { log_operation(cmd, rc, pid, next_device->id, output); schedule_stonith_command(cmd, next_device); /* Prevent cmd from being freed */ cmd = NULL; goto done; } stonith_send_async_reply(cmd, output, rc, pid); if (rc != 0) { goto done; } /* Check to see if any operations are scheduled to do the exact * same thing that just completed. If so, rather than * performing the same fencing operation twice, return the result * of this operation for all pending commands it matches. */ for (gIter = cmd_list; gIter != NULL; gIter = gIterNext) { async_command_t *cmd_other = gIter->data; gIterNext = gIter->next; if (cmd == cmd_other) { continue; } /* A pending scheduled command matches the command that just finished if. * 1. The client connections are different. * 2. The node victim is the same. * 3. The fencing action is the same. * 4. The device scheduled to execute the action is the same. */ if (safe_str_eq(cmd->client, cmd_other->client) || safe_str_neq(cmd->victim, cmd_other->victim) || safe_str_neq(cmd->action, cmd_other->action) || safe_str_neq(cmd->device, cmd_other->device)) { continue; } /* Duplicate merging will do the right thing for either type of remapped * reboot. If the executing fencer remapped an unsupported reboot to * off, then cmd->action will be reboot and will be merged with any * other reboot requests. If the originating fencer remapped a * topology reboot to off then on, we will get here once with * cmd->action "off" and once with "on", and they will be merged * separately with similar requests. */ crm_notice ("Merging stonith action %s for node %s originating from client %s with identical stonith request from client %s", cmd_other->action, cmd_other->victim, cmd_other->client_name, cmd->client_name); cmd_list = g_list_remove_link(cmd_list, gIter); stonith_send_async_reply(cmd_other, output, rc, pid); cancel_stonith_command(cmd_other); free_async_command(cmd_other); g_list_free_1(gIter); } done: free_async_command(cmd); } static gint sort_device_priority(gconstpointer a, gconstpointer b) { const stonith_device_t *dev_a = a; const stonith_device_t *dev_b = b; if (dev_a->priority > dev_b->priority) { return -1; } else if (dev_a->priority < dev_b->priority) { return 1; } return 0; } static void stonith_fence_get_devices_cb(GList * devices, void *user_data) { async_command_t *cmd = user_data; stonith_device_t *device = NULL; crm_info("Found %d matching devices for '%s'", g_list_length(devices), cmd->victim); if (g_list_length(devices) > 0) { /* Order based on priority */ devices = g_list_sort(devices, sort_device_priority); device = g_hash_table_lookup(device_list, devices->data); if (device) { cmd->device_list = devices; cmd->device_next = devices->next; devices = NULL; /* list owned by cmd now */ } } /* we have a device, schedule it for fencing. */ if (device) { schedule_stonith_command(cmd, device); /* in progress */ return; } /* no device found! */ stonith_send_async_reply(cmd, NULL, -ENODEV, 0); free_async_command(cmd); g_list_free_full(devices, free); } static int stonith_fence(xmlNode * msg) { const char *device_id = NULL; stonith_device_t *device = NULL; async_command_t *cmd = create_async_command(msg); xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, msg, LOG_ERR); if (cmd == NULL) { return -EPROTO; } device_id = crm_element_value(dev, F_STONITH_DEVICE); if (device_id) { device = g_hash_table_lookup(device_list, device_id); if (device == NULL) { crm_err("Requested device '%s' is not available", device_id); return -ENODEV; } schedule_stonith_command(cmd, device); } else { const char *host = crm_element_value(dev, F_STONITH_TARGET); if (cmd->options & st_opt_cs_nodeid) { int nodeid = crm_atoi(host, NULL); crm_node_t *node = crm_find_known_peer_full(nodeid, NULL, CRM_GET_PEER_ANY); if (node) { host = node->uname; } } /* If we get to here, then self-fencing is implicitly allowed */ get_capable_devices(host, cmd->action, cmd->default_timeout, TRUE, cmd, stonith_fence_get_devices_cb); } return -EINPROGRESS; } xmlNode * stonith_construct_reply(xmlNode * request, const char *output, xmlNode * data, int rc) { int lpc = 0; xmlNode *reply = NULL; const char *name = NULL; const char *value = NULL; const char *names[] = { F_STONITH_OPERATION, F_STONITH_CALLID, F_STONITH_CLIENTID, F_STONITH_CLIENTNAME, F_STONITH_REMOTE_OP_ID, F_STONITH_CALLOPTS }; crm_trace("Creating a basic reply"); reply = create_xml_node(NULL, T_STONITH_REPLY); crm_xml_add(reply, "st_origin", __FUNCTION__); crm_xml_add(reply, F_TYPE, T_STONITH_NG); crm_xml_add(reply, "st_output", output); crm_xml_add_int(reply, F_STONITH_RC, rc); CRM_CHECK(request != NULL, crm_warn("Can't create a sane reply"); return reply); for (lpc = 0; lpc < DIMOF(names); lpc++) { name = names[lpc]; value = crm_element_value(request, name); crm_xml_add(reply, name, value); } if (data != NULL) { crm_trace("Attaching reply output"); add_message_xml(reply, F_STONITH_CALLDATA, data); } return reply; } static xmlNode * stonith_construct_async_reply(async_command_t * cmd, const char *output, xmlNode * data, int rc) { xmlNode *reply = NULL; crm_trace("Creating a basic reply"); reply = create_xml_node(NULL, T_STONITH_REPLY); crm_xml_add(reply, "st_origin", __FUNCTION__); crm_xml_add(reply, F_TYPE, T_STONITH_NG); crm_xml_add(reply, F_STONITH_OPERATION, cmd->op); crm_xml_add(reply, F_STONITH_DEVICE, cmd->device); crm_xml_add(reply, F_STONITH_REMOTE_OP_ID, cmd->remote_op_id); crm_xml_add(reply, F_STONITH_CLIENTID, cmd->client); crm_xml_add(reply, F_STONITH_CLIENTNAME, cmd->client_name); crm_xml_add(reply, F_STONITH_TARGET, cmd->victim); crm_xml_add(reply, F_STONITH_ACTION, cmd->op); crm_xml_add(reply, F_STONITH_ORIGIN, cmd->origin); crm_xml_add_int(reply, F_STONITH_CALLID, cmd->id); crm_xml_add_int(reply, F_STONITH_CALLOPTS, cmd->options); crm_xml_add_int(reply, F_STONITH_RC, rc); crm_xml_add(reply, "st_output", output); if (data != NULL) { crm_info("Attaching reply output"); add_message_xml(reply, F_STONITH_CALLDATA, data); } return reply; } bool fencing_peer_active(crm_node_t *peer) { if (peer == NULL) { return FALSE; } else if (peer->uname == NULL) { return FALSE; } else if (is_set(peer->processes, crm_get_cluster_proc())) { return TRUE; } return FALSE; } /*! * \internal * \brief Determine if we need to use an alternate node to * fence the target. If so return that node's uname * * \retval NULL, no alternate host * \retval uname, uname of alternate host to use */ static const char * check_alternate_host(const char *target) { const char *alternate_host = NULL; crm_trace("Checking if we (%s) can fence %s", stonith_our_uname, target); if (find_topology_for_host(target) && safe_str_eq(target, stonith_our_uname)) { GHashTableIter gIter; crm_node_t *entry = NULL; g_hash_table_iter_init(&gIter, crm_peer_cache); while (g_hash_table_iter_next(&gIter, NULL, (void **)&entry)) { crm_trace("Checking for %s.%d != %s", entry->uname, entry->id, target); if (fencing_peer_active(entry) && safe_str_neq(entry->uname, target)) { alternate_host = entry->uname; break; } } if (alternate_host == NULL) { crm_err("No alternate host available to handle complex self fencing request"); g_hash_table_iter_init(&gIter, crm_peer_cache); while (g_hash_table_iter_next(&gIter, NULL, (void **)&entry)) { crm_notice("Peer[%d] %s", entry->id, entry->uname); } } } return alternate_host; } static void stonith_send_reply(xmlNode * reply, int call_options, const char *remote_peer, const char *client_id) { if (remote_peer) { send_cluster_message(crm_get_peer(0, remote_peer), crm_msg_stonith_ng, reply, FALSE); } else { do_local_reply(reply, client_id, is_set(call_options, st_opt_sync_call), remote_peer != NULL); } } static int handle_request(crm_client_t * client, uint32_t id, uint32_t flags, xmlNode * request, const char *remote_peer) { int call_options = 0; int rc = -EOPNOTSUPP; xmlNode *data = NULL; xmlNode *reply = NULL; char *output = NULL; const char *op = crm_element_value(request, F_STONITH_OPERATION); const char *client_id = crm_element_value(request, F_STONITH_CLIENTID); crm_element_value_int(request, F_STONITH_CALLOPTS, &call_options); if (is_set(call_options, st_opt_sync_call)) { CRM_ASSERT(client == NULL || client->request_id == id); } if (crm_str_eq(op, CRM_OP_REGISTER, TRUE)) { xmlNode *reply = create_xml_node(NULL, "reply"); CRM_ASSERT(client); crm_xml_add(reply, F_STONITH_OPERATION, CRM_OP_REGISTER); crm_xml_add(reply, F_STONITH_CLIENTID, client->id); crm_ipcs_send(client, id, reply, flags); client->request_id = 0; free_xml(reply); return 0; } else if (crm_str_eq(op, STONITH_OP_EXEC, TRUE)) { rc = stonith_device_action(request, &output); } else if (crm_str_eq(op, STONITH_OP_TIMEOUT_UPDATE, TRUE)) { const char *call_id = crm_element_value(request, F_STONITH_CALLID); const char *client_id = crm_element_value(request, F_STONITH_CLIENTID); int op_timeout = 0; crm_element_value_int(request, F_STONITH_TIMEOUT, &op_timeout); do_stonith_async_timeout_update(client_id, call_id, op_timeout); return 0; } else if (crm_str_eq(op, STONITH_OP_QUERY, TRUE)) { if (remote_peer) { create_remote_stonith_op(client_id, request, TRUE); /* Record it for the future notification */ } stonith_query(request, remote_peer, client_id, call_options); return 0; } else if (crm_str_eq(op, T_STONITH_NOTIFY, TRUE)) { const char *flag_name = NULL; CRM_ASSERT(client); flag_name = crm_element_value(request, F_STONITH_NOTIFY_ACTIVATE); if (flag_name) { crm_debug("Setting %s callbacks for %s (%s): ON", flag_name, client->name, client->id); client->options |= get_stonith_flag(flag_name); } flag_name = crm_element_value(request, F_STONITH_NOTIFY_DEACTIVATE); if (flag_name) { crm_debug("Setting %s callbacks for %s (%s): off", flag_name, client->name, client->id); client->options |= get_stonith_flag(flag_name); } if (flags & crm_ipc_client_response) { crm_ipcs_send_ack(client, id, flags, "ack", __FUNCTION__, __LINE__); } return 0; } else if (crm_str_eq(op, STONITH_OP_RELAY, TRUE)) { xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, request, LOG_TRACE); crm_notice("Peer %s has received a forwarded fencing request from %s to fence (%s) peer %s", stonith_our_uname, client ? client->name : remote_peer, crm_element_value(dev, F_STONITH_ACTION), crm_element_value(dev, F_STONITH_TARGET)); if (initiate_remote_stonith_op(NULL, request, FALSE) != NULL) { rc = -EINPROGRESS; } } else if (crm_str_eq(op, STONITH_OP_FENCE, TRUE)) { if (remote_peer || stand_alone) { rc = stonith_fence(request); } else if (call_options & st_opt_manual_ack) { remote_fencing_op_t *rop = NULL; xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, request, LOG_TRACE); const char *target = crm_element_value(dev, F_STONITH_TARGET); crm_notice("Received manual confirmation that %s is fenced", target); rop = initiate_remote_stonith_op(client, request, TRUE); rc = stonith_manual_ack(request, rop); } else { const char *alternate_host = NULL; xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, request, LOG_TRACE); const char *target = crm_element_value(dev, F_STONITH_TARGET); const char *action = crm_element_value(dev, F_STONITH_ACTION); const char *device = crm_element_value(dev, F_STONITH_DEVICE); if (client) { int tolerance = 0; crm_notice("Client %s.%.8s wants to fence (%s) '%s' with device '%s'", client->name, client->id, action, target, device ? device : "(any)"); crm_element_value_int(dev, F_STONITH_TOLERANCE, &tolerance); if (stonith_check_fence_tolerance(tolerance, target, action)) { rc = 0; goto done; } } else { crm_notice("Peer %s wants to fence (%s) '%s' with device '%s'", remote_peer, action, target, device ? device : "(any)"); } alternate_host = check_alternate_host(target); if (alternate_host && client) { const char *client_id = NULL; crm_notice("Forwarding complex self fencing request to peer %s", alternate_host); if (client->id) { client_id = client->id; } else { client_id = crm_element_value(request, F_STONITH_CLIENTID); } /* Create a record of it, otherwise call_id will be 0 if we need to notify of failures */ create_remote_stonith_op(client_id, request, FALSE); crm_xml_add(request, F_STONITH_OPERATION, STONITH_OP_RELAY); crm_xml_add(request, F_STONITH_CLIENTID, client->id); send_cluster_message(crm_get_peer(0, alternate_host), crm_msg_stonith_ng, request, FALSE); rc = -EINPROGRESS; } else if (initiate_remote_stonith_op(client, request, FALSE) != NULL) { rc = -EINPROGRESS; } } } else if (crm_str_eq(op, STONITH_OP_FENCE_HISTORY, TRUE)) { rc = stonith_fence_history(request, &data, remote_peer, call_options); if (call_options & st_opt_discard_reply) { /* we don't expect answers to the broadcast * we might have sent out */ free_xml(data); return pcmk_ok; } } else if (crm_str_eq(op, STONITH_OP_DEVICE_ADD, TRUE)) { const char *device_id = NULL; rc = stonith_device_register(request, &device_id, FALSE); do_stonith_notify_device(call_options, op, rc, device_id); } else if (crm_str_eq(op, STONITH_OP_DEVICE_DEL, TRUE)) { xmlNode *dev = get_xpath_object("//" F_STONITH_DEVICE, request, LOG_ERR); const char *device_id = crm_element_value(dev, XML_ATTR_ID); rc = stonith_device_remove(device_id, FALSE); do_stonith_notify_device(call_options, op, rc, device_id); } else if (crm_str_eq(op, STONITH_OP_LEVEL_ADD, TRUE)) { char *device_id = NULL; rc = stonith_level_register(request, &device_id); do_stonith_notify_level(call_options, op, rc, device_id); free(device_id); } else if (crm_str_eq(op, STONITH_OP_LEVEL_DEL, TRUE)) { char *device_id = NULL; rc = stonith_level_remove(request, &device_id); do_stonith_notify_level(call_options, op, rc, device_id); } else if(safe_str_eq(op, CRM_OP_RM_NODE_CACHE)) { int node_id = 0; const char *name = NULL; crm_element_value_int(request, XML_ATTR_ID, &node_id); name = crm_element_value(request, XML_ATTR_UNAME); reap_crm_member(node_id, name); return pcmk_ok; } else { crm_err("Unknown %s from %s", op, client ? client->name : remote_peer); crm_log_xml_warn(request, "UnknownOp"); } done: /* Always reply unless the request is in process still. * If in progress, a reply will happen async after the request * processing is finished */ if (rc != -EINPROGRESS) { crm_trace("Reply handling: %p %u %u %d %d %s", client, client?client->request_id:0, id, is_set(call_options, st_opt_sync_call), call_options, crm_element_value(request, F_STONITH_CALLOPTS)); if (is_set(call_options, st_opt_sync_call)) { CRM_ASSERT(client == NULL || client->request_id == id); } reply = stonith_construct_reply(request, output, data, rc); stonith_send_reply(reply, call_options, remote_peer, client_id); } free(output); free_xml(data); free_xml(reply); return rc; } static void handle_reply(crm_client_t * client, xmlNode * request, const char *remote_peer) { const char *op = crm_element_value(request, F_STONITH_OPERATION); if (crm_str_eq(op, STONITH_OP_QUERY, TRUE)) { process_remote_stonith_query(request); } else if (crm_str_eq(op, T_STONITH_NOTIFY, TRUE)) { process_remote_stonith_exec(request); } else if (crm_str_eq(op, STONITH_OP_FENCE, TRUE)) { /* Reply to a complex fencing op */ process_remote_stonith_exec(request); } else { crm_err("Unknown %s reply from %s", op, client ? client->name : remote_peer); crm_log_xml_warn(request, "UnknownOp"); } } void stonith_command(crm_client_t * client, uint32_t id, uint32_t flags, xmlNode * request, const char *remote_peer) { int call_options = 0; int rc = 0; gboolean is_reply = FALSE; /* Copy op for reporting. The original might get freed by handle_reply() * before we use it in crm_debug(): * handle_reply() * |- process_remote_stonith_exec() * |-- remote_op_done() * |--- handle_local_reply_and_notify() * |---- crm_xml_add(...F_STONITH_OPERATION...) * |--- free_xml(op->request) */ char *op = crm_element_value_copy(request, F_STONITH_OPERATION); if (get_xpath_object("//" T_STONITH_REPLY, request, LOG_TRACE)) { is_reply = TRUE; } crm_element_value_int(request, F_STONITH_CALLOPTS, &call_options); crm_debug("Processing %s%s %u from %s (%16x)", op, is_reply ? " reply" : "", id, client ? client->name : remote_peer, call_options); if (is_set(call_options, st_opt_sync_call)) { CRM_ASSERT(client == NULL || client->request_id == id); } if (is_reply) { handle_reply(client, request, remote_peer); } else { rc = handle_request(client, id, flags, request, remote_peer); } crm_debug("Processed %s%s from %s: %s (%d)", op, is_reply ? " reply" : "", client ? client->name : remote_peer, rc > 0 ? "" : pcmk_strerror(rc), rc); free(op); } diff --git a/doc/Clusters_from_Scratch/en-US/Clusters_from_Scratch.ent b/doc/Clusters_from_Scratch/en-US/Clusters_from_Scratch.ent index 4413d22505..ce8d7d1838 100644 --- a/doc/Clusters_from_Scratch/en-US/Clusters_from_Scratch.ent +++ b/doc/Clusters_from_Scratch/en-US/Clusters_from_Scratch.ent @@ -1,6 +1,6 @@ - + diff --git a/doc/Pacemaker_Administration/en-US/Pacemaker_Administration.ent b/doc/Pacemaker_Administration/en-US/Pacemaker_Administration.ent index f67f950ee5..5e1d8447fc 100644 --- a/doc/Pacemaker_Administration/en-US/Pacemaker_Administration.ent +++ b/doc/Pacemaker_Administration/en-US/Pacemaker_Administration.ent @@ -1,4 +1,4 @@ - + diff --git a/doc/Pacemaker_Administration/en-US/Revision_History.xml b/doc/Pacemaker_Administration/en-US/Revision_History.xml index dae578f0bb..1b94c3e81a 100644 --- a/doc/Pacemaker_Administration/en-US/Revision_History.xml +++ b/doc/Pacemaker_Administration/en-US/Revision_History.xml @@ -1,45 +1,57 @@ %BOOK_ENTITIES; ]> Revision History 1-0 Tue Jan 23 2018 KenGaillot kgaillot@redhat.com Move administration-oriented information from Pacemaker Explained into its own book 1-1 Mon Jan 28 2019 KenGaillot kgaillot@redhat.com JanPokorný jpokorny@redhat.com Add "Troubleshooting" chapter, minor clarifications and reformatting + + 1-2 + Mon May 13 2019 + + KenGaillot + kgaillot@redhat.com + + + Overhaul configuration how-to + + + diff --git a/doc/Pacemaker_Development/en-US/Ch-General.txt b/doc/Pacemaker_Development/en-US/Ch-General.txt index 65362c23d8..4186ec2d70 100644 --- a/doc/Pacemaker_Development/en-US/Ch-General.txt +++ b/doc/Pacemaker_Development/en-US/Ch-General.txt @@ -1,33 +1,35 @@ :compat-mode: legacy = General Guidelines for All Languages = == Copyright == indexterm:[copyright] When copyright notices are added to a file, they should look like this: + +.Copyright Notice Format ==== -Copyright the Pacemaker project contributors +Copyright +YYYY[-YYYY]+ the Pacemaker project contributors The version control history for this file may have further details. ==== -The first ++ is the year the file was originally published. The original +The first +YYYY+ is the year the file was originally published. The original date is important for two reasons: when two entities claim copyright ownership of the same work, the earlier claim generally prevails; and copyright expiration is generally calculated from the original publication date. footnote:[ See the U.S. Copyright Office's https://www.copyright.gov/comp3/["Compendium of U.S. Copyright Office Practices"], particularly "Chapter 2200: Notice of Copyright", sections 2205.1(A) and 2205.1(F), or https://techwhirl.com/updating-copyright-notices/["Updating Copyright Notices"] for a more readable summary. ] If the file is modified in later years, add +-YYYY+ with the most recent year of modification. Even though Pacemaker is an ongoing project, copyright notices are about the years of 'publication' of specific content. Copyright notices are intended to indicate, but do not affect, copyright 'ownership', which is determined by applicable laws and regulations. Authors may put more specific copyright notices in their commit messages if desired. diff --git a/doc/Pacemaker_Development/en-US/Pacemaker_Development.ent b/doc/Pacemaker_Development/en-US/Pacemaker_Development.ent index 1a9959792b..a2647c3f8b 100644 --- a/doc/Pacemaker_Development/en-US/Pacemaker_Development.ent +++ b/doc/Pacemaker_Development/en-US/Pacemaker_Development.ent @@ -1,4 +1,4 @@ - - + + diff --git a/doc/Pacemaker_Development/en-US/Revision_History.xml b/doc/Pacemaker_Development/en-US/Revision_History.xml index 6769c833ea..b9c1dea6ae 100644 --- a/doc/Pacemaker_Development/en-US/Revision_History.xml +++ b/doc/Pacemaker_Development/en-US/Revision_History.xml @@ -1,77 +1,95 @@ %BOOK_ENTITIES; ]> Revision History 1-0 Tue Jul 26 2016 KenGaillot kgaillot@redhat.com Convert coding guidelines and developer FAQ to Publican document 1-1 Mon Aug 29 2016 KenGaillot kgaillot@redhat.com Add Python coding guidelines, and more about licensing 2-0 Fri Jan 12 2018 KenGaillot kgaillot@redhat.com Drop support for Python 2.6 2-1 Tue Sep 18 2018 JanPokorný poki@redhat.com Start documenting notable evolutionary points 2-2 Fri Dec 7 2018 KenGaillot kgaillot@redhat.com Update FAQ and C guidelines + + 2-3 + Mon May 13 2019 + + KenGaillot + kgaillot@redhat.com + + + JanPokorný + poki@redhat.com + + + + Update copyright notice policy, and some external references + + + + diff --git a/doc/Pacemaker_Explained/en-US/Ch-Advanced-Resources.txt b/doc/Pacemaker_Explained/en-US/Ch-Advanced-Resources.txt index b8efd853fd..e431626f0b 100644 --- a/doc/Pacemaker_Explained/en-US/Ch-Advanced-Resources.txt +++ b/doc/Pacemaker_Explained/en-US/Ch-Advanced-Resources.txt @@ -1,1471 +1,1414 @@ :compat-mode: legacy = Advanced Resource Types = [[group-resources]] == Groups - A Syntactic Shortcut == indexterm:[Group Resources] indexterm:[Resource,Groups] One of the most common elements of a cluster is a set of resources that need to be located together, start sequentially, and stop in the reverse order. To simplify this configuration, we support the concept of groups. .A group of two primitive resources ====== [source,XML] ------- ------- ====== Although the example above contains only two resources, there is no limit to the number of resources a group can contain. The example is also sufficient to explain the fundamental properties of a group: * Resources are started in the order they appear in (+Public-IP+ first, then +Email+) * Resources are stopped in the reverse order to which they appear in (+Email+ first, then +Public-IP+) If a resource in the group can't run anywhere, then nothing after that is allowed to run, too. * If +Public-IP+ can't run anywhere, neither can +Email+; * but if +Email+ can't run anywhere, this does not affect +Public-IP+ in any way The group above is logically equivalent to writing: .How the cluster sees a group resource ====== [source,XML] ------- ------- ====== Obviously as the group grows bigger, the reduced configuration effort can become significant. Another (typical) example of a group is a DRBD volume, the filesystem mount, an IP address, and an application that uses them. === Group Properties === .Properties of a Group Resource [width="95%",cols="3m,<5",options="header",align="center"] |========================================================= |Field |Description |id |A unique name for the group indexterm:[id,Group Resource Property] indexterm:[Resource,Group Property,id] |========================================================= === Group Options === Groups inherit the +priority+, +target-role+, and +is-managed+ properties from primitive resources. See <> for information about those properties. === Group Instance Attributes === Groups have no instance attributes. However, any that are set for the group object will be inherited by the group's children. === Group Contents === Groups may only contain a collection of cluster resources (see <>). To refer to a child of a group resource, just use the child's +id+ instead of the group's. === Group Constraints === Although it is possible to reference a group's children in constraints, it is usually preferable to reference the group itself. .Some constraints involving groups ====== [source,XML] ------- ------- ====== === Group Stickiness === indexterm:[resource-stickiness,Groups] Stickiness, the measure of how much a resource wants to stay where it is, is additive in groups. Every active resource of the group will contribute its stickiness value to the group's total. So if the default +resource-stickiness+ is 100, and a group has seven members, five of which are active, then the group as a whole will prefer its current location with a score of 500. [[s-resource-clone]] == Clones - Resources That Can Have Multiple Active Instances == indexterm:[Clone Resources] indexterm:[Resource,Clones] 'Clone' resources are resources that can have more than one copy active at the same time. This allows you, for example, to run a copy of a daemon on every node. You can clone any primitive or group resource. footnote:[ Of course, the service must support running multiple instances. ] === Anonymous versus Unique Clones === A clone resource is configured to be either 'anonymous' or 'globally unique'. Anonymous clones are the simplest. These behave completely identically everywhere they are running. Because of this, there can be only one instance of an anonymous clone active per node. The instances of globally unique clones are distinct entities. All instances are launched identically, but one instance of the clone is not identical to any other instance, whether running on the same node or a different node. As an example, a cloned IP address can use special kernel functionality such that each instance handles a subset of requests for the same IP address. [[s-resource-promotable]] === Promotable clones === indexterm:[Promotable Clone Resources] indexterm:[Resource,Promotable] If a clone is 'promotable', its instances can perform a special role that Pacemaker will manage via the +promote+ and +demote+ actions of the resource agent. Services that support such a special role have various terms for the special role and the default role: primary and secondary, master and replica, controller and worker, etc. Pacemaker uses the terms 'master' and 'slave', footnote:[ These are historical terms that will eventually be replaced, but the extensive use of them and the need for backward compatibility makes it a long process. You may see examples using a +master+ tag instead of a +clone+ tag with the +promotable+ meta-attribute set to +true+; the +master+ tag is supported, but deprecated, and will be removed in a future version. You may also see such services referred to as 'multi-state' or 'stateful'; these means the same thing as 'promotable'. ] but is agnostic to what the service calls them or what they do. All that Pacemaker cares about is that an instance comes up in the default role when started, and the resource agent supports the +promote+ and +demote+ actions to manage entering and exiting the special role. === Clone Properties === .Properties of a Clone Resource [width="95%",cols="3m,<5",options="header",align="center"] |========================================================= |Field |Description |id |A unique name for the clone indexterm:[id,Clone Property] indexterm:[Clone,Property,id] |========================================================= === Clone Options === <> inherited from primitive resources: +priority, target-role, is-managed+ .Clone-specific configuration options [width="95%",cols="1m,1,<3",options="header",align="center"] |========================================================= |Field |Default |Description |globally-unique |false |If +true+, each clone instance performs a distinct function indexterm:[globally-unique,Clone Option] indexterm:[Clone,Option,globally-unique] |clone-max |number of nodes in cluster |The maximum number of clone instances that can be started across the entire cluster indexterm:[clone-max,Clone Option] indexterm:[Clone,Option,clone-max] |clone-node-max |1 |If +globally-unique+ is +true+, the maximum number of clone instances that can be started on a single node indexterm:[clone-node-max,Clone Option] indexterm:[Clone,Option,clone-node-max] |clone-min |0 |Require at least this number of clone instances to be runnable before allowing resources depending on the clone to be runnable. A value of 0 means require all clone instances to be runnable. indexterm:[clone-min,Clone Option] indexterm:[Clone,Option,clone-min] |notify |false |Call the resource agent's +notify+ action for all active instances, before and after starting or stopping any clone instance. The resource agent must support this action. Allowed values: +false+, +true+ indexterm:[notify,Clone Option] indexterm:[Clone,Option,notify] |ordered |false |If +true+, clone instances must be started sequentially instead of in parallel Allowed values: +false+, +true+ indexterm:[ordered,Clone Option] indexterm:[Clone,Option,ordered] |interleave |false |When this clone is ordered relative to another clone, if this option is +false+ (the default), the ordering is relative to 'all' instances of the other clone, whereas if this option is +true+, the ordering is relative only to instances on the same node. Allowed values: +false+, +true+ indexterm:[interleave,Clone Option] indexterm:[Clone,Option,interleave] |promotable |false |If +true+, clone instances can perform a special role that Pacemaker will manage via the resource agent's +promote+ and +demote+ actions. The resource agent must support these actions. Allowed values: +false+, +true+ indexterm:[promotable,Clone Option] indexterm:[Clone,Option,promotable] |promoted-max |1 |If +promotable+ is +true+, the number of instances that can be promoted at one time across the entire cluster indexterm:[promoted-max,Clone Option] indexterm:[Clone,Option,promoted-max] |promoted-node-max |1 |If +promotable+ is +true+ and +globally-unique+ is +false+, the number of clone instances can be promoted at one time on a single node indexterm:[promoted-node-max,Clone Option] indexterm:[Clone,Option,promoted-node-max] |========================================================= For backward compatibility, +master-max+ and +master-node-max+ are accepted as aliases for +promoted-max+ and +promoted-node-max+, but are deprecated since 2.0.0, and support for them will be removed in a future version. === Clone Contents === Clones must contain exactly one primitive or group resource. .A clone that runs a web server on all nodes ==== [source,XML] ---- ---- ==== [WARNING] You should never reference the name of a clone's child (the primitive or group resource being cloned). If you think you need to do this, you probably need to re-evaluate your design. === Clone Instance Attributes === Clones have no instance attributes; however, any that are set here will be inherited by the clone's child. === Clone Constraints === In most cases, a clone will have a single instance on each active cluster node. If this is not the case, you can indicate which nodes the cluster should preferentially assign copies to with resource location constraints. These constraints are written no differently from those for primitive resources except that the clone's +id+ is used. .Some constraints involving clones ====== [source,XML] ------- ------- ====== Ordering constraints behave slightly differently for clones. In the example above, +apache-stats+ will wait until all copies of +apache-clone+ that need to be started have done so before being started itself. Only if _no_ copies can be started will +apache-stats+ be prevented from being active. Additionally, the clone will wait for +apache-stats+ to be stopped before stopping itself. Colocation of a primitive or group resource with a clone means that the resource can run on any node with an active instance of the clone. The cluster will choose an instance based on where the clone is running and the resource's own location preferences. Colocation between clones is also possible. If one clone +A+ is colocated with another clone +B+, the set of allowed locations for +A+ is limited to nodes on which +B+ is (or will be) active. Placement is then performed normally. ==== Promotable Clone Constraints ==== For promotable clone resources, the +first-action+ and/or +then-action+ fields for ordering constraints may be set to +promote+ or +demote+ to constrain the master role, and colocation constraints may contain +rsc-role+ and/or +with-rsc-role+ fields. .Additional colocation constraint options for promotable clone resources [width="95%",cols="1m,1,<3",options="header",align="center"] |========================================================= |Field |Default |Description |rsc-role |Started |An additional attribute of colocation constraints that specifies the role that +rsc+ must be in. Allowed values: +Started+, +Master+, +Slave+. indexterm:[rsc-role,Ordering Constraints] indexterm:[Constraints,Ordering,rsc-role] |with-rsc-role |Started |An additional attribute of colocation constraints that specifies the role that +with-rsc+ must be in. Allowed values: +Started+, +Master+, +Slave+. indexterm:[with-rsc-role,Ordering Constraints] indexterm:[Constraints,Ordering,with-rsc-role] |========================================================= .Constraints involving promotable clone resources ====== [source,XML] ------- ------- ====== In the example above, +myApp+ will wait until one of the database copies has been started and promoted to master before being started itself on the same node. Only if no copies can be promoted will +myApp+ be prevented from being active. Additionally, the cluster will wait for +myApp+ to be stopped before demoting the database. Colocation of a primitive or group resource with a promotable clone resource means that it can run on any node with an active instance of the promotable clone resource that has the specified role (+master+ or +slave+). In the example above, the cluster will choose a location based on where database is running as a +master+, and if there are multiple +master+ instances it will also factor in +myApp+'s own location preferences when deciding which location to choose. Colocation with regular clones and other promotable clone resources is also possible. In such cases, the set of allowed locations for the +rsc+ clone is (after role filtering) limited to nodes on which the +with-rsc+ promotable clone resource is (or will be) in the specified role. Placement is then performed as normal. ==== Using Promotable Clone Resources in Colocation Sets ==== .Additional colocation set options relevant to promotable clone resources [width="95%",cols="1m,1,<6",options="header",align="center"] |========================================================= |Field |Default |Description |role |Started |The role that 'all members' of the set must be in. Allowed values: +Started+, +Master+, +Slave+. indexterm:[role,Ordering Constraints] indexterm:[Constraints,Ordering,role] |========================================================= In the following example +B+'s master must be located on the same node as +A+'s master. Additionally resources +C+ and +D+ must be located on the same node as +A+'s and +B+'s masters. .Colocate C and D with A's and B's master instances ====== [source,XML] ------- ------- ====== ==== Using Promotable Clone Resources in Ordered Sets ==== .Additional ordered set options relevant to promotable clone resources [width="95%",cols="1m,1,<3",options="header",align="center"] |========================================================= |Field |Default |Description |action |value of +first-action+ |An additional attribute of ordering constraint sets that specifies the action that applies to 'all members' of the set. Allowed values: +start+, +stop+, +promote+, +demote+. indexterm:[action,Ordering Constraints] indexterm:[Constraints,Ordering,action] |========================================================= .Start C and D after first promoting A and B ====== [source,XML] ------- ------- ====== In the above example, +B+ cannot be promoted to a master role until +A+ has been promoted. Additionally, resources +C+ and +D+ must wait until +A+ and +B+ have been promoted before they can start. [[s-clone-stickiness]] === Clone Stickiness === indexterm:[resource-stickiness,Clones] To achieve a stable allocation pattern, clones are slightly sticky by default. If no value for +resource-stickiness+ is provided, the clone will use a value of 1. Being a small value, it causes minimal disturbance to the score calculations of other resources but is enough to prevent Pacemaker from needlessly moving copies around the cluster. [NOTE] ==== For globally unique clones, this may result in multiple instances of the clone staying on a single node, even after another eligible node becomes active (for example, after being put into standby mode then made active again). If you do not want this behavior, specify a +resource-stickiness+ of 0 for the clone temporarily and let the cluster adjust, then set it back to 1 if you want the default behavior to apply again. ==== [IMPORTANT] ==== If +resource-stickiness+ is set in the +rsc_defaults+ section, it will apply to clone instances as well. This means an explicit +resource-stickiness+ of 0 in +rsc_defaults+ works differently from the implicit default used when +resource-stickiness+ is not specified. ==== === Clone Resource Agent Requirements === Any resource can be used as an anonymous clone, as it requires no additional support from the resource agent. Whether it makes sense to do so depends on your resource and its resource agent. ==== Resource Agent Requirements for Globally Unique Clones ==== Globally unique clones require additional support in the resource agent. In particular, it must only respond with +$\{OCF_SUCCESS}+ if the node has that exact instance active. All other probes for instances of the clone should result in +$\{OCF_NOT_RUNNING}+ (or one of the other OCF error codes if they are failed). Individual instances of a clone are identified by appending a colon and a numerical offset, e.g. +apache:2+. Resource agents can find out how many copies there are by examining the +OCF_RESKEY_CRM_meta_clone_max+ environment variable and which instance it is by examining +OCF_RESKEY_CRM_meta_clone+. The resource agent must not make any assumptions (based on +OCF_RESKEY_CRM_meta_clone+) about which numerical instances are active. In particular, the list of active copies will not always be an unbroken sequence, nor always start at 0. ==== Resource Agent Requirements for Promotable Clones ==== Promotable clone resources require two extra actions, +demote+ and +promote+, which are responsible for changing the state of the resource. Like +start+ and +stop+, they should return +$\{OCF_SUCCESS}+ if they completed successfully or a relevant error code if they did not. The states can mean whatever you wish, but when the resource is started, it must come up in the mode called +slave+. From there the cluster will decide which instances to promote to +master+. In addition to the clone requirements for monitor actions, agents must also _accurately_ report which state they are in. The cluster relies on the agent to report its status (including role) accurately and does not indicate to the agent what role it currently believes it to be in. .Role implications of OCF return codes [width="95%",cols="1,<1",options="header",align="center"] |========================================================= |Monitor Return Code |Description |OCF_NOT_RUNNING |Stopped indexterm:[Return Code,OCF_NOT_RUNNING] |OCF_SUCCESS |Running (Slave) indexterm:[Return Code,OCF_SUCCESS] |OCF_RUNNING_MASTER |Running (Master) indexterm:[Return Code,OCF_RUNNING_MASTER] |OCF_FAILED_MASTER |Failed (Master) indexterm:[Return Code,OCF_FAILED_MASTER] |Other |Failed (Slave) |========================================================= ==== Clone Notifications ==== If the clone has the +notify+ meta-attribute set to +true+, and the resource agent supports the +notify+ action, Pacemaker will call the action when appropriate, passing a number of extra variables which, when combined with additional context, can be used to calculate the current state of the cluster and what is about to happen to it. .Environment variables supplied with Clone notify actions [width="95%",cols="5,<3",options="header",align="center"] |========================================================= |Variable |Description |OCF_RESKEY_CRM_meta_notify_type |Allowed values: +pre+, +post+ indexterm:[Environment Variable,OCF_RESKEY_CRM_meta_notify_,type] indexterm:[type,Notification Environment Variable] |OCF_RESKEY_CRM_meta_notify_operation |Allowed values: +start+, +stop+ indexterm:[Environment Variable,OCF_RESKEY_CRM_meta_notify_,operation] indexterm:[operation,Notification Environment Variable] |OCF_RESKEY_CRM_meta_notify_start_resource |Resources to be started indexterm:[Environment Variable,OCF_RESKEY_CRM_meta_notify_,start_resource] indexterm:[start_resource,Notification Environment Variable] |OCF_RESKEY_CRM_meta_notify_stop_resource |Resources to be stopped indexterm:[Environment Variable,OCF_RESKEY_CRM_meta_notify_,stop_resource] indexterm:[stop_resource,Notification Environment Variable] |OCF_RESKEY_CRM_meta_notify_active_resource |Resources that are running indexterm:[Environment Variable,OCF_RESKEY_CRM_meta_notify_,active_resource] indexterm:[active_resource,Notification Environment Variable] |OCF_RESKEY_CRM_meta_notify_inactive_resource |Resources that are not running indexterm:[Environment Variable,OCF_RESKEY_CRM_meta_notify_,inactive_resource] indexterm:[inactive_resource,Notification Environment Variable] |OCF_RESKEY_CRM_meta_notify_start_uname |Nodes on which resources will be started indexterm:[Environment Variable,OCF_RESKEY_CRM_meta_notify_,start_uname] indexterm:[start_uname,Notification Environment Variable] |OCF_RESKEY_CRM_meta_notify_stop_uname |Nodes on which resources will be stopped indexterm:[Environment Variable,OCF_RESKEY_CRM_meta_notify_,stop_uname] indexterm:[stop_uname,Notification Environment Variable] |OCF_RESKEY_CRM_meta_notify_active_uname |Nodes on which resources are running indexterm:[Environment Variable,OCF_RESKEY_CRM_meta_notify_,active_uname] indexterm:[active_uname,Notification Environment Variable] |========================================================= The variables come in pairs, such as +OCF_RESKEY_CRM_meta_notify_start_resource+ and +OCF_RESKEY_CRM_meta_notify_start_uname+, and should be treated as an array of whitespace-separated elements. +OCF_RESKEY_CRM_meta_notify_inactive_resource+ is an exception, as the matching +uname+ variable does not exist since inactive resources are not running on any node. Thus, in order to indicate that +clone:0+ will be started on +sles-1+, +clone:2+ will be started on +sles-3+, and +clone:3+ will be started on +sles-2+, the cluster would set: .Notification variables ====== [source,Bash] ------- OCF_RESKEY_CRM_meta_notify_start_resource="clone:0 clone:2 clone:3" OCF_RESKEY_CRM_meta_notify_start_uname="sles-1 sles-3 sles-2" ------- ====== [NOTE] ==== Pacemaker will log but otherwise ignore failures of notify actions. ==== ==== Interpretation of Notification Variables ==== .Pre-notification (stop): * Active resources: +$OCF_RESKEY_CRM_meta_notify_active_resource+ * Inactive resources: +$OCF_RESKEY_CRM_meta_notify_inactive_resource+ * Resources to be started: +$OCF_RESKEY_CRM_meta_notify_start_resource+ * Resources to be stopped: +$OCF_RESKEY_CRM_meta_notify_stop_resource+ .Post-notification (stop) / Pre-notification (start): * Active resources ** +$OCF_RESKEY_CRM_meta_notify_active_resource+ ** minus +$OCF_RESKEY_CRM_meta_notify_stop_resource+ * Inactive resources ** +$OCF_RESKEY_CRM_meta_notify_inactive_resource+ ** plus +$OCF_RESKEY_CRM_meta_notify_stop_resource+ * Resources that were started: +$OCF_RESKEY_CRM_meta_notify_start_resource+ * Resources that were stopped: +$OCF_RESKEY_CRM_meta_notify_stop_resource+ .Post-notification (start): * Active resources: ** +$OCF_RESKEY_CRM_meta_notify_active_resource+ ** minus +$OCF_RESKEY_CRM_meta_notify_stop_resource+ ** plus +$OCF_RESKEY_CRM_meta_notify_start_resource+ * Inactive resources: ** +$OCF_RESKEY_CRM_meta_notify_inactive_resource+ ** plus +$OCF_RESKEY_CRM_meta_notify_stop_resource+ ** minus +$OCF_RESKEY_CRM_meta_notify_start_resource+ * Resources that were started: +$OCF_RESKEY_CRM_meta_notify_start_resource+ * Resources that were stopped: +$OCF_RESKEY_CRM_meta_notify_stop_resource+ ==== Extra Notifications for Promotable Clones ==== .Extra environment variables supplied for promotable clones [width="95%",cols="5,<3",options="header",align="center"] |========================================================= |Variable |Description |OCF_RESKEY_CRM_meta_notify_master_resource |Resources that are running in +Master+ mode indexterm:[Environment Variable,OCF_RESKEY_CRM_meta_notify_,master_resource] indexterm:[master_resource,Notification Environment Variable] |OCF_RESKEY_CRM_meta_notify_slave_resource |Resources that are running in +Slave+ mode indexterm:[Environment Variable,OCF_RESKEY_CRM_meta_notify_,slave_resource] indexterm:[slave_resource,Notification Environment Variable] |OCF_RESKEY_CRM_meta_notify_promote_resource |Resources to be promoted indexterm:[Environment Variable,OCF_RESKEY_CRM_meta_notify_,promote_resource] indexterm:[promote_resource,Notification Environment Variable] |OCF_RESKEY_CRM_meta_notify_demote_resource |Resources to be demoted indexterm:[Environment Variable,OCF_RESKEY_CRM_meta_notify_,demote_resource] indexterm:[demote_resource,Notification Environment Variable] |OCF_RESKEY_CRM_meta_notify_promote_uname |Nodes on which resources will be promoted indexterm:[Environment Variable,OCF_RESKEY_CRM_meta_notify_,promote_uname] indexterm:[promote_uname,Notification Environment Variable] |OCF_RESKEY_CRM_meta_notify_demote_uname |Nodes on which resources will be demoted indexterm:[Environment Variable,OCF_RESKEY_CRM_meta_notify_,demote_uname] indexterm:[demote_uname,Notification Environment Variable] |OCF_RESKEY_CRM_meta_notify_master_uname |Nodes on which resources are running in +Master+ mode indexterm:[Environment Variable,OCF_RESKEY_CRM_meta_notify_,master_uname] indexterm:[master_uname,Notification Environment Variable] |OCF_RESKEY_CRM_meta_notify_slave_uname |Nodes on which resources are running in +Slave+ mode indexterm:[Environment Variable,OCF_RESKEY_CRM_meta_notify_,slave_uname] indexterm:[slave_uname,Notification Environment Variable] |========================================================= ==== Interpretation of Promotable Notification Variables ==== .Pre-notification (demote): * +Active+ resources: +$OCF_RESKEY_CRM_meta_notify_active_resource+ * +Master+ resources: +$OCF_RESKEY_CRM_meta_notify_master_resource+ * +Slave+ resources: +$OCF_RESKEY_CRM_meta_notify_slave_resource+ * Inactive resources: +$OCF_RESKEY_CRM_meta_notify_inactive_resource+ * Resources to be started: +$OCF_RESKEY_CRM_meta_notify_start_resource+ * Resources to be promoted: +$OCF_RESKEY_CRM_meta_notify_promote_resource+ * Resources to be demoted: +$OCF_RESKEY_CRM_meta_notify_demote_resource+ * Resources to be stopped: +$OCF_RESKEY_CRM_meta_notify_stop_resource+ .Post-notification (demote) / Pre-notification (stop): * +Active+ resources: +$OCF_RESKEY_CRM_meta_notify_active_resource+ * +Master+ resources: ** +$OCF_RESKEY_CRM_meta_notify_master_resource+ ** minus +$OCF_RESKEY_CRM_meta_notify_demote_resource+ * +Slave+ resources: +$OCF_RESKEY_CRM_meta_notify_slave_resource+ * Inactive resources: +$OCF_RESKEY_CRM_meta_notify_inactive_resource+ * Resources to be started: +$OCF_RESKEY_CRM_meta_notify_start_resource+ * Resources to be promoted: +$OCF_RESKEY_CRM_meta_notify_promote_resource+ * Resources to be demoted: +$OCF_RESKEY_CRM_meta_notify_demote_resource+ * Resources to be stopped: +$OCF_RESKEY_CRM_meta_notify_stop_resource+ * Resources that were demoted: +$OCF_RESKEY_CRM_meta_notify_demote_resource+ .Post-notification (stop) / Pre-notification (start) * +Active+ resources: ** +$OCF_RESKEY_CRM_meta_notify_active_resource+ ** minus +$OCF_RESKEY_CRM_meta_notify_stop_resource+ * +Master+ resources: ** +$OCF_RESKEY_CRM_meta_notify_master_resource+ ** minus +$OCF_RESKEY_CRM_meta_notify_demote_resource+ * +Slave+ resources: ** +$OCF_RESKEY_CRM_meta_notify_slave_resource+ ** minus +$OCF_RESKEY_CRM_meta_notify_stop_resource+ * Inactive resources: ** +$OCF_RESKEY_CRM_meta_notify_inactive_resource+ ** plus +$OCF_RESKEY_CRM_meta_notify_stop_resource+ * Resources to be started: +$OCF_RESKEY_CRM_meta_notify_start_resource+ * Resources to be promoted: +$OCF_RESKEY_CRM_meta_notify_promote_resource+ * Resources to be demoted: +$OCF_RESKEY_CRM_meta_notify_demote_resource+ * Resources to be stopped: +$OCF_RESKEY_CRM_meta_notify_stop_resource+ * Resources that were demoted: +$OCF_RESKEY_CRM_meta_notify_demote_resource+ * Resources that were stopped: +$OCF_RESKEY_CRM_meta_notify_stop_resource+ .Post-notification (start) / Pre-notification (promote) * +Active+ resources: ** +$OCF_RESKEY_CRM_meta_notify_active_resource+ ** minus +$OCF_RESKEY_CRM_meta_notify_stop_resource+ ** plus +$OCF_RESKEY_CRM_meta_notify_start_resource+ * +Master+ resources: ** +$OCF_RESKEY_CRM_meta_notify_master_resource+ ** minus +$OCF_RESKEY_CRM_meta_notify_demote_resource+ * +Slave+ resources: ** +$OCF_RESKEY_CRM_meta_notify_slave_resource+ ** minus +$OCF_RESKEY_CRM_meta_notify_stop_resource+ ** plus +$OCF_RESKEY_CRM_meta_notify_start_resource+ * Inactive resources: ** +$OCF_RESKEY_CRM_meta_notify_inactive_resource+ ** plus +$OCF_RESKEY_CRM_meta_notify_stop_resource+ ** minus +$OCF_RESKEY_CRM_meta_notify_start_resource+ * Resources to be started: +$OCF_RESKEY_CRM_meta_notify_start_resource+ * Resources to be promoted: +$OCF_RESKEY_CRM_meta_notify_promote_resource+ * Resources to be demoted: +$OCF_RESKEY_CRM_meta_notify_demote_resource+ * Resources to be stopped: +$OCF_RESKEY_CRM_meta_notify_stop_resource+ * Resources that were started: +$OCF_RESKEY_CRM_meta_notify_start_resource+ * Resources that were demoted: +$OCF_RESKEY_CRM_meta_notify_demote_resource+ * Resources that were stopped: +$OCF_RESKEY_CRM_meta_notify_stop_resource+ .Post-notification (promote) * +Active+ resources: ** +$OCF_RESKEY_CRM_meta_notify_active_resource+ ** minus +$OCF_RESKEY_CRM_meta_notify_stop_resource+ ** plus +$OCF_RESKEY_CRM_meta_notify_start_resource+ * +Master+ resources: ** +$OCF_RESKEY_CRM_meta_notify_master_resource+ ** minus +$OCF_RESKEY_CRM_meta_notify_demote_resource+ ** plus +$OCF_RESKEY_CRM_meta_notify_promote_resource+ * +Slave+ resources: ** +$OCF_RESKEY_CRM_meta_notify_slave_resource+ ** minus +$OCF_RESKEY_CRM_meta_notify_stop_resource+ ** plus +$OCF_RESKEY_CRM_meta_notify_start_resource+ ** minus +$OCF_RESKEY_CRM_meta_notify_promote_resource+ * Inactive resources: ** +$OCF_RESKEY_CRM_meta_notify_inactive_resource+ ** plus +$OCF_RESKEY_CRM_meta_notify_stop_resource+ ** minus +$OCF_RESKEY_CRM_meta_notify_start_resource+ * Resources to be started: +$OCF_RESKEY_CRM_meta_notify_start_resource+ * Resources to be promoted: +$OCF_RESKEY_CRM_meta_notify_promote_resource+ * Resources to be demoted: +$OCF_RESKEY_CRM_meta_notify_demote_resource+ * Resources to be stopped: +$OCF_RESKEY_CRM_meta_notify_stop_resource+ * Resources that were started: +$OCF_RESKEY_CRM_meta_notify_start_resource+ * Resources that were promoted: +$OCF_RESKEY_CRM_meta_notify_promote_resource+ * Resources that were demoted: +$OCF_RESKEY_CRM_meta_notify_demote_resource+ * Resources that were stopped: +$OCF_RESKEY_CRM_meta_notify_stop_resource+ === Monitoring Promotable Clone Resources === The usual monitor actions are insufficient to monitor a promotable clone resource, because Pacemaker needs to verify not only that the resource is active, but also that its actual role matches its intended one. Define two monitoring actions: the usual one will cover the slave role, and an additional one with +role="master"+ will cover the master role. .Monitoring both states of a promotable clone resource ====== [source,XML] ------- ------- ====== [IMPORTANT] =========== It is crucial that _every_ monitor operation has a different interval! Pacemaker currently differentiates between operations only by resource and interval; so if (for example) a promotable clone resource had the same monitor interval for both roles, Pacemaker would ignore the role when checking the status -- which would cause unexpected return codes, and therefore unnecessary complications. =========== [[s-promotion-scores]] === Determining Which Instance is Promoted === Pacemaker can choose a promotable clone instance to be promoted in one of two ways: * Promotion scores: These are node attributes set via the `crm_master` utility, which generally would be called by the resource agent's start action if it supports promotable clones. This tool automatically detects both the resource and host, and should be used to set a preference for being promoted. Based on this, +promoted-max+, and +promoted-node-max+, the instance(s) with the highest preference will be promoted. * Constraints: Location constraints can indicate which nodes are most preferred as masters. .Explicitly preferring node1 to be promoted to master ====== [source,XML] ------- ------- ====== [[s-resource-bundle]] == Bundles - Isolated Environments == indexterm:[bundle] indexterm:[Resource,bundle] +indexterm:[container,bundle] indexterm:[Docker,bundle] +indexterm:[podman,bundle] indexterm:[rkt,bundle] Pacemaker supports a special syntax for launching a https://en.wikipedia.org/wiki/Operating-system-level_virtualization[container] with any infrastructure it requires: the 'bundle'. -Pacemaker bundles support https://www.docker.com/[Docker] and -https://coreos.com/rkt/[rkt] container technologies. +Pacemaker bundles support https://www.docker.com/[Docker], +https://podman.io/[podman], and https://coreos.com/rkt/[rkt] +container technologies. footnote:[Docker is a trademark of Docker, Inc. No endorsement by or association with Docker, Inc. is implied.] .A bundle for a containerized web server ==== [source,XML] ---- - + ---- ==== +=== Bundle Prerequisites === +indexterm:[bundle,prerequisites] + +Before configuring a bundle in Pacemaker, the user must install the appropriate +container launch technology (Docker, podman, or rkt), and supply a fully +configured container image, on every node allowed to run the bundle. + +Pacemaker will create an implicit resource of type +ocf:heartbeat:docker+, ++ocf:heartbeat:podman+, or +ocf:heartbeat:rkt+ to manage a bundle's +container. The user must ensure that the appropriate resource agent is +installed on every node allowed to run the bundle. + === Bundle Properties === .Properties of a Bundle [width="95%",cols="3m,<5",options="header",align="center"] |========================================================= |Field |Description |id |A unique name for the bundle (required) indexterm:[id,bundle] indexterm:[bundle,Property,id] |description |Arbitrary text (not used by Pacemaker) indexterm:[description,bundle] indexterm:[bundle,Property,description] |========================================================= -A bundle must contain exactly one ++ or ++ element. +A bundle must contain exactly one ++, ++, or ++ element. -=== Docker Properties === - -Before configuring a Docker bundle in Pacemaker, the user must install Docker -and supply a fully configured Docker image on every node allowed to run the -bundle. - -Pacemaker will create an implicit +ocf:heartbeat:docker+ resource to manage -a bundle's Docker container. The user must ensure that resource agent is -installed on every node allowed to run the bundle. +=== Bundle Container Properties === +indexterm:[container,Properties] +indexterm:[Docker,Properties] +indexterm:[podman,Properties] +indexterm:[rkt,Properties] -.Properties of a Bundle's Docker Element +.Properties of a Bundle's docker, podman, or rkt Element [width="95%",cols="3m,4,<5",options="header",align="center"] -|========================================================= +|==== |Field |Default |Description |image | -|Docker image tag (required) - indexterm:[image,Docker] - indexterm:[Docker,Property,image] +|Container image tag (required) + indexterm:[image,container] + indexterm:[container,Property,image] |replicas |Value of +promoted-max+ if that is positive, else 1 |A positive integer specifying the number of container instances to launch - indexterm:[replicas,Docker] - indexterm:[Docker,Property,replicas] + indexterm:[replicas,container] + indexterm:[container,Property,replicas] |replicas-per-host |1 |A positive integer specifying the number of container instances allowed to run on a single node - indexterm:[replicas-per-host,Docker] - indexterm:[Docker,Property,replicas-per-host] + indexterm:[replicas-per-host,container] + indexterm:[container,Property,replicas-per-host] |promoted-max |0 |A non-negative integer that, if positive, indicates that the containerized service should be treated as a promotable service, with this many replicas allowed to run the service in the master role - indexterm:[promoted-max,Docker] - indexterm:[Docker,Property,promoted-max] + indexterm:[promoted-max,container] + indexterm:[container,Property,promoted-max] |network | -|If specified, this will be passed to +docker run+ as the - https://docs.docker.com/engine/reference/run/#network-settings[network setting] - for the Docker container. - indexterm:[network,Docker] - indexterm:[Docker,Property,network] +|If specified, this will be passed to the `docker run`, `podman run`, or + `rkt run` command as the network setting for the container. + indexterm:[network,container] + indexterm:[container,Property,network] |run-command |`/usr/sbin/pacemaker-remoted` if bundle contains a +primitive+, otherwise none |This command will be run inside the container when launching it ("PID 1"). If the bundle contains a +primitive+, this command 'must' start pacemaker-remoted - (but could, for example, be a script that does other stuff, too). If the - container image has a pre-2.0.0 version of Pacemaker, set this to - +/usr/sbin/pacemaker_remoted+ (note the underbar instead of dash). - indexterm:[run-command,Docker] - indexterm:[Docker,Property,run-command] + (but could, for example, be a script that does other stuff, too). + indexterm:[run-command,container] + indexterm:[container,Property,run-command] |options | -|Extra command-line options to pass to `docker run` - indexterm:[options,Docker] - indexterm:[Docker,Property,options] - -|========================================================= - -For backward compatibility, +masters+ is accepted as an alias for -+promoted-max+, but is deprecated since 2.0.0, and support for it will be -removed in a future version. - -=== rkt Properties === - -Before configuring a rkt bundle in Pacemaker, the user must install rkt -and supply a fully configured container image on every node allowed to run the -bundle. +|Extra command-line options to pass to the `docker run`, `podman run`, or + `rkt run` command + indexterm:[options,container] + indexterm:[container,Property,options] -Pacemaker will create an implicit +ocf:heartbeat:rkt+ resource to manage -a bundle's rkt container. The user must ensure that resource agent is -installed on every node allowed to run the bundle. - -.Properties of a Bundle's rkt Element -[width="95%",cols="3m,4,<5",options="header",align="center"] -|========================================================= - -|Field -|Default -|Description +|==== -|image -| -|Container image tag (required) - indexterm:[image,rkt] - indexterm:[rkt,Property,image] - -|replicas -|Value of +promoted-max+ if that is positive, else 1 -|A positive integer specifying the number of container instances to launch - indexterm:[replicas,rkt] - indexterm:[rkt,Property,replicas] - -|replicas-per-host -|1 -|A positive integer specifying the number of container instances allowed to run - on a single node - indexterm:[replicas-per-host,rkt] - indexterm:[rkt,Property,replicas-per-host] - -|promoted-max -|0 -|A non-negative integer that, if positive, indicates that the containerized - service should be treated as a promotable service, with this many replicas - allowed to run the service in the master role - indexterm:[promoted-max,rkt] - indexterm:[rkt,Property,promoted-max] - -|network -| -|If specified, this will be passed to +rkt run+ as the - network setting for the rkt container. - indexterm:[network,rkt] - indexterm:[rkt,Property,network] - -|run-command -|`/usr/sbin/pacemaker-remoted` if bundle contains a +primitive+, otherwise none -|This command will be run inside the container when launching it ("PID 1"). If - the bundle contains a +primitive+, this command 'must' start pacemaker-remoted - (but could, for example, be a script that does other stuff, too). If the - container image has a pre-2.0.0 version of Pacemaker, set this to - +/usr/sbin/pacemaker_remoted+ (note the underbar instead of dash). - indexterm:[run-command,rkt] - indexterm:[rkt,Property,run-command] - -|options -| -|Extra command-line options to pass to `rkt run` - indexterm:[options,rkt] - indexterm:[rkt,Property,options] +[NOTE] +==== +Considerations when using cluster configurations or container images from +Pacemaker 1.1: -|========================================================= +- If the container image has a pre-2.0.0 version of Pacemaker, set +run-command+ +to +/usr/sbin/pacemaker_remoted+ (note the underbar instead of dash). -For backward compatibility, +masters+ is accepted as an alias for -+promoted-max+, but is deprecated since 2.0.0, and support for it will be -removed in a future version. +- +masters+ is accepted as an alias for +promoted-max+, but is deprecated since +2.0.0, and support for it will be removed in a future version. +==== === Bundle Network Properties === A bundle may optionally contain one ++ element. indexterm:[bundle,network] .Properties of a Bundle's Network Element [width="95%",cols="2m,1,<4",options="header",align="center"] |========================================================= |Field |Default |Description |add-host |TRUE |If TRUE, and +ip-range-start+ is used, Pacemaker will automatically ensure that +/etc/hosts+ inside the containers has entries for each <> and its assigned IP. indexterm:[add-host,network] indexterm:[network,Property,add-host] |ip-range-start | |If specified, Pacemaker will create an implicit +ocf:heartbeat:IPaddr2+ resource for each container instance, starting with this IP address, using up to +replicas+ sequential addresses. These addresses can be used from the host's network to reach the service inside the container, though it is not visible within the container itself. Only IPv4 addresses are currently supported. indexterm:[ip-range-start,network] indexterm:[network,Property,ip-range-start] |host-netmask |32 |If +ip-range-start+ is specified, the IP addresses are created with this CIDR netmask (as a number of bits). indexterm:[host-netmask,network] indexterm:[network,Property,host-netmask] |host-interface | |If +ip-range-start+ is specified, the IP addresses are created on this host interface (by default, it will be determined from the IP address). indexterm:[host-interface,network] indexterm:[network,Property,host-interface] |control-port |3121 |If the bundle contains a +primitive+, the cluster will use this integer TCP port for communication with Pacemaker Remote inside the container. Changing this is useful when the container is unable to listen on the default port, for example, when the container uses the host's network rather than +ip-range-start+ (in which case +replicas-per-host+ must be 1), or when the bundle may run on a Pacemaker Remote node that is already listening on the default port. Any PCMK_remote_port environment variable set on the host or in the container is ignored for bundle connections. indexterm:[control-port,network] indexterm:[network,Property,control-port] |========================================================= [[s-resource-bundle-note-replica-names]] [NOTE] ==== Replicas are named by the bundle id plus a dash and an integer counter starting with zero. For example, if a bundle named +httpd-bundle+ has +replicas=2+, its containers will be named +httpd-bundle-0+ and +httpd-bundle-1+. ==== Additionally, a ++ element may optionally contain one or more ++ elements. indexterm:[bundle,network,port-mapping] .Properties of a Bundle's Port-Mapping Element [width="95%",cols="2m,1,<4",options="header",align="center"] |========================================================= |Field |Default |Description |id | |A unique name for the port mapping (required) indexterm:[id,port-mapping] indexterm:[port-mapping,Property,id] |port | |If this is specified, connections to this TCP port number on the host network (on the container's assigned IP address, if +ip-range-start+ is specified) will be forwarded to the container network. Exactly one of +port+ or +range+ must be specified in a +port-mapping+. indexterm:[port,port-mapping] indexterm:[port-mapping,Property,port] |internal-port |value of +port+ |If +port+ and this are specified, connections to +port+ on the host's network will be forwarded to this port on the container network. indexterm:[internal-port,port-mapping] indexterm:[port-mapping,Property,internal-port] |range | |If this is specified, connections to these TCP port numbers (expressed as 'first_port'-'last_port') on the host network (on the container's assigned IP address, if +ip-range-start+ is specified) will be forwarded to the same ports in the container network. Exactly one of +port+ or +range+ must be specified in a +port-mapping+. indexterm:[range,port-mapping] indexterm:[port-mapping,Property,range] |========================================================= [NOTE] ==== If the bundle contains a +primitive+, Pacemaker will automatically map the +control-port+, so it is not necessary to specify that port in a +port-mapping+. ==== === Bundle Storage Properties === A bundle may optionally contain one ++ element. A ++ element has no properties of its own, but may contain one or more ++ elements. indexterm:[bundle,storage,storage-mapping] .Properties of a Bundle's Storage-Mapping Element [width="95%",cols="2m,1,<4",options="header",align="center"] |========================================================= |Field |Default |Description |id | |A unique name for the storage mapping (required) indexterm:[id,storage-mapping] indexterm:[storage-mapping,Property,id] |source-dir | |The absolute path on the host's filesystem that will be mapped into the container. Exactly one of +source-dir+ and +source-dir-root+ must be specified in a +storage-mapping+. indexterm:[source-dir,storage-mapping] indexterm:[storage-mapping,Property,source-dir] |source-dir-root | |The start of a path on the host's filesystem that will be mapped into the container, using a different subdirectory on the host for each container instance. The subdirectory will be named the same as the <>. Exactly one of +source-dir+ and +source-dir-root+ must be specified in a +storage-mapping+. indexterm:[source-dir-root,storage-mapping] indexterm:[storage-mapping,Property,source-dir-root] |target-dir | |The path name within the container where the host storage will be mapped (required) indexterm:[target-dir,storage-mapping] indexterm:[storage-mapping,Property,target-dir] |options | |File system mount options to use when mapping the storage indexterm:[options,storage-mapping] indexterm:[storage-mapping,Property,options] |========================================================= [NOTE] ==== Pacemaker does not define the behavior if the source directory does not already exist on the host. However, it is expected that the container technology and/or its resource agent will create the source directory in that case. ==== [NOTE] ==== If the bundle contains a +primitive+, Pacemaker will automatically map the equivalent of +source-dir=/etc/pacemaker/authkey target-dir=/etc/pacemaker/authkey+ and +source-dir-root=/var/log/pacemaker/bundles target-dir=/var/log+ into the container, so it is not necessary to specify those paths in a +storage-mapping+. ==== [IMPORTANT] ==== The +PCMK_authkey_location+ environment variable must not be set to anything other than the default of `/etc/pacemaker/authkey` on any node in the cluster. ==== === Bundle Primitive === A bundle may optionally contain one ++ resource (see <>). The primitive may have operations, instance attributes and meta-attributes defined, as usual. If a bundle contains a primitive resource, the container image must include the Pacemaker Remote daemon, and at least one of +ip-range-start+ or +control-port+ must be configured in the bundle. Pacemaker will create an implicit +ocf:pacemaker:remote+ resource for the connection, launch Pacemaker Remote within the container, and monitor and manage the primitive resource via Pacemaker Remote. If the bundle has more than one container instance (replica), the primitive resource will function as an implicit clone (see <>) -- a promotable clone if the bundle has +masters+ greater than zero (see <>). [IMPORTANT] ==== Containers in bundles with a +primitive+ must have an accessible networking environment, so that Pacemaker on the cluster nodes can contact Pacemaker Remote inside the container. For example, the Docker option `--net=none` should not be used with a +primitive+. The default (using a distinct network space inside the container) works in combination with +ip-range-start+. If the Docker option `--net=host` is used (making the container share the host's network space), a unique +control-port+ should be specified for each bundle. Any firewall must allow access to the +control-port+. ==== [[s-bundle-attributes]] === Bundle Node Attributes === If the bundle has a +primitive+, the primitive's resource agent may want to set node attributes such as <>. However, with containers, it is not apparent which node should get the attribute. If the container uses shared storage that is the same no matter which node the container is hosted on, then it is appropriate to use the promotion score on the bundle node itself. On the other hand, if the container uses storage exported from the underlying host, then it may be more appropriate to use the promotion score on the underlying host. Since this depends on the particular situation, the +container-attribute-target+ resource meta-attribute allows the user to specify which approach to use. If it is set to +host+, then user-defined node attributes will be checked on the underlying host. If it is anything else, the local node (in this case the bundle node) is used as usual. This only applies to user-defined attributes; the cluster will always check the local node for cluster-defined attributes such as +#uname+. If +container-attribute-target+ is +host+, the cluster will pass additional environment variables to the primitive's resource agent that allow it to set node attributes appropriately: +CRM_meta_container_attribute_target+ (identical to the meta-attribute value) and +CRM_meta_physical_host+ (the name of the underlying host). [NOTE] ==== When called by a resource agent, the attrd_updater and crm_attribute commands will automatically check those environment variables and set attributes appropriately. ==== === Bundle Meta-Attributes === Any meta-attribute set on a bundle will be inherited by the bundle's primitive and any resources implicitly created by Pacemaker for the bundle. This includes options such as +priority+, +target-role+, and +is-managed+. See <> for more information. === Limitations of Bundles === Restarting pacemaker while a bundle is unmanaged or the cluster is in maintenance mode may cause the bundle to fail. Bundles may not be explicitly cloned or included in groups. This includes the bundle's primitive and any resources implicitly created by Pacemaker for the bundle. (If +replicas+ is greater than 1, the bundle will behave like a clone implicitly.) Bundles do not have instance attributes, utilization attributes, or operations, though a bundle's primitive may have them. A bundle with a primitive can run on a Pacemaker Remote node only if the bundle uses a distinct +control-port+. diff --git a/doc/Pacemaker_Explained/en-US/Pacemaker_Explained.ent b/doc/Pacemaker_Explained/en-US/Pacemaker_Explained.ent index a79fcf33e8..6d5ae65169 100644 --- a/doc/Pacemaker_Explained/en-US/Pacemaker_Explained.ent +++ b/doc/Pacemaker_Explained/en-US/Pacemaker_Explained.ent @@ -1,4 +1,4 @@ - + diff --git a/doc/Pacemaker_Explained/en-US/Revision_History.xml b/doc/Pacemaker_Explained/en-US/Revision_History.xml index 71c5a5b005..270734f3ea 100644 --- a/doc/Pacemaker_Explained/en-US/Revision_History.xml +++ b/doc/Pacemaker_Explained/en-US/Revision_History.xml @@ -1,159 +1,174 @@ Revision History 1-0 19 Oct 2009 AndrewBeekhofandrew@beekhof.net Import from Pages.app 2-0 26 Oct 2009 AndrewBeekhofandrew@beekhof.net Cleanup and reformatting of docbook xml complete 3-0 Tue Nov 12 2009 AndrewBeekhofandrew@beekhof.net Split book into chapters and pass validation Re-organize book for use with Publican 4-0 Mon Oct 8 2012 AndrewBeekhofandrew@beekhof.net Converted to asciidoc (which is converted to docbook for use with Publican) 5-0 Mon Feb 23 2015 KenGaillotkgaillot@redhat.com Update for clarity, stylistic consistency and current command-line syntax 6-0 Tue Dec 8 2015 KenGaillotkgaillot@redhat.com Update for Pacemaker 1.1.14 7-0 Tue May 3 2016 KenGaillotkgaillot@redhat.com Update for Pacemaker 1.1.15 7-1 Fri Oct 28 2016 KenGaillotkgaillot@redhat.com Overhaul upgrade documentation, and document node health strategies 8-0 Tue Oct 25 2016 KenGaillotkgaillot@redhat.com Update for Pacemaker 1.1.16 9-0 Tue Jul 11 2017 KenGaillotkgaillot@redhat.com Update for Pacemaker 1.1.17 10-0 Fri Oct 6 2017 KenGaillotkgaillot@redhat.com Update for Pacemaker 1.1.18 11-0 Fri Jan 12 2018 KenGaillotkgaillot@redhat.com Update for Pacemaker 2.0.0 12-0 Mon Jan 28 2019 KenGaillotkgaillot@redhat.com ReidWahlnwahl@redhat.com JanPokornýjpokorny@redhat.com Update for Pacemaker 2.0.1, remove "Further Reading" and "FAQ" sections, and add minor clarifications and reformatting + + 12-1 + Mon May 13 2019 + KenGaillotkgaillot@redhat.com + MaciejSobkowiakmsobkowiak@egnyte.com + + + + Document podman support, cluster-name cluster option, and new + HealthIOWait agent, with other minor clarifications and + corrections + + + + diff --git a/doc/Pacemaker_Remote/en-US/Pacemaker_Remote.ent b/doc/Pacemaker_Remote/en-US/Pacemaker_Remote.ent index be2a282131..23aa4eb49e 100644 --- a/doc/Pacemaker_Remote/en-US/Pacemaker_Remote.ent +++ b/doc/Pacemaker_Remote/en-US/Pacemaker_Remote.ent @@ -1,6 +1,6 @@ - + diff --git a/include/crm/common/Makefile.am b/include/crm/common/Makefile.am index be19e5ee1d..deeff16fe6 100644 --- a/include/crm/common/Makefile.am +++ b/include/crm/common/Makefile.am @@ -1,18 +1,18 @@ # # Copyright 2004-2019 the Pacemaker project contributors # # The version control history for this file may have further details. # # This source code is licensed under the GNU General Public License version 2 # or later (GPLv2+) WITHOUT ANY WARRANTY. # MAINTAINERCLEANFILES = Makefile.in headerdir=$(pkgincludedir)/crm/common header_HEADERS = xml.h ipc.h util.h iso8601.h mainloop.h logging.h results.h \ nvpair.h noinst_HEADERS = cib_secrets.h ipcs.h internal.h alerts_internal.h \ iso8601_internal.h remote_internal.h xml_internal.h \ - ipc_internal.h + ipc_internal.h output.h diff --git a/include/crm/common/output.h b/include/crm/common/output.h index f31b78430f..a7fac89d80 100644 --- a/include/crm/common/output.h +++ b/include/crm/common/output.h @@ -1,513 +1,514 @@ /* * Copyright 2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef CRM_OUTPUT__H # define CRM_OUTPUT__H #ifdef __cplusplus extern "C" { #endif /** * \file * \brief Formatted output for pacemaker tools */ +# include # include # include # include # include # define PCMK__API_VERSION "1.0" /* Add to the long_options block in each tool to get the formatted output * command line options added. Then call pcmk__parse_output_args to handle * them. */ # define PCMK__OUTPUT_OPTIONS(fmts) \ { "output-as", required_argument, NULL, 0, \ "Specify the format for output, one of: " fmts \ }, \ { "output-to", required_argument, NULL, 0, \ "Specify the destination for formatted output, \"-\" for stdout or a filename" \ } typedef struct pcmk__output_s pcmk__output_t; /*! * \internal * \brief The type of a function that creates a ::pcmk__output_t. * * Instances of this type are passed to pcmk__register_format(), stored in an * internal data structure, and later accessed by pcmk__output_new(). For * examples, see pcmk__mk_xml_output() and pcmk__mk_text_output(). * * \param[in] argv The list of command line arguments. */ typedef pcmk__output_t * (*pcmk__output_factory_t)(char **argv); /*! * \internal * \brief The type of a custom message formatting function. * * These functions are defined by various libraries to support formatting of * types aside from the basic types provided by a ::pcmk__output_t. * * The meaning of the return value will be different for each message. * In general, however, 0 should be returned on success and a positive value * on error. * * \note These functions must not call va_start or va_end - that is done * automatically before the custom formatting function is called. */ typedef int (*pcmk__message_fn_t)(pcmk__output_t *out, va_list args); /*! * \internal * \brief Internal type for tracking custom messages. * * Each library can register functions that format custom message types. These * are commonly used to handle some library-specific type. Registration is * done by first defining a table of ::pcmk__message_entry_t structures and * then passing that table to pcmk__register_messages(). Separate handlers * can be defined for the same message, but for different formats (xml vs. * text). Unknown formats will be ignored. */ typedef struct pcmk__message_entry_s { /*! * \brief The message to be handled. * * This must be the same ID that is passed to the message function of * a ::pcmk__output_t. Unknown message IDs will be ignored. */ const char *message_id; /*! * \brief The format type this handler is for. * * This name must match the fmt_name of the currently active formatter in * order for the registered function to be called. It is valid to have * multiple entries for the same message_id but with different fmt_name * values. */ const char *fmt_name; /*! * \brief The function to be called for message_id given a match on * fmt_name. See comments on ::pcmk__message_fn_t. */ pcmk__message_fn_t fn; } pcmk__message_entry_t; /* Basic formatters everything supports. This block needs to be updated every * time a new base formatter is added. */ pcmk__output_t *pcmk__mk_text_output(char **argv); pcmk__output_t *pcmk__mk_xml_output(char **argv); /*! * \brief This structure contains everything that makes up a single output * formatter. * * Instances of this structure may be created by calling pcmk__output_new() * with the name of the desired formatter. They should later be freed with * pcmk__output_free(). */ struct pcmk__output_s { /*! * \brief The name of this output formatter. */ char *fmt_name; /*! * \brief A copy of the request that generated this output. * * In the case of command line usage, this would be the command line * arguments. For other use cases, it could be different. */ char *request; /*! * \brief Does this formatter support a special quiet mode? * * In this mode, most output can be supressed but some information is still * displayed to an interactive user. In general, machine-readable output * formats will not support this while user-oriented formats will. */ bool supports_quiet; /*! * \brief Where output should be written. * * This could be a file handle, or stdout or stderr. This is really only * useful internally. */ FILE *dest; /*! * \brief Custom messages that are currently registered on this formatter. * * Keys are the string message IDs, values are ::pcmk__message_fn_t function * pointers. */ GHashTable *messages; /*! * \brief Implementation-specific private data. * * Each individual formatter may have some private data useful in its * implementation. This points to that data. Callers should not rely on * its contents or structure. */ void *priv; /*! * \internal * \brief Take whatever actions are necessary to prepare out for use. This is * called by pcmk__output_new(). End users should not need to call this. * * \note For formatted output implementers - This function should be written in * such a way that it can be called repeatedly on an already initialized * object without causing problems, or on a previously finished object * without crashing. * * \param[in,out] out The output functions structure. * * \return true on success, false on error. */ bool (*init) (pcmk__output_t *out); /*! * \internal * \brief Free the private formatter-specific data. * * This is called from pcmk__output_free() and does not typically need to be * called directly. * * \param[in,out] out The output functions structure. */ void (*free_priv)(pcmk__output_t *out); /*! * \internal * \brief Take whatever actions are necessary to end formatted output. * * This could include flushing output to a file, but does not include freeing * anything. Note that pcmk__output_free() will automatically call this * function, so there is typically no need to do so manually. * * \note For formatted output implementers - This function should be written in * such a way that it can be called repeatedly on a previously finished * object without crashing. * * \param[in,out] out The output functions structure. * \param[in] exit_status The exit value of the whole program. */ void (*finish) (pcmk__output_t *out, crm_exit_t exit_status); /*! * \internal * \brief Finalize output and then immediately set back up to start a new set * of output. * * This is conceptually the same as calling finish and then init, though in * practice more be happening behind the scenes. * * \note This function differs from finish in that no exit_status is added. * The idea is that the program is not shutting down, so there is not * yet a final exit code. Call finish on the last time through if this * is needed. * * \param[in,out] out The output functions structure. */ void (*reset) (pcmk__output_t *out); /*! * \internal * \brief Register a custom message. * * \param[in,out] out The output functions structure. * \param[in] message_id The name of the message to register. This name * will be used as the message_id parameter to the * message function in order to call the custom * format function. * \param[in] fn The custom format function to call for message_id. */ void (*register_message) (pcmk__output_t *out, const char *message_id, pcmk__message_fn_t fn); /*! * \internal * \brief Call a previously registered custom message. * * \param[in,out] out The output functions structure. * \param[in] message_id The name of the message to call. This name must * be the same as the message_id parameter of some * previous call to register_message. * \param[in] ... Arguments to be passed to the registered function. * * \return 0 if a function was registered for the message, that function was * called, and returned successfully. A negative value is returned if * no function was registered. A positive value is returned if the * function was called but encountered an error. */ int (*message) (pcmk__output_t *out, const char *message_id, ...); /*! * \internal * \brief Format the output of a completed subprocess. * * \param[in,out] out The output functions structure. * \param[in] exit_status The exit value of the subprocess. * \param[in] proc_stdout stdout from the completed subprocess. * \param[in] proc_stderr stderr from the completed subprocess. */ void (*subprocess_output) (pcmk__output_t *out, int exit_status, const char *proc_stdout, const char *proc_stderr); /*! * \internal * \brief Format an informational message that should be shown to * to an interactive user. Not all formatters will do this. * * \note A newline will automatically be added to the end of the format * string, so callers should not include a newline. * * \param[in,out] out The output functions structure. * \param[in] buf The message to be printed. * \param[in] ... Arguments to be formatted. */ void (*info) (pcmk__output_t *out, const char *format, ...) G_GNUC_PRINTF(2, 3); /*! * \internal * \brief Format already formatted XML. * * \param[in,out] out The output functions structure. * \param[in] name A name to associate with the XML. * \param[in] buf The XML in a string. */ void (*output_xml) (pcmk__output_t *out, const char *name, const char *buf); /*! * \internal * \brief Start a new list of items. * * \note For text output, this corresponds to another level of indentation. For * XML output, this corresponds to wrapping any following output in another * layer of tags. * * \note If singular_noun and plural_noun are non-NULL, calling end_list will * result in a summary being added. * * \param[in,out] out The output functions structure. * \param[in] name A descriptive, user-facing name for this list. * \param[in] singular_noun When outputting the summary for a list with * one item, the noun to use. * \param[in] plural_noun When outputting the summary for a list with * more than one item, the noun to use. */ void (*begin_list) (pcmk__output_t *out, const char *name, const char *singular_noun, const char *plural_noun); /*! * \internal * \brief Format a single item in a list. * * \param[in,out] out The output functions structure. * \param[in] name A name to associate with this item. * \param[in] content The item to be formatted. */ void (*list_item) (pcmk__output_t *out, const char *name, const char *content); /*! * \internal * \brief Conclude a list. * * \note If begin_list was called with non-NULL for both the singular_noun * and plural_noun arguments, this function will output a summary. * Otherwise, no summary will be added. * * \param[in,out] out The output functions structure. */ void (*end_list) (pcmk__output_t *out); }; /*! * \internal * \brief Call a formatting function for a previously registered message. * * \note This function is for implementing custom formatters. It should not * be called directly. Instead, call out->message. * * \param[in,out] out The output functions structure. * \param[in] message_id The message to be handled. Unknown messages * will be ignored. * \param[in] ... Arguments to be passed to the registered function. */ int pcmk__call_message(pcmk__output_t *out, const char *message_id, ...); /*! * \internal * \brief Free a ::pcmk__output_t structure that was previously created by * pcmk__output_new(). This will first call the finish function. * * \note While the create and finish functions are designed in such a way that * they can be called repeatedly, this function will completely free the * memory of the object. Once this function has been called, producing * more output requires starting over from pcmk__output_new(). * * \param[in,out] out The output structure. * \param[in] exit_status The exit value of the whole program. */ void pcmk__output_free(pcmk__output_t *out, crm_exit_t exit_status); /*! * \internal * \brief Create a new ::pcmk__output_t structure. * * \param[in,out] out The destination of the new ::pcmk__output_t. * \param[in] fmt_name How should output be formatted? * \param[in] filename Where should formatted output be written to? This * can be a filename (which will be overwritten if it * already exists), or NULL or "-" for stdout. For no * output, pass a filename of "/dev/null". * \param[in] argv The list of command line arguments. * * \return 0 on success or an error code on error. */ int pcmk__output_new(pcmk__output_t **out, const char *fmt_name, const char *filename, char **argv); /*! * \internal * \brief Process formatted output related command line options. This should * be called wherever other long options are handled. * * \param[in] argname The long command line argument to process. * \param[in] argvalue The value of the command line argument. * \param[out] output_ty How should output be formatted? ("text", "xml", etc.) * \param[out] output_dest Where should formatted output be written to? This is * typically a filename, but could be NULL or "-". * * \return true if longname was handled, false otherwise. */ bool pcmk__parse_output_args(const char *argname, char *argvalue, char **output_ty, char **output_dest); /*! * \internal * \brief Register a new output formatter, making it available for use * the same as a base formatter. * * \param[in] fmt The new output formatter to register. * * \return 0 on success or an error code on error. */ int pcmk__register_format(const char *fmt_name, pcmk__output_factory_t create); /*! * \internal * \brief Register a function to handle a custom message. * * \note This function is for implementing custom formatters. It should not * be called directly. Instead, call out->register_message. * * \param[in,out] out The output functions structure. * \param[in] message_id The message to be handled. * \param[in] fn The custom format function to call for message_id. */ void pcmk__register_message(pcmk__output_t *out, const char *message_id, pcmk__message_fn_t fn); /*! * \internal * \brief Register an entire table of custom formatting functions at once. * * This table can contain multiple formatting functions for the same message ID * if they are for different format types. * * \param[in,out] out The output functions structure. * \param[in] table An array of ::pcmk__message_entry_t values which should * all be registered. This array must be NULL-terminated. */ void pcmk__register_messages(pcmk__output_t *out, pcmk__message_entry_t *table); /* Functions that are useful for implementing custom message formatters */ /*! * \internal * \brief A printf-like function. * * This function writes to out->dest and indents the text to the current level * of the text formatter's nesting. This should be used when implementing * custom message functions instead of printf. * * \param[in,out] out The output functions structure. */ void pcmk__indented_printf(pcmk__output_t *out, const char *format, ...) G_GNUC_PRINTF(2, 3); /*! * \internal * \brief Add the given node as a child of the current list parent. This is * used when implementing custom message functions. * * \param[in,out] out The output functions structure. * \param[in] node An XML node to be added as a child. */ void pcmk__xml_add_node(pcmk__output_t *out, xmlNodePtr node); /*! * \internal * \brief Push a parent XML node onto the stack. This is used when implementing * custom message functions. * * The XML output formatter maintains an internal stack to keep track of which nodes * are parents in order to build up the tree structure. This function can be used * to temporarily push a new node onto the stack. After calling this function, any * other formatting functions will have their nodes added as children of this new * parent. * * \param[in,out] out The output functions structure. * \param[in] node The node to be added/ */ void pcmk__xml_push_parent(pcmk__output_t *out, xmlNodePtr node); /*! * \internal * \brief Pop a parent XML node onto the stack. This is used when implementing * custom message functions. * * This function removes a parent node from the stack. See pcmk__xml_push_parent() * for more details. * * \note Little checking is done with this function. Be sure you only pop parents * that were previously pushed. In general, it is best to keep the code between * push and pop simple. * * \param[in,out] out The output functions structure. */ void pcmk__xml_pop_parent(pcmk__output_t *out); #ifdef __cplusplus } #endif #endif diff --git a/include/crm/fencing/internal.h b/include/crm/fencing/internal.h index e88a155384..0c0ac70fc4 100644 --- a/include/crm/fencing/internal.h +++ b/include/crm/fencing/internal.h @@ -1,146 +1,150 @@ /* - * Copyright 2011-2018 the Pacemaker project contributors + * Copyright 2011-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef STONITH_NG_INTERNAL__H # define STONITH_NG_INTERNAL__H # include +# include # include struct stonith_action_s; typedef struct stonith_action_s stonith_action_t; stonith_action_t *stonith_action_create(const char *agent, const char *_action, const char *victim, uint32_t victim_nodeid, int timeout, GHashTable * device_args, GHashTable * port_map); void stonith__destroy_action(stonith_action_t *action); void stonith__action_result(stonith_action_t *action, int *rc, char **output, char **error_output); -GPid +int stonith_action_execute_async(stonith_action_t * action, void *userdata, void (*done) (GPid pid, int rc, const char *output, - gpointer user_data)); + gpointer user_data), + void (*fork_cb) (GPid pid, gpointer user_data)); int stonith__execute(stonith_action_t *action); xmlNode *create_level_registration_xml(const char *node, const char *pattern, const char *attr, const char *value, int level, stonith_key_value_t *device_list); xmlNode *create_device_registration_xml(const char *id, enum stonith_namespace namespace, const char *agent, stonith_key_value_t *params, const char *rsc_provides); +void stonith__register_messages(pcmk__output_t *out); + # define ST_LEVEL_MAX 10 # define F_STONITH_CLIENTID "st_clientid" # define F_STONITH_CALLOPTS "st_callopt" # define F_STONITH_CALLID "st_callid" # define F_STONITH_CALLDATA "st_calldata" # define F_STONITH_OPERATION "st_op" # define F_STONITH_TARGET "st_target" # define F_STONITH_REMOTE_OP_ID "st_remote_op" # define F_STONITH_RC "st_rc" /*! Timeout period per a device execution */ # define F_STONITH_TIMEOUT "st_timeout" # define F_STONITH_TOLERANCE "st_tolerance" /*! Action specific timeout period returned in query of fencing devices. */ # define F_STONITH_ACTION_TIMEOUT "st_action_timeout" /*! Host in query result is not allowed to run this action */ # define F_STONITH_ACTION_DISALLOWED "st_action_disallowed" /*! Maximum of random fencing delay for a device */ # define F_STONITH_DELAY_MAX "st_delay_max" /*! Base delay used for a fencing delay */ # define F_STONITH_DELAY_BASE "st_delay_base" /*! Has this device been verified using a monitor type * operation (monitor, list, status) */ # define F_STONITH_DEVICE_VERIFIED "st_monitor_verified" /*! device is required for this action */ # define F_STONITH_DEVICE_REQUIRED "st_required" /*! number of available devices in query result */ # define F_STONITH_AVAILABLE_DEVICES "st-available-devices" # define F_STONITH_CALLBACK_TOKEN "st_async_id" # define F_STONITH_CLIENTNAME "st_clientname" # define F_STONITH_CLIENTNODE "st_clientnode" # define F_STONITH_NOTIFY_ACTIVATE "st_notify_activate" # define F_STONITH_NOTIFY_DEACTIVATE "st_notify_deactivate" # define F_STONITH_DELEGATE "st_delegate" /*! The node initiating the stonith operation. If an operation * is relayed, this is the last node the operation lands on. When * in standalone mode, origin is the client's id that originated the * operation. */ # define F_STONITH_ORIGIN "st_origin" # define F_STONITH_HISTORY_LIST "st_history" # define F_STONITH_DATE "st_date" # define F_STONITH_STATE "st_state" # define F_STONITH_ACTIVE "st_active" # define F_STONITH_DIFFERENTIAL "st_differential" # define F_STONITH_DEVICE "st_device_id" # define F_STONITH_ACTION "st_device_action" # define F_STONITH_MODE "st_mode" # define T_STONITH_NG "stonith-ng" # define T_STONITH_REPLY "st-reply" /*! For async operations, an event from the server containing * the total amount of time the server is allowing for the operation * to take place is returned to the client. */ # define T_STONITH_TIMEOUT_VALUE "st-async-timeout-value" # define T_STONITH_NOTIFY "st_notify" # define STONITH_ATTR_HOSTARG "pcmk_host_argument" # define STONITH_ATTR_HOSTMAP "pcmk_host_map" # define STONITH_ATTR_HOSTLIST "pcmk_host_list" # define STONITH_ATTR_HOSTCHECK "pcmk_host_check" # define STONITH_ATTR_DELAY_MAX "pcmk_delay_max" # define STONITH_ATTR_DELAY_BASE "pcmk_delay_base" # define STONITH_ATTR_ACTION_LIMIT "pcmk_action_limit" # define STONITH_ATTR_ACTION_OP "action" # define STONITH_OP_EXEC "st_execute" # define STONITH_OP_TIMEOUT_UPDATE "st_timeout_update" # define STONITH_OP_QUERY "st_query" # define STONITH_OP_FENCE "st_fence" # define STONITH_OP_RELAY "st_relay" # define STONITH_OP_DEVICE_ADD "st_device_register" # define STONITH_OP_DEVICE_DEL "st_device_remove" # define STONITH_OP_FENCE_HISTORY "st_fence_history" # define STONITH_OP_LEVEL_ADD "st_level_add" # define STONITH_OP_LEVEL_DEL "st_level_remove" # define STONITH_WATCHDOG_AGENT "#watchdog" # ifdef HAVE_STONITH_STONITH_H // utilities from st_lha.c int stonith__list_lha_agents(stonith_key_value_t **devices); int stonith__lha_metadata(const char *agent, int timeout, char **output); bool stonith__agent_is_lha(const char *agent); int stonith__lha_validate(stonith_t *st, int call_options, const char *target, const char *agent, GHashTable *params, int timeout, char **output, char **error_output); # endif // utilities from st_rhcs.c int stonith__list_rhcs_agents(stonith_key_value_t **devices); int stonith__rhcs_metadata(const char *agent, int timeout, char **output); bool stonith__agent_is_rhcs(const char *agent); int stonith__rhcs_validate(stonith_t *st, int call_options, const char *target, const char *agent, GHashTable *params, int timeout, char **output, char **error_output); #endif diff --git a/include/crm/stonith-ng.h b/include/crm/stonith-ng.h index 70c0233532..b7365a9700 100644 --- a/include/crm/stonith-ng.h +++ b/include/crm/stonith-ng.h @@ -1,560 +1,550 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifdef __cplusplus extern "C" { #endif /** * \file * \brief Fencing aka. STONITH * \ingroup fencing */ #ifndef STONITH_NG__H # define STONITH_NG__H # include # include # include // bool # include // uint32_t # include // time_t -# include - # define T_STONITH_NOTIFY_DISCONNECT "st_notify_disconnect" # define T_STONITH_NOTIFY_FENCE "st_notify_fence" # define T_STONITH_NOTIFY_HISTORY "st_notify_history" /* *INDENT-OFF* */ enum stonith_state { stonith_connected_command, stonith_connected_query, stonith_disconnected, }; enum stonith_call_options { st_opt_none = 0x00000000, st_opt_verbose = 0x00000001, st_opt_allow_suicide = 0x00000002, st_opt_manual_ack = 0x00000008, st_opt_discard_reply = 0x00000010, /* st_opt_all_replies = 0x00000020, */ st_opt_topology = 0x00000040, st_opt_scope_local = 0x00000100, st_opt_cs_nodeid = 0x00000200, st_opt_sync_call = 0x00001000, /*! Allow the timeout period for a callback to be adjusted * based on the time the server reports the operation will take. */ st_opt_timeout_updates = 0x00002000, /*! Only report back if operation is a success in callback */ st_opt_report_only_success = 0x00004000, /* used where ever apropriate - e.g. cleanup of history */ st_opt_cleanup = 0x000080000, /* used where ever apropriate - e.g. send out a history query to all nodes */ st_opt_broadcast = 0x000100000, }; /*! Order matters here, do not change values */ enum op_state { st_query, st_exec, st_done, st_duplicate, st_failed, }; // Supported fence agent interface standards enum stonith_namespace { st_namespace_invalid, st_namespace_any, st_namespace_internal, // Implemented internally by Pacemaker /* Neither of these projects are active any longer, but the fence agent * interfaces they created are still in use and supported by Pacemaker. */ st_namespace_rhcs, // Red Hat Cluster Suite compatible st_namespace_lha, // Linux-HA compatible }; enum stonith_namespace stonith_text2namespace(const char *namespace_s); const char *stonith_namespace2text(enum stonith_namespace st_namespace); enum stonith_namespace stonith_get_namespace(const char *agent, const char *namespace_s); typedef struct stonith_key_value_s { char *key; char *value; struct stonith_key_value_s *next; } stonith_key_value_t; typedef struct stonith_history_s { char *target; char *action; char *origin; char *delegate; char *client; int state; time_t completed; struct stonith_history_s *next; } stonith_history_t; typedef struct stonith_s stonith_t; typedef struct stonith_event_s { char *id; char *type; char *message; char *operation; int result; char *origin; char *target; char *action; char *executioner; char *device; /*! The name of the client that initiated the action. */ char *client_origin; } stonith_event_t; typedef struct stonith_callback_data_s { int rc; int call_id; void *userdata; } stonith_callback_data_t; typedef struct stonith_api_operations_s { /*! * \brief Destroy the stonith api structure. */ int (*free) (stonith_t *st); /*! * \brief Connect to the local stonith daemon. * * \retval 0, success * \retval negative error code on failure */ int (*connect) (stonith_t *st, const char *name, int *stonith_fd); /*! * \brief Disconnect from the local stonith daemon. * * \retval 0, success * \retval negative error code on failure */ int (*disconnect)(stonith_t *st); /*! * \brief Remove a registered stonith device with the local stonith daemon. * * \note Synchronous, guaranteed to occur in daemon before function returns. * * \retval 0, success * \retval negative error code on failure */ int (*remove_device)( stonith_t *st, int options, const char *name); /*! * \brief Register a stonith device with the local stonith daemon. * * \note Synchronous, guaranteed to occur in daemon before function returns. * * \retval 0, success * \retval negative error code on failure */ int (*register_device)( stonith_t *st, int options, const char *id, const char *provider, const char *agent, stonith_key_value_t *params); /*! * \brief Remove a fencing level for a specific node. * * \note This feature is not available when stonith is in standalone mode. * * \retval 0, success * \retval negative error code on failure */ int (*remove_level)( stonith_t *st, int options, const char *node, int level); /*! * \brief Register a fencing level containing the fencing devices to be used * at that level for a specific node. * * \note This feature is not available when stonith is in standalone mode. * * \retval 0, success * \retval negative error code on failure */ int (*register_level)( stonith_t *st, int options, const char *node, int level, stonith_key_value_t *device_list); /*! * \brief Get the metadata documentation for a resource. * * \note Value is returned in output. Output must be freed when set. * * \retval 0 success * \retval negative error code on failure */ int (*metadata)(stonith_t *st, int options, const char *device, const char *provider, char **output, int timeout); /*! * \brief Retrieve a list of installed stonith agents * * \note if provider is not provided, all known agents will be returned * \note list must be freed using stonith_key_value_freeall() * \note call_options parameter is not used, it is reserved for future use. * * \retval num items in list on success * \retval negative error code on failure */ int (*list_agents)(stonith_t *stonith, int call_options, const char *provider, stonith_key_value_t **devices, int timeout); /*! * \brief Retrieve string listing hosts and port assignments from a local stonith device. * * \retval 0 on success * \retval negative error code on failure */ int (*list)(stonith_t *st, int options, const char *id, char **list_output, int timeout); /*! * \brief Check to see if a local stonith device is reachable * * \retval 0 on success * \retval negative error code on failure */ int (*monitor)(stonith_t *st, int options, const char *id, int timeout); /*! * \brief Check to see if a local stonith device's port is reachable * * \retval 0 on success * \retval negative error code on failure */ int (*status)(stonith_t *st, int options, const char *id, const char *port, int timeout); /*! * \brief Retrieve a list of registered stonith devices. * * \note If node is provided, only devices that can fence the node id * will be returned. * * \retval num items in list on success * \retval negative error code on failure */ int (*query)(stonith_t *st, int options, const char *node, stonith_key_value_t **devices, int timeout); /*! * \brief Issue a fencing action against a node. * * \note Possible actions are, 'on', 'off', and 'reboot'. * * \param st, stonith connection * \param options, call options * \param node, The target node to fence * \param action, The fencing action to take * \param timeout, The default per device timeout to use with each device * capable of fencing the target. * * \retval 0 success * \retval negative error code on failure. */ int (*fence)(stonith_t *st, int options, const char *node, const char *action, int timeout, int tolerance); /*! * \brief Manually confirm that a node is down. * * \retval 0 success * \retval negative error code on failure. */ int (*confirm)(stonith_t *st, int options, const char *node); /*! * \brief Retrieve a list of fencing operations that have occurred for a specific node. * * \note History is not available in standalone mode. * * \retval 0 success * \retval negative error code on failure. */ int (*history)(stonith_t *st, int options, const char *node, stonith_history_t **output, int timeout); int (*register_notification)( stonith_t *st, const char *event, void (*notify)(stonith_t *st, stonith_event_t *e)); int (*remove_notification)(stonith_t *st, const char *event); /*! * \brief Register a callback to receive the result of an asynchronous call * * \param[in] call_id The call ID to register callback for * \param[in] timeout Default time to wait until callback expires * \param[in] options Bitmask of \c stonith_call_options (respects * \c st_opt_timeout_updates and * \c st_opt_report_only_success) * \param[in] userdata Pointer that will be given to callback * \param[in] callback_name Unique name to identify callback * \param[in] callback The callback function to register * * \return \c TRUE on success, \c FALSE if call_id is negative, -errno otherwise * * \todo This function should return \c pcmk_ok on success, and \c call_id * when negative, but that would break backward compatibility. */ int (*register_callback)(stonith_t *st, int call_id, int timeout, int options, void *userdata, const char *callback_name, void (*callback)(stonith_t *st, stonith_callback_data_t *data)); /*! * \brief Remove a registered callback for a given call id. */ int (*remove_callback)(stonith_t *st, int call_id, bool all_callbacks); /*! * \brief Remove fencing level for specific node, node regex or attribute * * \param[in] st Fencer connection to use * \param[in] options Bitmask of stonith_call_options to pass to the fencer * \param[in] node If not NULL, target level by this node name * \param[in] pattern If not NULL, target by node name using this regex * \param[in] attr If not NULL, target by this node attribute * \param[in] value If not NULL, target by this node attribute value * \param[in] level Index number of level to remove * * \return 0 on success, negative error code otherwise * * \note This feature is not available when stonith is in standalone mode. * The caller should set only one of node, pattern or attr/value. */ int (*remove_level_full)(stonith_t *st, int options, const char *node, const char *pattern, const char *attr, const char *value, int level); /*! * \brief Register fencing level for specific node, node regex or attribute * * \param[in] st Fencer connection to use * \param[in] options Bitmask of stonith_call_options to pass to fencer * \param[in] node If not NULL, target level by this node name * \param[in] pattern If not NULL, target by node name using this regex * \param[in] attr If not NULL, target by this node attribute * \param[in] value If not NULL, target by this node attribute value * \param[in] level Index number of level to add * \param[in] device_list Devices to use in level * * \return 0 on success, negative error code otherwise * * \note This feature is not available when stonith is in standalone mode. * The caller should set only one of node, pattern or attr/value. */ int (*register_level_full)(stonith_t *st, int options, const char *node, const char *pattern, const char *attr, const char *value, int level, stonith_key_value_t *device_list); /*! * \brief Validate an arbitrary stonith device configuration * * \param[in] st Stonithd connection to use * \param[in] call_options Bitmask of stonith_call_options to use with fencer * \param[in] rsc_id ID used to replace CIB secrets in params * \param[in] namespace_s Namespace of fence agent to validate (optional) * \param[in] agent Fence agent to validate * \param[in] params Configuration parameters to pass to fence agent * \param[in] timeout Fail if no response within this many seconds * \param[out] output If non-NULL, where to store any agent output * \param[out] error_output If non-NULL, where to store agent error output * * \return pcmk_ok if validation succeeds, -errno otherwise * * \note If pcmk_ok is returned, the caller is responsible for freeing * the output (if requested). */ int (*validate)(stonith_t *st, int call_options, const char *rsc_id, const char *namespace_s, const char *agent, stonith_key_value_t *params, int timeout, char **output, char **error_output); } stonith_api_operations_t; struct stonith_s { enum stonith_state state; int call_id; int call_timeout; void *st_private; stonith_api_operations_t *cmds; }; /* *INDENT-ON* */ /* Core functions */ stonith_t *stonith_api_new(void); void stonith_api_delete(stonith_t * st); void stonith_dump_pending_callbacks(stonith_t * st); // deprecated (use stonith_get_namespace() instead) const char *get_stonith_provider(const char *agent, const char *provider); bool stonith_dispatch(stonith_t * st); stonith_key_value_t *stonith_key_value_add(stonith_key_value_t * kvp, const char *key, const char *value); void stonith_key_value_freeall(stonith_key_value_t * kvp, int keys, int values); void stonith_history_free(stonith_history_t *history); /* Basic helpers that allows nodes to be fenced and the history to be * queried without mainloop or the caller understanding the full API * * At least one of nodeid and uname are required */ int stonith_api_kick(uint32_t nodeid, const char *uname, int timeout, bool off); time_t stonith_api_time(uint32_t nodeid, const char *uname, bool in_progress); /* * Helpers for using the above functions without install-time dependencies * * Usage: * #include * * To turn a node off by corosync nodeid: * stonith_api_kick_helper(nodeid, 120, 1); * * To check the last fence date/time (also by nodeid): * last = stonith_api_time_helper(nodeid, 0); * * To check if fencing is in progress: * if(stonith_api_time_helper(nodeid, 1) > 0) { ... } * * eg. #include #include #include int main(int argc, char ** argv) { int rc = 0; int nodeid = 102; rc = stonith_api_time_helper(nodeid, 0); printf("%d last fenced at %s\n", nodeid, ctime(rc)); rc = stonith_api_kick_helper(nodeid, 120, 1); printf("%d fence result: %d\n", nodeid, rc); rc = stonith_api_time_helper(nodeid, 0); printf("%d last fenced at %s\n", nodeid, ctime(rc)); return 0; } */ # define STONITH_LIBRARY "libstonithd.so.26" typedef int (*st_api_kick_fn) (int nodeid, const char *uname, int timeout, bool off); typedef time_t (*st_api_time_fn) (int nodeid, const char *uname, bool in_progress); static inline int stonith_api_kick_helper(uint32_t nodeid, int timeout, bool off) { static void *st_library = NULL; static st_api_kick_fn st_kick_fn; if (st_library == NULL) { st_library = dlopen(STONITH_LIBRARY, RTLD_LAZY); } if (st_library && st_kick_fn == NULL) { st_kick_fn = (st_api_kick_fn) dlsym(st_library, "stonith_api_kick"); } if (st_kick_fn == NULL) { #ifdef ELIBACC return -ELIBACC; #else return -ENOSYS; #endif } return (*st_kick_fn) (nodeid, NULL, timeout, off); } static inline time_t stonith_api_time_helper(uint32_t nodeid, bool in_progress) { static void *st_library = NULL; static st_api_time_fn st_time_fn; if (st_library == NULL) { st_library = dlopen(STONITH_LIBRARY, RTLD_LAZY); } if (st_library && st_time_fn == NULL) { st_time_fn = (st_api_time_fn) dlsym(st_library, "stonith_api_time"); } if (st_time_fn == NULL) { return 0; } return (*st_time_fn) (nodeid, NULL, in_progress); } /** * Does the given agent describe a stonith resource that can exist? * * \param[in] agent What is the name of the agent? * \param[in] timeout Timeout to use when querying. If 0 is given, * use a default of 120. * * \return A boolean */ bool stonith_agent_exists(const char *agent, int timeout); -/*! - * \internal - * \brief Register stonith-specific messages. - * - * \param out The output functions structure. - */ -void stonith__register_messages(pcmk__output_t *out); - /*! * \brief Turn stonith action into a more readable string. * * \param action Stonith action */ const char *stonith_action_str(const char *action); #ifdef __cplusplus } #endif #endif diff --git a/lib/common/logging.c b/lib/common/logging.c index 60745977ba..674a4b8c27 100644 --- a/lib/common/logging.c +++ b/lib/common/logging.c @@ -1,1035 +1,1040 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include unsigned int crm_log_priority = LOG_NOTICE; unsigned int crm_log_level = LOG_INFO; static gboolean crm_tracing_enabled(void); unsigned int crm_trace_nonlog = 0; bool crm_is_daemon = 0; GLogFunc glib_log_default; static void crm_glib_handler(const gchar * log_domain, GLogLevelFlags flags, const gchar * message, gpointer user_data) { int log_level = LOG_WARNING; GLogLevelFlags msg_level = (flags & G_LOG_LEVEL_MASK); static struct qb_log_callsite *glib_cs = NULL; if (glib_cs == NULL) { glib_cs = qb_log_callsite_get(__FUNCTION__, __FILE__, "glib-handler", LOG_DEBUG, __LINE__, crm_trace_nonlog); } switch (msg_level) { case G_LOG_LEVEL_CRITICAL: log_level = LOG_CRIT; if (crm_is_callsite_active(glib_cs, LOG_DEBUG, 0) == FALSE) { /* log and record how we got here */ crm_abort(__FILE__, __FUNCTION__, __LINE__, message, TRUE, TRUE); } break; case G_LOG_LEVEL_ERROR: log_level = LOG_ERR; break; case G_LOG_LEVEL_MESSAGE: log_level = LOG_NOTICE; break; case G_LOG_LEVEL_INFO: log_level = LOG_INFO; break; case G_LOG_LEVEL_DEBUG: log_level = LOG_DEBUG; break; case G_LOG_LEVEL_WARNING: case G_LOG_FLAG_RECURSION: case G_LOG_FLAG_FATAL: case G_LOG_LEVEL_MASK: log_level = LOG_WARNING; break; } do_crm_log(log_level, "%s: %s", log_domain, message); } #ifndef NAME_MAX # define NAME_MAX 256 #endif /*! * \internal * \brief Write out a blackbox (enabling blackboxes if needed) * * \param[in] nsig Signal number that was received * * \note This is a true signal handler, and so must be async-safe. */ static void crm_trigger_blackbox(int nsig) { if(nsig == SIGTRAP) { /* Turn it on if it wasn't already */ crm_enable_blackbox(nsig); } crm_write_blackbox(nsig, NULL); } const char * daemon_option(const char *option) { char env_name[NAME_MAX]; const char *value = NULL; snprintf(env_name, NAME_MAX, "PCMK_%s", option); value = getenv(env_name); if (value != NULL) { crm_trace("Found %s = %s", env_name, value); return value; } snprintf(env_name, NAME_MAX, "HA_%s", option); value = getenv(env_name); if (value != NULL) { crm_trace("Found %s = %s", env_name, value); return value; } crm_trace("Nothing found for %s", option); return NULL; } void set_daemon_option(const char *option, const char *value) { char env_name[NAME_MAX]; snprintf(env_name, NAME_MAX, "PCMK_%s", option); if (value) { crm_trace("Setting %s to %s", env_name, value); setenv(env_name, value, 1); } else { crm_trace("Unsetting %s", env_name); unsetenv(env_name); } snprintf(env_name, NAME_MAX, "HA_%s", option); if (value) { crm_trace("Setting %s to %s", env_name, value); setenv(env_name, value, 1); } else { crm_trace("Unsetting %s", env_name); unsetenv(env_name); } } gboolean daemon_option_enabled(const char *daemon, const char *option) { const char *value = daemon_option(option); if (value != NULL && crm_is_true(value)) { return TRUE; } else if (value != NULL && strstr(value, daemon)) { return TRUE; } return FALSE; } void crm_log_deinit(void) { g_log_set_default_handler(glib_log_default, NULL); } #define FMT_MAX 256 static void set_format_string(int method, const char *daemon) { if (method == QB_LOG_SYSLOG) { // The system log gets a simplified, user-friendly format crm_extended_logging(method, QB_FALSE); qb_log_format_set(method, "%g %p: %b"); } else { // Everything else gets more detail, for advanced troubleshooting int offset = 0; char fmt[FMT_MAX]; if (method > QB_LOG_STDERR) { struct utsname res; const char *nodename = "localhost"; if (uname(&res) == 0) { nodename = res.nodename; } // If logging to file, prefix with timestamp, node name, daemon ID offset += snprintf(fmt + offset, FMT_MAX - offset, "%%t %s %-20s[%lu] ", nodename, daemon, (unsigned long) getpid()); } // Add function name (in parentheses) offset += snprintf(fmt + offset, FMT_MAX - offset, "(%%n"); if (crm_tracing_enabled()) { // When tracing, add file and line number offset += snprintf(fmt + offset, FMT_MAX - offset, "@%%f:%%l"); } offset += snprintf(fmt + offset, FMT_MAX - offset, ")"); // Add tag (if any), severity, and actual message offset += snprintf(fmt + offset, FMT_MAX - offset, " %%g\t%%p: %%b"); CRM_LOG_ASSERT(offset > 0); qb_log_format_set(method, fmt); } } gboolean crm_add_logfile(const char *filename) { bool is_default = false; static int default_fd = -1; static gboolean have_logfile = FALSE; const char *default_logfile = CRM_LOG_DIR "/pacemaker.log"; struct stat parent; int fd = 0, rc = 0; FILE *logfile = NULL; char *parent_dir = NULL; char *filename_cp; if (filename == NULL && have_logfile == FALSE) { filename = default_logfile; } if (filename == NULL) { return FALSE; /* Nothing to do */ } else if(safe_str_eq(filename, "none")) { return FALSE; /* Nothing to do */ } else if(safe_str_eq(filename, "/dev/null")) { return FALSE; /* Nothing to do */ } else if(safe_str_eq(filename, default_logfile)) { is_default = TRUE; } if(is_default && default_fd >= 0) { return TRUE; /* Nothing to do */ } /* Check the parent directory */ filename_cp = strdup(filename); parent_dir = dirname(filename_cp); rc = stat(parent_dir, &parent); if (rc != 0) { crm_err("Directory '%s' does not exist: logging to '%s' is disabled", parent_dir, filename); free(filename_cp); return FALSE; } free(filename_cp); errno = 0; logfile = fopen(filename, "a"); if(logfile == NULL) { crm_err("%s (%d): Logging to '%s' as uid=%u, gid=%u is disabled", pcmk_strerror(errno), errno, filename, geteuid(), getegid()); return FALSE; } /* Check/Set permissions if we're root */ if (geteuid() == 0) { struct stat st; uid_t pcmk_uid = 0; gid_t pcmk_gid = 0; gboolean fix = FALSE; int logfd = fileno(logfile); rc = fstat(logfd, &st); if (rc < 0) { crm_perror(LOG_WARNING, "Cannot stat %s", filename); fclose(logfile); return FALSE; } if(crm_user_lookup(CRM_DAEMON_USER, &pcmk_uid, &pcmk_gid) == 0) { if (st.st_gid != pcmk_gid) { /* Wrong group */ fix = TRUE; } else if ((st.st_mode & S_IRWXG) != (S_IRGRP | S_IWGRP)) { /* Not read/writable by the correct group */ fix = TRUE; } } if (fix) { rc = fchown(logfd, pcmk_uid, pcmk_gid); if (rc < 0) { crm_warn("Cannot change the ownership of %s to user %s and gid %d", filename, CRM_DAEMON_USER, pcmk_gid); } rc = fchmod(logfd, S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP); if (rc < 0) { crm_warn("Cannot change the mode of %s to rw-rw----", filename); } fprintf(logfile, "Set r/w permissions for uid=%d, gid=%d on %s\n", pcmk_uid, pcmk_gid, filename); if (fflush(logfile) < 0 || fsync(logfd) < 0) { crm_err("Couldn't write out logfile: %s", filename); } } } /* Close and reopen with libqb */ fclose(logfile); fd = qb_log_file_open(filename); if (fd < 0) { crm_perror(LOG_WARNING, "Couldn't send additional logging to %s", filename); return FALSE; } if(is_default) { default_fd = fd; + // Some resource agents will log only if environment variable is set + if (daemon_option("logfile") == NULL) { + set_daemon_option("logfile", filename); + } + } else if(default_fd >= 0) { crm_notice("Switching to %s", filename); qb_log_ctl(default_fd, QB_LOG_CONF_ENABLED, QB_FALSE); } crm_notice("Additional logging available in %s", filename); qb_log_ctl(fd, QB_LOG_CONF_ENABLED, QB_TRUE); /* qb_log_ctl(fd, QB_LOG_CONF_FILE_SYNC, 1); Turn on synchronous writes */ #ifdef HAVE_qb_log_conf_QB_LOG_CONF_MAX_LINE_LEN // Longer than default, for logging long XML lines qb_log_ctl(QB_LOG_SYSLOG, QB_LOG_CONF_MAX_LINE_LEN, 800); #endif /* Enable callsites */ crm_update_callsites(); have_logfile = TRUE; return TRUE; } static int blackbox_trigger = 0; static volatile char *blackbox_file_prefix = NULL; #ifdef QB_FEATURE_LOG_HIRES_TIMESTAMPS typedef struct timespec *log_time_t; #else typedef time_t log_time_t; #endif static void blackbox_logger(int32_t t, struct qb_log_callsite *cs, log_time_t timestamp, const char *msg) { if(cs && cs->priority < LOG_ERR) { crm_write_blackbox(SIGTRAP, cs); /* Bypass the over-dumping logic */ } else { crm_write_blackbox(0, cs); } } static void crm_control_blackbox(int nsig, bool enable) { int lpc = 0; if (blackbox_file_prefix == NULL) { pid_t pid = getpid(); blackbox_file_prefix = crm_strdup_printf("%s/%s-%lu", CRM_BLACKBOX_DIR, crm_system_name, (unsigned long) pid); } if (enable && qb_log_ctl(QB_LOG_BLACKBOX, QB_LOG_CONF_STATE_GET, 0) != QB_LOG_STATE_ENABLED) { qb_log_ctl(QB_LOG_BLACKBOX, QB_LOG_CONF_SIZE, 5 * 1024 * 1024); /* Any size change drops existing entries */ qb_log_ctl(QB_LOG_BLACKBOX, QB_LOG_CONF_ENABLED, QB_TRUE); /* Setting the size seems to disable it */ /* Enable synchronous logging */ for (lpc = QB_LOG_BLACKBOX; lpc < QB_LOG_TARGET_MAX; lpc++) { qb_log_ctl(lpc, QB_LOG_CONF_FILE_SYNC, QB_TRUE); } crm_notice("Initiated blackbox recorder: %s", blackbox_file_prefix); /* Save to disk on abnormal termination */ crm_signal_handler(SIGSEGV, crm_trigger_blackbox); crm_signal_handler(SIGABRT, crm_trigger_blackbox); crm_signal_handler(SIGILL, crm_trigger_blackbox); crm_signal_handler(SIGBUS, crm_trigger_blackbox); crm_signal_handler(SIGFPE, crm_trigger_blackbox); crm_update_callsites(); blackbox_trigger = qb_log_custom_open(blackbox_logger, NULL, NULL, NULL); qb_log_ctl(blackbox_trigger, QB_LOG_CONF_ENABLED, QB_TRUE); crm_trace("Trigger: %d is %d %d", blackbox_trigger, qb_log_ctl(blackbox_trigger, QB_LOG_CONF_STATE_GET, 0), QB_LOG_STATE_ENABLED); crm_update_callsites(); } else if (!enable && qb_log_ctl(QB_LOG_BLACKBOX, QB_LOG_CONF_STATE_GET, 0) == QB_LOG_STATE_ENABLED) { qb_log_ctl(QB_LOG_BLACKBOX, QB_LOG_CONF_ENABLED, QB_FALSE); /* Disable synchronous logging again when the blackbox is disabled */ for (lpc = QB_LOG_BLACKBOX; lpc < QB_LOG_TARGET_MAX; lpc++) { qb_log_ctl(lpc, QB_LOG_CONF_FILE_SYNC, QB_FALSE); } } } void crm_enable_blackbox(int nsig) { crm_control_blackbox(nsig, TRUE); } void crm_disable_blackbox(int nsig) { crm_control_blackbox(nsig, FALSE); } /*! * \internal * \brief Write out a blackbox, if blackboxes are enabled * * \param[in] nsig Signal that was received * \param[in] cs libqb callsite * * \note This may be called via a true signal handler and so must be async-safe. * @TODO actually make this async-safe */ void crm_write_blackbox(int nsig, struct qb_log_callsite *cs) { static volatile int counter = 1; static volatile time_t last = 0; char buffer[NAME_MAX]; time_t now = time(NULL); if (blackbox_file_prefix == NULL) { return; } switch (nsig) { case 0: case SIGTRAP: /* The graceful case - such as assertion failure or user request */ if (nsig == 0 && now == last) { /* Prevent over-dumping */ return; } snprintf(buffer, NAME_MAX, "%s.%d", blackbox_file_prefix, counter++); if (nsig == SIGTRAP) { crm_notice("Blackbox dump requested, please see %s for contents", buffer); } else if (cs) { syslog(LOG_NOTICE, "Problem detected at %s:%d (%s), please see %s for additional details", cs->function, cs->lineno, cs->filename, buffer); } else { crm_notice("Problem detected, please see %s for additional details", buffer); } last = now; qb_log_blackbox_write_to_file(buffer); /* Flush the existing contents * A size change would also work */ qb_log_ctl(QB_LOG_BLACKBOX, QB_LOG_CONF_ENABLED, QB_FALSE); qb_log_ctl(QB_LOG_BLACKBOX, QB_LOG_CONF_ENABLED, QB_TRUE); break; default: /* Do as little as possible, just try to get what we have out * We logged the filename when the blackbox was enabled */ crm_signal_handler(nsig, SIG_DFL); qb_log_blackbox_write_to_file((const char *)blackbox_file_prefix); qb_log_ctl(QB_LOG_BLACKBOX, QB_LOG_CONF_ENABLED, QB_FALSE); raise(nsig); break; } } gboolean crm_log_cli_init(const char *entity) { return crm_log_init(entity, LOG_ERR, FALSE, FALSE, 0, NULL, TRUE); } static const char * crm_quark_to_string(uint32_t tag) { const char *text = g_quark_to_string(tag); if (text) { return text; } return ""; } static void crm_log_filter_source(int source, const char *trace_files, const char *trace_fns, const char *trace_fmts, const char *trace_tags, const char *trace_blackbox, struct qb_log_callsite *cs) { if (qb_log_ctl(source, QB_LOG_CONF_STATE_GET, 0) != QB_LOG_STATE_ENABLED) { return; } else if (cs->tags != crm_trace_nonlog && source == QB_LOG_BLACKBOX) { /* Blackbox gets everything if enabled */ qb_bit_set(cs->targets, source); } else if (source == blackbox_trigger && blackbox_trigger > 0) { /* Should this log message result in the blackbox being dumped */ if (cs->priority <= LOG_ERR) { qb_bit_set(cs->targets, source); } else if (trace_blackbox) { char *key = crm_strdup_printf("%s:%d", cs->function, cs->lineno); if (strstr(trace_blackbox, key) != NULL) { qb_bit_set(cs->targets, source); } free(key); } } else if (source == QB_LOG_SYSLOG) { /* No tracing to syslog */ if (cs->priority <= crm_log_priority && cs->priority <= crm_log_level) { qb_bit_set(cs->targets, source); } /* Log file tracing options... */ } else if (cs->priority <= crm_log_level) { qb_bit_set(cs->targets, source); } else if (trace_files && strstr(trace_files, cs->filename) != NULL) { qb_bit_set(cs->targets, source); } else if (trace_fns && strstr(trace_fns, cs->function) != NULL) { qb_bit_set(cs->targets, source); } else if (trace_fmts && strstr(trace_fmts, cs->format) != NULL) { qb_bit_set(cs->targets, source); } else if (trace_tags && cs->tags != 0 && cs->tags != crm_trace_nonlog && g_quark_to_string(cs->tags) != NULL) { qb_bit_set(cs->targets, source); } } static void crm_log_filter(struct qb_log_callsite *cs) { int lpc = 0; static int need_init = 1; static const char *trace_fns = NULL; static const char *trace_tags = NULL; static const char *trace_fmts = NULL; static const char *trace_files = NULL; static const char *trace_blackbox = NULL; if (need_init) { need_init = 0; trace_fns = getenv("PCMK_trace_functions"); trace_fmts = getenv("PCMK_trace_formats"); trace_tags = getenv("PCMK_trace_tags"); trace_files = getenv("PCMK_trace_files"); trace_blackbox = getenv("PCMK_trace_blackbox"); if (trace_tags != NULL) { uint32_t tag; char token[500]; const char *offset = NULL; const char *next = trace_tags; do { offset = next; next = strchrnul(offset, ','); snprintf(token, sizeof(token), "%.*s", (int)(next - offset), offset); tag = g_quark_from_string(token); crm_info("Created GQuark %u from token '%s' in '%s'", tag, token, trace_tags); if (next[0] != 0) { next++; } } while (next != NULL && next[0] != 0); } } cs->targets = 0; /* Reset then find targets to enable */ for (lpc = QB_LOG_SYSLOG; lpc < QB_LOG_TARGET_MAX; lpc++) { crm_log_filter_source(lpc, trace_files, trace_fns, trace_fmts, trace_tags, trace_blackbox, cs); } } gboolean crm_is_callsite_active(struct qb_log_callsite *cs, uint8_t level, uint32_t tags) { gboolean refilter = FALSE; if (cs == NULL) { return FALSE; } if (cs->priority != level) { cs->priority = level; refilter = TRUE; } if (cs->tags != tags) { cs->tags = tags; refilter = TRUE; } if (refilter) { crm_log_filter(cs); } if (cs->targets == 0) { return FALSE; } return TRUE; } void crm_update_callsites(void) { static gboolean log = TRUE; if (log) { log = FALSE; crm_debug ("Enabling callsites based on priority=%d, files=%s, functions=%s, formats=%s, tags=%s", crm_log_level, getenv("PCMK_trace_files"), getenv("PCMK_trace_functions"), getenv("PCMK_trace_formats"), getenv("PCMK_trace_tags")); } qb_log_filter_fn_set(crm_log_filter); } static gboolean crm_tracing_enabled(void) { if (crm_log_level >= LOG_TRACE) { return TRUE; } else if (getenv("PCMK_trace_files") || getenv("PCMK_trace_functions") || getenv("PCMK_trace_formats") || getenv("PCMK_trace_tags")) { return TRUE; } return FALSE; } static int crm_priority2int(const char *name) { struct syslog_names { const char *name; int priority; }; static struct syslog_names p_names[] = { {"emerg", LOG_EMERG}, {"alert", LOG_ALERT}, {"crit", LOG_CRIT}, {"error", LOG_ERR}, {"warning", LOG_WARNING}, {"notice", LOG_NOTICE}, {"info", LOG_INFO}, {"debug", LOG_DEBUG}, {NULL, -1} }; int lpc; for (lpc = 0; name != NULL && p_names[lpc].name != NULL; lpc++) { if (crm_str_eq(p_names[lpc].name, name, TRUE)) { return p_names[lpc].priority; } } return crm_log_priority; } static void crm_identity(const char *entity, int argc, char **argv) { if(crm_system_name != NULL) { /* Nothing to do */ } else if (entity) { free(crm_system_name); crm_system_name = strdup(entity); } else if (argc > 0 && argv != NULL) { char *mutable = strdup(argv[0]); char *modified = basename(mutable); if (strstr(modified, "lt-") == modified) { modified += 3; } free(crm_system_name); crm_system_name = strdup(modified); free(mutable); } else if (crm_system_name == NULL) { crm_system_name = strdup("Unknown"); } setenv("PCMK_service", crm_system_name, 1); } void crm_log_preinit(const char *entity, int argc, char **argv) { /* Configure libqb logging with nothing turned on */ int lpc = 0; int32_t qb_facility = 0; static bool have_logging = FALSE; if(have_logging == FALSE) { have_logging = TRUE; crm_xml_init(); /* Sets buffer allocation strategy */ if (crm_trace_nonlog == 0) { crm_trace_nonlog = g_quark_from_static_string("Pacemaker non-logging tracepoint"); } umask(S_IWGRP | S_IWOTH | S_IROTH); /* Redirect messages from glib functions to our handler */ glib_log_default = g_log_set_default_handler(crm_glib_handler, NULL); /* and for good measure... - this enum is a bit field (!) */ g_log_set_always_fatal((GLogLevelFlags) 0); /*value out of range */ /* Who do we log as */ crm_identity(entity, argc, argv); qb_facility = qb_log_facility2int("local0"); qb_log_init(crm_system_name, qb_facility, LOG_ERR); crm_log_level = LOG_CRIT; /* Nuke any syslog activity until it's asked for */ qb_log_ctl(QB_LOG_SYSLOG, QB_LOG_CONF_ENABLED, QB_FALSE); #ifdef HAVE_qb_log_conf_QB_LOG_CONF_MAX_LINE_LEN // Shorter than default, generous for what we *should* send to syslog qb_log_ctl(QB_LOG_SYSLOG, QB_LOG_CONF_MAX_LINE_LEN, 256); #endif /* Set format strings and disable threading * Pacemaker and threads do not mix well (due to the amount of forking) */ qb_log_tags_stringify_fn_set(crm_quark_to_string); for (lpc = QB_LOG_SYSLOG; lpc < QB_LOG_TARGET_MAX; lpc++) { qb_log_ctl(lpc, QB_LOG_CONF_THREADED, QB_FALSE); #ifdef HAVE_qb_log_conf_QB_LOG_CONF_ELLIPSIS // End truncated lines with '...' qb_log_ctl(lpc, QB_LOG_CONF_ELLIPSIS, QB_TRUE); #endif set_format_string(lpc, crm_system_name); } } } gboolean crm_log_init(const char *entity, uint8_t level, gboolean daemon, gboolean to_stderr, int argc, char **argv, gboolean quiet) { const char *syslog_priority = NULL; const char *logfile = daemon_option("logfile"); const char *facility = daemon_option("logfacility"); const char *f_copy = facility; crm_is_daemon = daemon; crm_log_preinit(entity, argc, argv); if(level > crm_log_level) { crm_log_level = level; } /* Should we log to syslog */ if (facility == NULL) { if(crm_is_daemon) { facility = "daemon"; } else { facility = "none"; } set_daemon_option("logfacility", facility); } if (safe_str_eq(facility, "none")) { quiet = TRUE; } else { qb_log_ctl(QB_LOG_SYSLOG, QB_LOG_CONF_FACILITY, qb_log_facility2int(facility)); } if (daemon_option_enabled(crm_system_name, "debug")) { /* Override the default setting */ crm_log_level = LOG_DEBUG; } /* What lower threshold do we have for sending to syslog */ syslog_priority = daemon_option("logpriority"); if(syslog_priority) { int priority = crm_priority2int(syslog_priority); crm_log_priority = priority; qb_log_filter_ctl(QB_LOG_SYSLOG, QB_LOG_FILTER_ADD, QB_LOG_FILTER_FILE, "*", priority); } else { qb_log_filter_ctl(QB_LOG_SYSLOG, QB_LOG_FILTER_ADD, QB_LOG_FILTER_FILE, "*", LOG_NOTICE); } // Log to syslog unless requested to be quiet if (!quiet) { qb_log_ctl(QB_LOG_SYSLOG, QB_LOG_CONF_ENABLED, QB_TRUE); } /* Should we log to stderr */ if (daemon_option_enabled(crm_system_name, "stderr")) { /* Override the default setting */ to_stderr = TRUE; } crm_enable_stderr(to_stderr); /* Should we log to a file */ if (safe_str_eq("none", logfile)) { /* No soup^Hlogs for you! */ } else if(crm_is_daemon) { // Daemons always get a log file, unless explicitly set to "none" crm_add_logfile(logfile); } else if(logfile) { crm_add_logfile(logfile); } if (crm_is_daemon && daemon_option_enabled(crm_system_name, "blackbox")) { crm_enable_blackbox(0); } /* Summary */ crm_trace("Quiet: %d, facility %s", quiet, f_copy); daemon_option("logfile"); daemon_option("logfacility"); crm_update_callsites(); /* Ok, now we can start logging... */ if (quiet == FALSE && crm_is_daemon == FALSE) { crm_log_args(argc, argv); } if (crm_is_daemon) { const char *user = getenv("USER"); if (user != NULL && safe_str_neq(user, "root") && safe_str_neq(user, CRM_DAEMON_USER)) { crm_trace("Not switching to corefile directory for %s", user); crm_is_daemon = FALSE; } } if (crm_is_daemon) { int user = getuid(); const char *base = CRM_CORE_DIR; struct passwd *pwent = getpwuid(user); if (pwent == NULL) { crm_perror(LOG_ERR, "Cannot get name for uid: %d", user); } else if (safe_str_neq(pwent->pw_name, "root") && safe_str_neq(pwent->pw_name, CRM_DAEMON_USER)) { crm_trace("Don't change active directory for regular user: %s", pwent->pw_name); } else if (chdir(base) < 0) { crm_perror(LOG_INFO, "Cannot change active directory to %s", base); } else { crm_info("Changed active directory to %s", base); #if 0 { char path[512]; snprintf(path, 512, "%s-%lu", crm_system_name, (unsigned long) getpid()); mkdir(path, 0750); chdir(path); crm_info("Changed active directory to %s/%s/%s", base, pwent->pw_name, path); } #endif } /* Original meanings from signal(7) * * Signal Value Action Comment * SIGTRAP 5 Core Trace/breakpoint trap * SIGUSR1 30,10,16 Term User-defined signal 1 * SIGUSR2 31,12,17 Term User-defined signal 2 * * Our usage is as similar as possible */ mainloop_add_signal(SIGUSR1, crm_enable_blackbox); mainloop_add_signal(SIGUSR2, crm_disable_blackbox); mainloop_add_signal(SIGTRAP, crm_trigger_blackbox); } return TRUE; } /* returns the old value */ unsigned int set_crm_log_level(unsigned int level) { unsigned int old = crm_log_level; crm_log_level = level; crm_update_callsites(); crm_trace("New log level: %d", level); return old; } void crm_enable_stderr(int enable) { if (enable && qb_log_ctl(QB_LOG_STDERR, QB_LOG_CONF_STATE_GET, 0) != QB_LOG_STATE_ENABLED) { qb_log_ctl(QB_LOG_STDERR, QB_LOG_CONF_ENABLED, QB_TRUE); crm_update_callsites(); } else if (enable == FALSE) { qb_log_ctl(QB_LOG_STDERR, QB_LOG_CONF_ENABLED, QB_FALSE); } } void crm_bump_log_level(int argc, char **argv) { static int args = TRUE; int level = crm_log_level; if (args && argc > 1) { crm_log_args(argc, argv); } if (qb_log_ctl(QB_LOG_STDERR, QB_LOG_CONF_STATE_GET, 0) == QB_LOG_STATE_ENABLED) { set_crm_log_level(level + 1); } /* Enable after potentially logging the argstring, not before */ crm_enable_stderr(TRUE); } unsigned int get_crm_log_level(void) { return crm_log_level; } #define ARGS_FMT "Invoked: %s" void crm_log_args(int argc, char **argv) { int lpc = 0; int len = 0; int existing_len = 0; int line = __LINE__; static int logged = 0; char *arg_string = NULL; if (argc == 0 || argv == NULL || logged) { return; } logged = 1; for (; lpc < argc; lpc++) { if (argv[lpc] == NULL) { break; } len = 2 + strlen(argv[lpc]); /* +1 space, +1 EOS */ arg_string = realloc_safe(arg_string, len + existing_len); existing_len += sprintf(arg_string + existing_len, "%s ", argv[lpc]); } qb_log_from_external_source(__func__, __FILE__, ARGS_FMT, LOG_NOTICE, line, 0, arg_string); free(arg_string); } void crm_log_output_fn(const char *file, const char *function, int line, int level, const char *prefix, const char *output) { const char *next = NULL; const char *offset = NULL; if (output == NULL) { level = LOG_TRACE; output = "-- empty --"; } next = output; do { offset = next; next = strchrnul(offset, '\n'); do_crm_log_alias(level, file, function, line, "%s [ %.*s ]", prefix, (int)(next - offset), offset); if (next[0] != 0) { next++; } } while (next != NULL && next[0] != 0); } diff --git a/lib/common/nvpair.c b/lib/common/nvpair.c index 1a1b77150d..53d34b2fe6 100644 --- a/lib/common/nvpair.c +++ b/lib/common/nvpair.c @@ -1,783 +1,784 @@ /* * Copyright 2004-2019 Andrew Beekhof * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include "crmcommon_private.h" /* * This file isolates handling of three types of name/value pairs: * * - pcmk_nvpair_t data type * - XML attributes () * - XML nvpair elements () */ // pcmk_nvpair_t handling /*! * \internal * \brief Allocate a new name/value pair * * \param[in] name New name (required) * \param[in] value New value * * \return Newly allocated name/value pair * \note The caller is responsible for freeing the result with * \c pcmk__free_nvpair(). */ static pcmk_nvpair_t * pcmk__new_nvpair(const char *name, const char *value) { pcmk_nvpair_t *nvpair = NULL; CRM_ASSERT(name); nvpair = calloc(1, sizeof(pcmk_nvpair_t)); CRM_ASSERT(nvpair); nvpair->name = strdup(name); nvpair->value = value? strdup(value) : NULL; return nvpair; } /*! * \internal * \brief Free a name/value pair * * \param[in] nvpair Name/value pair to free */ static void pcmk__free_nvpair(gpointer data) { if (data) { pcmk_nvpair_t *nvpair = data; free(nvpair->name); free(nvpair->value); free(nvpair); } } /*! * \brief Prepend a name/value pair to a list * * \param[in,out] nvpairs List to modify * \param[in] name New entry's name * \param[in] value New entry's value * * \return New head of list * \note The caller is responsible for freeing the list with * \c pcmk_free_nvpairs(). */ GSList * pcmk_prepend_nvpair(GSList *nvpairs, const char *name, const char *value) { return g_slist_prepend(nvpairs, pcmk__new_nvpair(name, value)); } /*! * \brief Free a list of name/value pairs * * \param[in] list List to free */ void pcmk_free_nvpairs(GSList *nvpairs) { g_slist_free_full(nvpairs, pcmk__free_nvpair); } /*! * \internal * \brief Compare two name/value pairs * * \param[in] a First name/value pair to compare * \param[in] b Second name/value pair to compare * * \return 0 if a == b, 1 if a > b, -1 if a < b */ static gint pcmk__compare_nvpair(gconstpointer a, gconstpointer b) { int rc = 0; const pcmk_nvpair_t *pair_a = a; const pcmk_nvpair_t *pair_b = b; CRM_ASSERT(a != NULL); CRM_ASSERT(pair_a->name != NULL); CRM_ASSERT(b != NULL); CRM_ASSERT(pair_b->name != NULL); rc = strcmp(pair_a->name, pair_b->name); if (rc < 0) { return -1; } else if (rc > 0) { return 1; } return 0; } /*! * \brief Sort a list of name/value pairs * * \param[in,out] list List to sort * * \return New head of list */ GSList * pcmk_sort_nvpairs(GSList *list) { return g_slist_sort(list, pcmk__compare_nvpair); } /*! * \brief Create a list of name/value pairs from an XML node's attributes * * \param[in] XML to parse * * \return New list of name/value pairs * \note It is the caller's responsibility to free the list with * \c pcmk_free_nvpairs(). */ GSList * pcmk_xml_attrs2nvpairs(xmlNode *xml) { GSList *result = NULL; for (xmlAttrPtr iter = pcmk__first_xml_attr(xml); iter != NULL; iter = iter->next) { result = pcmk_prepend_nvpair(result, (const char *) iter->name, (const char *) pcmk__xml_attr_value(iter)); } return result; } /*! * \internal * \brief Add an XML attribute corresponding to a name/value pair * * Suitable for glib list iterators, this function adds a NAME=VALUE * XML attribute based on a given name/value pair. * * \param[in] data Name/value pair * \param[out] user_data XML node to add attributes to */ static void pcmk__nvpair_add_xml_attr(gpointer data, gpointer user_data) { pcmk_nvpair_t *pair = data; xmlNode *parent = user_data; crm_xml_add(parent, pair->name, pair->value); } /*! * \brief Add XML attributes based on a list of name/value pairs * * \param[in] list List of name/value pairs * \param[in,out] xml XML node to add attributes to */ void pcmk_nvpairs2xml_attrs(GSList *list, xmlNode *xml) { g_slist_foreach(list, pcmk__nvpair_add_xml_attr, xml); } // convenience function for name=value strings /*! * \brief Extract the name and value from an input string formatted as "name=value". * If unable to extract them, they are returned as NULL. * * \param[in] input The input string, likely from the command line * \param[out] name Everything before the first '=' in the input string * \param[out] value Everything after the first '=' in the input string * * \return 2 if both name and value could be extracted, 1 if only one could, and * and error code otherwise */ int pcmk_scan_nvpair(const char *input, char **name, char **value) { #ifdef SSCANF_HAS_M *name = NULL; *value = NULL; if (sscanf(input, "%m[^=]=%m[^\n]", name, value) <= 0) { return -pcmk_err_bad_nvpair; } #else char *sep = NULL; *name = NULL; *value = NULL; sep = strstr(optarg, "="); if (sep == NULL) { return -pcmk_err_bad_nvpair; } *name = strndup(input, sep-input); if (*name == NULL) { return -ENOMEM; } /* If the last char in optarg is =, the user gave no * value for the option. Leave it as NULL. */ if (*(sep+1) != '\0') { *value = strdup(sep+1); if (*value == NULL) { return -ENOMEM; } } #endif if (*name != NULL && *value != NULL) { return 2; } else if (*name != NULL || *value != NULL) { return 1; } else { return -pcmk_err_bad_nvpair; } } // XML attribute handling /*! * \brief Create an XML attribute with specified name and value * * \param[in,out] node XML node to modify * \param[in] name Attribute name to set * \param[in] value Attribute value to set * * \return New value on success, \c NULL otherwise * \note This does nothing if node, name, or value are \c NULL or empty. */ const char * crm_xml_add(xmlNode *node, const char *name, const char *value) { bool dirty = FALSE; xmlAttr *attr = NULL; CRM_CHECK(node != NULL, return NULL); CRM_CHECK(name != NULL, return NULL); if (value == NULL) { return NULL; } #if XML_PARANOIA_CHECKS { const char *old_value = NULL; old_value = crm_element_value(node, name); /* Could be re-setting the same value */ CRM_CHECK(old_value != value, crm_err("Cannot reset %s with crm_xml_add(%s)", name, value); return value); } #endif if (pcmk__tracking_xml_changes(node, FALSE)) { const char *old = crm_element_value(node, name); if (old == NULL || value == NULL || strcmp(old, value) != 0) { dirty = TRUE; } } if (dirty && (pcmk__check_acl(node, name, xpf_acl_create) == FALSE)) { crm_trace("Cannot add %s=%s to %s", name, value, node->name); return NULL; } attr = xmlSetProp(node, (pcmkXmlStr) name, (pcmkXmlStr) value); if (dirty) { pcmk__mark_xml_attr_dirty(attr); } CRM_CHECK(attr && attr->children && attr->children->content, return NULL); return (char *)attr->children->content; } /*! * \brief Replace an XML attribute with specified name and (possibly NULL) value * * \param[in,out] node XML node to modify * \param[in] name Attribute name to set * \param[in] value Attribute value to set * * \return New value on success, \c NULL otherwise * \note This does nothing if node or name is \c NULL or empty. */ const char * crm_xml_replace(xmlNode *node, const char *name, const char *value) { bool dirty = FALSE; xmlAttr *attr = NULL; const char *old_value = NULL; CRM_CHECK(node != NULL, return NULL); CRM_CHECK(name != NULL && name[0] != 0, return NULL); old_value = crm_element_value(node, name); /* Could be re-setting the same value */ CRM_CHECK(old_value != value, return value); if (pcmk__check_acl(node, name, xpf_acl_write) == FALSE) { /* Create a fake object linked to doc->_private instead? */ crm_trace("Cannot replace %s=%s to %s", name, value, node->name); return NULL; } else if (old_value && !value) { xml_remove_prop(node, name); return NULL; } if (pcmk__tracking_xml_changes(node, FALSE)) { if (!old_value || !value || !strcmp(old_value, value)) { dirty = TRUE; } } attr = xmlSetProp(node, (pcmkXmlStr) name, (pcmkXmlStr) value); if (dirty) { pcmk__mark_xml_attr_dirty(attr); } CRM_CHECK(attr && attr->children && attr->children->content, return NULL); return (char *) attr->children->content; } /*! * \brief Create an XML attribute with specified name and integer value * * This is like \c crm_xml_add() but taking an integer value. * * \param[in,out] node XML node to modify * \param[in] name Attribute name to set * \param[in] value Attribute value to set * * \return New value as string on success, \c NULL otherwise * \note This does nothing if node or name are \c NULL or empty. */ const char * crm_xml_add_int(xmlNode *node, const char *name, int value) { char *number = crm_itoa(value); const char *added = crm_xml_add(node, name, number); free(number); return added; } /*! * \brief Create an XML attribute with specified name and unsigned value * * This is like \c crm_xml_add() but taking a guint value. * * \param[in,out] node XML node to modify * \param[in] name Attribute name to set * \param[in] ms Attribute value to set * * \return New value as string on success, \c NULL otherwise * \note This does nothing if node or name are \c NULL or empty. */ const char * crm_xml_add_ms(xmlNode *node, const char *name, guint ms) { char *number = crm_strdup_printf("%u", ms); const char *added = crm_xml_add(node, name, number); free(number); return added; } /*! * \brief Retrieve the value of an XML attribute * * \param[in] data XML node to check * \param[in] name Attribute name to check * * \return Value of specified attribute (may be \c NULL) */ const char * crm_element_value(const xmlNode *data, const char *name) { xmlAttr *attr = NULL; if (data == NULL) { crm_err("Couldn't find %s in NULL", name ? name : ""); CRM_LOG_ASSERT(data != NULL); return NULL; } else if (name == NULL) { crm_err("Couldn't find NULL in %s", crm_element_name(data)); return NULL; } /* The first argument to xmlHasProp() has always been const, * but libxml2 <2.9.2 didn't declare that, so cast it */ attr = xmlHasProp((xmlNode *) data, (pcmkXmlStr) name); if (!attr || !attr->children) { return NULL; } return (const char *) attr->children->content; } /*! * \brief Retrieve the integer value of an XML attribute * - * This is like \c crm_element_value() but returning the value as an integer. + * This is like \c crm_element_value() but getting the value as an integer. * * \param[in] data XML node to check * \param[in] name Attribute name to check + * \param[in] dest Where to store element value * - * \return Integer value of specified attribute on success, -1 otherwise + * \return 0 on success, -1 otherwise */ int crm_element_value_int(const xmlNode *data, const char *name, int *dest) { const char *value = NULL; CRM_CHECK(dest != NULL, return -1); value = crm_element_value(data, name); if (value) { *dest = crm_int_helper(value, NULL); return 0; } return -1; } /*! * \brief Retrieve the millisecond value of an XML attribute * * This is like \c crm_element_value() but returning the value as a guint. * * \param[in] data XML node to check * \param[in] name Attribute name to check * \param[out] dest Where to store attribute value * * \return \c pcmk_ok on success, -1 otherwise */ int crm_element_value_ms(const xmlNode *data, const char *name, guint *dest) { const char *value = NULL; CRM_CHECK(dest != NULL, return -1); value = crm_element_value(data, name); *dest = crm_parse_ms(value); return errno? -1 : 0; } /*! * \brief Retrieve the value of XML second/microsecond attributes as time * * This is like \c crm_element_value() but returning value as a struct timeval. * * \param[in] xml XML to parse * \param[in] name_sec Name of XML attribute for seconds * \param[in] name_usec Name of XML attribute for microseconds * \param[out] dest Where to store result * * \return \c pcmk_ok on success, -errno on error * \note Values default to 0 if XML or XML attribute does not exist */ int crm_element_value_timeval(const xmlNode *xml, const char *name_sec, const char *name_usec, struct timeval *dest) { const char *value_s = NULL; long long value_i = 0; CRM_CHECK(dest != NULL, return -EINVAL); dest->tv_sec = 0; dest->tv_usec = 0; if (xml == NULL) { return 0; } // Parse seconds value_s = crm_element_value(xml, name_sec); if (value_s) { value_i = crm_parse_ll(value_s, NULL); if (errno) { return -errno; } dest->tv_sec = (time_t) value_i; } // Parse microseconds value_s = crm_element_value(xml, name_usec); if (value_s) { value_i = crm_parse_ll(value_s, NULL); if (errno) { return -errno; } dest->tv_usec = (suseconds_t) value_i; } return 0; } /*! * \brief Retrieve a copy of the value of an XML attribute * * This is like \c crm_element_value() but allocating new memory for the result. * * \param[in] data XML node to check * \param[in] name Attribute name to check * * \return Value of specified attribute (may be \c NULL) * \note The caller is responsible for freeing the result. */ char * crm_element_value_copy(const xmlNode *data, const char *name) { char *value_copy = NULL; const char *value = crm_element_value(data, name); if (value != NULL) { value_copy = strdup(value); } return value_copy; } /*! * \brief Add hash table entry to XML as (possibly legacy) name/value * * Suitable for \c g_hash_table_foreach(), this function takes a hash table key * and value, with an XML node passed as user data, and adds an XML attribute * with the specified name and value if it does not already exist. If the key * name starts with a digit, this will instead add a \ child to the XML (for legacy compatibility with heartbeat). * * \param[in] key Key of hash table entry * \param[in] value Value of hash table entry * \param[in] user_data XML node */ void hash2smartfield(gpointer key, gpointer value, gpointer user_data) { const char *name = key; const char *s_value = value; xmlNode *xml_node = user_data; if (isdigit(name[0])) { xmlNode *tmp = create_xml_node(xml_node, XML_TAG_PARAM); crm_xml_add(tmp, XML_NVPAIR_ATTR_NAME, name); crm_xml_add(tmp, XML_NVPAIR_ATTR_VALUE, s_value); } else if (crm_element_value(xml_node, name) == NULL) { crm_xml_add(xml_node, name, s_value); crm_trace("dumped: %s=%s", name, s_value); } else { crm_trace("duplicate: %s=%s", name, s_value); } } /*! * \brief Set XML attribute based on hash table entry * * Suitable for \c g_hash_table_foreach(), this function takes a hash table key * and value, with an XML node passed as user data, and adds an XML attribute * with the specified name and value if it does not already exist. * * \param[in] key Key of hash table entry * \param[in] value Value of hash table entry * \param[in] user_data XML node */ void hash2field(gpointer key, gpointer value, gpointer user_data) { const char *name = key; const char *s_value = value; xmlNode *xml_node = user_data; if (crm_element_value(xml_node, name) == NULL) { crm_xml_add(xml_node, name, s_value); } else { crm_trace("duplicate: %s=%s", name, s_value); } } /*! * \brief Set XML attribute based on hash table entry, as meta-attribute name * * Suitable for \c g_hash_table_foreach(), this function takes a hash table key * and value, with an XML node passed as user data, and adds an XML attribute * with the meta-attribute version of the specified name and value if it does * not already exist and if the name does not appear to be cluster-internal. * * \param[in] key Key of hash table entry * \param[in] value Value of hash table entry * \param[in] user_data XML node */ void hash2metafield(gpointer key, gpointer value, gpointer user_data) { char *crm_name = NULL; if (key == NULL || value == NULL) { return; } /* Filter out cluster-generated attributes that contain a '#' or ':' * (like fail-count and last-failure). */ for (crm_name = key; *crm_name; ++crm_name) { if ((*crm_name == '#') || (*crm_name == ':')) { return; } } crm_name = crm_meta_name(key); hash2field(crm_name, value, user_data); free(crm_name); } // nvpair handling /*! * \brief Create an XML name/value pair * * \param[in] parent If not \c NULL, make new XML node a child of this one * \param[in] id If not \c NULL, use this as ID (otherwise auto-generate) * \param[in] name Name to use * \param[in] value Value to use * * \return New XML object on success, \c NULL otherwise */ xmlNode * crm_create_nvpair_xml(xmlNode *parent, const char *id, const char *name, const char *value) { xmlNode *nvp; /* id can be NULL so we auto-generate one, and name can be NULL if this * will be used to delete a name/value pair by ID, but both can't be NULL */ CRM_CHECK(id || name, return NULL); nvp = create_xml_node(parent, XML_CIB_TAG_NVPAIR); CRM_CHECK(nvp, return NULL); if (id) { crm_xml_add(nvp, XML_ATTR_ID, id); } else { const char *parent_id = ID(parent); crm_xml_set_id(nvp, "%s-%s", (parent_id? parent_id : XML_CIB_TAG_NVPAIR), name); } crm_xml_add(nvp, XML_NVPAIR_ATTR_NAME, name); crm_xml_add(nvp, XML_NVPAIR_ATTR_VALUE, value); return nvp; } /*! * \brief Add XML nvpair element based on hash table entry * * Suitable for \c g_hash_table_foreach(), this function takes a hash table key * and value, with an XML node passed as the user data, and adds an \c nvpair * XML element with the specified name and value. * * \param[in] key Key of hash table entry * \param[in] value Value of hash table entry * \param[in] user_data XML node */ void hash2nvpair(gpointer key, gpointer value, gpointer user_data) { const char *name = key; const char *s_value = value; xmlNode *xml_node = user_data; crm_create_nvpair_xml(xml_node, name, name, s_value); crm_trace("dumped: name=%s value=%s", name, s_value); } /*! * \brief Retrieve XML attributes as a hash table * * Given an XML element, this will look for any \ element child, * creating a hash table of (newly allocated string) name/value pairs taken * first from the attributes element's NAME=VALUE XML attributes, and then * from any \ children of attributes. * * \param[in] XML node to parse * * \return Hash table with name/value pairs * \note It is the caller's responsibility to free the result using * \c g_hash_table_destroy(). */ GHashTable * xml2list(xmlNode *parent) { xmlNode *child = NULL; xmlAttrPtr pIter = NULL; xmlNode *nvpair_list = NULL; GHashTable *nvpair_hash = crm_str_table_new(); CRM_CHECK(parent != NULL, return nvpair_hash); nvpair_list = find_xml_node(parent, XML_TAG_ATTRS, FALSE); if (nvpair_list == NULL) { crm_trace("No attributes in %s", crm_element_name(parent)); crm_log_xml_trace(parent, "No attributes for resource op"); } crm_log_xml_trace(nvpair_list, "Unpacking"); for (pIter = pcmk__first_xml_attr(nvpair_list); pIter != NULL; pIter = pIter->next) { const char *p_name = (const char *)pIter->name; const char *p_value = pcmk__xml_attr_value(pIter); crm_trace("Added %s=%s", p_name, p_value); g_hash_table_insert(nvpair_hash, strdup(p_name), strdup(p_value)); } for (child = __xml_first_child(nvpair_list); child != NULL; child = __xml_next(child)) { if (strcmp((const char *)child->name, XML_TAG_PARAM) == 0) { const char *key = crm_element_value(child, XML_NVPAIR_ATTR_NAME); const char *value = crm_element_value(child, XML_NVPAIR_ATTR_VALUE); crm_trace("Added %s=%s", key, value); if (key != NULL && value != NULL) { g_hash_table_insert(nvpair_hash, strdup(key), strdup(value)); } } } return nvpair_hash; } diff --git a/lib/common/output_text.c b/lib/common/output_text.c index 76ae4ca4e1..06a51d10a2 100644 --- a/lib/common/output_text.c +++ b/lib/common/output_text.c @@ -1,223 +1,223 @@ /* * Copyright 2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include /* Disabled for the moment, but we can enable it (or remove it entirely) * when we make a decision on whether this is preferred output. */ #define FANCY_TEXT_OUTPUT 0 typedef struct text_list_data_s { unsigned int len; char *singular_noun; char *plural_noun; } text_list_data_t; typedef struct text_private_s { GQueue *parent_q; } text_private_t; static void text_free_priv(pcmk__output_t *out) { text_private_t *priv = out->priv; if (priv == NULL) { return; } g_queue_free(priv->parent_q); free(priv); } static bool text_init(pcmk__output_t *out) { text_private_t *priv = NULL; /* If text_init was previously called on this output struct, just return. */ if (out->priv != NULL) { return true; } else { out->priv = calloc(1, sizeof(text_private_t)); if (out->priv == NULL) { return false; } priv = out->priv; } priv->parent_q = g_queue_new(); return true; } static void text_finish(pcmk__output_t *out, crm_exit_t exit_status) { /* This function intentionally left blank */ } static void text_reset(pcmk__output_t *out) { CRM_ASSERT(out->priv != NULL); text_free_priv(out); text_init(out); } static void text_subprocess_output(pcmk__output_t *out, int exit_status, const char *proc_stdout, const char *proc_stderr) { if (proc_stdout != NULL) { fprintf(out->dest, "%s\n", proc_stdout); } if (proc_stderr != NULL) { fprintf(out->dest, "%s\n", proc_stderr); } } G_GNUC_PRINTF(2, 3) static void text_info(pcmk__output_t *out, const char *format, ...) { va_list ap; int len = 0; va_start(ap, format); /* Informational output does not get indented, to separate it from other * potentially indented list output. */ len = vfprintf(out->dest, format, ap); CRM_ASSERT(len > 0); va_end(ap); /* Add a newline. */ fprintf(out->dest, "\n"); } static void text_output_xml(pcmk__output_t *out, const char *name, const char *buf) { text_private_t *priv = out->priv; CRM_ASSERT(priv != NULL); pcmk__indented_printf(out, "%s", buf); } static void text_begin_list(pcmk__output_t *out, const char *name, const char *singular_noun, const char *plural_noun) { text_private_t *priv = out->priv; text_list_data_t *new_list = NULL; CRM_ASSERT(priv != NULL); #if FANCY_TEXT_OUTPUT > 0 pcmk__indented_printf(out, "%s:\n", name); #endif new_list = calloc(1, sizeof(text_list_data_t)); new_list->len = 0; - new_list->singular_noun = strdup(singular_noun); - new_list->plural_noun = strdup(plural_noun); + new_list->singular_noun = singular_noun == NULL ? NULL : strdup(singular_noun); + new_list->plural_noun = plural_noun == NULL ? NULL : strdup(plural_noun); g_queue_push_tail(priv->parent_q, new_list); } static void text_list_item(pcmk__output_t *out, const char *id, const char *content) { text_private_t *priv = out->priv; CRM_ASSERT(priv != NULL); #if FANCY_TEXT_OUTPUT > 0 if (id != NULL) { pcmk__indented_printf(out, "* %s: %s\n", id, content); } else { pcmk__indented_printf(out, "* %s\n", content); } #else fprintf(out->dest, "%s\n", content); #endif ((text_list_data_t *) g_queue_peek_tail(priv->parent_q))->len++; } static void text_end_list(pcmk__output_t *out) { text_private_t *priv = out->priv; text_list_data_t *node = NULL; CRM_ASSERT(priv != NULL); node = g_queue_pop_tail(priv->parent_q); if (node->singular_noun != NULL && node->plural_noun != NULL) { if (node->len == 1) { pcmk__indented_printf(out, "%d %s found\n", node->len, node->singular_noun); } else { pcmk__indented_printf(out, "%d %s found\n", node->len, node->plural_noun); } } free(node); } pcmk__output_t * pcmk__mk_text_output(char **argv) { pcmk__output_t *retval = calloc(1, sizeof(pcmk__output_t)); if (retval == NULL) { return NULL; } retval->request = g_strjoinv(" ", argv); retval->supports_quiet = true; retval->init = text_init; retval->free_priv = text_free_priv; retval->finish = text_finish; retval->reset = text_reset; retval->register_message = pcmk__register_message; retval->message = pcmk__call_message; retval->subprocess_output = text_subprocess_output; retval->info = text_info; retval->output_xml = text_output_xml; retval->begin_list = text_begin_list; retval->list_item = text_list_item; retval->end_list = text_end_list; return retval; } G_GNUC_PRINTF(2, 3) void pcmk__indented_printf(pcmk__output_t *out, const char *format, ...) { va_list ap; int len = 0; #if FANCY_TEXT_OUTPUT > 0 int level = 0; text_private_t *priv = out->priv; CRM_ASSERT(priv != NULL); level = g_queue_get_length(priv->parent_q); for (int i = 0; i < level; i++) { putc('\t', out->dest); } #endif va_start(ap, format); len = vfprintf(out->dest, format, ap); CRM_ASSERT(len > 0); va_end(ap); } diff --git a/lib/common/pid.c b/lib/common/pid.c index 4fbf2dd261..ccee03fc1e 100644 --- a/lib/common/pid.c +++ b/lib/common/pid.c @@ -1,201 +1,201 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #ifndef _GNU_SOURCE # define _GNU_SOURCE #endif #include #include #include #include int crm_pid_active(long pid, const char *daemon) { static int last_asked_pid = 0; /* log spam prevention */ #if SUPPORT_PROCFS static int have_proc_pid = 0; #else static int have_proc_pid = -1; #endif int rc = 0; if (have_proc_pid == 0) { /* evaluation of /proc/PID/exe applicability via self-introspection */ char proc_path[PATH_MAX], exe_path[PATH_MAX]; snprintf(proc_path, sizeof(proc_path), "/proc/%lu/exe", (long unsigned int) getpid()); have_proc_pid = 1; if (readlink(proc_path, exe_path, sizeof(exe_path) - 1) < 0) { have_proc_pid = -1; } } if (pid <= 0) { return -1; } else if ((rc = kill(pid, 0)) < 0 && errno == ESRCH) { return 0; /* no such PID detected */ - } else if (rc < 0 && have_proc_pid == -1) { + } else if (rc < 0 && (daemon == NULL || have_proc_pid == -1)) { if (last_asked_pid != pid) { crm_info("Cannot examine PID %ld: %s", pid, strerror(errno)); last_asked_pid = pid; } return -2; /* errno != ESRCH */ } else if (rc == 0 && (daemon == NULL || have_proc_pid == -1)) { return 1; /* kill as the only indicator, cannot double check */ } else if (daemon != NULL) { /* make sure PID hasn't been reused by another process XXX: might still be just a zombie, which could confuse decisions */ bool checked_through_kill = (rc == 0); char proc_path[PATH_MAX], exe_path[PATH_MAX], myexe_path[PATH_MAX]; snprintf(proc_path, sizeof(proc_path), "/proc/%ld/exe", pid); rc = readlink(proc_path, exe_path, sizeof(exe_path) - 1); if ((rc < 0) && (errno == EACCES)) { if (last_asked_pid != pid) { crm_info("Could not read from %s: %s", proc_path, strerror(errno)); last_asked_pid = pid; } return checked_through_kill ? 1 : -2; } else if (rc < 0) { if (last_asked_pid != pid) { crm_err("Could not read from %s: %s (%d)", proc_path, strerror(errno), errno); last_asked_pid = pid; } return 0; /* most likely errno == ENOENT */ } exe_path[rc] = '\0'; if (daemon[0] != '/') { rc = snprintf(myexe_path, sizeof(myexe_path), CRM_DAEMON_DIR"/%s", daemon); } else { rc = snprintf(myexe_path, sizeof(myexe_path), "%s", daemon); } if (rc > 0 && rc < sizeof(myexe_path) && !strcmp(exe_path, myexe_path)) { return 1; } } return 0; } #define LOCKSTRLEN 11 long crm_read_pidfile(const char *filename) { int fd; struct stat sbuf; long pid = -ENOENT; char buf[LOCKSTRLEN + 1]; fd = open(filename, O_RDONLY); if (fd < 0) { goto bail; } if ((fstat(fd, &sbuf) >= 0) && (sbuf.st_size < LOCKSTRLEN)) { sleep(2); /* if someone was about to create one, * give'm a sec to do so */ } if (read(fd, buf, sizeof(buf)) < 1) { goto bail; } if (sscanf(buf, "%ld", &pid) > 0) { if (pid <= 0) { pid = -ESRCH; } else { crm_trace("Got pid %lu from %s\n", pid, filename); } } bail: if (fd >= 0) { close(fd); } return pid; } long crm_pidfile_inuse(const char *filename, long mypid, const char *daemon) { long pid = crm_read_pidfile(filename); if (pid < 2) { // Invalid pid pid = -ENOENT; unlink(filename); } else if (mypid && (pid == mypid)) { // In use by us pid = pcmk_ok; } else if (crm_pid_active(pid, daemon) == FALSE) { // Contains a stale value unlink(filename); pid = -ENOENT; } else if (mypid && (pid != mypid)) { // Locked by existing process pid = -EEXIST; } return pid; } int crm_lock_pidfile(const char *filename, const char *name) { long mypid = 0; int fd = 0; int rc = 0; char buf[LOCKSTRLEN + 2]; mypid = (unsigned long) getpid(); rc = crm_pidfile_inuse(filename, 0, name); if (rc == -ENOENT) { // Exists, but the process is not active } else if (rc != pcmk_ok) { // Locked by existing process return rc; } fd = open(filename, O_CREAT | O_WRONLY | O_EXCL, 0644); if (fd < 0) { return -errno; } snprintf(buf, sizeof(buf), "%*ld\n", LOCKSTRLEN - 1, mypid); rc = write(fd, buf, LOCKSTRLEN); close(fd); if (rc != LOCKSTRLEN) { crm_perror(LOG_ERR, "Incomplete write to %s", filename); return -errno; } return crm_pidfile_inuse(filename, mypid, name); } diff --git a/lib/fencing/st_client.c b/lib/fencing/st_client.c index f4b7addc2a..270ef8d196 100644 --- a/lib/fencing/st_client.c +++ b/lib/fencing/st_client.c @@ -1,2267 +1,2281 @@ /* * Copyright 2004-2018 Andrew Beekhof * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #if SUPPORT_CIBSECRETS # include #endif CRM_TRACE_INIT_DATA(stonith); struct stonith_action_s { /*! user defined data */ char *agent; char *action; char *victim; GHashTable *args; int timeout; int async; void *userdata; void (*done_cb) (GPid pid, gint status, const char *output, gpointer user_data); + void (*fork_cb) (GPid pid, gpointer user_data); svc_action_t *svc_action; /*! internal timing information */ time_t initial_start_time; int tries; int remaining_timeout; int max_retries; /* device output data */ GPid pid; int rc; char *output; char *error; }; typedef struct stonith_private_s { char *token; crm_ipc_t *ipc; mainloop_io_t *source; GHashTable *stonith_op_callback_table; GList *notify_list; void (*op_callback) (stonith_t * st, stonith_callback_data_t * data); } stonith_private_t; typedef struct stonith_notify_client_s { const char *event; const char *obj_id; /* implement one day */ const char *obj_type; /* implement one day */ void (*notify) (stonith_t * st, stonith_event_t * e); } stonith_notify_client_t; typedef struct stonith_callback_client_s { void (*callback) (stonith_t * st, stonith_callback_data_t * data); const char *id; void *user_data; gboolean only_success; gboolean allow_timeout_updates; struct timer_rec_s *timer; } stonith_callback_client_t; struct notify_blob_s { stonith_t *stonith; xmlNode *xml; }; struct timer_rec_s { int call_id; int timeout; guint ref; stonith_t *stonith; }; typedef int (*stonith_op_t) (const char *, int, const char *, xmlNode *, xmlNode *, xmlNode *, xmlNode **, xmlNode **); bool stonith_dispatch(stonith_t * st); xmlNode *stonith_create_op(int call_id, const char *token, const char *op, xmlNode * data, int call_options); int stonith_send_command(stonith_t * stonith, const char *op, xmlNode * data, xmlNode ** output_data, int call_options, int timeout); static void stonith_connection_destroy(gpointer user_data); static void stonith_send_notification(gpointer data, gpointer user_data); static int internal_stonith_action_execute(stonith_action_t * action); static void log_action(stonith_action_t *action, pid_t pid); /*! * \brief Get agent namespace by name * * \param[in] namespace_s Name of namespace as string * * \return Namespace as enum value */ enum stonith_namespace stonith_text2namespace(const char *namespace_s) { if ((namespace_s == NULL) || !strcmp(namespace_s, "any")) { return st_namespace_any; } else if (!strcmp(namespace_s, "redhat") || !strcmp(namespace_s, "stonith-ng")) { return st_namespace_rhcs; } else if (!strcmp(namespace_s, "internal")) { return st_namespace_internal; } else if (!strcmp(namespace_s, "heartbeat")) { return st_namespace_lha; } return st_namespace_invalid; } /*! * \brief Get agent namespace name * * \param[in] namespace Namespace as enum value * * \return Namespace name as string */ const char * stonith_namespace2text(enum stonith_namespace st_namespace) { switch (st_namespace) { case st_namespace_any: return "any"; case st_namespace_rhcs: return "stonith-ng"; case st_namespace_internal: return "internal"; case st_namespace_lha: return "heartbeat"; default: break; } return "unsupported"; } /*! * \brief Determine namespace of a fence agent * * \param[in] agent Fence agent type * \param[in] namespace_s Name of agent namespace as string, if known * * \return Namespace of specified agent, as enum value */ enum stonith_namespace stonith_get_namespace(const char *agent, const char *namespace_s) { if (safe_str_eq(namespace_s, "internal")) { return st_namespace_internal; } if (stonith__agent_is_rhcs(agent)) { return st_namespace_rhcs; } #if HAVE_STONITH_STONITH_H if (stonith__agent_is_lha(agent)) { return st_namespace_lha; } #endif crm_err("Unknown fence agent: %s", agent); return st_namespace_invalid; } static void log_action(stonith_action_t *action, pid_t pid) { if (action->output) { /* Logging the whole string confuses syslog when the string is xml */ char *prefix = crm_strdup_printf("%s[%d] stdout:", action->agent, pid); crm_log_output(LOG_TRACE, prefix, action->output); free(prefix); } if (action->error) { /* Logging the whole string confuses syslog when the string is xml */ char *prefix = crm_strdup_printf("%s[%d] stderr:", action->agent, pid); crm_log_output(LOG_WARNING, prefix, action->error); free(prefix); } } static void stonith_connection_destroy(gpointer user_data) { stonith_t *stonith = user_data; stonith_private_t *native = NULL; struct notify_blob_s blob; crm_trace("Sending destroyed notification"); blob.stonith = stonith; blob.xml = create_xml_node(NULL, "notify"); native = stonith->st_private; native->ipc = NULL; native->source = NULL; stonith->state = stonith_disconnected; crm_xml_add(blob.xml, F_TYPE, T_STONITH_NOTIFY); crm_xml_add(blob.xml, F_SUBTYPE, T_STONITH_NOTIFY_DISCONNECT); g_list_foreach(native->notify_list, stonith_send_notification, &blob); free_xml(blob.xml); } xmlNode * create_device_registration_xml(const char *id, enum stonith_namespace namespace, const char *agent, stonith_key_value_t *params, const char *rsc_provides) { xmlNode *data = create_xml_node(NULL, F_STONITH_DEVICE); xmlNode *args = create_xml_node(data, XML_TAG_ATTRS); #if HAVE_STONITH_STONITH_H if (namespace == st_namespace_any) { namespace = stonith_get_namespace(agent, NULL); } if (namespace == st_namespace_lha) { hash2field((gpointer) "plugin", (gpointer) agent, args); agent = "fence_legacy"; } #endif crm_xml_add(data, XML_ATTR_ID, id); crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__); crm_xml_add(data, "agent", agent); if ((namespace != st_namespace_any) && (namespace != st_namespace_invalid)) { crm_xml_add(data, "namespace", stonith_namespace2text(namespace)); } if (rsc_provides) { crm_xml_add(data, "rsc_provides", rsc_provides); } for (; params; params = params->next) { hash2field((gpointer) params->key, (gpointer) params->value, args); } return data; } static int stonith_api_register_device(stonith_t * st, int call_options, const char *id, const char *namespace, const char *agent, stonith_key_value_t * params) { int rc = 0; xmlNode *data = NULL; data = create_device_registration_xml(id, stonith_text2namespace(namespace), agent, params, NULL); rc = stonith_send_command(st, STONITH_OP_DEVICE_ADD, data, NULL, call_options, 0); free_xml(data); return rc; } static int stonith_api_remove_device(stonith_t * st, int call_options, const char *name) { int rc = 0; xmlNode *data = NULL; data = create_xml_node(NULL, F_STONITH_DEVICE); crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__); crm_xml_add(data, XML_ATTR_ID, name); rc = stonith_send_command(st, STONITH_OP_DEVICE_DEL, data, NULL, call_options, 0); free_xml(data); return rc; } static int stonith_api_remove_level_full(stonith_t *st, int options, const char *node, const char *pattern, const char *attr, const char *value, int level) { int rc = 0; xmlNode *data = NULL; CRM_CHECK(node || pattern || (attr && value), return -EINVAL); data = create_xml_node(NULL, XML_TAG_FENCING_LEVEL); crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__); if (node) { crm_xml_add(data, XML_ATTR_STONITH_TARGET, node); } else if (pattern) { crm_xml_add(data, XML_ATTR_STONITH_TARGET_PATTERN, pattern); } else { crm_xml_add(data, XML_ATTR_STONITH_TARGET_ATTRIBUTE, attr); crm_xml_add(data, XML_ATTR_STONITH_TARGET_VALUE, value); } crm_xml_add_int(data, XML_ATTR_STONITH_INDEX, level); rc = stonith_send_command(st, STONITH_OP_LEVEL_DEL, data, NULL, options, 0); free_xml(data); return rc; } static int stonith_api_remove_level(stonith_t * st, int options, const char *node, int level) { return stonith_api_remove_level_full(st, options, node, NULL, NULL, NULL, level); } /*! * \internal * \brief Create XML for fence topology level registration request * * \param[in] node If not NULL, target level by this node name * \param[in] pattern If not NULL, target by node name using this regex * \param[in] attr If not NULL, target by this node attribute * \param[in] value If not NULL, target by this node attribute value * \param[in] level Index number of level to register * \param[in] device_list List of devices in level * * \return Newly allocated XML tree on success, NULL otherwise * * \note The caller should set only one of node, pattern or attr/value. */ xmlNode * create_level_registration_xml(const char *node, const char *pattern, const char *attr, const char *value, int level, stonith_key_value_t *device_list) { int len = 0; char *list = NULL; xmlNode *data; CRM_CHECK(node || pattern || (attr && value), return NULL); data = create_xml_node(NULL, XML_TAG_FENCING_LEVEL); CRM_CHECK(data, return NULL); crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__); crm_xml_add_int(data, XML_ATTR_ID, level); crm_xml_add_int(data, XML_ATTR_STONITH_INDEX, level); if (node) { crm_xml_add(data, XML_ATTR_STONITH_TARGET, node); } else if (pattern) { crm_xml_add(data, XML_ATTR_STONITH_TARGET_PATTERN, pattern); } else { crm_xml_add(data, XML_ATTR_STONITH_TARGET_ATTRIBUTE, attr); crm_xml_add(data, XML_ATTR_STONITH_TARGET_VALUE, value); } for (; device_list; device_list = device_list->next) { int adding = strlen(device_list->value); if(list) { adding++; /* +1 space */ } crm_trace("Adding %s (%dc) at offset %d", device_list->value, adding, len); list = realloc_safe(list, len + adding + 1); /* +1 EOS */ if (list == NULL) { crm_perror(LOG_CRIT, "Could not create device list"); free_xml(data); return NULL; } sprintf(list + len, "%s%s", len?",":"", device_list->value); len += adding; } crm_xml_add(data, XML_ATTR_STONITH_DEVICES, list); free(list); return data; } static int stonith_api_register_level_full(stonith_t * st, int options, const char *node, const char *pattern, const char *attr, const char *value, int level, stonith_key_value_t *device_list) { int rc = 0; xmlNode *data = create_level_registration_xml(node, pattern, attr, value, level, device_list); CRM_CHECK(data != NULL, return -EINVAL); rc = stonith_send_command(st, STONITH_OP_LEVEL_ADD, data, NULL, options, 0); free_xml(data); return rc; } static int stonith_api_register_level(stonith_t * st, int options, const char *node, int level, stonith_key_value_t * device_list) { return stonith_api_register_level_full(st, options, node, NULL, NULL, NULL, level, device_list); } static void append_arg(const char *key, const char *value, GHashTable **args) { CRM_CHECK(key != NULL, return); CRM_CHECK(value != NULL, return); CRM_CHECK(args != NULL, return); if (strstr(key, "pcmk_")) { return; } else if (strstr(key, CRM_META)) { return; } else if (safe_str_eq(key, "crm_feature_set")) { return; } if (!*args) { *args = crm_str_table_new(); } CRM_CHECK(*args != NULL, return); crm_trace("Appending: %s=%s", key, value); g_hash_table_replace(*args, strdup(key), strdup(value)); } static void append_config_arg(gpointer key, gpointer value, gpointer user_data) { /* The fencer will filter action out when it registers the device, * but ignore it here just in case any other library callers * fail to do so. */ if (safe_str_neq(key, STONITH_ATTR_ACTION_OP)) { append_arg(key, value, user_data); return; } } static GHashTable * make_args(const char *agent, const char *action, const char *victim, uint32_t victim_nodeid, GHashTable * device_args, GHashTable * port_map) { char buffer[512]; GHashTable *arg_list = NULL; const char *value = NULL; CRM_CHECK(action != NULL, return NULL); snprintf(buffer, sizeof(buffer), "pcmk_%s_action", action); if (device_args) { value = g_hash_table_lookup(device_args, buffer); } if (value) { crm_info("Substituting action '%s' for requested operation '%s'", value, action); action = value; } append_arg(STONITH_ATTR_ACTION_OP, action, &arg_list); if (victim && device_args) { const char *alias = victim; const char *param = g_hash_table_lookup(device_args, STONITH_ATTR_HOSTARG); if (port_map && g_hash_table_lookup(port_map, victim)) { alias = g_hash_table_lookup(port_map, victim); } /* Always supply the node's name, too: * https://github.com/ClusterLabs/fence-agents/blob/master/doc/FenceAgentAPI.md */ append_arg("nodename", victim, &arg_list); if (victim_nodeid) { char nodeid_str[33] = { 0, }; if (snprintf(nodeid_str, 33, "%u", (unsigned int)victim_nodeid)) { crm_info("For stonith action (%s) for victim %s, adding nodeid (%s) to parameters", action, victim, nodeid_str); append_arg("nodeid", nodeid_str, &arg_list); } } /* Check if we need to supply the victim in any other form */ if(safe_str_eq(agent, "fence_legacy")) { value = agent; } else if (param == NULL) { param = "port"; value = g_hash_table_lookup(device_args, param); } else if (safe_str_eq(param, "none")) { value = param; /* Nothing more to do */ } else { value = g_hash_table_lookup(device_args, param); } /* Don't overwrite explictly set values for $param */ if (value == NULL || safe_str_eq(value, "dynamic")) { crm_debug("Performing %s action for node '%s' as '%s=%s'", action, victim, param, alias); append_arg(param, alias, &arg_list); } } if (device_args) { g_hash_table_foreach(device_args, append_config_arg, &arg_list); } return arg_list; } /*! * \internal * \brief Free all memory used by a stonith action * * \param[in,out] action Action to free */ void stonith__destroy_action(stonith_action_t *action) { if (action) { free(action->agent); if (action->args) { g_hash_table_destroy(action->args); } free(action->action); free(action->victim); if (action->svc_action) { services_action_free(action->svc_action); } free(action->output); free(action->error); free(action); } } /*! * \internal * \brief Get the result of an executed stonith action * * \param[in,out] action Executed action * \param[out] rc Where to store result code (or NULL) * \param[out] output Where to store standard output (or NULL) * \param[out] error_output Where to store standard error output (or NULL) * * \note If output or error_output is not NULL, the caller is responsible for * freeing the memory. */ void stonith__action_result(stonith_action_t *action, int *rc, char **output, char **error_output) { if (rc) { *rc = pcmk_ok; } if (output) { *output = NULL; } if (error_output) { *error_output = NULL; } if (action != NULL) { if (rc) { *rc = action->rc; } if (output && action->output) { *output = action->output; action->output = NULL; // hand off memory management to caller } if (error_output && action->error) { *error_output = action->error; action->error = NULL; // hand off memory management to caller } } } #define FAILURE_MAX_RETRIES 2 stonith_action_t * stonith_action_create(const char *agent, const char *_action, const char *victim, uint32_t victim_nodeid, int timeout, GHashTable * device_args, GHashTable * port_map) { stonith_action_t *action; action = calloc(1, sizeof(stonith_action_t)); action->args = make_args(agent, _action, victim, victim_nodeid, device_args, port_map); crm_debug("Preparing '%s' action for %s using agent %s", _action, (victim? victim : "no target"), agent); action->agent = strdup(agent); action->action = strdup(_action); if (victim) { action->victim = strdup(victim); } action->timeout = action->remaining_timeout = timeout; action->max_retries = FAILURE_MAX_RETRIES; if (device_args) { char buffer[512]; const char *value = NULL; snprintf(buffer, sizeof(buffer), "pcmk_%s_retries", _action); value = g_hash_table_lookup(device_args, buffer); if (value) { action->max_retries = atoi(value); } } return action; } static gboolean update_remaining_timeout(stonith_action_t * action) { int diff = time(NULL) - action->initial_start_time; if (action->tries >= action->max_retries) { crm_info("Attempted to execute agent %s (%s) the maximum number of times (%d) allowed", action->agent, action->action, action->max_retries); action->remaining_timeout = 0; } else if ((action->rc != -ETIME) && diff < (action->timeout * 0.7)) { /* only set remaining timeout period if there is 30% * or greater of the original timeout period left */ action->remaining_timeout = action->timeout - diff; } else { action->remaining_timeout = 0; } return action->remaining_timeout ? TRUE : FALSE; } static int svc_action_to_errno(svc_action_t *svc_action) { int rv = pcmk_ok; if (svc_action->rc > 0) { /* Try to provide a useful error code based on the fence agent's * error output. */ if (svc_action->rc == PCMK_OCF_TIMEOUT) { rv = -ETIME; } else if (svc_action->stderr_data == NULL) { rv = -ENODATA; } else if (strstr(svc_action->stderr_data, "imed out")) { /* Some agents have their own internal timeouts */ rv = -ETIME; } else if (strstr(svc_action->stderr_data, "Unrecognised action")) { rv = -EOPNOTSUPP; } else { rv = -pcmk_err_generic; } } return rv; } static void stonith_action_async_done(svc_action_t *svc_action) { stonith_action_t *action = (stonith_action_t *) svc_action->cb_data; action->rc = svc_action_to_errno(svc_action); action->output = svc_action->stdout_data; svc_action->stdout_data = NULL; action->error = svc_action->stderr_data; svc_action->stderr_data = NULL; svc_action->params = NULL; crm_debug("Child process %d performing action '%s' exited with rc %d", action->pid, action->action, svc_action->rc); log_action(action, action->pid); if (action->rc != pcmk_ok && update_remaining_timeout(action)) { int rc = internal_stonith_action_execute(action); if (rc == pcmk_ok) { return; } } if (action->done_cb) { action->done_cb(action->pid, action->rc, action->output, action->userdata); } action->svc_action = NULL; // don't remove our caller stonith__destroy_action(action); } static void stonith_action_async_forked(svc_action_t *svc_action) { stonith_action_t *action = (stonith_action_t *) svc_action->cb_data; action->pid = svc_action->pid; action->svc_action = svc_action; + if (action->fork_cb) { + (action->fork_cb) (svc_action->pid, action->userdata); + } + crm_trace("Child process %d performing action '%s' successfully forked", action->pid, action->action); } static int internal_stonith_action_execute(stonith_action_t * action) { int rc = -EPROTO; int is_retry = 0; svc_action_t *svc_action = NULL; static int stonith_sequence = 0; char *buffer = NULL; if (!action->tries) { action->initial_start_time = time(NULL); } action->tries++; if (action->tries > 1) { crm_info("Attempt %d to execute %s (%s). remaining timeout is %d", action->tries, action->agent, action->action, action->remaining_timeout); is_retry = 1; } if (action->args == NULL || action->agent == NULL) goto fail; buffer = crm_strdup_printf(RH_STONITH_DIR "/%s", basename(action->agent)); svc_action = services_action_create_generic(buffer, NULL); free(buffer); svc_action->timeout = 1000 * action->remaining_timeout; svc_action->standard = strdup(PCMK_RESOURCE_CLASS_STONITH); svc_action->id = crm_strdup_printf("%s_%s_%d", basename(action->agent), action->action, action->tries); svc_action->agent = strdup(action->agent); svc_action->sequence = stonith_sequence++; svc_action->params = action->args; svc_action->cb_data = (void *) action; /* keep retries from executing out of control and free previous results */ if (is_retry) { free(action->output); action->output = NULL; free(action->error); action->error = NULL; sleep(1); } if (action->async) { /* async */ if(services_action_async_fork_notify(svc_action, &stonith_action_async_done, &stonith_action_async_forked) == FALSE) { services_action_free(svc_action); svc_action = NULL; } else { rc = 0; } } else { /* sync */ if (services_action_sync(svc_action)) { rc = 0; action->rc = svc_action_to_errno(svc_action); action->output = svc_action->stdout_data; svc_action->stdout_data = NULL; action->error = svc_action->stderr_data; svc_action->stderr_data = NULL; } else { action->rc = -ECONNABORTED; rc = action->rc; } svc_action->params = NULL; services_action_free(svc_action); } fail: return rc; } -GPid +/*! + * \internal + * \brief Kick off execution of an async stonith action + * + * \param[in,out] action Action to be executed + * \param[in,out] userdata Datapointer to be passed to callbacks + * \param[in] done Callback to notify action has failed/succeeded + * \param[in] fork_callback Callback to notify successful fork of child + * + * \return pcmk_ok if ownership of action has been taken, -errno otherwise + */ +int stonith_action_execute_async(stonith_action_t * action, void *userdata, void (*done) (GPid pid, int rc, const char *output, - gpointer user_data)) + gpointer user_data), + void (*fork_cb) (GPid pid, gpointer user_data)) { - int rc = 0; - if (!action) { return -1; } action->userdata = userdata; action->done_cb = done; + action->fork_cb = fork_cb; action->async = 1; - rc = internal_stonith_action_execute(action); - - return rc < 0 ? rc : action->pid; + return internal_stonith_action_execute(action); } /*! * \internal * \brief Execute a stonith action * * \param[in,out] action Action to execute * * \return pcmk_ok on success, -errno otherwise */ int stonith__execute(stonith_action_t *action) { int rc = pcmk_ok; CRM_CHECK(action != NULL, return -EINVAL); // Keep trying until success, max retries, or timeout do { rc = internal_stonith_action_execute(action); } while ((rc != pcmk_ok) && update_remaining_timeout(action)); return rc; } static int stonith_api_device_list(stonith_t * stonith, int call_options, const char *namespace, stonith_key_value_t ** devices, int timeout) { int count = 0; enum stonith_namespace ns = stonith_text2namespace(namespace); if (devices == NULL) { crm_err("Parameter error: stonith_api_device_list"); return -EFAULT; } #if HAVE_STONITH_STONITH_H // Include Linux-HA agents if requested if ((ns == st_namespace_any) || (ns == st_namespace_lha)) { count += stonith__list_lha_agents(devices); } #endif // Include Red Hat agents if requested if ((ns == st_namespace_any) || (ns == st_namespace_rhcs)) { count += stonith__list_rhcs_agents(devices); } return count; } static int stonith_api_device_metadata(stonith_t * stonith, int call_options, const char *agent, const char *namespace, char **output, int timeout) { /* By executing meta-data directly, we can get it from stonith_admin when * the cluster is not running, which is important for higher-level tools. */ enum stonith_namespace ns = stonith_get_namespace(agent, namespace); crm_trace("Looking up metadata for %s agent %s", stonith_namespace2text(ns), agent); switch (ns) { case st_namespace_rhcs: return stonith__rhcs_metadata(agent, timeout, output); #if HAVE_STONITH_STONITH_H case st_namespace_lha: return stonith__lha_metadata(agent, timeout, output); #endif default: errno = EINVAL; crm_perror(LOG_ERR, "Agent %s not found or does not support meta-data", agent); break; } return -EINVAL; } static int stonith_api_query(stonith_t * stonith, int call_options, const char *target, stonith_key_value_t ** devices, int timeout) { int rc = 0, lpc = 0, max = 0; xmlNode *data = NULL; xmlNode *output = NULL; xmlXPathObjectPtr xpathObj = NULL; CRM_CHECK(devices != NULL, return -EINVAL); data = create_xml_node(NULL, F_STONITH_DEVICE); crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__); crm_xml_add(data, F_STONITH_TARGET, target); crm_xml_add(data, F_STONITH_ACTION, "off"); rc = stonith_send_command(stonith, STONITH_OP_QUERY, data, &output, call_options, timeout); if (rc < 0) { return rc; } xpathObj = xpath_search(output, "//@agent"); if (xpathObj) { max = numXpathResults(xpathObj); for (lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpathObj, lpc); CRM_LOG_ASSERT(match != NULL); if(match != NULL) { xmlChar *match_path = xmlGetNodePath(match); crm_info("%s[%d] = %s", "//@agent", lpc, match_path); free(match_path); *devices = stonith_key_value_add(*devices, NULL, crm_element_value(match, XML_ATTR_ID)); } } freeXpathObject(xpathObj); } free_xml(output); free_xml(data); return max; } static int stonith_api_call(stonith_t * stonith, int call_options, const char *id, const char *action, const char *victim, int timeout, xmlNode ** output) { int rc = 0; xmlNode *data = NULL; data = create_xml_node(NULL, F_STONITH_DEVICE); crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__); crm_xml_add(data, F_STONITH_DEVICE, id); crm_xml_add(data, F_STONITH_ACTION, action); crm_xml_add(data, F_STONITH_TARGET, victim); rc = stonith_send_command(stonith, STONITH_OP_EXEC, data, output, call_options, timeout); free_xml(data); return rc; } static int stonith_api_list(stonith_t * stonith, int call_options, const char *id, char **list_info, int timeout) { int rc; xmlNode *output = NULL; rc = stonith_api_call(stonith, call_options, id, "list", NULL, timeout, &output); if (output && list_info) { const char *list_str; list_str = crm_element_value(output, "st_output"); if (list_str) { *list_info = strdup(list_str); } } if (output) { free_xml(output); } return rc; } static int stonith_api_monitor(stonith_t * stonith, int call_options, const char *id, int timeout) { return stonith_api_call(stonith, call_options, id, "monitor", NULL, timeout, NULL); } static int stonith_api_status(stonith_t * stonith, int call_options, const char *id, const char *port, int timeout) { return stonith_api_call(stonith, call_options, id, "status", port, timeout, NULL); } static int stonith_api_fence(stonith_t * stonith, int call_options, const char *node, const char *action, int timeout, int tolerance) { int rc = 0; xmlNode *data = NULL; data = create_xml_node(NULL, __FUNCTION__); crm_xml_add(data, F_STONITH_TARGET, node); crm_xml_add(data, F_STONITH_ACTION, action); crm_xml_add_int(data, F_STONITH_TIMEOUT, timeout); crm_xml_add_int(data, F_STONITH_TOLERANCE, tolerance); rc = stonith_send_command(stonith, STONITH_OP_FENCE, data, NULL, call_options, timeout); free_xml(data); return rc; } static int stonith_api_confirm(stonith_t * stonith, int call_options, const char *target) { return stonith_api_fence(stonith, call_options | st_opt_manual_ack, target, "off", 0, 0); } static int stonith_api_history(stonith_t * stonith, int call_options, const char *node, stonith_history_t ** history, int timeout) { int rc = 0; xmlNode *data = NULL; xmlNode *output = NULL; stonith_history_t *last = NULL; *history = NULL; if (node) { data = create_xml_node(NULL, __FUNCTION__); crm_xml_add(data, F_STONITH_TARGET, node); } rc = stonith_send_command(stonith, STONITH_OP_FENCE_HISTORY, data, &output, call_options | st_opt_sync_call, timeout); free_xml(data); if (rc == 0) { xmlNode *op = NULL; xmlNode *reply = get_xpath_object("//" F_STONITH_HISTORY_LIST, output, LOG_TRACE); for (op = __xml_first_child(reply); op != NULL; op = __xml_next(op)) { stonith_history_t *kvp; int completed; kvp = calloc(1, sizeof(stonith_history_t)); kvp->target = crm_element_value_copy(op, F_STONITH_TARGET); kvp->action = crm_element_value_copy(op, F_STONITH_ACTION); kvp->origin = crm_element_value_copy(op, F_STONITH_ORIGIN); kvp->delegate = crm_element_value_copy(op, F_STONITH_DELEGATE); kvp->client = crm_element_value_copy(op, F_STONITH_CLIENTNAME); crm_element_value_int(op, F_STONITH_DATE, &completed); kvp->completed = (time_t) completed; crm_element_value_int(op, F_STONITH_STATE, &kvp->state); if (last) { last->next = kvp; } else { *history = kvp; } last = kvp; } } free_xml(output); return rc; } void stonith_history_free(stonith_history_t *history) { stonith_history_t *hp, *hp_old; for (hp = history; hp; hp_old = hp, hp = hp->next, free(hp_old)) { free(hp->target); free(hp->action); free(hp->origin); free(hp->delegate); free(hp->client); } } /*! * \brief Deprecated (use stonith_get_namespace() instead) */ const char * get_stonith_provider(const char *agent, const char *provider) { return stonith_namespace2text(stonith_get_namespace(agent, provider)); } static gint stonithlib_GCompareFunc(gconstpointer a, gconstpointer b) { int rc = 0; const stonith_notify_client_t *a_client = a; const stonith_notify_client_t *b_client = b; CRM_CHECK(a_client->event != NULL && b_client->event != NULL, return 0); rc = strcmp(a_client->event, b_client->event); if (rc == 0) { if (a_client->notify == NULL || b_client->notify == NULL) { return 0; } else if (a_client->notify == b_client->notify) { return 0; } else if (((long)a_client->notify) < ((long)b_client->notify)) { crm_err("callbacks for %s are not equal: %p vs. %p", a_client->event, a_client->notify, b_client->notify); return -1; } crm_err("callbacks for %s are not equal: %p vs. %p", a_client->event, a_client->notify, b_client->notify); return 1; } return rc; } xmlNode * stonith_create_op(int call_id, const char *token, const char *op, xmlNode * data, int call_options) { xmlNode *op_msg = create_xml_node(NULL, "stonith_command"); CRM_CHECK(op_msg != NULL, return NULL); CRM_CHECK(token != NULL, return NULL); crm_xml_add(op_msg, F_XML_TAGNAME, "stonith_command"); crm_xml_add(op_msg, F_TYPE, T_STONITH_NG); crm_xml_add(op_msg, F_STONITH_CALLBACK_TOKEN, token); crm_xml_add(op_msg, F_STONITH_OPERATION, op); crm_xml_add_int(op_msg, F_STONITH_CALLID, call_id); crm_trace("Sending call options: %.8lx, %d", (long)call_options, call_options); crm_xml_add_int(op_msg, F_STONITH_CALLOPTS, call_options); if (data != NULL) { add_message_xml(op_msg, F_STONITH_CALLDATA, data); } return op_msg; } static void stonith_destroy_op_callback(gpointer data) { stonith_callback_client_t *blob = data; if (blob->timer && blob->timer->ref > 0) { g_source_remove(blob->timer->ref); } free(blob->timer); free(blob); } static int stonith_api_signoff(stonith_t * stonith) { stonith_private_t *native = stonith->st_private; crm_debug("Disconnecting from the fencer"); if (native->source != NULL) { /* Attached to mainloop */ mainloop_del_ipc_client(native->source); native->source = NULL; native->ipc = NULL; } else if (native->ipc) { /* Not attached to mainloop */ crm_ipc_t *ipc = native->ipc; native->ipc = NULL; crm_ipc_close(ipc); crm_ipc_destroy(ipc); } free(native->token); native->token = NULL; stonith->state = stonith_disconnected; return pcmk_ok; } static int stonith_api_del_callback(stonith_t * stonith, int call_id, bool all_callbacks) { stonith_private_t *private = stonith->st_private; if (all_callbacks) { private->op_callback = NULL; g_hash_table_destroy(private->stonith_op_callback_table); private->stonith_op_callback_table = g_hash_table_new_full(g_direct_hash, g_direct_equal, NULL, stonith_destroy_op_callback); } else if (call_id == 0) { private->op_callback = NULL; } else { g_hash_table_remove(private->stonith_op_callback_table, GINT_TO_POINTER(call_id)); } return pcmk_ok; } static void invoke_callback(stonith_t * st, int call_id, int rc, void *userdata, void (*callback) (stonith_t * st, stonith_callback_data_t * data)) { stonith_callback_data_t data = { 0, }; data.call_id = call_id; data.rc = rc; data.userdata = userdata; callback(st, &data); } static void stonith_perform_callback(stonith_t * stonith, xmlNode * msg, int call_id, int rc) { stonith_private_t *private = NULL; stonith_callback_client_t *blob = NULL; stonith_callback_client_t local_blob; CRM_CHECK(stonith != NULL, return); CRM_CHECK(stonith->st_private != NULL, return); private = stonith->st_private; local_blob.id = NULL; local_blob.callback = NULL; local_blob.user_data = NULL; local_blob.only_success = FALSE; if (msg != NULL) { crm_element_value_int(msg, F_STONITH_RC, &rc); crm_element_value_int(msg, F_STONITH_CALLID, &call_id); } CRM_CHECK(call_id > 0, crm_log_xml_err(msg, "Bad result")); blob = g_hash_table_lookup(private->stonith_op_callback_table, GINT_TO_POINTER(call_id)); if (blob != NULL) { local_blob = *blob; blob = NULL; stonith_api_del_callback(stonith, call_id, FALSE); } else { crm_trace("No callback found for call %d", call_id); local_blob.callback = NULL; } if (local_blob.callback != NULL && (rc == pcmk_ok || local_blob.only_success == FALSE)) { crm_trace("Invoking callback %s for call %d", crm_str(local_blob.id), call_id); invoke_callback(stonith, call_id, rc, local_blob.user_data, local_blob.callback); } else if (private->op_callback == NULL && rc != pcmk_ok) { crm_warn("Fencing command failed: %s", pcmk_strerror(rc)); crm_log_xml_debug(msg, "Failed fence update"); } if (private->op_callback != NULL) { crm_trace("Invoking global callback for call %d", call_id); invoke_callback(stonith, call_id, rc, NULL, private->op_callback); } crm_trace("OP callback activated."); } static gboolean stonith_async_timeout_handler(gpointer data) { struct timer_rec_s *timer = data; crm_err("Async call %d timed out after %dms", timer->call_id, timer->timeout); stonith_perform_callback(timer->stonith, NULL, timer->call_id, -ETIME); /* Always return TRUE, never remove the handler * We do that in stonith_del_callback() */ return TRUE; } static void set_callback_timeout(stonith_callback_client_t * callback, stonith_t * stonith, int call_id, int timeout) { struct timer_rec_s *async_timer = callback->timer; if (timeout <= 0) { return; } if (!async_timer) { async_timer = calloc(1, sizeof(struct timer_rec_s)); callback->timer = async_timer; } async_timer->stonith = stonith; async_timer->call_id = call_id; /* Allow a fair bit of grace to allow the server to tell us of a timeout * This is only a fallback */ async_timer->timeout = (timeout + 60) * 1000; if (async_timer->ref) { g_source_remove(async_timer->ref); } async_timer->ref = g_timeout_add(async_timer->timeout, stonith_async_timeout_handler, async_timer); } static void update_callback_timeout(int call_id, int timeout, stonith_t * st) { stonith_callback_client_t *callback = NULL; stonith_private_t *private = st->st_private; callback = g_hash_table_lookup(private->stonith_op_callback_table, GINT_TO_POINTER(call_id)); if (!callback || !callback->allow_timeout_updates) { return; } set_callback_timeout(callback, st, call_id, timeout); } static int stonith_dispatch_internal(const char *buffer, ssize_t length, gpointer userdata) { const char *type = NULL; struct notify_blob_s blob; stonith_t *st = userdata; stonith_private_t *private = NULL; CRM_ASSERT(st != NULL); private = st->st_private; blob.stonith = st; blob.xml = string2xml(buffer); if (blob.xml == NULL) { crm_warn("Received malformed message from fencer: %s", buffer); return 0; } /* do callbacks */ type = crm_element_value(blob.xml, F_TYPE); crm_trace("Activating %s callbacks...", type); if (safe_str_eq(type, T_STONITH_NG)) { stonith_perform_callback(st, blob.xml, 0, 0); } else if (safe_str_eq(type, T_STONITH_NOTIFY)) { g_list_foreach(private->notify_list, stonith_send_notification, &blob); } else if (safe_str_eq(type, T_STONITH_TIMEOUT_VALUE)) { int call_id = 0; int timeout = 0; crm_element_value_int(blob.xml, F_STONITH_TIMEOUT, &timeout); crm_element_value_int(blob.xml, F_STONITH_CALLID, &call_id); update_callback_timeout(call_id, timeout, st); } else { crm_err("Unknown message type: %s", type); crm_log_xml_warn(blob.xml, "BadReply"); } free_xml(blob.xml); return 1; } static int stonith_api_signon(stonith_t * stonith, const char *name, int *stonith_fd) { int rc = pcmk_ok; stonith_private_t *native = stonith->st_private; static struct ipc_client_callbacks st_callbacks = { .dispatch = stonith_dispatch_internal, .destroy = stonith_connection_destroy }; crm_trace("Connecting command channel"); stonith->state = stonith_connected_command; if (stonith_fd) { /* No mainloop */ native->ipc = crm_ipc_new("stonith-ng", 0); if (native->ipc && crm_ipc_connect(native->ipc)) { *stonith_fd = crm_ipc_get_fd(native->ipc); } else if (native->ipc) { crm_perror(LOG_ERR, "Connection to fencer failed"); rc = -ENOTCONN; } } else { /* With mainloop */ native->source = mainloop_add_ipc_client("stonith-ng", G_PRIORITY_MEDIUM, 0, stonith, &st_callbacks); native->ipc = mainloop_get_ipc_client(native->source); } if (native->ipc == NULL) { crm_debug("Could not connect to the Stonith API"); rc = -ENOTCONN; } if (rc == pcmk_ok) { xmlNode *reply = NULL; xmlNode *hello = create_xml_node(NULL, "stonith_command"); crm_xml_add(hello, F_TYPE, T_STONITH_NG); crm_xml_add(hello, F_STONITH_OPERATION, CRM_OP_REGISTER); crm_xml_add(hello, F_STONITH_CLIENTNAME, name); rc = crm_ipc_send(native->ipc, hello, crm_ipc_client_response, -1, &reply); if (rc < 0) { crm_perror(LOG_DEBUG, "Couldn't complete registration with the fencing API: %d", rc); rc = -ECOMM; } else if (reply == NULL) { crm_err("Did not receive registration reply"); rc = -EPROTO; } else { const char *msg_type = crm_element_value(reply, F_STONITH_OPERATION); const char *tmp_ticket = crm_element_value(reply, F_STONITH_CLIENTID); if (safe_str_neq(msg_type, CRM_OP_REGISTER)) { crm_err("Invalid registration message: %s", msg_type); crm_log_xml_err(reply, "Bad reply"); rc = -EPROTO; } else if (tmp_ticket == NULL) { crm_err("No registration token provided"); crm_log_xml_err(reply, "Bad reply"); rc = -EPROTO; } else { crm_trace("Obtained registration token: %s", tmp_ticket); native->token = strdup(tmp_ticket); rc = pcmk_ok; } } free_xml(reply); free_xml(hello); } if (rc == pcmk_ok) { #if HAVE_MSGFROMIPC_TIMEOUT stonith->call_timeout = MAX_IPC_DELAY; #endif crm_debug("Connection to fencer successful"); return pcmk_ok; } crm_debug("Connection to fencer failed: %s", pcmk_strerror(rc)); stonith->cmds->disconnect(stonith); return rc; } static int stonith_set_notification(stonith_t * stonith, const char *callback, int enabled) { int rc = pcmk_ok; xmlNode *notify_msg = create_xml_node(NULL, __FUNCTION__); stonith_private_t *native = stonith->st_private; if (stonith->state != stonith_disconnected) { crm_xml_add(notify_msg, F_STONITH_OPERATION, T_STONITH_NOTIFY); if (enabled) { crm_xml_add(notify_msg, F_STONITH_NOTIFY_ACTIVATE, callback); } else { crm_xml_add(notify_msg, F_STONITH_NOTIFY_DEACTIVATE, callback); } rc = crm_ipc_send(native->ipc, notify_msg, crm_ipc_client_response, -1, NULL); if (rc < 0) { crm_perror(LOG_DEBUG, "Couldn't register for fencing notifications: %d", rc); rc = -ECOMM; } else { rc = pcmk_ok; } } free_xml(notify_msg); return rc; } static int stonith_api_add_notification(stonith_t * stonith, const char *event, void (*callback) (stonith_t * stonith, stonith_event_t * e)) { GList *list_item = NULL; stonith_notify_client_t *new_client = NULL; stonith_private_t *private = NULL; private = stonith->st_private; crm_trace("Adding callback for %s events (%d)", event, g_list_length(private->notify_list)); new_client = calloc(1, sizeof(stonith_notify_client_t)); new_client->event = event; new_client->notify = callback; list_item = g_list_find_custom(private->notify_list, new_client, stonithlib_GCompareFunc); if (list_item != NULL) { crm_warn("Callback already present"); free(new_client); return -ENOTUNIQ; } else { private->notify_list = g_list_append(private->notify_list, new_client); stonith_set_notification(stonith, event, 1); crm_trace("Callback added (%d)", g_list_length(private->notify_list)); } return pcmk_ok; } static int stonith_api_del_notification(stonith_t * stonith, const char *event) { GList *list_item = NULL; stonith_notify_client_t *new_client = NULL; stonith_private_t *private = NULL; crm_debug("Removing callback for %s events", event); private = stonith->st_private; new_client = calloc(1, sizeof(stonith_notify_client_t)); new_client->event = event; new_client->notify = NULL; list_item = g_list_find_custom(private->notify_list, new_client, stonithlib_GCompareFunc); stonith_set_notification(stonith, event, 0); if (list_item != NULL) { stonith_notify_client_t *list_client = list_item->data; private->notify_list = g_list_remove(private->notify_list, list_client); free(list_client); crm_trace("Removed callback"); } else { crm_trace("Callback not present"); } free(new_client); return pcmk_ok; } static int stonith_api_add_callback(stonith_t * stonith, int call_id, int timeout, int options, void *user_data, const char *callback_name, void (*callback) (stonith_t * st, stonith_callback_data_t * data)) { stonith_callback_client_t *blob = NULL; stonith_private_t *private = NULL; CRM_CHECK(stonith != NULL, return -EINVAL); CRM_CHECK(stonith->st_private != NULL, return -EINVAL); private = stonith->st_private; if (call_id == 0) { private->op_callback = callback; } else if (call_id < 0) { if (!(options & st_opt_report_only_success)) { crm_trace("Call failed, calling %s: %s", callback_name, pcmk_strerror(call_id)); invoke_callback(stonith, call_id, call_id, user_data, callback); } else { crm_warn("Fencer call failed: %s", pcmk_strerror(call_id)); } return FALSE; } blob = calloc(1, sizeof(stonith_callback_client_t)); blob->id = callback_name; blob->only_success = (options & st_opt_report_only_success) ? TRUE : FALSE; blob->user_data = user_data; blob->callback = callback; blob->allow_timeout_updates = (options & st_opt_timeout_updates) ? TRUE : FALSE; if (timeout > 0) { set_callback_timeout(blob, stonith, call_id, timeout); } g_hash_table_insert(private->stonith_op_callback_table, GINT_TO_POINTER(call_id), blob); crm_trace("Added callback to %s for call %d", callback_name, call_id); return TRUE; } static void stonith_dump_pending_op(gpointer key, gpointer value, gpointer user_data) { int call = GPOINTER_TO_INT(key); stonith_callback_client_t *blob = value; crm_debug("Call %d (%s): pending", call, crm_str(blob->id)); } void stonith_dump_pending_callbacks(stonith_t * stonith) { stonith_private_t *private = stonith->st_private; if (private->stonith_op_callback_table == NULL) { return; } return g_hash_table_foreach(private->stonith_op_callback_table, stonith_dump_pending_op, NULL); } /* */ static stonith_event_t * xml_to_event(xmlNode * msg) { stonith_event_t *event = calloc(1, sizeof(stonith_event_t)); const char *ntype = crm_element_value(msg, F_SUBTYPE); char *data_addr = crm_strdup_printf("//%s", ntype); xmlNode *data = get_xpath_object(data_addr, msg, LOG_DEBUG); crm_log_xml_trace(msg, "stonith_notify"); crm_element_value_int(msg, F_STONITH_RC, &(event->result)); if (safe_str_eq(ntype, T_STONITH_NOTIFY_FENCE)) { event->operation = crm_element_value_copy(msg, F_STONITH_OPERATION); if (data) { event->origin = crm_element_value_copy(data, F_STONITH_ORIGIN); event->action = crm_element_value_copy(data, F_STONITH_ACTION); event->target = crm_element_value_copy(data, F_STONITH_TARGET); event->executioner = crm_element_value_copy(data, F_STONITH_DELEGATE); event->id = crm_element_value_copy(data, F_STONITH_REMOTE_OP_ID); event->client_origin = crm_element_value_copy(data, F_STONITH_CLIENTNAME); event->device = crm_element_value_copy(data, F_STONITH_DEVICE); } else { crm_err("No data for %s event", ntype); crm_log_xml_notice(msg, "BadEvent"); } } free(data_addr); return event; } static void event_free(stonith_event_t * event) { free(event->id); free(event->type); free(event->message); free(event->operation); free(event->origin); free(event->action); free(event->target); free(event->executioner); free(event->device); free(event->client_origin); free(event); } static void stonith_send_notification(gpointer data, gpointer user_data) { struct notify_blob_s *blob = user_data; stonith_notify_client_t *entry = data; stonith_event_t *st_event = NULL; const char *event = NULL; if (blob->xml == NULL) { crm_warn("Skipping callback - NULL message"); return; } event = crm_element_value(blob->xml, F_SUBTYPE); if (entry == NULL) { crm_warn("Skipping callback - NULL callback client"); return; } else if (entry->notify == NULL) { crm_warn("Skipping callback - NULL callback"); return; } else if (safe_str_neq(entry->event, event)) { crm_trace("Skipping callback - event mismatch %p/%s vs. %s", entry, entry->event, event); return; } st_event = xml_to_event(blob->xml); crm_trace("Invoking callback for %p/%s event...", entry, event); entry->notify(blob->stonith, st_event); crm_trace("Callback invoked..."); event_free(st_event); } int stonith_send_command(stonith_t * stonith, const char *op, xmlNode * data, xmlNode ** output_data, int call_options, int timeout) { int rc = 0; int reply_id = -1; enum crm_ipc_flags ipc_flags = crm_ipc_flags_none; xmlNode *op_msg = NULL; xmlNode *op_reply = NULL; stonith_private_t *native = stonith->st_private; if (stonith->state == stonith_disconnected) { return -ENOTCONN; } if (output_data != NULL) { *output_data = NULL; } if (op == NULL) { crm_err("No operation specified"); return -EINVAL; } if (call_options & st_opt_sync_call) { ipc_flags |= crm_ipc_client_response; } stonith->call_id++; /* prevent call_id from being negative (or zero) and conflicting * with the stonith_errors enum * use 2 because we use it as (stonith->call_id - 1) below */ if (stonith->call_id < 1) { stonith->call_id = 1; } CRM_CHECK(native->token != NULL,; ); op_msg = stonith_create_op(stonith->call_id, native->token, op, data, call_options); if (op_msg == NULL) { return -EINVAL; } crm_xml_add_int(op_msg, F_STONITH_TIMEOUT, timeout); crm_trace("Sending %s message to fencer with timeout %ds", op, timeout); rc = crm_ipc_send(native->ipc, op_msg, ipc_flags, 1000 * (timeout + 60), &op_reply); free_xml(op_msg); if (rc < 0) { crm_perror(LOG_ERR, "Couldn't perform %s operation (timeout=%ds): %d", op, timeout, rc); rc = -ECOMM; goto done; } crm_log_xml_trace(op_reply, "Reply"); if (!(call_options & st_opt_sync_call)) { crm_trace("Async call %d, returning", stonith->call_id); CRM_CHECK(stonith->call_id != 0, return -EPROTO); free_xml(op_reply); return stonith->call_id; } rc = pcmk_ok; crm_element_value_int(op_reply, F_STONITH_CALLID, &reply_id); if (reply_id == stonith->call_id) { crm_trace("Synchronous reply %d received", reply_id); if (crm_element_value_int(op_reply, F_STONITH_RC, &rc) != 0) { rc = -ENOMSG; } if ((call_options & st_opt_discard_reply) || output_data == NULL) { crm_trace("Discarding reply"); } else { *output_data = op_reply; op_reply = NULL; /* Prevent subsequent free */ } } else if (reply_id <= 0) { crm_err("Received bad reply: No id set"); crm_log_xml_err(op_reply, "Bad reply"); free_xml(op_reply); rc = -ENOMSG; } else { crm_err("Received bad reply: %d (wanted %d)", reply_id, stonith->call_id); crm_log_xml_err(op_reply, "Old reply"); free_xml(op_reply); rc = -ENOMSG; } done: if (crm_ipc_connected(native->ipc) == FALSE) { crm_err("Fencer disconnected"); stonith->state = stonith_disconnected; } free_xml(op_reply); return rc; } /* Not used with mainloop */ bool stonith_dispatch(stonith_t * st) { gboolean stay_connected = TRUE; stonith_private_t *private = NULL; CRM_ASSERT(st != NULL); private = st->st_private; while (crm_ipc_ready(private->ipc)) { if (crm_ipc_read(private->ipc) > 0) { const char *msg = crm_ipc_buffer(private->ipc); stonith_dispatch_internal(msg, strlen(msg), st); } if (crm_ipc_connected(private->ipc) == FALSE) { crm_err("Connection closed"); stay_connected = FALSE; } } return stay_connected; } static int stonith_api_free(stonith_t * stonith) { int rc = pcmk_ok; crm_trace("Destroying %p", stonith); if (stonith->state != stonith_disconnected) { crm_trace("Disconnecting %p first", stonith); rc = stonith->cmds->disconnect(stonith); } if (stonith->state == stonith_disconnected) { stonith_private_t *private = stonith->st_private; crm_trace("Removing %d callbacks", g_hash_table_size(private->stonith_op_callback_table)); g_hash_table_destroy(private->stonith_op_callback_table); crm_trace("Destroying %d notification clients", g_list_length(private->notify_list)); g_list_free_full(private->notify_list, free); free(stonith->st_private); free(stonith->cmds); free(stonith); } else { crm_err("Not free'ing active connection: %s (%d)", pcmk_strerror(rc), rc); } return rc; } void stonith_api_delete(stonith_t * stonith) { crm_trace("Destroying %p", stonith); if(stonith) { stonith->cmds->free(stonith); } } static int stonith_api_validate(stonith_t *st, int call_options, const char *rsc_id, const char *namespace_s, const char *agent, stonith_key_value_t *params, int timeout, char **output, char **error_output) { /* Validation should be done directly via the agent, so we can get it from * stonith_admin when the cluster is not running, which is important for * higher-level tools. */ int rc = pcmk_ok; /* Use a dummy node name in case the agent requires a target. We assume the * actual target doesn't matter for validation purposes (if in practice, * that is incorrect, we will need to allow the caller to pass the target). */ const char *target = "node1"; GHashTable *params_table = crm_str_table_new(); // Convert parameter list to a hash table for (; params; params = params->next) { // Strip out Pacemaker-implemented parameters if (!crm_starts_with(params->key, "pcmk_") && strcmp(params->key, "provides") && strcmp(params->key, "stonith-timeout")) { g_hash_table_insert(params_table, strdup(params->key), strdup(params->value)); } } #if SUPPORT_CIBSECRETS rc = replace_secret_params(rsc_id, params_table); if (rc < 0) { crm_warn("Could not replace secret parameters for validation of %s: %s", agent, pcmk_strerror(rc)); } #endif if (output) { *output = NULL; } if (error_output) { *error_output = NULL; } switch (stonith_get_namespace(agent, namespace_s)) { case st_namespace_rhcs: rc = stonith__rhcs_validate(st, call_options, target, agent, params_table, timeout, output, error_output); break; #if HAVE_STONITH_STONITH_H case st_namespace_lha: rc = stonith__lha_validate(st, call_options, target, agent, params_table, timeout, output, error_output); break; #endif default: rc = -EINVAL; errno = EINVAL; crm_perror(LOG_ERR, "Agent %s not found or does not support validation", agent); break; } g_hash_table_destroy(params_table); return rc; } stonith_t * stonith_api_new(void) { stonith_t *new_stonith = NULL; stonith_private_t *private = NULL; new_stonith = calloc(1, sizeof(stonith_t)); private = calloc(1, sizeof(stonith_private_t)); new_stonith->st_private = private; private->stonith_op_callback_table = g_hash_table_new_full(g_direct_hash, g_direct_equal, NULL, stonith_destroy_op_callback); private->notify_list = NULL; new_stonith->call_id = 1; new_stonith->state = stonith_disconnected; new_stonith->cmds = calloc(1, sizeof(stonith_api_operations_t)); /* *INDENT-OFF* */ new_stonith->cmds->free = stonith_api_free; new_stonith->cmds->connect = stonith_api_signon; new_stonith->cmds->disconnect = stonith_api_signoff; new_stonith->cmds->list = stonith_api_list; new_stonith->cmds->monitor = stonith_api_monitor; new_stonith->cmds->status = stonith_api_status; new_stonith->cmds->fence = stonith_api_fence; new_stonith->cmds->confirm = stonith_api_confirm; new_stonith->cmds->history = stonith_api_history; new_stonith->cmds->list_agents = stonith_api_device_list; new_stonith->cmds->metadata = stonith_api_device_metadata; new_stonith->cmds->query = stonith_api_query; new_stonith->cmds->remove_device = stonith_api_remove_device; new_stonith->cmds->register_device = stonith_api_register_device; new_stonith->cmds->remove_level = stonith_api_remove_level; new_stonith->cmds->remove_level_full = stonith_api_remove_level_full; new_stonith->cmds->register_level = stonith_api_register_level; new_stonith->cmds->register_level_full = stonith_api_register_level_full; new_stonith->cmds->remove_callback = stonith_api_del_callback; new_stonith->cmds->register_callback = stonith_api_add_callback; new_stonith->cmds->remove_notification = stonith_api_del_notification; new_stonith->cmds->register_notification = stonith_api_add_notification; new_stonith->cmds->validate = stonith_api_validate; /* *INDENT-ON* */ return new_stonith; } stonith_key_value_t * stonith_key_value_add(stonith_key_value_t * head, const char *key, const char *value) { stonith_key_value_t *p, *end; p = calloc(1, sizeof(stonith_key_value_t)); if (key) { p->key = strdup(key); } if (value) { p->value = strdup(value); } end = head; while (end && end->next) { end = end->next; } if (end) { end->next = p; } else { head = p; } return head; } void stonith_key_value_freeall(stonith_key_value_t * head, int keys, int values) { stonith_key_value_t *p; while (head) { p = head->next; if (keys) { free(head->key); } if (values) { free(head->value); } free(head); head = p; } } #define api_log_open() openlog("stonith-api", LOG_CONS | LOG_NDELAY | LOG_PID, LOG_DAEMON) #define api_log(level, fmt, args...) syslog(level, "%s: "fmt, __FUNCTION__, args) int stonith_api_kick(uint32_t nodeid, const char *uname, int timeout, bool off) { char *name = NULL; const char *action = "reboot"; int rc = -EPROTO; stonith_t *st = NULL; enum stonith_call_options opts = st_opt_sync_call | st_opt_allow_suicide; api_log_open(); st = stonith_api_new(); if (st) { rc = st->cmds->connect(st, "stonith-api", NULL); if(rc != pcmk_ok) { api_log(LOG_ERR, "Connection failed, could not kick (%s) node %u/%s : %s (%d)", action, nodeid, uname, pcmk_strerror(rc), rc); } } if (uname != NULL) { name = strdup(uname); } else if (nodeid > 0) { opts |= st_opt_cs_nodeid; name = crm_itoa(nodeid); } if (off) { action = "off"; } if (rc == pcmk_ok) { rc = st->cmds->fence(st, opts, name, action, timeout, 0); if(rc != pcmk_ok) { api_log(LOG_ERR, "Could not kick (%s) node %u/%s : %s (%d)", action, nodeid, uname, pcmk_strerror(rc), rc); } else { api_log(LOG_NOTICE, "Node %u/%s kicked: %s ", nodeid, uname, action); } } if (st) { st->cmds->disconnect(st); stonith_api_delete(st); } free(name); return rc; } time_t stonith_api_time(uint32_t nodeid, const char *uname, bool in_progress) { int rc = 0; char *name = NULL; time_t when = 0; stonith_t *st = NULL; stonith_history_t *history = NULL, *hp = NULL; enum stonith_call_options opts = st_opt_sync_call; st = stonith_api_new(); if (st) { rc = st->cmds->connect(st, "stonith-api", NULL); if(rc != pcmk_ok) { api_log(LOG_NOTICE, "Connection failed: %s (%d)", pcmk_strerror(rc), rc); } } if (uname != NULL) { name = strdup(uname); } else if (nodeid > 0) { opts |= st_opt_cs_nodeid; name = crm_itoa(nodeid); } if (st && rc == pcmk_ok) { int entries = 0; int progress = 0; int completed = 0; rc = st->cmds->history(st, opts, name, &history, 120); for (hp = history; hp; hp = hp->next) { entries++; if (in_progress) { progress++; if (hp->state != st_done && hp->state != st_failed) { when = time(NULL); } } else if (hp->state == st_done) { completed++; if (hp->completed > when) { when = hp->completed; } } } stonith_history_free(history); if(rc == pcmk_ok) { api_log(LOG_INFO, "Found %d entries for %u/%s: %d in progress, %d completed", entries, nodeid, uname, progress, completed); } else { api_log(LOG_ERR, "Could not retrieve fence history for %u/%s: %s (%d)", nodeid, uname, pcmk_strerror(rc), rc); } } if (st) { st->cmds->disconnect(st); stonith_api_delete(st); } if(when) { api_log(LOG_INFO, "Node %u/%s last kicked at: %ld", nodeid, uname, (long int)when); } free(name); return when; } bool stonith_agent_exists(const char *agent, int timeout) { stonith_t *st = NULL; stonith_key_value_t *devices = NULL; stonith_key_value_t *dIter = NULL; bool rc = FALSE; if (agent == NULL) { return rc; } st = stonith_api_new(); st->cmds->list_agents(st, st_opt_sync_call, NULL, &devices, timeout == 0 ? 120 : timeout); for (dIter = devices; dIter != NULL; dIter = dIter->next) { if (crm_str_eq(dIter->value, agent, TRUE)) { rc = TRUE; break; } } stonith_key_value_freeall(devices, 1, 1); stonith_api_delete(st); return rc; } const char * stonith_action_str(const char *action) { if (action == NULL) { return "fencing"; } else if (!strcmp(action, "on")) { return "unfencing"; } else if (!strcmp(action, "off")) { return "turning off"; } else { return action; } } diff --git a/lib/fencing/st_output.c b/lib/fencing/st_output.c index ae3d046fee..a8d0a60d6b 100644 --- a/lib/fencing/st_output.c +++ b/lib/fencing/st_output.c @@ -1,235 +1,239 @@ /* * Copyright 2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include +#include static int fence_target_text(pcmk__output_t *out, va_list args) { const char *hostname = va_arg(args, const char *); const char *uuid = va_arg(args, const char *); const char *status = va_arg(args, const char *); pcmk__indented_printf(out, "%s\t%s\t%s\n", hostname, uuid, status); return 0; } static int fence_target_xml(pcmk__output_t *out, va_list args) { xmlNodePtr node = NULL; const char *hostname = va_arg(args, const char *); const char *uuid = va_arg(args, const char *); const char *status = va_arg(args, const char *); node = xmlNewNode(NULL, (pcmkXmlStr) "target"); xmlSetProp(node, (pcmkXmlStr) "hostname", (pcmkXmlStr) hostname); xmlSetProp(node, (pcmkXmlStr) "uuid", (pcmkXmlStr) uuid); xmlSetProp(node, (pcmkXmlStr) "status", (pcmkXmlStr) status); pcmk__xml_add_node(out, node); return 0; } static int last_fenced_text(pcmk__output_t *out, va_list args) { const char *target = va_arg(args, const char *); time_t when = va_arg(args, time_t); if (when) { pcmk__indented_printf(out, "Node %s last fenced at: %s", target, ctime(&when)); } else { pcmk__indented_printf(out, "Node %s has never been fenced\n", target); } return 0; } static int last_fenced_xml(pcmk__output_t *out, va_list args) { const char *target = va_arg(args, const char *); time_t when = va_arg(args, time_t); if (when) { crm_time_t *crm_when = crm_time_new(NULL); xmlNodePtr node = xmlNewNode(NULL, (pcmkXmlStr) "last-fenced"); char *buf = NULL; crm_time_set_timet(crm_when, &when); buf = crm_time_as_string(crm_when, crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone); xmlSetProp(node, (pcmkXmlStr) "target", (pcmkXmlStr) target); xmlSetProp(node, (pcmkXmlStr) "when", (pcmkXmlStr) buf); pcmk__xml_add_node(out, node); crm_time_free(crm_when); free(buf); } return 0; } static int stonith_event_text(pcmk__output_t *out, va_list args) { stonith_history_t *event = va_arg(args, stonith_history_t *); switch (event->state) { case st_failed: pcmk__indented_printf(out, "%s failed %s node %s on behalf of %s from %s at %s", event->delegate ? event->delegate : "We", stonith_action_str(event->action), event->target, event->client, event->origin, ctime(&event->completed)); break; case st_done: pcmk__indented_printf(out, "%s succeeded %s node %s on behalf of %s from %s at %s", event->delegate ? event->delegate : "This node", stonith_action_str(event->action), event->target, event->client, event->origin, ctime(&event->completed)); break; default: /* ocf:pacemaker:controld depends on "wishes to" being * in this output, when used with older versions of DLM * that don't report stateful_merge_wait */ pcmk__indented_printf(out, "%s at %s wishes to %s node %s - %d %lld\n", event->client, event->origin, stonith_action_str(event->action), event->target, event->state, (long long) event->completed); break; } return 0; } static int stonith_event_xml(pcmk__output_t *out, va_list args) { xmlNodePtr node = NULL; stonith_history_t *event = va_arg(args, stonith_history_t *); crm_time_t *crm_when = crm_time_new(NULL); char *buf = NULL; - node = xmlNewNode(NULL, (pcmkXmlStr) "stonith-event"); + node = xmlNewNode(NULL, (pcmkXmlStr) "fence_event"); switch (event->state) { case st_failed: xmlSetProp(node, (pcmkXmlStr) "status", (pcmkXmlStr) "failed"); break; case st_done: - xmlSetProp(node, (pcmkXmlStr) "status", (pcmkXmlStr) "done"); + xmlSetProp(node, (pcmkXmlStr) "status", (pcmkXmlStr) "success"); break; default: { char *state = crm_itoa(event->state); - xmlSetProp(node, (pcmkXmlStr) "status", (pcmkXmlStr) state); + xmlSetProp(node, (pcmkXmlStr) "status", (pcmkXmlStr) "pending"); + xmlSetProp(node, (pcmkXmlStr) "extended-status", (pcmkXmlStr) state); free(state); break; } } if (event->delegate != NULL) { xmlSetProp(node, (pcmkXmlStr) "delegate", (pcmkXmlStr) event->delegate); } xmlSetProp(node, (pcmkXmlStr) "action", (pcmkXmlStr) event->action); xmlSetProp(node, (pcmkXmlStr) "target", (pcmkXmlStr) event->target); xmlSetProp(node, (pcmkXmlStr) "client", (pcmkXmlStr) event->client); xmlSetProp(node, (pcmkXmlStr) "origin", (pcmkXmlStr) event->origin); - crm_time_set_timet(crm_when, &event->completed); - buf = crm_time_as_string(crm_when, crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone); - xmlSetProp(node, (pcmkXmlStr) "when", (pcmkXmlStr) buf); + if (event->state == st_failed || event->state == st_done) { + crm_time_set_timet(crm_when, &event->completed); + buf = crm_time_as_string(crm_when, crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone); + xmlSetProp(node, (pcmkXmlStr) "completed", (pcmkXmlStr) buf); + free(buf); + } pcmk__xml_add_node(out, node); crm_time_free(crm_when); - free(buf); return 0; } static int validate_agent_text(pcmk__output_t *out, va_list args) { const char *agent = va_arg(args, const char *); const char *device = va_arg(args, const char *); const char *output = va_arg(args, const char *); const char *error_output = va_arg(args, const char *); int rc = va_arg(args, int); if (device) { pcmk__indented_printf(out, "Validation of %s on %s %s\n", agent, device, rc ? "failed" : "succeeded"); } else { pcmk__indented_printf(out, "Validation of %s %s\n", agent, rc ? "failed" : "succeeded"); } if (output) { puts(output); } if (error_output) { puts(error_output); } return rc; } static int validate_agent_xml(pcmk__output_t *out, va_list args) { xmlNodePtr node = NULL; const char *agent = va_arg(args, const char *); const char *device = va_arg(args, const char *); const char *output = va_arg(args, const char *); const char *error_output = va_arg(args, const char *); int rc = va_arg(args, int); node = xmlNewNode(NULL, (pcmkXmlStr) "validate"); xmlSetProp(node, (pcmkXmlStr) "agent", (pcmkXmlStr) agent); if (device != NULL) { xmlSetProp(node, (pcmkXmlStr) "device", (pcmkXmlStr) device); } xmlSetProp(node, (pcmkXmlStr) "valid", (pcmkXmlStr) (rc ? "false" : "true")); pcmk__xml_push_parent(out, node); out->subprocess_output(out, rc, output, error_output); pcmk__xml_pop_parent(out); pcmk__xml_add_node(out, node); return rc; } static pcmk__message_entry_t fmt_functions[] = { { "fence-target", "text", fence_target_text }, { "fence-target", "xml", fence_target_xml }, { "last-fenced", "text", last_fenced_text }, { "last-fenced", "xml", last_fenced_xml }, { "stonith-event", "text", stonith_event_text }, { "stonith-event", "xml", stonith_event_xml }, { "validate", "text", validate_agent_text }, { "validate", "xml", validate_agent_xml }, { NULL, NULL, NULL } }; void stonith__register_messages(pcmk__output_t *out) { static bool registered = FALSE; if (!registered) { pcmk__register_messages(out, fmt_functions); registered = TRUE; } } diff --git a/pacemaker.spec.in b/pacemaker.spec.in index 09c925c6be..262fcca6e2 100644 --- a/pacemaker.spec.in +++ b/pacemaker.spec.in @@ -1,839 +1,839 @@ # Globals and defines to control package behavior (configure these as desired) ## User and group to use for nonprivileged services %global uname hacluster %global gname haclient ## Where to install Pacemaker documentation %global pcmk_docdir %{_docdir}/%{name} ## GitHub entity that distributes source (for ease of using a fork) %global github_owner ClusterLabs ## Upstream pacemaker version, and its package version (specversion ## can be incremented to build packages reliably considered "newer" ## than previously built packages with the same pcmkversion) %global pcmkversion 2.0.2 %global specversion 1 ## Upstream commit (or git tag, such as "Pacemaker-" plus the ## {pcmkversion} macro for an official release) to use for this package %global commit HEAD ## Since git v2.11, the extent of abbreviation is autoscaled by default ## (used to be constant of 7), so we need to convey it for non-tags, too. %global commit_abbrev 7 ## Python major version to use (2, 3, or 0 for auto-detect) %global python_major 0 # Define globals for convenient use later ## Workaround to use parentheses in other globals %global lparen ( %global rparen ) ## Short version of git commit %define shortcommit %(c=%{commit}; case ${c} in Pacemaker-*%{rparen} echo ${c:10};; *%{rparen} echo ${c:0:%{commit_abbrev}};; esac) ## Whether this is a tagged release %define tag_release %([ %{commit} != Pacemaker-%{shortcommit} ]; echo $?) ## Whether this is a release candidate (in case of a tagged release) %define pre_release %([ "%{tag_release}" -eq 0 ] || { case "%{shortcommit}" in *-rc[[:digit:]]*%{rparen} false;; esac; }; echo $?) ## Heuristic used to infer bleeding-edge deployments that are ## less likely to have working versions of the documentation tools %define bleeding %(test ! -e /etc/yum.repos.d/fedora-rawhide.repo; echo $?) ## Whether this platform defaults to using systemd as an init system ## (needs to be evaluated prior to BuildRequires being enumerated and ## installed as it's intended to conditionally select some of these, and ## for that there are only few indicators with varying reliability: ## - presence of systemd-defined macros (when building in a full-fledged ## environment, which is not the case with ordinary mock-based builds) ## - systemd-aware rpm as manifested with the presence of particular ## macro (rpm itself will trivially always be present when building) ## - existence of /usr/lib/os-release file, which is something heavily ## propagated by systemd project ## - when not good enough, there's always a possibility to check ## particular distro-specific macros (incl. version comparison) %define systemd_native (%{?_unitdir:1}%{!?_unitdir:0}%{nil \ } || %{?__transaction_systemd_inhibit:1}%{!?__transaction_systemd_inhibit:0}%{nil \ } || %(test -f /usr/lib/os-release; test $? -ne 0; echo $?)) %if 0%{?fedora} > 20 || 0%{?rhel} > 7 ## Base GnuTLS cipher priorities (presumably only the initial, required keyword) ## overridable with "rpmbuild --define 'pcmk_gnutls_priorities PRIORITY-SPEC'" %define gnutls_priorities %{?pcmk_gnutls_priorities}%{!?pcmk_gnutls_priorities:@SYSTEM} %endif # Python-related definitions ## Use Python 3 on certain platforms if major version not specified %if %{?python_major} == 0 %if 0%{?fedora} > 26 || 0%{?rhel} > 7 %global python_major 3 %endif %endif ## Turn off auto-compilation of Python files outside Python specific paths, ## so there's no risk that unexpected "__python" macro gets picked to do the ## RPM-native byte-compiling there (only "{_datadir}/pacemaker/tests" affected) ## -- distro-dependent tricks or automake's fallback to be applied there %if %{defined _python_bytecompile_extra} %global _python_bytecompile_extra 0 %else ### the statement effectively means no RPM-native byte-compiling will occur at ### all, so distro-dependent tricks for Python-specific packages to be applied %global __os_install_post %(echo '%{__os_install_post}' | { sed -e 's!/usr/lib[^[:space:]]*/brp-python-bytecompile[[:space:]].*$!!g'; }) %endif ## Values that differ by Python major version %if 0%{?python_major} > 2 %global python_name python3 %global python_path %{?__python3}%{!?__python3:/usr/bin/python%{?python3_pkgversion}%{!?python3_pkgversion:3}} %define python_site %{?python3_sitelib}%{!?python3_sitelib:%( %{python_path} -c 'from distutils.sysconfig import get_python_lib as gpl; print(gpl(1))' 2>/dev/null)} %else %if 0%{?python_major} > 1 %global python_name python2 %global python_path %{?__python2}%{!?__python2:/usr/bin/python%{?python2_pkgversion}%{!?python2_pkgversion:2}} %define python_site %{?python2_sitelib}%{!?python2_sitelib:%( %{python_path} -c 'from distutils.sysconfig import get_python_lib as gpl; print(gpl(1))' 2>/dev/null)} %else %global python_name python %global python_path %{?__python}%{!?__python:/usr/bin/python%{?python_pkgversion}} %define python_site %{?python_sitelib}%{!?python_sitelib:%( python -c 'from distutils.sysconfig import get_python_lib as gpl; print(gpl(1))' 2>/dev/null)} %endif %endif # Definitions for backward compatibility with older RPM versions ## Ensure the license macro behaves consistently (older RPM will otherwise ## overwrite it once it encounters "License:"). Courtesy Jason Tibbitts: ## https://pkgs.fedoraproject.org/cgit/rpms/epel-rpm-macros.git/tree/macros.zzz-epel?h=el6&id=e1adcb77 %if !%{defined _licensedir} %define description %{lua: rpm.define("license %doc") print("%description") } %endif # Define conditionals so that "rpmbuild --with " and # "rpmbuild --without " can enable and disable specific features ## Add option to enable support for stonith/external fencing agents %bcond_with stonithd ## Add option to create binaries suitable for use with profiling tools %bcond_with profiling ## Add option to create binaries with coverage analysis %bcond_with coverage ## Add option to skip generating documentation ## (the build tools aren't available everywhere) %bcond_without doc ## Add option to prefix package version with "0." ## (so later "official" packages will be considered updates) %bcond_with pre_release ## Add option to ship Upstart job files %bcond_with upstart_job ## Add option to turn off hardening of libraries and daemon executables %bcond_without hardening ## Add option to disable links for legacy daemon names %bcond_without legacy_links # Keep sane profiling data if requested %if %{with profiling} ## Disable -debuginfo package and stripping binaries/libraries %define debug_package %{nil} %endif # Define the release version # (do not look at externally enforced pre-release flag for tagged releases # as only -rc tags, captured with the second condition, implies that then) %if (!%{tag_release} && %{with pre_release}) || 0%{pre_release} %if 0%{pre_release} %define pcmk_release 0.%{specversion}.%(s=%{shortcommit}; echo ${s: -3}) %else %define pcmk_release 0.%{specversion}.%{shortcommit}.git %endif %else %if 0%{tag_release} %define pcmk_release %{specversion} %else %define pcmk_release %{specversion}.%{shortcommit}.git %endif %endif Name: pacemaker Summary: Scalable High-Availability cluster resource manager Version: %{pcmkversion} Release: %{pcmk_release}%{?dist} %if %{defined _unitdir} License: GPLv2+ and LGPLv2+ %else # initscript is Revised BSD License: GPLv2+ and LGPLv2+ and BSD %endif Url: http://www.clusterlabs.org Group: System Environment/Daemons # Hint: use "spectool -s 0 pacemaker.spec" (rpmdevtools) to check the final URL: # https://github.com/ClusterLabs/pacemaker/archive/e91769e5a39f5cb2f7b097d3c612368f0530535e/pacemaker-e91769e.tar.gz Source0: https://github.com/%{github_owner}/%{name}/archive/%{commit}/%{name}-%{shortcommit}.tar.gz Requires: resource-agents Requires: %{name}-libs%{?_isa} = %{version}-%{release} Requires: %{name}-cluster-libs%{?_isa} = %{version}-%{release} Requires: %{name}-cli = %{version}-%{release} %if !%{defined _unitdir} Requires: procps-ng Requires: psmisc %endif %{?systemd_requires} Requires: %{python_path} BuildRequires: %{python_name}-devel # Pacemaker requires a minimum libqb functionality Requires: libqb >= 0.13.0 BuildRequires: libqb-devel >= 0.13.0 # Basics required for the build (even if usually satisfied through other BRs) BuildRequires: coreutils findutils grep sed # Required for core functionality BuildRequires: automake autoconf gcc libtool pkgconfig libtool-ltdl-devel BuildRequires: pkgconfig(glib-2.0) >= 2.16 BuildRequires: libxml2-devel libxslt-devel libuuid-devel BuildRequires: bzip2-devel # Enables optional functionality BuildRequires: ncurses-devel docbook-style-xsl BuildRequires: help2man gnutls-devel pam-devel pkgconfig(dbus-1) %if %{systemd_native} BuildRequires: pkgconfig(systemd) %endif Requires: corosync >= 2.0.0 BuildRequires: corosynclib-devel >= 2.0.0 %if %{with stonithd} BuildRequires: cluster-glue-libs-devel %endif ## (note no avoiding effect when building through non-customized mock) %if !%{bleeding} %if %{with doc} BuildRequires: inkscape asciidoc publican %endif %endif Provides: pcmk-cluster-manager = %{version}-%{release} Provides: pcmk-cluster-manager%{?_isa} = %{version}-%{release} # Bundled bits ## Pacemaker uses the crypto/md5-buffer module from gnulib %if 0%{?fedora} || 0%{?rhel} Provides: bundled(gnulib) %endif %description Pacemaker is an advanced, scalable High-Availability cluster resource manager. It supports more than 16 node clusters with significant capabilities for managing resources and dependencies. It will run scripts at initialization, when machines go up or down, when related resources fail and can be configured to periodically check resource health. Available rpmbuild rebuild options: --with(out) : coverage doc stonithd hardening pre_release profiling upstart_job %package cli License: GPLv2+ and LGPLv2+ Summary: Command line tools for controlling Pacemaker clusters Group: System Environment/Daemons Requires: %{name}-libs%{?_isa} = %{version}-%{release} %if 0%{?fedora} > 22 || 0%{?rhel} > 7 Recommends: pcmk-cluster-manager = %{version}-%{release} %endif Requires: perl-TimeDate Requires: procps-ng Requires: psmisc Requires(post):coreutils %description cli Pacemaker is an advanced, scalable High-Availability cluster resource manager. The %{name}-cli package contains command line tools that can be used to query and control the cluster from machines that may, or may not, be part of the cluster. %package libs License: GPLv2+ and LGPLv2+ Summary: Core Pacemaker libraries Group: System Environment/Daemons Requires(pre): shadow-utils Requires: %{name}-schemas = %{version}-%{release} # sbd 1.4.0+ supports the libpe_status API for pe_working_set_t Conflicts: sbd < 1.4.0 %description libs Pacemaker is an advanced, scalable High-Availability cluster resource manager. The %{name}-libs package contains shared libraries needed for cluster nodes and those just running the CLI tools. %package cluster-libs License: GPLv2+ and LGPLv2+ Summary: Cluster Libraries used by Pacemaker Group: System Environment/Daemons Requires: %{name}-libs%{?_isa} = %{version}-%{release} %description cluster-libs Pacemaker is an advanced, scalable High-Availability cluster resource manager. The %{name}-cluster-libs package contains cluster-aware shared libraries needed for nodes that will form part of the cluster nodes. %package remote %if %{defined _unitdir} License: GPLv2+ and LGPLv2+ %else # initscript is Revised BSD License: GPLv2+ and LGPLv2+ and BSD %endif Summary: Pacemaker remote daemon for non-cluster nodes Group: System Environment/Daemons Requires: %{name}-libs%{?_isa} = %{version}-%{release} Requires: %{name}-cli = %{version}-%{release} Requires: resource-agents %if !%{defined _unitdir} Requires: procps-ng %endif # -remote can be fully independent of systemd %{?systemd_ordering}%{!?systemd_ordering:%{?systemd_requires}} Provides: pcmk-cluster-manager = %{version}-%{release} Provides: pcmk-cluster-manager%{?_isa} = %{version}-%{release} %description remote Pacemaker is an advanced, scalable High-Availability cluster resource manager. The %{name}-remote package contains the Pacemaker Remote daemon which is capable of extending pacemaker functionality to remote nodes not running the full corosync/cluster stack. %package libs-devel License: GPLv2+ and LGPLv2+ Summary: Pacemaker development package Group: Development/Libraries Requires: %{name}-libs%{?_isa} = %{version}-%{release} Requires: %{name}-cluster-libs%{?_isa} = %{version}-%{release} Requires: libuuid-devel%{?_isa} libtool-ltdl-devel%{?_isa} Requires: libxml2-devel%{?_isa} libxslt-devel%{?_isa} Requires: bzip2-devel%{?_isa} glib2-devel%{?_isa} Requires: libqb-devel%{?_isa} Requires: corosynclib-devel%{?_isa} >= 2.0.0 %description libs-devel Pacemaker is an advanced, scalable High-Availability cluster resource manager. The %{name}-libs-devel package contains headers and shared libraries for developing tools for Pacemaker. %package cts License: GPLv2+ and LGPLv2+ Summary: Test framework for cluster-related technologies like Pacemaker Group: System Environment/Daemons Requires: %{python_path} Requires: %{name}-libs = %{version}-%{release} Requires: procps-ng Requires: psmisc BuildArch: noarch # systemd python bindings are separate package in some distros %if %{defined systemd_requires} %if 0%{?fedora} > 22 || 0%{?rhel} > 7 Requires: %{python_name}-systemd %else %if 0%{?fedora} > 20 || 0%{?rhel} > 6 Requires: systemd-python %endif %endif %endif %description cts Test framework for cluster-related technologies like Pacemaker %package doc License: CC-BY-SA-4.0 Summary: Documentation for Pacemaker Group: Documentation BuildArch: noarch %description doc Documentation for Pacemaker. Pacemaker is an advanced, scalable High-Availability cluster resource manager. %package schemas License: GPLv2+ Summary: Schemas and upgrade stylesheets for Pacemaker BuildArch: noarch %description schemas Schemas and upgrade stylesheets for Pacemaker Pacemaker is an advanced, scalable High-Availability cluster resource manager. %prep %setup -q -n %{name}-%{commit} %build # Early versions of autotools (e.g. RHEL <= 5) do not support --docdir export docdir=%{pcmk_docdir} export systemdunitdir=%{?_unitdir}%{!?_unitdir:no} %if %{with hardening} # prefer distro-provided hardening flags in case they are defined # through _hardening_{c,ld}flags macros, configure script will # use its own defaults otherwise; if such hardenings are completely # undesired, rpmbuild using "--without hardening" # (or "--define '_without_hardening 1'") export CFLAGS_HARDENED_EXE="%{?_hardening_cflags}" export CFLAGS_HARDENED_LIB="%{?_hardening_cflags}" export LDFLAGS_HARDENED_EXE="%{?_hardening_ldflags}" export LDFLAGS_HARDENED_LIB="%{?_hardening_ldflags}" %endif ./autogen.sh %{configure} \ PYTHON=%{python_path} \ %{!?with_hardening: --disable-hardening} \ %{!?with_legacy_links: --disable-legacy-links} \ %{?with_profiling: --with-profiling} \ %{?with_coverage: --with-coverage} \ %{!?with_doc: --with-brand=} \ %{?gnutls_priorities: --with-gnutls-priorities="%{gnutls_priorities}"} \ --with-initdir=%{_initrddir} \ --localstatedir=%{_var} \ --with-version=%{version}-%{release} %if 0%{?suse_version} >= 1200 # Fedora handles rpath removal automagically sed -i 's|^hardcode_libdir_flag_spec=.*|hardcode_libdir_flag_spec=""|g' libtool sed -i 's|^runpath_var=LD_RUN_PATH|runpath_var=DIE_RPATH_DIE|g' libtool %endif make %{_smp_mflags} V=1 all %check { cts/cts-scheduler --run load-stopped-loop \ && cts/cts-cli \ && touch .CHECKED } 2>&1 | sed 's/[fF]ail/faiil/g' # prevent false positives in rpmlint [ -f .CHECKED ] && rm -f -- .CHECKED exit $? # TODO remove when rpm<4.14 compatibility irrelevant %install # skip automake-native Python byte-compilation, since RPM-native one (possibly # distro-confined to Python-specific directories, which is currently the only # relevant place, anyway) assures proper intrinsic alignment with wider system # (such as with py_byte_compile macro, which is concurrent Fedora/EL specific) make install \ DESTDIR=%{buildroot} V=1 docdir=%{pcmk_docdir} \ %{?_python_bytecompile_extra:%{?py_byte_compile:am__py_compile=true}} mkdir -p ${RPM_BUILD_ROOT}%{_sysconfdir}/sysconfig install -m 644 daemons/pacemakerd/pacemaker.sysconfig ${RPM_BUILD_ROOT}%{_sysconfdir}/sysconfig/pacemaker install -m 644 tools/crm_mon.sysconfig ${RPM_BUILD_ROOT}%{_sysconfdir}/sysconfig/crm_mon %if %{with upstart_job} mkdir -p ${RPM_BUILD_ROOT}%{_sysconfdir}/init install -m 644 pacemakerd/pacemaker.upstart ${RPM_BUILD_ROOT}%{_sysconfdir}/init/pacemaker.conf install -m 644 pacemakerd/pacemaker.combined.upstart ${RPM_BUILD_ROOT}%{_sysconfdir}/init/pacemaker.combined.conf install -m 644 tools/crm_mon.upstart ${RPM_BUILD_ROOT}%{_sysconfdir}/init/crm_mon.conf %endif %if %{defined _unitdir} mkdir -p ${RPM_BUILD_ROOT}%{_localstatedir}/lib/rpm-state/%{name} %endif # Don't package static libs find %{buildroot} -name '*.a' -type f -print0 | xargs -0 rm -f find %{buildroot} -name '*.la' -type f -print0 | xargs -0 rm -f # For now, don't package the servicelog-related binaries built only for # ppc64le when certain dependencies are installed. If they get more exercise by # advanced users, we can reconsider. rm -f %{buildroot}/%{_sbindir}/notifyServicelogEvent rm -f %{buildroot}/%{_sbindir}/ipmiservicelogd # Don't ship init scripts for systemd based platforms %if %{defined _unitdir} rm -f %{buildroot}/%{_initrddir}/pacemaker rm -f %{buildroot}/%{_initrddir}/pacemaker_remote %endif # Byte-compile Python sources where suitable and the distro procedures known %if %{defined py_byte_compile} %{py_byte_compile %{python_path} %{buildroot}%{_datadir}/pacemaker/tests} %if !%{defined _python_bytecompile_extra} %{py_byte_compile %{python_path} %{buildroot}%{python_site}/cts} %endif %endif %if %{with coverage} GCOV_BASE=%{buildroot}/%{_var}/lib/pacemaker/gcov mkdir -p $GCOV_BASE find . -name '*.gcno' -type f | while read F ; do D=`dirname $F` mkdir -p ${GCOV_BASE}/$D cp $F ${GCOV_BASE}/$D done %endif %post %if %{defined _unitdir} %systemd_post pacemaker.service %else /sbin/chkconfig --add pacemaker || : %endif %preun %if %{defined _unitdir} %systemd_preun pacemaker.service %else /sbin/service pacemaker stop >/dev/null 2>&1 || : if [ "$1" -eq 0 ]; then # Package removal, not upgrade /sbin/chkconfig --del pacemaker || : fi %endif %postun %if %{defined _unitdir} %systemd_postun_with_restart pacemaker.service %endif %pre remote %if %{defined _unitdir} # Stop the service before anything is touched, and remember to restart # it as one of the last actions (compared to using systemd_postun_with_restart, # this avoids suicide when sbd is in use) systemctl --quiet is-active pacemaker_remote if [ $? -eq 0 ] ; then mkdir -p %{_localstatedir}/lib/rpm-state/%{name} touch %{_localstatedir}/lib/rpm-state/%{name}/restart_pacemaker_remote systemctl stop pacemaker_remote >/dev/null 2>&1 else rm -f %{_localstatedir}/lib/rpm-state/%{name}/restart_pacemaker_remote fi %endif %post remote %if %{defined _unitdir} %systemd_post pacemaker_remote.service %else /sbin/chkconfig --add pacemaker_remote || : %endif %preun remote %if %{defined _unitdir} %systemd_preun pacemaker_remote.service %else /sbin/service pacemaker_remote stop >/dev/null 2>&1 || : if [ "$1" -eq 0 ]; then # Package removal, not upgrade /sbin/chkconfig --del pacemaker_remote || : fi %endif %postun remote %if %{defined _unitdir} # This next line is a no-op, because we stopped the service earlier, but # we leave it here because it allows us to revert to the standard behavior # in the future if desired %systemd_postun_with_restart pacemaker_remote.service # Explicitly take care of removing the flag-file(s) upon final removal if [ "$1" -eq 0 ] ; then rm -f %{_localstatedir}/lib/rpm-state/%{name}/restart_pacemaker_remote fi %endif %posttrans remote %if %{defined _unitdir} if [ -e %{_localstatedir}/lib/rpm-state/%{name}/restart_pacemaker_remote ] ; then systemctl start pacemaker_remote >/dev/null 2>&1 rm -f %{_localstatedir}/lib/rpm-state/%{name}/restart_pacemaker_remote fi %endif %post cli %if %{defined _unitdir} %systemd_post crm_mon.service %endif if [ "$1" -eq 2 ]; then # Package upgrade, not initial install: # Move any pre-2.0 logs to new location to ensure they get rotated { mv -fbS.rpmsave %{_var}/log/pacemaker.log* %{_var}/log/pacemaker \ || mv -f %{_var}/log/pacemaker.log* %{_var}/log/pacemaker } >/dev/null 2>/dev/null || : fi %preun cli %if %{defined _unitdir} %systemd_preun crm_mon.service %endif %postun cli %if %{defined _unitdir} %systemd_postun_with_restart crm_mon.service %endif %pre libs getent group %{gname} >/dev/null || groupadd -r %{gname} -g 189 getent passwd %{uname} >/dev/null || useradd -r -g %{gname} -u 189 -s /sbin/nologin -c "cluster user" %{uname} exit 0 %if %{defined ldconfig_scriptlets} %ldconfig_scriptlets libs %ldconfig_scriptlets cluster-libs %else %post libs -p /sbin/ldconfig %postun libs -p /sbin/ldconfig %post cluster-libs -p /sbin/ldconfig %postun cluster-libs -p /sbin/ldconfig %endif %files ########################################################### %config(noreplace) %{_sysconfdir}/sysconfig/pacemaker %{_sbindir}/pacemakerd %if %{defined _unitdir} %{_unitdir}/pacemaker.service %else %{_initrddir}/pacemaker %endif %exclude %{_libexecdir}/pacemaker/cts-log-watcher %exclude %{_libexecdir}/pacemaker/cts-support %exclude %{_sbindir}/pacemaker-remoted %if %{with legacy_links} %exclude %{_sbindir}/pacemaker_remoted %endif %{_libexecdir}/pacemaker/* %{_sbindir}/crm_attribute %{_sbindir}/crm_master %{_sbindir}/fence_legacy %doc %{_mandir}/man7/pacemaker-controld.* %doc %{_mandir}/man7/pacemaker-schedulerd.* %doc %{_mandir}/man7/pacemaker-fenced.* %doc %{_mandir}/man7/ocf_pacemaker_controld.* %doc %{_mandir}/man7/ocf_pacemaker_o2cb.* %doc %{_mandir}/man7/ocf_pacemaker_remote.* %doc %{_mandir}/man8/crm_attribute.* %doc %{_mandir}/man8/crm_master.* %doc %{_mandir}/man8/fence_legacy.* %doc %{_mandir}/man8/pacemakerd.* %doc %{_datadir}/pacemaker/alerts %license licenses/GPLv2 %doc COPYING %doc ChangeLog %dir %attr (750, %{uname}, %{gname}) %{_var}/lib/pacemaker/cib %dir %attr (750, %{uname}, %{gname}) %{_var}/lib/pacemaker/pengine /usr/lib/ocf/resource.d/pacemaker/controld /usr/lib/ocf/resource.d/pacemaker/o2cb /usr/lib/ocf/resource.d/pacemaker/remote %if %{with upstart_job} %config(noreplace) %{_sysconfdir}/init/pacemaker.conf %config(noreplace) %{_sysconfdir}/init/pacemaker.combined.conf %endif %files cli %dir %attr (750, root, %{gname}) %{_sysconfdir}/pacemaker %config(noreplace) %{_sysconfdir}/logrotate.d/pacemaker %config(noreplace) %{_sysconfdir}/sysconfig/crm_mon %if %{defined _unitdir} %{_unitdir}/crm_mon.service %endif %if %{with upstart_job} %config(noreplace) %{_sysconfdir}/init/crm_mon.conf %endif %{_sbindir}/attrd_updater %{_sbindir}/cibadmin %{_sbindir}/crm_diff %{_sbindir}/crm_error %{_sbindir}/crm_failcount %{_sbindir}/crm_mon %{_sbindir}/crm_node %{_sbindir}/crm_resource %{_sbindir}/crm_rule %{_sbindir}/crm_standby %{_sbindir}/crm_verify %{_sbindir}/crmadmin %{_sbindir}/iso8601 %{_sbindir}/crm_shadow %{_sbindir}/crm_simulate %{_sbindir}/crm_report %{_sbindir}/crm_ticket %{_sbindir}/stonith_admin -%exclude %{_datadir}/pacemaker/alerts -%exclude %{_datadir}/pacemaker/tests -%{_datadir}/pacemaker -%exclude %{_datadir}/pacemaker/*.rng -%exclude %{_datadir}/pacemaker/*.xsl +# "dirname" is owned by -schemas, which is a prerequisite +%{_datadir}/pacemaker/report.collector +%{_datadir}/pacemaker/report.common +# XXX "dirname" is not owned by any prerequisite %{_datadir}/snmp/mibs/PCMK-MIB.txt %exclude /usr/lib/ocf/resource.d/pacemaker/controld %exclude /usr/lib/ocf/resource.d/pacemaker/o2cb %exclude /usr/lib/ocf/resource.d/pacemaker/remote %dir /usr/lib/ocf %dir /usr/lib/ocf/resource.d /usr/lib/ocf/resource.d/pacemaker %doc %{_mandir}/man7/* %exclude %{_mandir}/man7/pacemaker-controld.* %exclude %{_mandir}/man7/pacemaker-schedulerd.* %exclude %{_mandir}/man7/pacemaker-fenced.* %exclude %{_mandir}/man7/ocf_pacemaker_controld.* %exclude %{_mandir}/man7/ocf_pacemaker_o2cb.* %exclude %{_mandir}/man7/ocf_pacemaker_remote.* %doc %{_mandir}/man8/* %exclude %{_mandir}/man8/crm_attribute.* %exclude %{_mandir}/man8/crm_master.* %exclude %{_mandir}/man8/fence_legacy.* %exclude %{_mandir}/man8/pacemakerd.* %exclude %{_mandir}/man8/pacemaker-remoted.* %license licenses/GPLv2 %doc COPYING %doc ChangeLog %dir %attr (750, %{uname}, %{gname}) %{_var}/lib/pacemaker %dir %attr (750, %{uname}, %{gname}) %{_var}/lib/pacemaker/blackbox %dir %attr (750, %{uname}, %{gname}) %{_var}/lib/pacemaker/cores %dir %attr (770, %{uname}, %{gname}) %{_var}/log/pacemaker %dir %attr (770, %{uname}, %{gname}) %{_var}/log/pacemaker/bundles %files libs %{_libdir}/libcib.so.* %{_libdir}/liblrmd.so.* %{_libdir}/libcrmservice.so.* %{_libdir}/libcrmcommon.so.* %{_libdir}/libpe_status.so.* %{_libdir}/libpe_rules.so.* %{_libdir}/libpacemaker.so.* %{_libdir}/libstonithd.so.* %license licenses/LGPLv2.1 %doc COPYING %doc ChangeLog %files cluster-libs %{_libdir}/libcrmcluster.so.* %license licenses/LGPLv2.1 %doc COPYING %doc ChangeLog %files remote %config(noreplace) %{_sysconfdir}/sysconfig/pacemaker %if %{defined _unitdir} # state directory is shared between the subpackets # let rpm take care of removing it once it isn't # referenced anymore and empty %ghost %dir %{_localstatedir}/lib/rpm-state/%{name} %{_unitdir}/pacemaker_remote.service %else %{_initrddir}/pacemaker_remote %endif %{_sbindir}/pacemaker-remoted %if %{with legacy_links} %{_sbindir}/pacemaker_remoted %endif %{_mandir}/man8/pacemaker-remoted.* %license licenses/GPLv2 %doc COPYING %doc ChangeLog %files doc %doc %{pcmk_docdir} %license licenses/CC-BY-SA-4.0 %files cts %{python_site}/cts %{_datadir}/pacemaker/tests %{_libexecdir}/pacemaker/cts-log-watcher %{_libexecdir}/pacemaker/cts-support %license licenses/GPLv2 %doc COPYING %doc ChangeLog %files libs-devel %{_includedir}/pacemaker %{_libdir}/*.so %if %{with coverage} %{_var}/lib/pacemaker/gcov %endif %{_libdir}/pkgconfig/*.pc %license licenses/LGPLv2.1 %doc COPYING %doc ChangeLog %files schemas %license licenses/GPLv2 +%dir %{_datadir}/pacemaker %{_datadir}/pacemaker/*.rng %{_datadir}/pacemaker/*.xsl -%{_datadir}/pacemaker/api/*.rng +%{_datadir}/pacemaker/api %changelog diff --git a/tools/stonith_admin.c b/tools/stonith_admin.c index 9cfd8adaa9..d960fb13e4 100644 --- a/tools/stonith_admin.c +++ b/tools/stonith_admin.c @@ -1,785 +1,786 @@ /* * Copyright 2009-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include #include #include /* *INDENT-OFF* */ static struct crm_option long_options[] = { { "help", no_argument, NULL, '?', "\tDisplay this text and exit." }, { "version", no_argument, NULL, '$', "\tDisplay version information and exit." }, { "verbose", no_argument, NULL, 'V', "\tIncrease debug output (may be specified multiple times)." }, { "quiet", no_argument, NULL, 'q', "\tBe less descriptive in output." }, { "cleanup", no_argument, NULL, 'c', "\tCleanup wherever appropriate. Requires: --history." }, { "broadcast", no_argument, NULL, 'b', "Broadcast wherever appropriate." }, PCMK__OUTPUT_OPTIONS("text, xml"), { "-spacer-", no_argument, NULL, '-', "\nDevice definition commands:" }, { "register", required_argument, NULL, 'R', "Register the named stonith device. Requires: --agent.\n" "\t\t\tOptional: --option, --env-option." }, { "deregister", required_argument, NULL, 'D', "De-register the named stonith device." }, { "register-level", required_argument, NULL, 'r', "Register a stonith level for the named target,\n" "\t\t\tspecified as one of NAME, @PATTERN, or ATTR=VALUE.\n" "\t\t\tRequires: --index and one or more --device entries." }, { "deregister-level", required_argument, NULL, 'd', "Unregister a stonith level for the named target,\n" "\t\t\tspecified as for --register-level. Requires: --index." }, { "-spacer-", no_argument, NULL, '-', "\nQueries:" }, { "list", required_argument, NULL, 'l', "List devices that can terminate the specified host.\n" "\t\t\tOptional: --timeout." }, { "list-registered", no_argument, NULL, 'L', "List all registered devices. Optional: --timeout." }, { "list-installed", no_argument, NULL, 'I', "List all installed devices. Optional: --timeout." }, { "list-targets", required_argument, NULL, 's', "List the targets that can be fenced by the\n" "\t\t\tnamed device. Optional: --timeout." }, { "metadata", no_argument, NULL, 'M', "\tShow agent metadata. Requires: --agent.\n" "\t\t\tOptional: --timeout." }, { "query", required_argument, NULL, 'Q', "Check the named device's status. Optional: --timeout." }, { "history", required_argument, NULL, 'H', "Show last successful fencing operation for named node\n" "\t\t\t(or '*' for all nodes). Optional: --timeout, --cleanup,\n" "\t\t\t--quiet (show only the operation's epoch timestamp),\n" "\t\t\t--verbose (show all recorded and pending operations),\n" "\t\t\t--broadcast (update history from all nodes available)." }, { "last", required_argument, NULL, 'h', "Indicate when the named node was last fenced.\n" "\t\t\tOptional: --as-node-id." }, { "validate", no_argument, NULL, 'K', "\tValidate a fence device configuration.\n" "\t\t\tRequires: --agent. Optional: --option, --env-option,\n" "\t\t\t--quiet (print no output, only return status).\n" }, { "-spacer-", no_argument, NULL, '-', "\nFencing Commands:" }, { "fence", required_argument, NULL, 'F', "Fence named host. Optional: --timeout, --tolerance." }, { "unfence", required_argument, NULL, 'U', "Unfence named host. Optional: --timeout, --tolerance." }, { "reboot", required_argument, NULL, 'B', "Reboot named host. Optional: --timeout, --tolerance." }, { "confirm", required_argument, NULL, 'C', "Tell cluster that named host is now safely down." }, { "-spacer-", no_argument, NULL, '-', "\nAdditional Options:" }, { "agent", required_argument, NULL, 'a', "The agent to use (for example, fence_xvm;\n" "\t\t\twith --register, --metadata, --validate)." }, { "option", required_argument, NULL, 'o', "Specify a device configuration parameter as NAME=VALUE\n" "\t\t\t(may be specified multiple times; with --register,\n" "\t\t\t--validate)." }, { "env-option", required_argument, NULL, 'e', "Specify a device configuration parameter with the\n" "\t\t\tspecified name, using the value of the\n" "\t\t\tenvironment variable of the same name prefixed with\n" "\t\t\tOCF_RESKEY_ (may be specified multiple times;\n" "\t\t\twith --register, --validate)." }, { "tag", required_argument, NULL, 'T', "Identify fencing operations in logs with the specified\n" "\t\t\ttag; useful when multiple entities might invoke\n" "\t\t\tstonith_admin (used with most commands)." }, { "device", required_argument, NULL, 'v', "Device ID (with --register-level, device to associate with\n" "\t\t\ta given host and level; may be specified multiple times)" #if SUPPORT_CIBSECRETS "\n\t\t\t(with --validate, name to use to load CIB secrets)" #endif "." }, { "index", required_argument, NULL, 'i', "The stonith level (1-9) (with --register-level,\n" "\t\t\t--deregister-level)." }, { "timeout", required_argument, NULL, 't', "Operation timeout in seconds (default 120;\n" "\t\t\tused with most commands)." }, { "as-node-id", no_argument, NULL, 'n', "(Advanced) The supplied node is the corosync node ID\n" "\t\t\t(with --last)." }, { "tolerance", required_argument, NULL, 0, "(Advanced) Do nothing if an equivalent --fence request\n" "\t\t\tsucceeded less than this many seconds earlier\n" "\t\t\t(with --fence, --unfence, --reboot)." }, { 0, 0, 0, 0 } }; /* *INDENT-ON* */ static int st_opts = st_opt_sync_call | st_opt_allow_suicide; static GMainLoop *mainloop = NULL; struct { stonith_t *st; const char *target; const char *action; char *name; int timeout; int tolerance; int rc; } async_fence_data; static int try_mainloop_connect(void) { stonith_t *st = async_fence_data.st; int tries = 10; int i = 0; int rc = 0; for (i = 0; i < tries; i++) { crm_debug("Connecting as %s", async_fence_data.name); rc = st->cmds->connect(st, async_fence_data.name, NULL); if (!rc) { crm_debug("stonith client connection established"); return 0; } else { crm_debug("stonith client connection failed"); } sleep(1); } crm_err("Could not connect to the fencer"); return -1; } static void notify_callback(stonith_t * st, stonith_event_t * e) { if (e->result != pcmk_ok) { return; } if (safe_str_eq(async_fence_data.target, e->target) && safe_str_eq(async_fence_data.action, e->action)) { async_fence_data.rc = e->result; g_main_loop_quit(mainloop); } } static void fence_callback(stonith_t * stonith, stonith_callback_data_t * data) { async_fence_data.rc = data->rc; g_main_loop_quit(mainloop); } static gboolean async_fence_helper(gpointer user_data) { stonith_t *st = async_fence_data.st; int call_id = 0; if (try_mainloop_connect()) { g_main_loop_quit(mainloop); return TRUE; } st->cmds->register_notification(st, T_STONITH_NOTIFY_FENCE, notify_callback); call_id = st->cmds->fence(st, st_opt_allow_suicide, async_fence_data.target, async_fence_data.action, async_fence_data.timeout, async_fence_data.tolerance); if (call_id < 0) { g_main_loop_quit(mainloop); return TRUE; } st->cmds->register_callback(st, call_id, async_fence_data.timeout, st_opt_timeout_updates, NULL, "callback", fence_callback); return TRUE; } static int mainloop_fencing(stonith_t * st, const char *target, const char *action, int timeout, int tolerance) { crm_trigger_t *trig; async_fence_data.st = st; async_fence_data.target = target; async_fence_data.action = action; async_fence_data.timeout = timeout; async_fence_data.tolerance = tolerance; async_fence_data.rc = -1; trig = mainloop_add_trigger(G_PRIORITY_HIGH, async_fence_helper, NULL); mainloop_set_trigger(trig); mainloop = g_main_loop_new(NULL, FALSE); g_main_loop_run(mainloop); return async_fence_data.rc; } static int handle_level(stonith_t *st, char *target, int fence_level, stonith_key_value_t *devices, bool added) { char *node = NULL; char *pattern = NULL; char *name = NULL; char *value = strchr(target, '='); /* Determine if targeting by attribute, node name pattern or node name */ if (value != NULL) { name = target; *value++ = '\0'; } else if (*target == '@') { pattern = target + 1; } else { node = target; } /* Register or unregister level as appropriate */ if (added) { return st->cmds->register_level_full(st, st_opts, node, pattern, name, value, fence_level, devices); } return st->cmds->remove_level_full(st, st_opts, node, pattern, name, value, fence_level); } static int handle_history(stonith_t *st, const char *target, int timeout, int quiet, int verbose, int cleanup, int broadcast, pcmk__output_t *out) { stonith_history_t *history = NULL, *hp, *latest = NULL; int rc = 0; if (!quiet) { if (cleanup) { out->info(out, "cleaning up fencing-history%s%s", target ? " for node " : "", target ? target : ""); } if (broadcast) { out->info(out, "gather fencing-history from all nodes"); } } rc = st->cmds->history(st, st_opts | (cleanup?st_opt_cleanup:0) | (broadcast?st_opt_broadcast:0), (safe_str_eq(target, "*")? NULL : target), &history, timeout); out->begin_list(out, "Fencing history", "event", "events"); for (hp = history; hp; hp = hp->next) { if (hp->state == st_done) { latest = hp; } if (quiet || !verbose) { continue; } out->message(out, "stonith-event", hp); } if (latest) { if (quiet && out->supports_quiet) { out->info(out, "%lld", (long long) latest->completed); } else if (!verbose) { // already printed if verbose out->message(out, "stonith-event", latest); } } out->end_list(out); stonith_history_free(history); return rc; } static int validate(stonith_t *st, const char *agent, const char *id, stonith_key_value_t *params, int timeout, int quiet, pcmk__output_t *out) { int rc = 1; char *output = NULL; char *error_output = NULL; rc = st->cmds->validate(st, st_opt_sync_call, id, NULL, agent, params, timeout, &output, &error_output); if (quiet) { return rc; } out->message(out, "validate", agent, id, output, error_output, rc); return rc; } int main(int argc, char **argv) { int flag; int rc = 0; int quiet = 0; int cleanup = 0; int broadcast = 0; int verbose = 0; int argerr = 0; int timeout = 120; int option_index = 0; int fence_level = 0; int no_connect = 0; int tolerance = 0; int as_nodeid = FALSE; bool required_agent = false; char *name = NULL; char *value = NULL; char *target = NULL; char *lists = NULL; const char *agent = NULL; const char *device = NULL; const char *longname = NULL; char action = 0; crm_exit_t exit_code = CRM_EX_OK; stonith_t *st = NULL; stonith_key_value_t *params = NULL; stonith_key_value_t *devices = NULL; stonith_key_value_t *dIter = NULL; char *output_ty = NULL; char *output_dest = NULL; pcmk__output_t *out = NULL; crm_log_cli_init("stonith_admin"); crm_set_options(NULL, " []", long_options, "access the Pacemaker fencing API"); async_fence_data.name = strdup(crm_system_name); while (1) { flag = crm_get_option_long(argc, argv, &option_index, &longname); if (flag == -1) break; switch (flag) { case 'V': verbose = 1; crm_bump_log_level(argc, argv); break; case '$': case '?': crm_help(flag, CRM_EX_OK); break; case 'K': required_agent = true; /* fall through */ case 'I': no_connect = 1; /* fall through */ case 'L': action = flag; break; case 'q': quiet = 1; break; case 'c': cleanup = 1; break; case 'b': broadcast = 1; break; case 'R': required_agent = true; /* fall through */ case 'Q': case 'D': case 's': action = flag; device = optarg; break; case 'T': free(async_fence_data.name); async_fence_data.name = crm_strdup_printf("%s.%s", crm_system_name, optarg); break; case 'a': agent = optarg; break; case 'l': target = optarg; action = 'L'; break; case 'M': no_connect = 1; action = flag; required_agent = true; break; case 't': timeout = crm_atoi(optarg, NULL); break; case 'B': case 'F': case 'U': /* using mainloop here */ no_connect = 1; /* fall through */ case 'C': /* Always log the input arguments */ crm_log_args(argc, argv); target = optarg; action = flag; break; case 'n': as_nodeid = TRUE; break; case 'h': case 'H': case 'r': case 'd': target = optarg; action = flag; break; case 'i': fence_level = crm_atoi(optarg, NULL); break; case 'v': devices = stonith_key_value_add(devices, NULL, optarg); break; case 'o': crm_info("Scanning: -o %s", optarg); rc = pcmk_scan_nvpair(optarg, &name, &value); if (rc != 2) { crm_err("Invalid option: -o %s: %s", optarg, pcmk_strerror(rc)); ++argerr; } else { crm_info("Got: '%s'='%s'", name, value); params = stonith_key_value_add(params, name, value); } free(value); value = NULL; free(name); name = NULL; break; case 'e': { char *key = crm_concat("OCF_RESKEY", optarg, '_'); const char *env = getenv(key); if (env == NULL) { crm_err("Invalid option: -e %s", optarg); ++argerr; } else { crm_info("Got: '%s'='%s'", optarg, env); params = stonith_key_value_add(params, optarg, env); } free(key); } break; case 0: if (safe_str_eq("tolerance", longname)) { tolerance = crm_get_msec(optarg) / 1000; /* Send in seconds */ } else if (pcmk__parse_output_args(longname, optarg, &output_ty, &output_dest) == false) { fprintf(stderr, "Unknown long option used: %s\n", longname); ++argerr; } break; default: ++argerr; break; } } if (optind > argc || action == 0) { ++argerr; } if (required_agent && agent == NULL) { fprintf(stderr, "Please specify an agent to query using -a,--agent [value]\n"); ++argerr; } if (argerr) { crm_help('?', CRM_EX_USAGE); } CRM_ASSERT(pcmk__register_format("text", pcmk__mk_text_output) == 0); CRM_ASSERT(pcmk__register_format("xml", pcmk__mk_xml_output) == 0); rc = pcmk__output_new(&out, output_ty, output_dest, argv); if (rc != 0) { fprintf(stderr, "Error creating output format %s: %s\n", output_ty, pcmk_strerror(rc)); exit_code = CRM_EX_ERROR; goto done; } stonith__register_messages(out); st = stonith_api_new(); if (!no_connect) { rc = st->cmds->connect(st, async_fence_data.name, NULL); if (rc < 0) { fprintf(stderr, "Could not connect to fencer: %s\n", pcmk_strerror(rc)); exit_code = CRM_EX_DISCONNECT; goto done; } } switch (action) { case 'I': rc = st->cmds->list_agents(st, st_opt_sync_call, NULL, &devices, timeout); if (rc < 0) { fprintf(stderr, "Failed to list installed devices: %s\n", pcmk_strerror(rc)); break; } out->begin_list(out, "Installed fence devices", "fence device", "fence devices"); for (dIter = devices; dIter; dIter = dIter->next) { out->list_item(out, "device", dIter->value); } out->end_list(out); rc = 0; stonith_key_value_freeall(devices, 1, 1); break; case 'L': rc = st->cmds->query(st, st_opts, target, &devices, timeout); if (rc < 0) { fprintf(stderr, "Failed to list registered devices: %s\n", pcmk_strerror(rc)); break; } out->begin_list(out, "Registered fence devices", "fence device", "fence devices"); for (dIter = devices; dIter; dIter = dIter->next) { out->list_item(out, "device", dIter->value); } out->end_list(out); rc = 0; stonith_key_value_freeall(devices, 1, 1); break; case 'Q': rc = st->cmds->monitor(st, st_opts, device, timeout); if (rc < 0) { rc = st->cmds->list(st, st_opts, device, NULL, timeout); } break; case 's': rc = st->cmds->list(st, st_opts, device, &lists, timeout); if (rc == 0 && lists) { char *head = lists; char *eol = NULL; out->begin_list(out, "Fence targets", "fence target", "fence targets"); do { char *line = NULL; char *elem = NULL; char *hostname = NULL; char *uuid = NULL; char *status = NULL; eol = strstr(head, "\\n"); line = strndup(head, eol-head); while ((elem = strsep(&line, " ")) != NULL) { if (strcmp(elem, "") == 0) { continue; } if (hostname == NULL) { hostname = elem; } else if (uuid == NULL) { uuid = elem; } else if (status == NULL) { char *end = NULL; status = elem; end = strchr(status, '\n'); if (end != NULL) { *end = '\0'; } } } if (hostname != NULL && uuid != NULL && status != NULL) { out->message(out, "fence-target", hostname, uuid, status); } free(line); head = eol+2; } while (eol != NULL); out->end_list(out); } else if (rc != 0) { fprintf(stderr, "List command returned error. rc : %d\n", rc); } break; case 'R': rc = st->cmds->register_device(st, st_opts, device, NULL, agent, params); break; case 'D': rc = st->cmds->remove_device(st, st_opts, device); break; case 'd': case 'r': rc = handle_level(st, target, fence_level, devices, action == 'r'); break; case 'M': { char *buffer = NULL; rc = st->cmds->metadata(st, st_opt_sync_call, agent, NULL, &buffer, timeout); if (rc == pcmk_ok) { out->output_xml(out, "metadata", buffer); } free(buffer); } break; case 'C': rc = st->cmds->confirm(st, st_opts, target); break; case 'B': rc = mainloop_fencing(st, target, "reboot", timeout, tolerance); break; case 'F': rc = mainloop_fencing(st, target, "off", timeout, tolerance); break; case 'U': rc = mainloop_fencing(st, target, "on", timeout, tolerance); break; case 'h': { time_t when = 0; if(as_nodeid) { uint32_t nodeid = atol(target); when = stonith_api_time(nodeid, NULL, FALSE); } else { when = stonith_api_time(0, target, FALSE); } out->message(out, "last-fenced", target, when); } break; case 'H': rc = handle_history(st, target, timeout, quiet, verbose, cleanup, broadcast, out); break; case 'K': device = (devices? devices->key : NULL); rc = validate(st, agent, device, params, timeout, quiet, out); break; } crm_info("Command returned: %s (%d)", pcmk_strerror(rc), rc); exit_code = crm_errno2exit(rc); pcmk__output_free(out, exit_code); done: free(async_fence_data.name); stonith_key_value_freeall(params, 1, 1); if (st != NULL) { st->cmds->disconnect(st); stonith_api_delete(st); } return exit_code; } diff --git a/xml/api/stonith_admin-1.0.rng b/xml/api/stonith_admin-1.0.rng index 4c12c4173c..dd52a63a5c 100644 --- a/xml/api/stonith_admin-1.0.rng +++ b/xml/api/stonith_admin-1.0.rng @@ -1,81 +1,92 @@ device - - + + + + failed + success + pending + + + + + - + + +