diff --git a/cts/cts-scheduler.in b/cts/cts-scheduler.in index 7fc76cce44..b8595186cb 100644 --- a/cts/cts-scheduler.in +++ b/cts/cts-scheduler.in @@ -1,1608 +1,1610 @@ #!@PYTHON@ """ Regression tests for Pacemaker's scheduler """ __copyright__ = "Copyright 2004-2022 the Pacemaker project contributors" __license__ = "GNU General Public License version 2 or later (GPLv2+) WITHOUT ANY WARRANTY" import io import os import re import sys import stat import shlex import shutil import argparse import subprocess import platform import tempfile DESC = """Regression tests for Pacemaker's scheduler""" # Each entry in TESTS is a group of tests, where each test consists of a # test base name, test description, and additional test arguments. # Test groups will be separated by newlines in output. TESTS = [ [ [ "simple1", "Offline" ], [ "simple2", "Start" ], [ "simple3", "Start 2" ], [ "simple4", "Start Failed" ], [ "simple6", "Stop Start" ], [ "simple7", "Shutdown" ], #[ "simple8", "Stonith" ], #[ "simple9", "Lower version" ], #[ "simple10", "Higher version" ], [ "simple11", "Priority (ne)" ], [ "simple12", "Priority (eq)" ], [ "simple8", "Stickiness" ], ], [ [ "group1", "Group" ], [ "group2", "Group + Native" ], [ "group3", "Group + Group" ], [ "group4", "Group + Native (nothing)" ], [ "group5", "Group + Native (move)" ], [ "group6", "Group + Group (move)" ], [ "group7", "Group colocation" ], [ "group13", "Group colocation (cant run)" ], [ "group8", "Group anti-colocation" ], [ "group9", "Group recovery" ], [ "group10", "Group partial recovery" ], [ "group11", "Group target_role" ], [ "group14", "Group stop (graph terminated)" ], [ "group15", "Negative group colocation" ], [ "bug-1573", "Partial stop of a group with two children" ], [ "bug-1718", "Mandatory group ordering - Stop group_FUN" ], [ "bug-lf-2613", "Move group on failure" ], [ "bug-lf-2619", "Move group on clone failure" ], [ "group-fail", "Ensure stop order is preserved for partially active groups" ], [ "group-unmanaged", "No need to restart r115 because r114 is unmanaged" ], [ "group-unmanaged-stopped", "Make sure r115 is stopped when r114 fails" ], [ "group-dependents", "Account for the location preferences of things colocated with a group" ], [ "group-stop-ordering", "Ensure blocked group member stop does not force other member stops" ], [ "colocate-unmanaged-group", "Respect mandatory colocations even if earlier group member is unmanaged" ], ], [ [ "rsc_dep1", "Must not" ], [ "rsc_dep3", "Must" ], [ "rsc_dep5", "Must not 3" ], [ "rsc_dep7", "Must 3" ], [ "rsc_dep10", "Must (but cant)" ], [ "rsc_dep2", "Must (running)" ], [ "rsc_dep8", "Must (running : alt)" ], [ "rsc_dep4", "Must (running + move)" ], [ "asymmetric", "Asymmetric - require explicit location constraints" ], ], [ [ "orphan-0", "Orphan ignore" ], [ "orphan-1", "Orphan stop" ], [ "orphan-2", "Orphan stop, remove failcount" ], ], [ [ "params-0", "Params: No change" ], [ "params-1", "Params: Changed" ], [ "params-2", "Params: Resource definition" ], [ "params-3", "Params: Restart instead of reload if start pending" ], [ "params-4", "Params: Reload" ], [ "params-5", "Params: Restart based on probe digest" ], [ "novell-251689", "Resource definition change + target_role=stopped" ], [ "bug-lf-2106", "Restart all anonymous clone instances after config change" ], [ "params-6", "Params: Detect reload in previously migrated resource" ], [ "nvpair-id-ref", "Support id-ref in nvpair with optional name" ], [ "not-reschedule-unneeded-monitor", "Do not reschedule unneeded monitors while resource definitions have changed" ], [ "reload-becomes-restart", "Cancel reload if restart becomes required" ], ], [ [ "target-0", "Target Role : baseline" ], [ "target-1", "Target Role : promoted" ], [ "target-2", "Target Role : invalid" ], ], [ [ "base-score", "Set a node's default score for all nodes" ], ], [ [ "date-1", "Dates", [ "-t", "2005-020" ] ], [ "date-2", "Date Spec - Pass", [ "-t", "2005-020T12:30" ] ], [ "date-3", "Date Spec - Fail", [ "-t", "2005-020T11:30" ] ], [ "origin", "Timing of recurring operations", [ "-t", "2014-05-07 00:28:00" ] ], [ "probe-0", "Probe (anon clone)" ], [ "probe-1", "Pending Probe" ], [ "probe-2", "Correctly re-probe cloned groups" ], [ "probe-3", "Probe (pending node)" ], [ "probe-4", "Probe (pending node + stopped resource)" ], [ "probe-pending-node", "Probe (pending node + unmanaged resource)" ], [ "failed-probe-primitive", "Maskable vs. unmaskable probe failures on primitive resources" ], [ "failed-probe-clone", "Maskable vs. unmaskable probe failures on cloned resources" ], [ "expired-failed-probe-primitive", "Maskable, expired probe failure on primitive resources" ], [ "standby", "Standby" ], [ "comments", "Comments" ], ], [ [ "one-or-more-0", "Everything starts" ], [ "one-or-more-1", "Nothing starts because of A" ], [ "one-or-more-2", "D can start because of C" ], [ "one-or-more-3", "D cannot start because of B and C" ], [ "one-or-more-4", "D cannot start because of target-role" ], [ "one-or-more-5", "Start A and F even though C and D are stopped" ], [ "one-or-more-6", "Leave A running even though B is stopped" ], [ "one-or-more-7", "Leave A running even though C is stopped" ], [ "bug-5140-require-all-false", "Allow basegrp:0 to stop" ], [ "clone-require-all-1", "clone B starts node 3 and 4" ], [ "clone-require-all-2", "clone B remains stopped everywhere" ], [ "clone-require-all-3", "clone B stops everywhere because A stops everywhere" ], [ "clone-require-all-4", "clone B remains on node 3 and 4 with only one instance of A remaining" ], [ "clone-require-all-5", "clone B starts on node 1 3 and 4" ], [ "clone-require-all-6", "clone B remains active after shutting down instances of A" ], [ "clone-require-all-7", "clone A and B both start at the same time. all instances of A start before B" ], [ "clone-require-all-no-interleave-1", "C starts everywhere after A and B" ], [ "clone-require-all-no-interleave-2", "C starts on nodes 1, 2, and 4 with only one active instance of B" ], [ "clone-require-all-no-interleave-3", "C remains active when instance of B is stopped on one node and started on another" ], [ "one-or-more-unrunnable-instances", "Avoid dependencies on instances that won't ever be started" ], ], [ [ "location-date-rules-1", "Use location constraints with ineffective date-based rules" ], [ "location-date-rules-2", "Use location constraints with effective date-based rules" ], [ "nvpair-date-rules-1", "Use nvpair blocks with a variety of date-based rules" ], [ "value-source", "Use location constraints with node attribute expressions using value-source" ], [ "rule-dbl-as-auto-number-match", "Floating-point rule values default to number comparison: match" ], [ "rule-dbl-as-auto-number-no-match", "Floating-point rule values default to number comparison: no " "match" ], [ "rule-dbl-as-integer-match", "Floating-point rule values set to integer comparison: match" ], [ "rule-dbl-as-integer-no-match", "Floating-point rule values set to integer comparison: no match" ], [ "rule-dbl-as-number-match", "Floating-point rule values set to number comparison: match" ], [ "rule-dbl-as-number-no-match", "Floating-point rule values set to number comparison: no match" ], [ "rule-dbl-parse-fail-default-str-match", "Floating-point rule values fail to parse, default to string " "comparison: match" ], [ "rule-dbl-parse-fail-default-str-no-match", "Floating-point rule values fail to parse, default to string " "comparison: no match" ], [ "rule-int-as-auto-integer-match", "Integer rule values default to integer comparison: match" ], [ "rule-int-as-auto-integer-no-match", "Integer rule values default to integer comparison: no match" ], [ "rule-int-as-integer-match", "Integer rule values set to integer comparison: match" ], [ "rule-int-as-integer-no-match", "Integer rule values set to integer comparison: no match" ], [ "rule-int-as-number-match", "Integer rule values set to number comparison: match" ], [ "rule-int-as-number-no-match", "Integer rule values set to number comparison: no match" ], [ "rule-int-parse-fail-default-str-match", "Integer rule values fail to parse, default to string " "comparison: match" ], [ "rule-int-parse-fail-default-str-no-match", "Integer rule values fail to parse, default to string " "comparison: no match" ], ], [ [ "order1", "Order start 1" ], [ "order2", "Order start 2" ], [ "order3", "Order stop" ], [ "order4", "Order (multiple)" ], [ "order5", "Order (move)" ], [ "order6", "Order (move w/ restart)" ], [ "order7", "Order (mandatory)" ], [ "order-optional", "Order (score=0)" ], [ "order-required", "Order (score=INFINITY)" ], [ "bug-lf-2171", "Prevent group start when clone is stopped" ], [ "order-clone", "Clone ordering should be able to prevent startup of dependent clones" ], [ "order-sets", "Ordering for resource sets" ], [ "order-serialize", "Serialize resources without inhibiting migration" ], [ "order-serialize-set", "Serialize a set of resources without inhibiting migration" ], [ "clone-order-primitive", "Order clone start after a primitive" ], [ "clone-order-16instances", "Verify ordering of 16 cloned resources" ], [ "order-optional-keyword", "Order (optional keyword)" ], [ "order-mandatory", "Order (mandatory keyword)" ], [ "bug-lf-2493", "Don't imply colocation requirements when applying ordering constraints with clones" ], [ "ordered-set-basic-startup", "Constraint set with default order settings" ], [ "ordered-set-natural", "Allow natural set ordering" ], [ "order-wrong-kind", "Order (error)" ], ], [ [ "coloc-loop", "Colocation - loop" ], [ "coloc-many-one", "Colocation - many-to-one" ], [ "coloc-list", "Colocation - many-to-one with list" ], [ "coloc-group", "Colocation - groups" ], [ "coloc-unpromoted-anti", "Anti-colocation with unpromoted shouldn't prevent promoted colocation" ], [ "coloc-attr", "Colocation based on node attributes" ], [ "coloc-negative-group", "Negative colocation with a group" ], [ "coloc-intra-set", "Intra-set colocation" ], [ "bug-lf-2435", "Colocation sets with a negative score" ], [ "coloc-clone-stays-active", "Ensure clones don't get stopped/demoted because a dependent must stop" ], [ "coloc_fp_logic", "Verify floating point calculations in colocation are working" ], [ "colo_promoted_w_native", "cl#5070 - Verify promotion order is affected when colocating promoted with primitive" ], [ "colo_unpromoted_w_native", "cl#5070 - Verify promotion order is affected when colocating unpromoted with primitive" ], [ "anti-colocation-order", "cl#5187 - Prevent resources in an anti-colocation from even temporarily running on a same node" ], [ "anti-colocation-promoted", "Organize order of actions for promoted resources in anti-colocations" ], [ "anti-colocation-unpromoted", "Organize order of actions for unpromoted resources in anti-colocations" ], [ "enforce-colo1", "Always enforce B with A INFINITY" ], [ "complex_enforce_colo", "Always enforce B with A INFINITY. (make sure heat-engine stops)" ], [ "coloc-dependee-should-stay", "Stickiness outweighs group colocation" ], [ "coloc-dependee-should-move", "Group colocation outweighs stickiness" ], [ "colocation-influence", "Respect colocation influence" ], ], [ [ "rsc-sets-seq-true", "Resource Sets - sequential=false" ], [ "rsc-sets-seq-false", "Resource Sets - sequential=true" ], [ "rsc-sets-clone", "Resource Sets - Clone" ], [ "rsc-sets-promoted", "Resource Sets - Promoted" ], [ "rsc-sets-clone-1", "Resource Sets - Clone (lf#2404)" ], ], [ [ "attrs1", "string: eq (and)" ], [ "attrs2", "string: lt / gt (and)" ], [ "attrs3", "string: ne (or)" ], [ "attrs4", "string: exists" ], [ "attrs5", "string: not_exists" ], [ "attrs6", "is_dc: true" ], [ "attrs7", "is_dc: false" ], [ "attrs8", "score_attribute" ], [ "per-node-attrs", "Per node resource parameters" ], ], [ [ "mon-rsc-1", "Schedule Monitor - start" ], [ "mon-rsc-2", "Schedule Monitor - move" ], [ "mon-rsc-3", "Schedule Monitor - pending start" ], [ "mon-rsc-4", "Schedule Monitor - move/pending start" ], ], [ [ "rec-rsc-0", "Resource Recover - no start" ], [ "rec-rsc-1", "Resource Recover - start" ], [ "rec-rsc-2", "Resource Recover - monitor" ], [ "rec-rsc-3", "Resource Recover - stop - ignore" ], [ "rec-rsc-4", "Resource Recover - stop - block" ], [ "rec-rsc-5", "Resource Recover - stop - fence" ], [ "rec-rsc-6", "Resource Recover - multiple - restart" ], [ "rec-rsc-7", "Resource Recover - multiple - stop" ], [ "rec-rsc-8", "Resource Recover - multiple - block" ], [ "rec-rsc-9", "Resource Recover - group/group" ], [ "stop-unexpected", "Recover multiply active group with stop_unexpected" ], [ "stop-unexpected-2", "Resource multiply active primitve with stop_unexpected" ], [ "monitor-recovery", "on-fail=block + resource recovery detected by recurring monitor" ], [ "stop-failure-no-quorum", "Stop failure without quorum" ], [ "stop-failure-no-fencing", "Stop failure without fencing available" ], [ "stop-failure-with-fencing", "Stop failure with fencing available" ], [ "multiple-active-block-group", "Support of multiple-active=block for resource groups" ], [ "multiple-monitor-one-failed", "Consider resource failed if any of the configured monitor operations failed" ], ], [ [ "quorum-1", "No quorum - ignore" ], [ "quorum-2", "No quorum - freeze" ], [ "quorum-3", "No quorum - stop" ], [ "quorum-4", "No quorum - start anyway" ], [ "quorum-5", "No quorum - start anyway (group)" ], [ "quorum-6", "No quorum - start anyway (clone)" ], [ "bug-cl-5212", "No promotion with no-quorum-policy=freeze" ], [ "suicide-needed-inquorate", "no-quorum-policy=suicide: suicide necessary" ], [ "suicide-not-needed-initial-quorum", "no-quorum-policy=suicide: suicide not necessary at initial quorum" ], [ "suicide-not-needed-never-quorate", "no-quorum-policy=suicide: suicide not necessary if never quorate" ], [ "suicide-not-needed-quorate", "no-quorum-policy=suicide: suicide necessary if quorate" ], ], [ [ "rec-node-1", "Node Recover - Startup - no fence" ], [ "rec-node-2", "Node Recover - Startup - fence" ], [ "rec-node-3", "Node Recover - HA down - no fence" ], [ "rec-node-4", "Node Recover - HA down - fence" ], [ "rec-node-5", "Node Recover - CRM down - no fence" ], [ "rec-node-6", "Node Recover - CRM down - fence" ], [ "rec-node-7", "Node Recover - no quorum - ignore" ], [ "rec-node-8", "Node Recover - no quorum - freeze" ], [ "rec-node-9", "Node Recover - no quorum - stop" ], [ "rec-node-10", "Node Recover - no quorum - stop w/fence" ], [ "rec-node-11", "Node Recover - CRM down w/ group - fence" ], [ "rec-node-12", "Node Recover - nothing active - fence" ], [ "rec-node-13", "Node Recover - failed resource + shutdown - fence" ], [ "rec-node-15", "Node Recover - unknown lrm section" ], [ "rec-node-14", "Serialize all stonith's" ], ], [ [ "multi1", "Multiple Active (stop/start)" ], ], [ [ "migrate-begin", "Normal migration" ], [ "migrate-success", "Completed migration" ], [ "migrate-partial-1", "Completed migration, missing stop on source" ], [ "migrate-partial-2", "Successful migrate_to only" ], [ "migrate-partial-3", "Successful migrate_to only, target down" ], [ "migrate-partial-4", "Migrate from the correct host after migrate_to+migrate_from" ], [ "bug-5186-partial-migrate", "Handle partial migration when src node loses membership" ], [ "migrate-fail-2", "Failed migrate_from" ], [ "migrate-fail-3", "Failed migrate_from + stop on source" ], [ "migrate-fail-4", "Failed migrate_from + stop on target - ideally we wouldn't need to re-stop on target" ], [ "migrate-fail-5", "Failed migrate_from + stop on source and target" ], [ "migrate-fail-6", "Failed migrate_to" ], [ "migrate-fail-7", "Failed migrate_to + stop on source" ], [ "migrate-fail-8", "Failed migrate_to + stop on target - ideally we wouldn't need to re-stop on target" ], [ "migrate-fail-9", "Failed migrate_to + stop on source and target" ], [ "migration-ping-pong", "Old migrate_to failure + successful migrate_from on same node" ], [ "migrate-stop", "Migration in a stopping stack" ], [ "migrate-start", "Migration in a starting stack" ], [ "migrate-stop_start", "Migration in a restarting stack" ], [ "migrate-stop-complex", "Migration in a complex stopping stack" ], [ "migrate-start-complex", "Migration in a complex starting stack" ], [ "migrate-stop-start-complex", "Migration in a complex moving stack" ], [ "migrate-shutdown", "Order the post-migration 'stop' before node shutdown" ], [ "migrate-1", "Migrate (migrate)" ], [ "migrate-2", "Migrate (stable)" ], [ "migrate-3", "Migrate (failed migrate_to)" ], [ "migrate-4", "Migrate (failed migrate_from)" ], [ "novell-252693", "Migration in a stopping stack" ], [ "novell-252693-2", "Migration in a starting stack" ], [ "novell-252693-3", "Non-Migration in a starting and stopping stack" ], [ "bug-1820", "Migration in a group" ], [ "bug-1820-1", "Non-migration in a group" ], [ "migrate-5", "Primitive migration with a clone" ], [ "migrate-fencing", "Migration after Fencing" ], [ "migrate-both-vms", "Migrate two VMs that have no colocation" ], [ "migration-behind-migrating-remote", "Migrate resource behind migrating remote connection" ], [ "1-a-then-bm-move-b", "Advanced migrate logic. A then B. migrate B" ], [ "2-am-then-b-move-a", "Advanced migrate logic, A then B, migrate A without stopping B" ], [ "3-am-then-bm-both-migrate", "Advanced migrate logic. A then B. migrate both" ], [ "4-am-then-bm-b-not-migratable", "Advanced migrate logic, A then B, B not migratable" ], [ "5-am-then-bm-a-not-migratable", "Advanced migrate logic. A then B. move both, a not migratable" ], [ "6-migrate-group", "Advanced migrate logic, migrate a group" ], [ "7-migrate-group-one-unmigratable", "Advanced migrate logic, migrate group mixed with allow-migrate true/false" ], [ "8-am-then-bm-a-migrating-b-stopping", "Advanced migrate logic, A then B, A migrating, B stopping" ], [ "9-am-then-bm-b-migrating-a-stopping", "Advanced migrate logic, A then B, B migrate, A stopping" ], [ "10-a-then-bm-b-move-a-clone", "Advanced migrate logic, A clone then B, migrate B while stopping A" ], [ "11-a-then-bm-b-move-a-clone-starting", "Advanced migrate logic, A clone then B, B moving while A is start/stopping" ], [ "a-promote-then-b-migrate", "A promote then B start. migrate B" ], [ "a-demote-then-b-migrate", "A demote then B stop. migrate B" ], + [ "probe-target-of-failed-migrate_to-1", "Failed migrate_to, target rejoins" ], + [ "probe-target-of-failed-migrate_to-2", "Failed migrate_to, target rejoined and probed" ], # @TODO: If pacemaker implements versioned attributes, uncomment this test #[ "migrate-versioned", "Disable migration for versioned resources" ], [ "bug-lf-2422", "Dependency on partially active group - stop ocfs:*" ], ], [ [ "clone-anon-probe-1", "Probe the correct (anonymous) clone instance for each node" ], [ "clone-anon-probe-2", "Avoid needless re-probing of anonymous clones" ], [ "clone-anon-failcount", "Merge failcounts for anonymous clones" ], [ "force-anon-clone-max", "Update clone-max properly when forcing a clone to be anonymous" ], [ "anon-instance-pending", "Assign anonymous clone instance numbers properly when action pending" ], [ "inc0", "Incarnation start" ], [ "inc1", "Incarnation start order" ], [ "inc2", "Incarnation silent restart, stop, move" ], [ "inc3", "Inter-incarnation ordering, silent restart, stop, move" ], [ "inc4", "Inter-incarnation ordering, silent restart, stop, move (ordered)" ], [ "inc5", "Inter-incarnation ordering, silent restart, stop, move (restart 1)" ], [ "inc6", "Inter-incarnation ordering, silent restart, stop, move (restart 2)" ], [ "inc7", "Clone colocation" ], [ "inc8", "Clone anti-colocation" ], [ "inc9", "Non-unique clone" ], [ "inc10", "Non-unique clone (stop)" ], [ "inc11", "Primitive colocation with clones" ], [ "inc12", "Clone shutdown" ], [ "cloned-group", "Make sure only the correct number of cloned groups are started" ], [ "cloned-group-stop", "Ensure stopping qpidd also stops glance and cinder" ], [ "clone-no-shuffle", "Don't prioritize allocation of instances that must be moved" ], [ "clone-max-zero", "Orphan processing with clone-max=0" ], [ "clone-anon-dup", "Bug LF#2087 - Correctly parse the state of anonymous clones that are active more than once per node" ], [ "bug-lf-2160", "Don't shuffle clones due to colocation" ], [ "bug-lf-2213", "clone-node-max enforcement for cloned groups" ], [ "bug-lf-2153", "Clone ordering constraints" ], [ "bug-lf-2361", "Ensure clones observe mandatory ordering constraints if the LHS is unrunnable" ], [ "bug-lf-2317", "Avoid needless restart of primitive depending on a clone" ], [ "clone-colocate-instance-1", "Colocation with a specific clone instance (negative example)" ], [ "clone-colocate-instance-2", "Colocation with a specific clone instance" ], [ "clone-order-instance", "Ordering with specific clone instances" ], [ "bug-lf-2453", "Enforce mandatory clone ordering without colocation" ], [ "bug-lf-2508", "Correctly reconstruct the status of anonymous cloned groups" ], [ "bug-lf-2544", "Balanced clone placement" ], [ "bug-lf-2445", "Redistribute clones with node-max > 1 and stickiness = 0" ], [ "bug-lf-2574", "Avoid clone shuffle" ], [ "bug-lf-2581", "Avoid group restart due to unrelated clone (re)start" ], [ "bug-cl-5168", "Don't shuffle clones" ], [ "bug-cl-5170", "Prevent clone from starting with on-fail=block" ], [ "clone-fail-block-colocation", "Move colocated group when failed clone has on-fail=block" ], [ "clone-interleave-1", "Clone-3 cannot start on pcmk-1 due to interleaved ordering (no colocation)" ], [ "clone-interleave-2", "Clone-3 must stop on pcmk-1 due to interleaved ordering (no colocation)" ], [ "clone-interleave-3", "Clone-3 must be recovered on pcmk-1 due to interleaved ordering (no colocation)" ], [ "rebalance-unique-clones", "Rebalance unique clone instances with no stickiness" ], [ "clone-requires-quorum-recovery", "Clone with requires=quorum on failed node needing recovery" ], [ "clone-requires-quorum", "Clone with requires=quorum with presumed-inactive instance on failed node" ], ], [ [ "cloned_start_one", "order first clone then clone... first clone_min=2" ], [ "cloned_start_two", "order first clone then clone... first clone_min=2" ], [ "cloned_stop_one", "order first clone then clone... first clone_min=2" ], [ "cloned_stop_two", "order first clone then clone... first clone_min=2" ], [ "clone_min_interleave_start_one", "order first clone then clone... first clone_min=2 and then has interleave=true" ], [ "clone_min_interleave_start_two", "order first clone then clone... first clone_min=2 and then has interleave=true" ], [ "clone_min_interleave_stop_one", "order first clone then clone... first clone_min=2 and then has interleave=true" ], [ "clone_min_interleave_stop_two", "order first clone then clone... first clone_min=2 and then has interleave=true" ], [ "clone_min_start_one", "order first clone then primitive... first clone_min=2" ], [ "clone_min_start_two", "order first clone then primitive... first clone_min=2" ], [ "clone_min_stop_all", "order first clone then primitive... first clone_min=2" ], [ "clone_min_stop_one", "order first clone then primitive... first clone_min=2" ], [ "clone_min_stop_two", "order first clone then primitive... first clone_min=2" ], ], [ [ "unfence-startup", "Clean unfencing" ], [ "unfence-definition", "Unfencing when the agent changes" ], [ "unfence-parameters", "Unfencing when the agent parameters changes" ], [ "unfence-device", "Unfencing when a cluster has only fence devices" ], ], [ [ "promoted-0", "Stopped -> Unpromoted" ], [ "promoted-1", "Stopped -> Promote" ], [ "promoted-2", "Stopped -> Promote : notify" ], [ "promoted-3", "Stopped -> Promote : promoted location" ], [ "promoted-4", "Started -> Promote : promoted location" ], [ "promoted-5", "Promoted -> Promoted" ], [ "promoted-6", "Promoted -> Promoted (2)" ], [ "promoted-7", "Promoted -> Fenced" ], [ "promoted-8", "Promoted -> Fenced -> Moved" ], [ "promoted-9", "Stopped + Promotable + No quorum" ], [ "promoted-10", "Stopped -> Promotable : notify with monitor" ], [ "promoted-11", "Stopped -> Promote : colocation" ], [ "novell-239082", "Demote/Promote ordering" ], [ "novell-239087", "Stable promoted placement" ], [ "promoted-12", "Promotion based solely on rsc_location constraints" ], [ "promoted-13", "Include preferences of colocated resources when placing promoted" ], [ "promoted-demote", "Ordering when actions depends on demoting an unpromoted resource" ], [ "promoted-ordering", "Prevent resources from starting that need a promoted" ], [ "bug-1765", "Verify promoted-with-promoted colocation does not stop unpromoted instances" ], [ "promoted-group", "Promotion of cloned groups" ], [ "bug-lf-1852", "Don't shuffle promotable instances unnecessarily" ], [ "promoted-failed-demote", "Don't retry failed demote actions" ], [ "promoted-failed-demote-2", "Don't retry failed demote actions (notify=false)" ], [ "promoted-depend", "Ensure resources that depend on promoted instance don't get allocated until that does" ], [ "promoted-reattach", "Re-attach to a running promoted" ], [ "promoted-allow-start", "Don't include promoted score if it would prevent allocation" ], [ "promoted-colocation", "Allow promoted instances placemaker to be influenced by colocation constraints" ], [ "promoted-pseudo", "Make sure promote/demote pseudo actions are created correctly" ], [ "promoted-role", "Prevent target-role from promoting more than promoted-max instances" ], [ "bug-lf-2358", "Anti-colocation of promoted instances" ], [ "promoted-promotion-constraint", "Mandatory promoted colocation constraints" ], [ "unmanaged-promoted", "Ensure role is preserved for unmanaged resources" ], [ "promoted-unmanaged-monitor", "Start correct monitor for unmanaged promoted instances" ], [ "promoted-demote-2", "Demote does not clear past failure" ], [ "promoted-move", "Move promoted based on failure of colocated group" ], [ "promoted-probed-score", "Observe the promotion score of probed resources" ], [ "colocation_constraint_stops_promoted", "cl#5054 - Ensure promoted is demoted when stopped by colocation constraint" ], [ "colocation_constraint_stops_unpromoted", "cl#5054 - Ensure unpromoted is not demoted when stopped by colocation constraint" ], [ "order_constraint_stops_promoted", "cl#5054 - Ensure promoted is demoted when stopped by order constraint" ], [ "order_constraint_stops_unpromoted", "cl#5054 - Ensure unpromoted is not demoted when stopped by order constraint" ], [ "promoted_monitor_restart", "cl#5072 - Ensure promoted monitor operation will start after promotion" ], [ "bug-rh-880249", "Handle replacement of an m/s resource with a primitive" ], [ "bug-5143-ms-shuffle", "Prevent promoted instance shuffling due to promotion score" ], [ "promoted-demote-block", "Block promotion if demote fails with on-fail=block" ], [ "promoted-dependent-ban", "Don't stop instances from being active because a dependent is banned from that host" ], [ "promoted-stop", "Stop instances due to location constraint with role=Started" ], [ "promoted-partially-demoted-group", "Allow partially demoted group to finish demoting" ], [ "bug-cl-5213", "Ensure role colocation with -INFINITY is enforced" ], [ "bug-cl-5219", "Allow unrelated resources with a common colocation target to remain promoted" ], [ "promoted-asymmetrical-order", "Fix the behaviors of multi-state resources with asymmetrical ordering" ], [ "promoted-notify", "Promotion with notifications" ], [ "promoted-score-startup", "Use permanent promoted scores without LRM history" ], [ "failed-demote-recovery", "Recover resource in unpromoted role after demote fails" ], [ "failed-demote-recovery-promoted", "Recover resource in promoted role after demote fails" ], [ "on_fail_demote1", "Recovery with on-fail=\"demote\" on healthy cluster, remote, guest, and bundle nodes" ], [ "on_fail_demote2", "Recovery with on-fail=\"demote\" with promotion on different node" ], [ "on_fail_demote3", "Recovery with on-fail=\"demote\" with no promotion" ], [ "on_fail_demote4", "Recovery with on-fail=\"demote\" on failed cluster, remote, guest, and bundle nodes" ], [ "no_quorum_demote", "Promotable demotion and primitive stop with no-quorum-policy=\"demote\"" ], [ "no-promote-on-unrunnable-guest", "Don't select bundle instance for promotion when container can't run" ], ], [ [ "history-1", "Correctly parse stateful-1 resource state" ], ], [ [ "managed-0", "Managed (reference)" ], [ "managed-1", "Not managed - down" ], [ "managed-2", "Not managed - up" ], [ "bug-5028", "Shutdown should block if anything depends on an unmanaged resource" ], [ "bug-5028-detach", "Ensure detach still works" ], [ "bug-5028-bottom", "Ensure shutdown still blocks if the blocked resource is at the bottom of the stack" ], [ "unmanaged-stop-1", "cl#5155 - Block the stop of resources if any depending resource is unmanaged" ], [ "unmanaged-stop-2", "cl#5155 - Block the stop of resources if the first resource in a mandatory stop order is unmanaged" ], [ "unmanaged-stop-3", "cl#5155 - Block the stop of resources if any depending resource in a group is unmanaged" ], [ "unmanaged-stop-4", "cl#5155 - Block the stop of resources if any depending resource in the middle of a group is unmanaged" ], [ "unmanaged-block-restart", "Block restart of resources if any dependent resource in a group is unmanaged" ], ], [ [ "interleave-0", "Interleave (reference)" ], [ "interleave-1", "coloc - not interleaved" ], [ "interleave-2", "coloc - interleaved" ], [ "interleave-3", "coloc - interleaved (2)" ], [ "interleave-pseudo-stop", "Interleaved clone during stonith" ], [ "interleave-stop", "Interleaved clone during stop" ], [ "interleave-restart", "Interleaved clone during dependency restart" ], ], [ [ "notify-0", "Notify reference" ], [ "notify-1", "Notify simple" ], [ "notify-2", "Notify simple, confirm" ], [ "notify-3", "Notify move, confirm" ], [ "novell-239079", "Notification priority" ], #[ "notify-2", "Notify - 764" ], [ "notifs-for-unrunnable", "Don't schedule notifications for an unrunnable action" ], [ "route-remote-notify", "Route remote notify actions through correct cluster node" ], [ "notify-behind-stopping-remote", "Don't schedule notifications behind stopped remote" ], ], [ [ "594", "OSDL #594 - Unrunnable actions scheduled in transition" ], [ "662", "OSDL #662 - Two resources start on one node when incarnation_node_max = 1" ], [ "696", "OSDL #696 - CRM starts stonith RA without monitor" ], [ "726", "OSDL #726 - Attempting to schedule rsc_posic041_monitor_5000 _after_ a stop" ], [ "735", "OSDL #735 - Correctly detect that rsc_hadev1 is stopped on hadev3" ], [ "764", "OSDL #764 - Missing monitor op for DoFencing:child_DoFencing:1" ], [ "797", "OSDL #797 - Assert triggered: task_id_i > max_call_id" ], [ "829", "OSDL #829" ], [ "994", "OSDL #994 - Stopping the last resource in a resource group causes the entire group to be restarted" ], [ "994-2", "OSDL #994 - with a dependent resource" ], [ "1360", "OSDL #1360 - Clone stickiness" ], [ "1484", "OSDL #1484 - on_fail=stop" ], [ "1494", "OSDL #1494 - Clone stability" ], [ "unrunnable-1", "Unrunnable" ], [ "unrunnable-2", "Unrunnable 2" ], [ "stonith-0", "Stonith loop - 1" ], [ "stonith-1", "Stonith loop - 2" ], [ "stonith-2", "Stonith loop - 3" ], [ "stonith-3", "Stonith startup" ], [ "stonith-4", "Stonith node state" ], [ "dc-fence-ordering", "DC needs fencing while other nodes are shutting down" ], [ "bug-1572-1", "Recovery of groups depending on promotable role" ], [ "bug-1572-2", "Recovery of groups depending on promotable role when promoted is not re-promoted" ], [ "bug-1685", "Depends-on-promoted ordering" ], [ "bug-1822", "Don't promote partially active groups" ], [ "bug-pm-11", "New resource added to a m/s group" ], [ "bug-pm-12", "Recover only the failed portion of a cloned group" ], [ "bug-n-387749", "Don't shuffle clone instances" ], [ "bug-n-385265", "Don't ignore the failure stickiness of group children - resource_idvscommon should stay stopped" ], [ "bug-n-385265-2", "Ensure groups are migrated instead of remaining partially active on the current node" ], [ "bug-lf-1920", "Correctly handle probes that find active resources" ], [ "bnc-515172", "Location constraint with multiple expressions" ], [ "colocate-primitive-with-clone", "Optional colocation with a clone" ], [ "use-after-free-merge", "Use-after-free in native_merge_weights" ], [ "bug-lf-2551", "STONITH ordering for stop" ], [ "bug-lf-2606", "Stonith implies demote" ], [ "bug-lf-2474", "Ensure resource op timeout takes precedence over op_defaults" ], [ "bug-suse-707150", "Prevent vm-01 from starting due to colocation/ordering" ], [ "bug-5014-A-start-B-start", "Verify when A starts B starts using symmetrical=false" ], [ "bug-5014-A-stop-B-started", "Verify when A stops B does not stop if it has already started using symmetric=false" ], [ "bug-5014-A-stopped-B-stopped", "Verify when A is stopped and B has not started, B does not start before A using symmetric=false" ], [ "bug-5014-CthenAthenB-C-stopped", "Verify when C then A is symmetrical=true, A then B is symmetric=false, and C is stopped that nothing starts" ], [ "bug-5014-CLONE-A-start-B-start", "Verify when A starts B starts using clone resources with symmetric=false" ], [ "bug-5014-CLONE-A-stop-B-started", "Verify when A stops B does not stop if it has already started using clone resources with symmetric=false" ], [ "bug-5014-GROUP-A-start-B-start", "Verify when A starts B starts when using group resources with symmetric=false" ], [ "bug-5014-GROUP-A-stopped-B-started", "Verify when A stops B does not stop if it has already started using group resources with symmetric=false" ], [ "bug-5014-GROUP-A-stopped-B-stopped", "Verify when A is stopped and B has not started, B does not start before A using group resources with symmetric=false" ], [ "bug-5014-ordered-set-symmetrical-false", "Verify ordered sets work with symmetrical=false" ], [ "bug-5014-ordered-set-symmetrical-true", "Verify ordered sets work with symmetrical=true" ], [ "clbz5007-promotable-colocation", "Verify use of colocation scores other than INFINITY and -INFINITY work on multi-state resources" ], [ "bug-5038", "Prevent restart of anonymous clones when clone-max decreases" ], [ "bug-5025-1", "Automatically clean up failcount after resource config change with reload" ], [ "bug-5025-2", "Make sure clear failcount action isn't set when config does not change" ], [ "bug-5025-3", "Automatically clean up failcount after resource config change with restart" ], [ "bug-5025-4", "Clear failcount when last failure is a start op and rsc attributes changed" ], [ "failcount", "Ensure failcounts are correctly expired" ], [ "failcount-block", "Ensure failcounts are not expired when on-fail=block is present" ], [ "per-op-failcount", "Ensure per-operation failcount is handled and not passed to fence agent" ], [ "on-fail-ignore", "Ensure on-fail=ignore works even beyond migration-threshold" ], [ "monitor-onfail-restart", "bug-5058 - Monitor failure with on-fail set to restart" ], [ "monitor-onfail-stop", "bug-5058 - Monitor failure wiht on-fail set to stop" ], [ "bug-5059", "No need to restart p_stateful1:*" ], [ "bug-5069-op-enabled", "Test on-fail=ignore with failure when monitor is enabled" ], [ "bug-5069-op-disabled", "Test on-fail-ignore with failure when monitor is disabled" ], [ "obsolete-lrm-resource", "cl#5115 - Do not use obsolete lrm_resource sections" ], [ "expire-non-blocked-failure", "Ignore failure-timeout only if the failed operation has on-fail=block" ], [ "asymmetrical-order-move", "Respect asymmetrical ordering when trying to move resources" ], [ "asymmetrical-order-restart", "Respect asymmetrical ordering when restarting dependent resource" ], [ "start-then-stop-with-unfence", "Avoid graph loop with start-then-stop constraint plus unfencing" ], [ "order-expired-failure", "Order failcount cleanup after remote fencing" ], [ "ignore_stonith_rsc_order1", "cl#5056- Ignore order constraint between stonith and non-stonith rsc" ], [ "ignore_stonith_rsc_order2", "cl#5056- Ignore order constraint with group rsc containing mixed stonith and non-stonith" ], [ "ignore_stonith_rsc_order3", "cl#5056- Ignore order constraint, stonith clone and mixed group" ], [ "ignore_stonith_rsc_order4", "cl#5056- Ignore order constraint, stonith clone and clone with nested mixed group" ], [ "honor_stonith_rsc_order1", "cl#5056- Honor order constraint, stonith clone and pure stonith group(single rsc)" ], [ "honor_stonith_rsc_order2", "cl#5056- Honor order constraint, stonith clone and pure stonith group(multiple rsc)" ], [ "honor_stonith_rsc_order3", "cl#5056- Honor order constraint, stonith clones with nested pure stonith group" ], [ "honor_stonith_rsc_order4", "cl#5056- Honor order constraint, between two native stonith rscs" ], [ "multiply-active-stonith", "Multiply active stonith" ], [ "probe-timeout", "cl#5099 - Default probe timeout" ], [ "order-first-probes", "cl#5301 - respect order constraints when relevant resources are being probed" ], [ "concurrent-fencing", "Allow performing fencing operations in parallel" ], [ "priority-fencing-delay", "Delay fencing targeting the more significant node" ], ], [ [ "systemhealth1", "System Health () #1" ], [ "systemhealth2", "System Health () #2" ], [ "systemhealth3", "System Health () #3" ], [ "systemhealthn1", "System Health (None) #1" ], [ "systemhealthn2", "System Health (None) #2" ], [ "systemhealthn3", "System Health (None) #3" ], [ "systemhealthm1", "System Health (Migrate On Red) #1" ], [ "systemhealthm2", "System Health (Migrate On Red) #2" ], [ "systemhealthm3", "System Health (Migrate On Red) #3" ], [ "systemhealtho1", "System Health (Only Green) #1" ], [ "systemhealtho2", "System Health (Only Green) #2" ], [ "systemhealtho3", "System Health (Only Green) #3" ], [ "systemhealthp1", "System Health (Progessive) #1" ], [ "systemhealthp2", "System Health (Progessive) #2" ], [ "systemhealthp3", "System Health (Progessive) #3" ], [ "allow-unhealthy-nodes", "System Health (migrate-on-red + allow-unhealth-nodes)" ], ], [ [ "utilization", "Placement Strategy - utilization" ], [ "minimal", "Placement Strategy - minimal" ], [ "balanced", "Placement Strategy - balanced" ], ], [ [ "placement-stickiness", "Optimized Placement Strategy - stickiness" ], [ "placement-priority", "Optimized Placement Strategy - priority" ], [ "placement-location", "Optimized Placement Strategy - location" ], [ "placement-capacity", "Optimized Placement Strategy - capacity" ], ], [ [ "utilization-order1", "Utilization Order - Simple" ], [ "utilization-order2", "Utilization Order - Complex" ], [ "utilization-order3", "Utilization Order - Migrate" ], [ "utilization-order4", "Utilization Order - Live Migration (bnc#695440)" ], [ "utilization-complex", "Utilization with complex relationships" ], [ "utilization-shuffle", "Don't displace prmExPostgreSQLDB2 on act2, Start prmExPostgreSQLDB1 on act3" ], [ "load-stopped-loop", "Avoid transition loop due to load_stopped (cl#5044)" ], [ "load-stopped-loop-2", "cl#5235 - Prevent graph loops that can be introduced by load_stopped -> migrate_to ordering" ], ], [ [ "colocated-utilization-primitive-1", "Colocated Utilization - Primitive" ], [ "colocated-utilization-primitive-2", "Colocated Utilization - Choose the most capable node" ], [ "colocated-utilization-group", "Colocated Utilization - Group" ], [ "colocated-utilization-clone", "Colocated Utilization - Clone" ], [ "utilization-check-allowed-nodes", "Only check the capacities of the nodes that can run the resource" ], ], [ [ "reprobe-target_rc", "Ensure correct target_rc for reprobe of inactive resources" ], [ "node-maintenance-1", "cl#5128 - Node maintenance" ], [ "node-maintenance-2", "cl#5128 - Node maintenance (coming out of maintenance mode)" ], [ "shutdown-maintenance-node", "Do not fence a maintenance node if it shuts down cleanly" ], [ "rsc-maintenance", "Per-resource maintenance" ], ], [ [ "not-installed-agent", "The resource agent is missing" ], [ "not-installed-tools", "Something the resource agent needs is missing" ], ], [ [ "stopped-monitor-00", "Stopped Monitor - initial start" ], [ "stopped-monitor-01", "Stopped Monitor - failed started" ], [ "stopped-monitor-02", "Stopped Monitor - started multi-up" ], [ "stopped-monitor-03", "Stopped Monitor - stop started" ], [ "stopped-monitor-04", "Stopped Monitor - failed stop" ], [ "stopped-monitor-05", "Stopped Monitor - start unmanaged" ], [ "stopped-monitor-06", "Stopped Monitor - unmanaged multi-up" ], [ "stopped-monitor-07", "Stopped Monitor - start unmanaged multi-up" ], [ "stopped-monitor-08", "Stopped Monitor - migrate" ], [ "stopped-monitor-09", "Stopped Monitor - unmanage started" ], [ "stopped-monitor-10", "Stopped Monitor - unmanaged started multi-up" ], [ "stopped-monitor-11", "Stopped Monitor - stop unmanaged started" ], [ "stopped-monitor-12", "Stopped Monitor - unmanaged started multi-up (target-role=Stopped)" ], [ "stopped-monitor-20", "Stopped Monitor - initial stop" ], [ "stopped-monitor-21", "Stopped Monitor - stopped single-up" ], [ "stopped-monitor-22", "Stopped Monitor - stopped multi-up" ], [ "stopped-monitor-23", "Stopped Monitor - start stopped" ], [ "stopped-monitor-24", "Stopped Monitor - unmanage stopped" ], [ "stopped-monitor-25", "Stopped Monitor - unmanaged stopped multi-up" ], [ "stopped-monitor-26", "Stopped Monitor - start unmanaged stopped" ], [ "stopped-monitor-27", "Stopped Monitor - unmanaged stopped multi-up (target-role=Started)" ], [ "stopped-monitor-30", "Stopped Monitor - new node started" ], [ "stopped-monitor-31", "Stopped Monitor - new node stopped" ], ], [ # This is a combo test to check: # - probe timeout defaults to the minimum-interval monitor's # - duplicate recurring operations are ignored # - if timeout spec is bad, the default timeout is used # - failure is blocked with on-fail=block even if ISO8601 interval is specified # - started/stopped role monitors are started/stopped on right nodes [ "intervals", "Recurring monitor interval handling" ], ], [ [ "ticket-primitive-1", "Ticket - Primitive (loss-policy=stop, initial)" ], [ "ticket-primitive-2", "Ticket - Primitive (loss-policy=stop, granted)" ], [ "ticket-primitive-3", "Ticket - Primitive (loss-policy-stop, revoked)" ], [ "ticket-primitive-4", "Ticket - Primitive (loss-policy=demote, initial)" ], [ "ticket-primitive-5", "Ticket - Primitive (loss-policy=demote, granted)" ], [ "ticket-primitive-6", "Ticket - Primitive (loss-policy=demote, revoked)" ], [ "ticket-primitive-7", "Ticket - Primitive (loss-policy=fence, initial)" ], [ "ticket-primitive-8", "Ticket - Primitive (loss-policy=fence, granted)" ], [ "ticket-primitive-9", "Ticket - Primitive (loss-policy=fence, revoked)" ], [ "ticket-primitive-10", "Ticket - Primitive (loss-policy=freeze, initial)" ], [ "ticket-primitive-11", "Ticket - Primitive (loss-policy=freeze, granted)" ], [ "ticket-primitive-12", "Ticket - Primitive (loss-policy=freeze, revoked)" ], [ "ticket-primitive-13", "Ticket - Primitive (loss-policy=stop, standby, granted)" ], [ "ticket-primitive-14", "Ticket - Primitive (loss-policy=stop, granted, standby)" ], [ "ticket-primitive-15", "Ticket - Primitive (loss-policy=stop, standby, revoked)" ], [ "ticket-primitive-16", "Ticket - Primitive (loss-policy=demote, standby, granted)" ], [ "ticket-primitive-17", "Ticket - Primitive (loss-policy=demote, granted, standby)" ], [ "ticket-primitive-18", "Ticket - Primitive (loss-policy=demote, standby, revoked)" ], [ "ticket-primitive-19", "Ticket - Primitive (loss-policy=fence, standby, granted)" ], [ "ticket-primitive-20", "Ticket - Primitive (loss-policy=fence, granted, standby)" ], [ "ticket-primitive-21", "Ticket - Primitive (loss-policy=fence, standby, revoked)" ], [ "ticket-primitive-22", "Ticket - Primitive (loss-policy=freeze, standby, granted)" ], [ "ticket-primitive-23", "Ticket - Primitive (loss-policy=freeze, granted, standby)" ], [ "ticket-primitive-24", "Ticket - Primitive (loss-policy=freeze, standby, revoked)" ], ], [ [ "ticket-group-1", "Ticket - Group (loss-policy=stop, initial)" ], [ "ticket-group-2", "Ticket - Group (loss-policy=stop, granted)" ], [ "ticket-group-3", "Ticket - Group (loss-policy-stop, revoked)" ], [ "ticket-group-4", "Ticket - Group (loss-policy=demote, initial)" ], [ "ticket-group-5", "Ticket - Group (loss-policy=demote, granted)" ], [ "ticket-group-6", "Ticket - Group (loss-policy=demote, revoked)" ], [ "ticket-group-7", "Ticket - Group (loss-policy=fence, initial)" ], [ "ticket-group-8", "Ticket - Group (loss-policy=fence, granted)" ], [ "ticket-group-9", "Ticket - Group (loss-policy=fence, revoked)" ], [ "ticket-group-10", "Ticket - Group (loss-policy=freeze, initial)" ], [ "ticket-group-11", "Ticket - Group (loss-policy=freeze, granted)" ], [ "ticket-group-12", "Ticket - Group (loss-policy=freeze, revoked)" ], [ "ticket-group-13", "Ticket - Group (loss-policy=stop, standby, granted)" ], [ "ticket-group-14", "Ticket - Group (loss-policy=stop, granted, standby)" ], [ "ticket-group-15", "Ticket - Group (loss-policy=stop, standby, revoked)" ], [ "ticket-group-16", "Ticket - Group (loss-policy=demote, standby, granted)" ], [ "ticket-group-17", "Ticket - Group (loss-policy=demote, granted, standby)" ], [ "ticket-group-18", "Ticket - Group (loss-policy=demote, standby, revoked)" ], [ "ticket-group-19", "Ticket - Group (loss-policy=fence, standby, granted)" ], [ "ticket-group-20", "Ticket - Group (loss-policy=fence, granted, standby)" ], [ "ticket-group-21", "Ticket - Group (loss-policy=fence, standby, revoked)" ], [ "ticket-group-22", "Ticket - Group (loss-policy=freeze, standby, granted)" ], [ "ticket-group-23", "Ticket - Group (loss-policy=freeze, granted, standby)" ], [ "ticket-group-24", "Ticket - Group (loss-policy=freeze, standby, revoked)" ], ], [ [ "ticket-clone-1", "Ticket - Clone (loss-policy=stop, initial)" ], [ "ticket-clone-2", "Ticket - Clone (loss-policy=stop, granted)" ], [ "ticket-clone-3", "Ticket - Clone (loss-policy-stop, revoked)" ], [ "ticket-clone-4", "Ticket - Clone (loss-policy=demote, initial)" ], [ "ticket-clone-5", "Ticket - Clone (loss-policy=demote, granted)" ], [ "ticket-clone-6", "Ticket - Clone (loss-policy=demote, revoked)" ], [ "ticket-clone-7", "Ticket - Clone (loss-policy=fence, initial)" ], [ "ticket-clone-8", "Ticket - Clone (loss-policy=fence, granted)" ], [ "ticket-clone-9", "Ticket - Clone (loss-policy=fence, revoked)" ], [ "ticket-clone-10", "Ticket - Clone (loss-policy=freeze, initial)" ], [ "ticket-clone-11", "Ticket - Clone (loss-policy=freeze, granted)" ], [ "ticket-clone-12", "Ticket - Clone (loss-policy=freeze, revoked)" ], [ "ticket-clone-13", "Ticket - Clone (loss-policy=stop, standby, granted)" ], [ "ticket-clone-14", "Ticket - Clone (loss-policy=stop, granted, standby)" ], [ "ticket-clone-15", "Ticket - Clone (loss-policy=stop, standby, revoked)" ], [ "ticket-clone-16", "Ticket - Clone (loss-policy=demote, standby, granted)" ], [ "ticket-clone-17", "Ticket - Clone (loss-policy=demote, granted, standby)" ], [ "ticket-clone-18", "Ticket - Clone (loss-policy=demote, standby, revoked)" ], [ "ticket-clone-19", "Ticket - Clone (loss-policy=fence, standby, granted)" ], [ "ticket-clone-20", "Ticket - Clone (loss-policy=fence, granted, standby)" ], [ "ticket-clone-21", "Ticket - Clone (loss-policy=fence, standby, revoked)" ], [ "ticket-clone-22", "Ticket - Clone (loss-policy=freeze, standby, granted)" ], [ "ticket-clone-23", "Ticket - Clone (loss-policy=freeze, granted, standby)" ], [ "ticket-clone-24", "Ticket - Clone (loss-policy=freeze, standby, revoked)" ], ], [ [ "ticket-promoted-1", "Ticket - Promoted (loss-policy=stop, initial)" ], [ "ticket-promoted-2", "Ticket - Promoted (loss-policy=stop, granted)" ], [ "ticket-promoted-3", "Ticket - Promoted (loss-policy-stop, revoked)" ], [ "ticket-promoted-4", "Ticket - Promoted (loss-policy=demote, initial)" ], [ "ticket-promoted-5", "Ticket - Promoted (loss-policy=demote, granted)" ], [ "ticket-promoted-6", "Ticket - Promoted (loss-policy=demote, revoked)" ], [ "ticket-promoted-7", "Ticket - Promoted (loss-policy=fence, initial)" ], [ "ticket-promoted-8", "Ticket - Promoted (loss-policy=fence, granted)" ], [ "ticket-promoted-9", "Ticket - Promoted (loss-policy=fence, revoked)" ], [ "ticket-promoted-10", "Ticket - Promoted (loss-policy=freeze, initial)" ], [ "ticket-promoted-11", "Ticket - Promoted (loss-policy=freeze, granted)" ], [ "ticket-promoted-12", "Ticket - Promoted (loss-policy=freeze, revoked)" ], [ "ticket-promoted-13", "Ticket - Promoted (loss-policy=stop, standby, granted)" ], [ "ticket-promoted-14", "Ticket - Promoted (loss-policy=stop, granted, standby)" ], [ "ticket-promoted-15", "Ticket - Promoted (loss-policy=stop, standby, revoked)" ], [ "ticket-promoted-16", "Ticket - Promoted (loss-policy=demote, standby, granted)" ], [ "ticket-promoted-17", "Ticket - Promoted (loss-policy=demote, granted, standby)" ], [ "ticket-promoted-18", "Ticket - Promoted (loss-policy=demote, standby, revoked)" ], [ "ticket-promoted-19", "Ticket - Promoted (loss-policy=fence, standby, granted)" ], [ "ticket-promoted-20", "Ticket - Promoted (loss-policy=fence, granted, standby)" ], [ "ticket-promoted-21", "Ticket - Promoted (loss-policy=fence, standby, revoked)" ], [ "ticket-promoted-22", "Ticket - Promoted (loss-policy=freeze, standby, granted)" ], [ "ticket-promoted-23", "Ticket - Promoted (loss-policy=freeze, granted, standby)" ], [ "ticket-promoted-24", "Ticket - Promoted (loss-policy=freeze, standby, revoked)" ], ], [ [ "ticket-rsc-sets-1", "Ticket - Resource sets (1 ticket, initial)" ], [ "ticket-rsc-sets-2", "Ticket - Resource sets (1 ticket, granted)" ], [ "ticket-rsc-sets-3", "Ticket - Resource sets (1 ticket, revoked)" ], [ "ticket-rsc-sets-4", "Ticket - Resource sets (2 tickets, initial)" ], [ "ticket-rsc-sets-5", "Ticket - Resource sets (2 tickets, granted)" ], [ "ticket-rsc-sets-6", "Ticket - Resource sets (2 tickets, granted)" ], [ "ticket-rsc-sets-7", "Ticket - Resource sets (2 tickets, revoked)" ], [ "ticket-rsc-sets-8", "Ticket - Resource sets (1 ticket, standby, granted)" ], [ "ticket-rsc-sets-9", "Ticket - Resource sets (1 ticket, granted, standby)" ], [ "ticket-rsc-sets-10", "Ticket - Resource sets (1 ticket, standby, revoked)" ], [ "ticket-rsc-sets-11", "Ticket - Resource sets (2 tickets, standby, granted)" ], [ "ticket-rsc-sets-12", "Ticket - Resource sets (2 tickets, standby, granted)" ], [ "ticket-rsc-sets-13", "Ticket - Resource sets (2 tickets, granted, standby)" ], [ "ticket-rsc-sets-14", "Ticket - Resource sets (2 tickets, standby, revoked)" ], [ "cluster-specific-params", "Cluster-specific instance attributes based on rules" ], [ "site-specific-params", "Site-specific instance attributes based on rules" ], ], [ [ "template-1", "Template - 1" ], [ "template-2", "Template - 2" ], [ "template-3", "Template - 3 (merge operations)" ], [ "template-coloc-1", "Template - Colocation 1" ], [ "template-coloc-2", "Template - Colocation 2" ], [ "template-coloc-3", "Template - Colocation 3" ], [ "template-order-1", "Template - Order 1" ], [ "template-order-2", "Template - Order 2" ], [ "template-order-3", "Template - Order 3" ], [ "template-ticket", "Template - Ticket" ], [ "template-rsc-sets-1", "Template - Resource Sets 1" ], [ "template-rsc-sets-2", "Template - Resource Sets 2" ], [ "template-rsc-sets-3", "Template - Resource Sets 3" ], [ "template-rsc-sets-4", "Template - Resource Sets 4" ], [ "template-clone-primitive", "Cloned primitive from template" ], [ "template-clone-group", "Cloned group from template" ], [ "location-sets-templates", "Resource sets and templates - Location" ], [ "tags-coloc-order-1", "Tags - Colocation and Order (Simple)" ], [ "tags-coloc-order-2", "Tags - Colocation and Order (Resource Sets with Templates)" ], [ "tags-location", "Tags - Location" ], [ "tags-ticket", "Tags - Ticket" ], ], [ [ "container-1", "Container - initial" ], [ "container-2", "Container - monitor failed" ], [ "container-3", "Container - stop failed" ], [ "container-4", "Container - reached migration-threshold" ], [ "container-group-1", "Container in group - initial" ], [ "container-group-2", "Container in group - monitor failed" ], [ "container-group-3", "Container in group - stop failed" ], [ "container-group-4", "Container in group - reached migration-threshold" ], [ "container-is-remote-node", "Place resource within container when container is remote-node" ], [ "bug-rh-1097457", "Kill user defined container/contents ordering" ], [ "bug-cl-5247", "Graph loop when recovering m/s resource in a container" ], [ "bundle-order-startup", "Bundle startup ordering" ], [ "bundle-order-partial-start", "Bundle startup ordering when some dependencies are already running" ], [ "bundle-order-partial-start-2", "Bundle startup ordering when some dependencies and the container are already running" ], [ "bundle-order-stop", "Bundle stop ordering" ], [ "bundle-order-partial-stop", "Bundle startup ordering when some dependencies are already stopped" ], [ "bundle-order-stop-on-remote", "Stop nested resource after bringing up the connection" ], [ "bundle-order-startup-clone", "Prevent startup because bundle isn't promoted" ], [ "bundle-order-startup-clone-2", "Bundle startup with clones" ], [ "bundle-order-stop-clone", "Stop bundle because clone is stopping" ], [ "bundle-nested-colocation", "Colocation of nested connection resources" ], [ "bundle-order-fencing", "Order pseudo bundle fencing after parent node fencing if both are happening" ], [ "bundle-probe-order-1", "order 1" ], [ "bundle-probe-order-2", "order 2" ], [ "bundle-probe-order-3", "order 3" ], [ "bundle-probe-remotes", "Ensure remotes get probed too" ], [ "bundle-replicas-change", "Change bundle from 1 replica to multiple" ], [ "nested-remote-recovery", "Recover bundle's container hosted on remote node" ], ], [ [ "whitebox-fail1", "Fail whitebox container rsc" ], [ "whitebox-fail2", "Fail cluster connection to guest node" ], [ "whitebox-fail3", "Failed containers should not run nested on remote nodes" ], [ "whitebox-start", "Start whitebox container with resources assigned to it" ], [ "whitebox-stop", "Stop whitebox container with resources assigned to it" ], [ "whitebox-move", "Move whitebox container with resources assigned to it" ], [ "whitebox-asymmetric", "Verify connection rsc opts-in based on container resource" ], [ "whitebox-ms-ordering", "Verify promote/demote can not occur before connection is established" ], [ "whitebox-ms-ordering-move", "Stop/Start cycle within a moving container" ], [ "whitebox-orphaned", "Properly shutdown orphaned whitebox container" ], [ "whitebox-orphan-ms", "Properly tear down orphan ms resources on remote-nodes" ], [ "whitebox-unexpectedly-running", "Recover container nodes the cluster did not start" ], [ "whitebox-migrate1", "Migrate both container and connection resource" ], [ "whitebox-imply-stop-on-fence", "imply stop action on container node rsc when host node is fenced" ], [ "whitebox-nested-group", "Verify guest remote-node works nested in a group" ], [ "guest-node-host-dies", "Verify guest node is recovered if host goes away" ], [ "guest-node-cleanup", "Order guest node connection recovery after container probe" ], [ "guest-host-not-fenceable", "Actions on guest node are unrunnable if host is unclean and cannot be fenced" ], ], [ [ "remote-startup-probes", "Baremetal remote-node startup probes" ], [ "remote-startup", "Startup a newly discovered remote-nodes with no status" ], [ "remote-fence-unclean", "Fence unclean baremetal remote-node" ], [ "remote-fence-unclean2", "Fence baremetal remote-node after cluster node fails and connection can not be recovered" ], [ "remote-fence-unclean-3", "Probe failed remote nodes (triggers fencing)" ], [ "remote-move", "Move remote-node connection resource" ], [ "remote-disable", "Disable a baremetal remote-node" ], [ "remote-probe-disable", "Probe then stop a baremetal remote-node" ], [ "remote-orphaned", "Properly shutdown orphaned connection resource" ], [ "remote-orphaned2", "verify we can handle orphaned remote connections with active resources on the remote" ], [ "remote-recover", "Recover connection resource after cluster-node fails" ], [ "remote-stale-node-entry", "Make sure we properly handle leftover remote-node entries in the node section" ], [ "remote-partial-migrate", "Make sure partial migrations are handled before ops on the remote node" ], [ "remote-partial-migrate2", "Make sure partial migration target is prefered for remote connection" ], [ "remote-recover-fail", "Make sure start failure causes fencing if rsc are active on remote" ], [ "remote-start-fail", "Make sure a start failure does not result in fencing if no active resources are on remote" ], [ "remote-unclean2", "Make monitor failure always results in fencing, even if no rsc are active on remote" ], [ "remote-fence-before-reconnect", "Fence before clearing recurring monitor failure" ], [ "remote-recovery", "Recover remote connections before attempting demotion" ], [ "remote-recover-connection", "Optimistically recovery of only the connection" ], [ "remote-recover-all", "Fencing when the connection has no home" ], [ "remote-recover-no-resources", "Fencing when the connection has no home and no active resources" ], [ "remote-recover-unknown", "Fencing when the connection has no home and the remote has no operation history" ], [ "remote-reconnect-delay", "Waiting for remote reconnect interval to expire" ], [ "remote-connection-unrecoverable", "Remote connection host must be fenced, with connection unrecoverable" ], [ "cancel-behind-moving-remote", "Route recurring monitor cancellations through original node of a moving remote connection" ], ], [ [ "resource-discovery", "Exercises resource-discovery location constraint option" ], [ "rsc-discovery-per-node", "Disable resource discovery per node" ], [ "shutdown-lock", "Ensure shutdown lock works properly" ], [ "shutdown-lock-expiration", "Ensure shutdown lock expiration works properly" ], ], [ [ "op-defaults", "Test op_defaults conditional expressions" ], [ "op-defaults-2", "Test op_defaults AND'ed conditional expressions" ], [ "op-defaults-3", "Test op_defaults precedence" ], [ "rsc-defaults", "Test rsc_defaults conditional expressions" ], [ "rsc-defaults-2", "Test rsc_defaults conditional expressions without type" ], ], [ [ "stop-all-resources", "Test stop-all-resources=true "], ], [ [ "ocf_degraded-remap-ocf_ok", "Test degraded remapped to OK" ], [ "ocf_degraded_promoted-remap-ocf_ok", "Test degraded promoted remapped to OK"], ], # @TODO: If pacemaker implements versioned attributes, uncomment these tests #[ # [ "versioned-resources", "Start resources with #ra-version rules" ], # [ "restart-versioned", "Restart resources on #ra-version change" ], # [ "reload-versioned", "Reload resources on #ra-version change" ], #], #[ # [ "versioned-operations-1", "Use #ra-version to configure operations of native resources" ], # [ "versioned-operations-2", "Use #ra-version to configure operations of stonith resources" ], # [ "versioned-operations-3", "Use #ra-version to configure operations of promotable resources" ], # [ "versioned-operations-4", "Use #ra-version to configure operations of groups of the resources" ], #], ] TESTS_64BIT = [ [ [ "year-2038", "Check handling of timestamps beyond 2038-01-19 03:14:08 UTC" ], ], ] # Constants substituted in the build process class BuildVars(object): SBINDIR = "@sbindir@" BUILDDIR = "@abs_top_builddir@" CRM_SCHEMA_DIRECTORY = "@CRM_SCHEMA_DIRECTORY@" # These values must be kept in sync with crm_exit_t class CrmExit(object): OK = 0 ERROR = 1 NOT_INSTALLED = 5 NOINPUT = 66 CANTCREAT = 73 def is_executable(path): """ Check whether a file at a given path is executable. """ try: return os.stat(path)[stat.ST_MODE] & stat.S_IXUSR except OSError: return False def diff(file1, file2, **kwargs): """ Call diff on two files """ return subprocess.call([ "diff", "-u", "-N", "--ignore-all-space", "--ignore-blank-lines", file1, file2 ], **kwargs) def sort_file(filename): """ Sort a file alphabetically """ with io.open(filename, "rt") as f: lines = sorted(f) with io.open(filename, "wt") as f: f.writelines(lines) def remove_files(filenames): """ Remove a list of files """ for filename in filenames: try: os.remove(filename) except OSError: pass def normalize(filename): """ Remove text from a file that isn't important for comparison """ if not hasattr(normalize, "patterns"): normalize.patterns = [ re.compile(r'crm_feature_set="[^"]*"'), re.compile(r'batch-limit="[0-9]*"') ] if os.path.isfile(filename): with io.open(filename, "rt") as f: lines = f.readlines() with io.open(filename, "wt") as f: for line in lines: for pattern in normalize.patterns: line = pattern.sub("", line) f.write(line) def cat(filename, dest=sys.stdout): """ Copy a file to a destination file descriptor """ with io.open(filename, "rt") as f: shutil.copyfileobj(f, dest) class CtsScheduler(object): """ Regression tests for Pacemaker's scheduler """ def _parse_args(self, argv): """ Parse command-line arguments """ parser = argparse.ArgumentParser(description=DESC) parser.add_argument('-V', '--verbose', action='count', help='Display any differences from expected output') parser.add_argument('--run', metavar='TEST', help=('Run only single specified test (any further ' 'arguments will be passed to crm_simulate)')) parser.add_argument('--update', action='store_true', help='Update expected results with actual results') parser.add_argument('-b', '--binary', metavar='PATH', help='Specify path to crm_simulate') parser.add_argument('-i', '--io-dir', metavar='PATH', help='Specify path to regression test data directory') parser.add_argument('-o', '--out-dir', metavar='PATH', help='Specify where intermediate and output files should go') parser.add_argument('-v', '--valgrind', action='store_true', help='Run all commands under valgrind') parser.add_argument('--valgrind-dhat', action='store_true', help='Run all commands under valgrind with heap analyzer') parser.add_argument('--valgrind-skip-output', action='store_true', help='If running under valgrind, do not display output') parser.add_argument('--testcmd-options', metavar='OPTIONS', default='', help='Additional options for command under test') # argparse can't handle "everything after --run TEST", so grab that self.single_test_args = [] narg = 0 for arg in argv: narg = narg + 1 if arg == '--run': (argv, self.single_test_args) = (argv[:narg+1], argv[narg+1:]) break self.args = parser.parse_args(argv[1:]) def _error(self, s): print(" * ERROR: %s" % s) def _failed(self, s): print(" * FAILED: %s" % s) def _get_valgrind_cmd(self): """ Return command arguments needed (or not) to run valgrind """ if self.args.valgrind: os.environ['G_SLICE'] = "always-malloc" return [ "valgrind", "-q", "--gen-suppressions=all", "--time-stamp=yes", "--trace-children=no", "--show-reachable=no", "--leak-check=full", "--num-callers=20", "--suppressions=%s/valgrind-pcmk.suppressions" % (self.test_home) ] if self.args.valgrind_dhat: os.environ['G_SLICE'] = "always-malloc" return [ "valgrind", "--tool=exp-dhat", "--time-stamp=yes", "--trace-children=no", "--show-top-n=100", "--num-callers=4" ] return [] def _get_simulator_cmd(self): """ Locate the simulation binary """ if self.args.binary is None: self.args.binary = BuildVars.BUILDDIR + "/tools/crm_simulate" if not is_executable(self.args.binary): self.args.binary = BuildVars.SBINDIR + "/crm_simulate" if not is_executable(self.args.binary): # @TODO it would be more pythonic to raise an exception self._error("Test binary " + self.args.binary + " not found") sys.exit(CrmExit.NOT_INSTALLED) return [ self.args.binary ] + shlex.split(self.args.testcmd_options) def set_schema_env(self): """ Ensure schema directory environment variable is set, if possible """ try: return os.environ['PCMK_schema_directory'] except KeyError: for d in [ os.path.join(BuildVars.BUILDDIR, "xml"), BuildVars.CRM_SCHEMA_DIRECTORY ]: if os.path.isdir(d): os.environ['PCMK_schema_directory'] = d return d return None def __init__(self, argv=sys.argv): # Ensure all command output is in portable locale for comparison os.environ['LC_ALL'] = "C" self._parse_args(argv) # Where this executable lives self.test_home = os.path.dirname(os.path.realpath(argv[0])) # Where test data resides if self.args.io_dir is None: self.args.io_dir = os.path.join(self.test_home, "scheduler") self.xml_input_dir = os.path.join(self.args.io_dir, "xml") self.expected_dir = os.path.join(self.args.io_dir, "exp") self.dot_expected_dir = os.path.join(self.args.io_dir, "dot") self.scores_dir = os.path.join(self.args.io_dir, "scores") self.summary_dir = os.path.join(self.args.io_dir, "summary") self.stderr_expected_dir = os.path.join(self.args.io_dir, "stderr") # Create a temporary directory to store diff file self.failed_dir = tempfile.mkdtemp(prefix='cts-scheduler_') # Where to store generated files if self.args.out_dir is None: self.args.out_dir = self.args.io_dir self.failed_filename = os.path.join(self.failed_dir, "test-output.diff") else: self.failed_filename = os.path.join(self.args.out_dir, "test-output.diff") os.environ['CIB_shadow_dir'] = self.args.out_dir self.failed_file = None self.outfile_out_dir = os.path.join(self.args.out_dir, "out") self.dot_out_dir = os.path.join(self.args.out_dir, "dot") self.scores_out_dir = os.path.join(self.args.out_dir, "scores") self.summary_out_dir = os.path.join(self.args.out_dir, "summary") self.stderr_out_dir = os.path.join(self.args.out_dir, "stderr") self.valgrind_out_dir = os.path.join(self.args.out_dir, "valgrind") # Single test mode (if requested) try: # User can give test base name or file name of a test input self.args.run = os.path.splitext(os.path.basename(self.args.run))[0] except (AttributeError, TypeError): pass # --run was not specified self.set_schema_env() # Arguments needed (or not) to run commands self.valgrind_args = self._get_valgrind_cmd() self.simulate_args = self._get_simulator_cmd() # Test counters self.num_failed = 0 self.num_tests = 0 # Ensure that the main output directory exists # We don't want to create it with os.makedirs below if not os.path.isdir(self.args.out_dir): self._error("Output directory missing; can't create output files") sys.exit(CrmExit.CANTCREAT) # Create output subdirectories if they don't exist try: os.makedirs(self.outfile_out_dir, 0o755, True) os.makedirs(self.dot_out_dir, 0o755, True) os.makedirs(self.scores_out_dir, 0o755, True) os.makedirs(self.summary_out_dir, 0o755, True) os.makedirs(self.stderr_out_dir, 0o755, True) if self.valgrind_args: os.makedirs(self.valgrind_out_dir, 0o755, True) except OSError as ex: self._error("Unable to create output subdirectory: %s" % ex) remove_files([ self.outfile_out_dir, self.dot_out_dir, self.scores_out_dir, self.summary_out_dir, self.stderr_out_dir, ]) sys.exit(CrmExit.CANTCREAT) def _compare_files(self, filename1, filename2): """ Add any file differences to failed results """ if diff(filename1, filename2, stdout=subprocess.DEVNULL) != 0: diff(filename1, filename2, stdout=self.failed_file, stderr=subprocess.DEVNULL) self.failed_file.write("\n") return True return False def run_one(self, test_name, test_desc, test_args=[]): """ Run one scheduler test """ print(" Test %-25s %s" % ((test_name + ":"), test_desc)) did_fail = False self.num_tests = self.num_tests + 1 # Test inputs input_filename = os.path.join( self.xml_input_dir, "%s.xml" % test_name) expected_filename = os.path.join( self.expected_dir, "%s.exp" % test_name) dot_expected_filename = os.path.join( self.dot_expected_dir, "%s.dot" % test_name) scores_filename = os.path.join( self.scores_dir, "%s.scores" % test_name) summary_filename = os.path.join( self.summary_dir, "%s.summary" % test_name) stderr_expected_filename = os.path.join( self.stderr_expected_dir, "%s.stderr" % test_name) # (Intermediate) test outputs output_filename = os.path.join( self.outfile_out_dir, "%s.out" % test_name) dot_output_filename = os.path.join( self.dot_out_dir, "%s.dot.pe" % test_name) score_output_filename = os.path.join( self.scores_out_dir, "%s.scores.pe" % test_name) summary_output_filename = os.path.join( self.summary_out_dir, "%s.summary.pe" % test_name) stderr_output_filename = os.path.join( self.stderr_out_dir, "%s.stderr.pe" % test_name) valgrind_output_filename = os.path.join( self.valgrind_out_dir, "%s.valgrind" % test_name) # Common arguments for running test test_cmd = [] if self.valgrind_args: test_cmd = self.valgrind_args + [ "--log-file=%s" % valgrind_output_filename ] test_cmd = test_cmd + self.simulate_args # @TODO It would be more pythonic to raise exceptions for errors, # then perhaps it would be nice to make a single-test class # Ensure necessary test inputs exist if not os.path.isfile(input_filename): self._error("No input") self.num_failed = self.num_failed + 1 return CrmExit.NOINPUT if not self.args.update and not os.path.isfile(expected_filename): self._error("no stored output") return CrmExit.NOINPUT # Run simulation to generate summary output if self.args.run: # Single test mode test_cmd_full = test_cmd + [ '-x', input_filename, '-S' ] + test_args print(" ".join(test_cmd_full)) else: # @TODO Why isn't test_args added here? test_cmd_full = test_cmd + [ '-x', input_filename, '-S' ] with io.open(summary_output_filename, "wt") as f: simulation = subprocess.Popen(test_cmd_full, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, env=os.environ) # This makes diff happy regardless of --enable-compat-2.0. # Use sed -E to make Linux and BSD special characters more compatible. sed = subprocess.Popen(["sed", "-E", "-e", "s/ocf::/ocf:/g", "-e", r"s/Masters:/Promoted:/", "-e", r"s/Slaves:/Unpromoted:/", "-e", r"s/ Master( |\[|$)/ Promoted\1/", "-e", r"s/ Slave / Unpromoted /", ], stdin=simulation.stdout, stdout=f, stderr=subprocess.STDOUT) simulation.stdout.close() sed.communicate() if self.args.run: cat(summary_output_filename) # Re-run simulation to generate dot, graph, and scores test_cmd_full = test_cmd + [ '-x', input_filename, '-D', dot_output_filename, '-G', output_filename, '-sSQ' ] + test_args with io.open(stderr_output_filename, "wt") as f_stderr, \ io.open(score_output_filename, "wt") as f_score: rc = subprocess.call(test_cmd_full, stdout=f_score, stderr=f_stderr, env=os.environ) # Check for test command failure if rc != CrmExit.OK: self._failed("Test returned: %d" % rc) did_fail = True print(" ".join(test_cmd_full)) # Check for valgrind errors if self.valgrind_args and not self.args.valgrind_skip_output: if os.stat(valgrind_output_filename).st_size > 0: self._failed("Valgrind reported errors") did_fail = True cat(valgrind_output_filename) remove_files([ valgrind_output_filename ]) # Check for core dump if os.path.isfile("core"): self._failed("Core-file detected: core." + test_name) did_fail = True os.rename("core", "%s/core.%s" % (self.test_home, test_name)) # Check any stderr output if os.path.isfile(stderr_expected_filename): if self._compare_files(stderr_expected_filename, stderr_output_filename): self._failed("stderr changed") did_fail = True elif os.stat(stderr_output_filename).st_size > 0: self._failed("Output was written to stderr") did_fail = True cat(stderr_output_filename) remove_files([ stderr_output_filename ]) # Check whether output graph exists, and normalize it if (not os.path.isfile(output_filename) or os.stat(output_filename).st_size == 0): self._error("No graph produced") did_fail = True self.num_failed = self.num_failed + 1 remove_files([ output_filename ]) return CrmExit.ERROR normalize(output_filename) # Check whether dot output exists, and sort it if (not os.path.isfile(dot_output_filename) or os.stat(dot_output_filename).st_size == 0): self._error("No dot-file summary produced") did_fail = True self.num_failed = self.num_failed + 1 remove_files([ dot_output_filename, output_filename ]) return CrmExit.ERROR with io.open(dot_output_filename, "rt") as f: first_line = f.readline() # "digraph" line with opening brace lines = f.readlines() last_line = lines[-1] # closing brace del lines[-1] lines = sorted(set(lines)) # unique sort with io.open(dot_output_filename, "wt") as f: f.write(first_line) f.writelines(lines) f.write(last_line) # Check whether score output exists, and sort it if (not os.path.isfile(score_output_filename) or os.stat(score_output_filename).st_size == 0): self._error("No allocation scores produced") did_fail = True self.num_failed = self.num_failed + 1 remove_files([ score_output_filename, output_filename ]) return CrmExit.ERROR else: sort_file(score_output_filename) if self.args.update: shutil.copyfile(output_filename, expected_filename) shutil.copyfile(dot_output_filename, dot_expected_filename) shutil.copyfile(score_output_filename, scores_filename) shutil.copyfile(summary_output_filename, summary_filename) print(" Updated expected outputs") if self._compare_files(summary_filename, summary_output_filename): self._failed("summary changed") did_fail = True if self._compare_files(dot_expected_filename, dot_output_filename): self._failed("dot-file summary changed") did_fail = True else: remove_files([ dot_output_filename ]) if self._compare_files(expected_filename, output_filename): self._failed("xml-file changed") did_fail = True if self._compare_files(scores_filename, score_output_filename): self._failed("scores-file changed") did_fail = True remove_files([ output_filename, score_output_filename, summary_output_filename]) if did_fail: self.num_failed = self.num_failed + 1 return CrmExit.ERROR return CrmExit.OK def run_all(self): """ Run all defined tests """ if platform.architecture()[0] == "64bit": TESTS.extend(TESTS_64BIT) for group in TESTS: for test in group: try: args = test[2] except IndexError: args = [] self.run_one(test[0], test[1], args) print() def _print_summary(self): """ Print a summary of parameters for this test run """ print("Test home is:\t" + self.test_home) print("Test binary is:\t" + self.args.binary) if 'PCMK_schema_directory' in os.environ: print("Schema home is:\t" + os.environ['PCMK_schema_directory']) if self.valgrind_args != []: print("Activating memory testing with valgrind") print() def _test_results(self): if self.num_failed == 0: shutil.rmtree(self.failed_dir) return CrmExit.OK if os.path.isfile(self.failed_filename) and os.stat(self.failed_filename).st_size != 0: if self.args.verbose: self._error("Results of %d failed tests (out of %d):" % (self.num_failed, self.num_tests)) cat(self.failed_filename) else: self._error("Results of %d failed tests (out of %d) are in %s" % (self.num_failed, self.num_tests, self.failed_filename)) self._error("Use -V to display them after running the tests") else: self._error("%d (of %d) tests failed (no diff results)" % (self.num_failed, self.num_tests)) if os.path.isfile(self.failed_filename): shutil.rmtree(self.failed_dir) return CrmExit.ERROR def run(self): """ Run test(s) as specified """ self._print_summary() # Zero out the error log self.failed_file = io.open(self.failed_filename, "wt") if self.args.run is None: print("Performing the following tests from " + self.args.io_dir) print() self.run_all() print() self.failed_file.close() rc = self._test_results() else: rc = self.run_one(self.args.run, "Single shot", self.single_test_args) self.failed_file.close() if self.num_failed > 0: print("\nFailures:\nThese have also been written to: " + self.failed_filename + "\n") cat(self.failed_filename) shutil.rmtree(self.failed_dir) return rc if __name__ == "__main__": sys.exit(CtsScheduler().run()) # vim: set filetype=python expandtab tabstop=4 softtabstop=4 shiftwidth=4 textwidth=120: diff --git a/cts/scheduler/dot/probe-target-of-failed-migrate_to-1.dot b/cts/scheduler/dot/probe-target-of-failed-migrate_to-1.dot new file mode 100644 index 0000000000..9453f238c9 --- /dev/null +++ b/cts/scheduler/dot/probe-target-of-failed-migrate_to-1.dot @@ -0,0 +1,4 @@ + digraph "g" { +"dummy1_monitor_0 node2" [ style=bold color="green" fontcolor="black"] +"st-sbd_monitor_0 node2" [ style=bold color="green" fontcolor="black"] +} diff --git a/cts/scheduler/dot/probe-target-of-failed-migrate_to-2.dot b/cts/scheduler/dot/probe-target-of-failed-migrate_to-2.dot new file mode 100644 index 0000000000..d8f1c9f22b --- /dev/null +++ b/cts/scheduler/dot/probe-target-of-failed-migrate_to-2.dot @@ -0,0 +1,2 @@ + digraph "g" { +} diff --git a/cts/scheduler/exp/probe-target-of-failed-migrate_to-1.exp b/cts/scheduler/exp/probe-target-of-failed-migrate_to-1.exp new file mode 100644 index 0000000000..4c01d683d3 --- /dev/null +++ b/cts/scheduler/exp/probe-target-of-failed-migrate_to-1.exp @@ -0,0 +1,20 @@ + + + + + + + + + + + + + + + + + + + + diff --git a/cts/scheduler/exp/probe-target-of-failed-migrate_to-2.exp b/cts/scheduler/exp/probe-target-of-failed-migrate_to-2.exp new file mode 100644 index 0000000000..56e315ff01 --- /dev/null +++ b/cts/scheduler/exp/probe-target-of-failed-migrate_to-2.exp @@ -0,0 +1 @@ + diff --git a/cts/scheduler/scores/probe-target-of-failed-migrate_to-1.scores b/cts/scheduler/scores/probe-target-of-failed-migrate_to-1.scores new file mode 100644 index 0000000000..6561a9fdf0 --- /dev/null +++ b/cts/scheduler/scores/probe-target-of-failed-migrate_to-1.scores @@ -0,0 +1,5 @@ + +pcmk__native_allocate: dummy1 allocation score on node1: INFINITY +pcmk__native_allocate: dummy1 allocation score on node2: 0 +pcmk__native_allocate: st-sbd allocation score on node1: 0 +pcmk__native_allocate: st-sbd allocation score on node2: 0 diff --git a/cts/scheduler/scores/probe-target-of-failed-migrate_to-2.scores b/cts/scheduler/scores/probe-target-of-failed-migrate_to-2.scores new file mode 100644 index 0000000000..6561a9fdf0 --- /dev/null +++ b/cts/scheduler/scores/probe-target-of-failed-migrate_to-2.scores @@ -0,0 +1,5 @@ + +pcmk__native_allocate: dummy1 allocation score on node1: INFINITY +pcmk__native_allocate: dummy1 allocation score on node2: 0 +pcmk__native_allocate: st-sbd allocation score on node1: 0 +pcmk__native_allocate: st-sbd allocation score on node2: 0 diff --git a/cts/scheduler/summary/probe-target-of-failed-migrate_to-1.summary b/cts/scheduler/summary/probe-target-of-failed-migrate_to-1.summary new file mode 100644 index 0000000000..1be0ea6036 --- /dev/null +++ b/cts/scheduler/summary/probe-target-of-failed-migrate_to-1.summary @@ -0,0 +1,23 @@ +Using the original execution date of: 2022-05-09 10:28:56Z +Current cluster status: + * Node List: + * Online: [ node1 node2 ] + + * Full List of Resources: + * st-sbd (stonith:external/sbd): Started node1 + * dummy1 (ocf:pacemaker:Dummy): Started node1 + +Transition Summary: + +Executing Cluster Transition: + * Resource action: st-sbd monitor on node2 + * Resource action: dummy1 monitor on node2 +Using the original execution date of: 2022-05-09 10:28:56Z + +Revised Cluster Status: + * Node List: + * Online: [ node1 node2 ] + + * Full List of Resources: + * st-sbd (stonith:external/sbd): Started node1 + * dummy1 (ocf:pacemaker:Dummy): Started node1 diff --git a/cts/scheduler/summary/probe-target-of-failed-migrate_to-2.summary b/cts/scheduler/summary/probe-target-of-failed-migrate_to-2.summary new file mode 100644 index 0000000000..6346e383fa --- /dev/null +++ b/cts/scheduler/summary/probe-target-of-failed-migrate_to-2.summary @@ -0,0 +1,19 @@ +Current cluster status: + * Node List: + * Online: [ node1 node2 ] + + * Full List of Resources: + * st-sbd (stonith:external/sbd): Started node1 + * dummy1 (ocf:pacemaker:Dummy): Started node1 + +Transition Summary: + +Executing Cluster Transition: + +Revised Cluster Status: + * Node List: + * Online: [ node1 node2 ] + + * Full List of Resources: + * st-sbd (stonith:external/sbd): Started node1 + * dummy1 (ocf:pacemaker:Dummy): Started node1 diff --git a/cts/scheduler/xml/clone-require-all-6.xml b/cts/scheduler/xml/clone-require-all-6.xml index d7fb6102b5..9a2a996976 100644 --- a/cts/scheduler/xml/clone-require-all-6.xml +++ b/cts/scheduler/xml/clone-require-all-6.xml @@ -1,150 +1,150 @@ - + diff --git a/cts/scheduler/xml/migration-behind-migrating-remote.xml b/cts/scheduler/xml/migration-behind-migrating-remote.xml index f7859f25a7..a6fe40038a 100644 --- a/cts/scheduler/xml/migration-behind-migrating-remote.xml +++ b/cts/scheduler/xml/migration-behind-migrating-remote.xml @@ -1,116 +1,116 @@ - + diff --git a/cts/scheduler/xml/probe-target-of-failed-migrate_to-1.xml b/cts/scheduler/xml/probe-target-of-failed-migrate_to-1.xml new file mode 100644 index 0000000000..f9ebb90b45 --- /dev/null +++ b/cts/scheduler/xml/probe-target-of-failed-migrate_to-1.xml @@ -0,0 +1,48 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/cts/scheduler/xml/probe-target-of-failed-migrate_to-2.xml b/cts/scheduler/xml/probe-target-of-failed-migrate_to-2.xml new file mode 100644 index 0000000000..63efce3ff9 --- /dev/null +++ b/cts/scheduler/xml/probe-target-of-failed-migrate_to-2.xml @@ -0,0 +1,55 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/include/crm/pengine/internal.h b/include/crm/pengine/internal.h index fe9a23b7e2..7f47e190d3 100644 --- a/include/crm/pengine/internal.h +++ b/include/crm/pengine/internal.h @@ -1,599 +1,600 @@ /* * Copyright 2004-2022 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef PE_INTERNAL__H # define PE_INTERNAL__H # include # include # include # include # include # include # include # define pe_rsc_info(rsc, fmt, args...) crm_log_tag(LOG_INFO, rsc ? rsc->id : "", fmt, ##args) # define pe_rsc_debug(rsc, fmt, args...) crm_log_tag(LOG_DEBUG, rsc ? rsc->id : "", fmt, ##args) # define pe_rsc_trace(rsc, fmt, args...) crm_log_tag(LOG_TRACE, rsc ? rsc->id : "", fmt, ##args) # define pe_err(fmt...) do { \ was_processing_error = TRUE; \ pcmk__config_err(fmt); \ } while (0) # define pe_warn(fmt...) do { \ was_processing_warning = TRUE; \ pcmk__config_warn(fmt); \ } while (0) # define pe_proc_err(fmt...) { was_processing_error = TRUE; crm_err(fmt); } # define pe_proc_warn(fmt...) { was_processing_warning = TRUE; crm_warn(fmt); } #define pe__set_working_set_flags(working_set, flags_to_set) do { \ (working_set)->flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, "Working set", crm_system_name, \ (working_set)->flags, (flags_to_set), #flags_to_set); \ } while (0) #define pe__clear_working_set_flags(working_set, flags_to_clear) do { \ (working_set)->flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_TRACE, "Working set", crm_system_name, \ (working_set)->flags, (flags_to_clear), #flags_to_clear); \ } while (0) #define pe__set_resource_flags(resource, flags_to_set) do { \ (resource)->flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, "Resource", (resource)->id, (resource)->flags, \ (flags_to_set), #flags_to_set); \ } while (0) #define pe__clear_resource_flags(resource, flags_to_clear) do { \ (resource)->flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_TRACE, "Resource", (resource)->id, (resource)->flags, \ (flags_to_clear), #flags_to_clear); \ } while (0) #define pe__set_action_flags(action, flags_to_set) do { \ (action)->flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, \ "Action", (action)->uuid, \ (action)->flags, \ (flags_to_set), \ #flags_to_set); \ } while (0) #define pe__clear_action_flags(action, flags_to_clear) do { \ (action)->flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_TRACE, \ "Action", (action)->uuid, \ (action)->flags, \ (flags_to_clear), \ #flags_to_clear); \ } while (0) #define pe__set_raw_action_flags(action_flags, action_name, flags_to_set) do { \ action_flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, "Action", action_name, \ (action_flags), \ (flags_to_set), #flags_to_set); \ } while (0) #define pe__clear_raw_action_flags(action_flags, action_name, flags_to_clear) do { \ action_flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_TRACE, \ "Action", action_name, \ (action_flags), \ (flags_to_clear), \ #flags_to_clear); \ } while (0) #define pe__set_action_flags_as(function, line, action, flags_to_set) do { \ (action)->flags = pcmk__set_flags_as((function), (line), \ LOG_TRACE, \ "Action", (action)->uuid, \ (action)->flags, \ (flags_to_set), \ #flags_to_set); \ } while (0) #define pe__clear_action_flags_as(function, line, action, flags_to_clear) do { \ (action)->flags = pcmk__clear_flags_as((function), (line), \ LOG_TRACE, \ "Action", (action)->uuid, \ (action)->flags, \ (flags_to_clear), \ #flags_to_clear); \ } while (0) #define pe__set_order_flags(order_flags, flags_to_set) do { \ order_flags = pcmk__set_flags_as(__func__, __LINE__, LOG_TRACE, \ "Ordering", "constraint", \ order_flags, (flags_to_set), \ #flags_to_set); \ } while (0) #define pe__clear_order_flags(order_flags, flags_to_clear) do { \ order_flags = pcmk__clear_flags_as(__func__, __LINE__, LOG_TRACE, \ "Ordering", "constraint", \ order_flags, (flags_to_clear), \ #flags_to_clear); \ } while (0) #define pe__set_graph_flags(graph_flags, gr_action, flags_to_set) do { \ graph_flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, "Graph", \ (gr_action)->uuid, graph_flags, \ (flags_to_set), #flags_to_set); \ } while (0) #define pe__clear_graph_flags(graph_flags, gr_action, flags_to_clear) do { \ graph_flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_TRACE, "Graph", \ (gr_action)->uuid, graph_flags, \ (flags_to_clear), #flags_to_clear); \ } while (0) // Some warnings we don't want to print every transition enum pe_warn_once_e { pe_wo_blind = (1 << 0), pe_wo_restart_type = (1 << 1), pe_wo_role_after = (1 << 2), pe_wo_poweroff = (1 << 3), pe_wo_require_all = (1 << 4), pe_wo_order_score = (1 << 5), pe_wo_neg_threshold = (1 << 6), pe_wo_remove_after = (1 << 7), pe_wo_ping_node = (1 << 8), }; extern uint32_t pe_wo; #define pe_warn_once(pe_wo_bit, fmt...) do { \ if (!pcmk_is_set(pe_wo, pe_wo_bit)) { \ if (pe_wo_bit == pe_wo_blind) { \ crm_warn(fmt); \ } else { \ pe_warn(fmt); \ } \ pe_wo = pcmk__set_flags_as(__func__, __LINE__, LOG_TRACE, \ "Warn-once", "logging", pe_wo, \ (pe_wo_bit), #pe_wo_bit); \ } \ } while (0); typedef struct pe__location_constraint_s { char *id; // Constraint XML ID pe_resource_t *rsc_lh; // Resource being located enum rsc_role_e role_filter; // Role to locate enum pe_discover_e discover_mode; // Resource discovery GList *node_list_rh; // List of pe_node_t* } pe__location_t; typedef struct pe__order_constraint_s { int id; enum pe_ordering type; void *lh_opaque; pe_resource_t *lh_rsc; pe_action_t *lh_action; char *lh_action_task; void *rh_opaque; pe_resource_t *rh_rsc; pe_action_t *rh_action; char *rh_action_task; } pe__ordering_t; typedef struct notify_data_s { GSList *keys; // Environment variable name/value pairs const char *action; pe_action_t *pre; pe_action_t *post; pe_action_t *pre_done; pe_action_t *post_done; GList *active; /* notify_entry_t* */ GList *inactive; /* notify_entry_t* */ GList *start; /* notify_entry_t* */ GList *stop; /* notify_entry_t* */ GList *demote; /* notify_entry_t* */ GList *promote; /* notify_entry_t* */ GList *promoted; /* notify_entry_t* */ GList *unpromoted; /* notify_entry_t* */ GHashTable *allowed_nodes; } notify_data_t; bool pe_can_fence(pe_working_set_t *data_set, pe_node_t *node); void add_hash_param(GHashTable * hash, const char *name, const char *value); char *native_parameter(pe_resource_t * rsc, pe_node_t * node, gboolean create, const char *name, pe_working_set_t * data_set); pe_node_t *native_location(const pe_resource_t *rsc, GList **list, int current); void pe_metadata(pcmk__output_t *out); void verify_pe_options(GHashTable * options); void common_update_score(pe_resource_t * rsc, const char *id, int score); void native_add_running(pe_resource_t * rsc, pe_node_t * node, pe_working_set_t * data_set, gboolean failed); gboolean native_unpack(pe_resource_t * rsc, pe_working_set_t * data_set); gboolean group_unpack(pe_resource_t * rsc, pe_working_set_t * data_set); gboolean clone_unpack(pe_resource_t * rsc, pe_working_set_t * data_set); gboolean pe__unpack_bundle(pe_resource_t *rsc, pe_working_set_t *data_set); pe_resource_t *native_find_rsc(pe_resource_t *rsc, const char *id, const pe_node_t *node, int flags); gboolean native_active(pe_resource_t * rsc, gboolean all); gboolean group_active(pe_resource_t * rsc, gboolean all); gboolean clone_active(pe_resource_t * rsc, gboolean all); gboolean pe__bundle_active(pe_resource_t *rsc, gboolean all); void native_print(pe_resource_t * rsc, const char *pre_text, long options, void *print_data); void group_print(pe_resource_t * rsc, const char *pre_text, long options, void *print_data); void clone_print(pe_resource_t * rsc, const char *pre_text, long options, void *print_data); void pe__print_bundle(pe_resource_t *rsc, const char *pre_text, long options, void *print_data); gchar * pcmk__native_output_string(pe_resource_t *rsc, const char *name, pe_node_t *node, uint32_t show_opts, const char *target_role, bool show_nodes); int pe__name_and_nvpairs_xml(pcmk__output_t *out, bool is_list, const char *tag_name , size_t pairs_count, ...); char *pe__node_display_name(pe_node_t *node, bool print_detail); static inline const char * pe__rsc_bool_str(pe_resource_t *rsc, uint64_t rsc_flag) { return pcmk__btoa(pcmk_is_set(rsc->flags, rsc_flag)); } int pe__clone_xml(pcmk__output_t *out, va_list args); int pe__clone_default(pcmk__output_t *out, va_list args); int pe__group_xml(pcmk__output_t *out, va_list args); int pe__group_default(pcmk__output_t *out, va_list args); int pe__bundle_xml(pcmk__output_t *out, va_list args); int pe__bundle_html(pcmk__output_t *out, va_list args); int pe__bundle_text(pcmk__output_t *out, va_list args); int pe__node_html(pcmk__output_t *out, va_list args); int pe__node_text(pcmk__output_t *out, va_list args); int pe__node_xml(pcmk__output_t *out, va_list args); int pe__resource_xml(pcmk__output_t *out, va_list args); int pe__resource_html(pcmk__output_t *out, va_list args); int pe__resource_text(pcmk__output_t *out, va_list args); void native_free(pe_resource_t * rsc); void group_free(pe_resource_t * rsc); void clone_free(pe_resource_t * rsc); void pe__free_bundle(pe_resource_t *rsc); enum rsc_role_e native_resource_state(const pe_resource_t * rsc, gboolean current); enum rsc_role_e group_resource_state(const pe_resource_t * rsc, gboolean current); enum rsc_role_e clone_resource_state(const pe_resource_t * rsc, gboolean current); enum rsc_role_e pe__bundle_resource_state(const pe_resource_t *rsc, gboolean current); void pe__count_common(pe_resource_t *rsc); void pe__count_bundle(pe_resource_t *rsc); gboolean common_unpack(xmlNode * xml_obj, pe_resource_t ** rsc, pe_resource_t * parent, pe_working_set_t * data_set); void common_free(pe_resource_t * rsc); pe_node_t *pe__copy_node(const pe_node_t *this_node); extern time_t get_effective_time(pe_working_set_t * data_set); /* Failure handling utilities (from failcounts.c) */ // bit flags for fail count handling options enum pe_fc_flags_e { pe_fc_default = (1 << 0), pe_fc_effective = (1 << 1), // don't count expired failures pe_fc_fillers = (1 << 2), // if container, include filler failures in count }; int pe_get_failcount(pe_node_t *node, pe_resource_t *rsc, time_t *last_failure, uint32_t flags, xmlNode *xml_op, pe_working_set_t *data_set); pe_action_t *pe__clear_failcount(pe_resource_t *rsc, pe_node_t *node, const char *reason, pe_working_set_t *data_set); /* Functions for finding/counting a resource's active nodes */ pe_node_t *pe__find_active_on(const pe_resource_t *rsc, unsigned int *count_all, unsigned int *count_clean); pe_node_t *pe__find_active_requires(const pe_resource_t *rsc, unsigned int *count); static inline pe_node_t * pe__current_node(const pe_resource_t *rsc) { return pe__find_active_on(rsc, NULL, NULL); } /* Binary like operators for lists of nodes */ extern void node_list_exclude(GHashTable * list, GList *list2, gboolean merge_scores); GHashTable *pe__node_list2table(GList *list); static inline gpointer pe_hash_table_lookup(GHashTable * hash, gconstpointer key) { if (hash) { return g_hash_table_lookup(hash, key); } return NULL; } extern pe_action_t *get_pseudo_op(const char *name, pe_working_set_t * data_set); extern gboolean order_actions(pe_action_t * lh_action, pe_action_t * rh_action, enum pe_ordering order); /* Printing functions for debug */ extern void print_str_str(gpointer key, gpointer value, gpointer user_data); extern void pe__output_node(pe_node_t * node, gboolean details, pcmk__output_t *out); void pe__show_node_weights_as(const char *file, const char *function, int line, bool to_log, pe_resource_t *rsc, const char *comment, GHashTable *nodes, pe_working_set_t *data_set); #define pe__show_node_weights(level, rsc, text, nodes, data_set) \ pe__show_node_weights_as(__FILE__, __func__, __LINE__, \ (level), (rsc), (text), (nodes), (data_set)) /* Sorting functions */ extern gint sort_rsc_priority(gconstpointer a, gconstpointer b); extern gint sort_rsc_index(gconstpointer a, gconstpointer b); extern xmlNode *find_rsc_op_entry(pe_resource_t * rsc, const char *key); extern pe_action_t *custom_action(pe_resource_t * rsc, char *key, const char *task, pe_node_t * on_node, gboolean optional, gboolean foo, pe_working_set_t * data_set); # define delete_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_DELETE, 0) # define delete_action(rsc, node, optional) custom_action( \ rsc, delete_key(rsc), CRMD_ACTION_DELETE, node, \ optional, TRUE, data_set); # define stopped_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_STOPPED, 0) # define stopped_action(rsc, node, optional) custom_action( \ rsc, stopped_key(rsc), CRMD_ACTION_STOPPED, node, \ optional, TRUE, data_set); # define stop_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_STOP, 0) # define stop_action(rsc, node, optional) custom_action( \ rsc, stop_key(rsc), CRMD_ACTION_STOP, node, \ optional, TRUE, data_set); # define reload_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_RELOAD_AGENT, 0) # define start_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_START, 0) # define start_action(rsc, node, optional) custom_action( \ rsc, start_key(rsc), CRMD_ACTION_START, node, \ optional, TRUE, data_set) # define started_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_STARTED, 0) # define started_action(rsc, node, optional) custom_action( \ rsc, started_key(rsc), CRMD_ACTION_STARTED, node, \ optional, TRUE, data_set) # define promote_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_PROMOTE, 0) # define promote_action(rsc, node, optional) custom_action( \ rsc, promote_key(rsc), CRMD_ACTION_PROMOTE, node, \ optional, TRUE, data_set) # define promoted_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_PROMOTED, 0) # define promoted_action(rsc, node, optional) custom_action( \ rsc, promoted_key(rsc), CRMD_ACTION_PROMOTED, node, \ optional, TRUE, data_set) # define demote_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_DEMOTE, 0) # define demote_action(rsc, node, optional) custom_action( \ rsc, demote_key(rsc), CRMD_ACTION_DEMOTE, node, \ optional, TRUE, data_set) # define demoted_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_DEMOTED, 0) # define demoted_action(rsc, node, optional) custom_action( \ rsc, demoted_key(rsc), CRMD_ACTION_DEMOTED, node, \ optional, TRUE, data_set) extern int pe_get_configured_timeout(pe_resource_t *rsc, const char *action, pe_working_set_t *data_set); extern pe_action_t *find_first_action(GList *input, const char *uuid, const char *task, pe_node_t * on_node); extern enum action_tasks get_complex_task(pe_resource_t * rsc, const char *name, gboolean allow_non_atomic); extern GList *find_actions(GList *input, const char *key, const pe_node_t *on_node); GList *find_actions_exact(GList *input, const char *key, const pe_node_t *on_node); extern GList *find_recurring_actions(GList *input, pe_node_t * not_on_node); GList *pe__resource_actions(const pe_resource_t *rsc, const pe_node_t *node, const char *task, bool require_node); extern void pe_free_action(pe_action_t * action); extern void resource_location(pe_resource_t * rsc, pe_node_t * node, int score, const char *tag, pe_working_set_t * data_set); +extern int pe__is_newer_op(const xmlNode *xml_a, const xmlNode *xml_b); extern gint sort_op_by_callid(gconstpointer a, gconstpointer b); extern gboolean get_target_role(pe_resource_t * rsc, enum rsc_role_e *role); void pe__set_next_role(pe_resource_t *rsc, enum rsc_role_e role, const char *why); extern pe_resource_t *find_clone_instance(pe_resource_t * rsc, const char *sub_id, pe_working_set_t * data_set); extern void destroy_ticket(gpointer data); extern pe_ticket_t *ticket_new(const char *ticket_id, pe_working_set_t * data_set); // Resources for manipulating resource names const char *pe_base_name_end(const char *id); char *clone_strip(const char *last_rsc_id); char *clone_zero(const char *last_rsc_id); static inline bool pe_base_name_eq(pe_resource_t *rsc, const char *id) { if (id && rsc && rsc->id) { // Number of characters in rsc->id before any clone suffix size_t base_len = pe_base_name_end(rsc->id) - rsc->id + 1; return (strlen(id) == base_len) && !strncmp(id, rsc->id, base_len); } return false; } int pe__target_rc_from_xml(xmlNode *xml_op); gint sort_node_uname(gconstpointer a, gconstpointer b); bool is_set_recursive(pe_resource_t * rsc, long long flag, bool any); enum rsc_digest_cmp_val { /*! Digests are the same */ RSC_DIGEST_MATCH = 0, /*! Params that require a restart changed */ RSC_DIGEST_RESTART, /*! Some parameter changed. */ RSC_DIGEST_ALL, /*! rsc op didn't have a digest associated with it, so * it is unknown if parameters changed or not. */ RSC_DIGEST_UNKNOWN, }; typedef struct op_digest_cache_s { enum rsc_digest_cmp_val rc; xmlNode *params_all; xmlNode *params_secure; xmlNode *params_restart; char *digest_all_calc; char *digest_secure_calc; char *digest_restart_calc; } op_digest_cache_t; op_digest_cache_t *pe__calculate_digests(pe_resource_t *rsc, const char *task, guint *interval_ms, pe_node_t *node, xmlNode *xml_op, GHashTable *overrides, bool calc_secure, pe_working_set_t *data_set); void pe__free_digests(gpointer ptr); op_digest_cache_t *rsc_action_digest_cmp(pe_resource_t * rsc, xmlNode * xml_op, pe_node_t * node, pe_working_set_t * data_set); pe_action_t *pe_fence_op(pe_node_t * node, const char *op, bool optional, const char *reason, bool priority_delay, pe_working_set_t * data_set); void trigger_unfencing( pe_resource_t * rsc, pe_node_t *node, const char *reason, pe_action_t *dependency, pe_working_set_t * data_set); char *pe__action2reason(pe_action_t *action, enum pe_action_flags flag); void pe_action_set_reason(pe_action_t *action, const char *reason, bool overwrite); void pe__set_resource_flags_recursive(pe_resource_t *rsc, uint64_t flags); void pe__clear_resource_flags_recursive(pe_resource_t *rsc, uint64_t flags); void pe__clear_resource_flags_on_all(pe_working_set_t *data_set, uint64_t flag); gboolean add_tag_ref(GHashTable * tags, const char * tag_name, const char * obj_ref); void print_rscs_brief(GList *rsc_list, const char * pre_text, long options, void * print_data, gboolean print_all); int pe__rscs_brief_output(pcmk__output_t *out, GList *rsc_list, unsigned int options); void pe_fence_node(pe_working_set_t * data_set, pe_node_t * node, const char *reason, bool priority_delay); pe_node_t *pe_create_node(const char *id, const char *uname, const char *type, const char *score, pe_working_set_t * data_set); void common_print(pe_resource_t * rsc, const char *pre_text, const char *name, pe_node_t *node, long options, void *print_data); int pe__common_output_text(pcmk__output_t *out, pe_resource_t * rsc, const char *name, pe_node_t *node, unsigned int options); int pe__common_output_html(pcmk__output_t *out, pe_resource_t * rsc, const char *name, pe_node_t *node, unsigned int options); pe_resource_t *pe__find_bundle_replica(const pe_resource_t *bundle, const pe_node_t *node); bool pe__bundle_needs_remote_name(pe_resource_t *rsc, pe_working_set_t *data_set); const char *pe__add_bundle_remote_name(pe_resource_t *rsc, pe_working_set_t *data_set, xmlNode *xml, const char *field); const char *pe_node_attribute_calculated(const pe_node_t *node, const char *name, const pe_resource_t *rsc); const char *pe_node_attribute_raw(pe_node_t *node, const char *name); bool pe__is_universal_clone(pe_resource_t *rsc, pe_working_set_t *data_set); void pe__add_param_check(xmlNode *rsc_op, pe_resource_t *rsc, pe_node_t *node, enum pe_check_parameters, pe_working_set_t *data_set); void pe__foreach_param_check(pe_working_set_t *data_set, void (*cb)(pe_resource_t*, pe_node_t*, xmlNode*, enum pe_check_parameters, pe_working_set_t*)); void pe__free_param_checks(pe_working_set_t *data_set); bool pe__shutdown_requested(pe_node_t *node); void pe__update_recheck_time(time_t recheck, pe_working_set_t *data_set); /*! * \internal * \brief Register xml formatting message functions. */ void pe__register_messages(pcmk__output_t *out); void pe__unpack_dataset_nvpairs(xmlNode *xml_obj, const char *set_name, pe_rule_eval_data_t *rule_data, GHashTable *hash, const char *always_first, gboolean overwrite, pe_working_set_t *data_set); bool pe__resource_is_disabled(pe_resource_t *rsc); pe_action_t *pe__clear_resource_history(pe_resource_t *rsc, pe_node_t *node, pe_working_set_t *data_set); GList *pe__rscs_with_tag(pe_working_set_t *data_set, const char *tag_name); GList *pe__unames_with_tag(pe_working_set_t *data_set, const char *tag_name); bool pe__rsc_has_tag(pe_working_set_t *data_set, const char *rsc, const char *tag); bool pe__uname_has_tag(pe_working_set_t *data_set, const char *node, const char *tag); bool pe__rsc_running_on_any(pe_resource_t *rsc, GList *node_list); GList *pe__filter_rsc_list(GList *rscs, GList *filter); GList * pe__build_node_name_list(pe_working_set_t *data_set, const char *s); GList * pe__build_rsc_list(pe_working_set_t *data_set, const char *s); bool pcmk__rsc_filtered_by_node(pe_resource_t *rsc, GList *only_node); gboolean pe__bundle_is_filtered(pe_resource_t *rsc, GList *only_rsc, gboolean check_parent); gboolean pe__clone_is_filtered(pe_resource_t *rsc, GList *only_rsc, gboolean check_parent); gboolean pe__group_is_filtered(pe_resource_t *rsc, GList *only_rsc, gboolean check_parent); gboolean pe__native_is_filtered(pe_resource_t *rsc, GList *only_rsc, gboolean check_parent); xmlNode *pe__failed_probe_for_rsc(pe_resource_t *rsc, const char *name); const char *pe__clone_child_id(pe_resource_t *rsc); int pe__sum_node_health_scores(const pe_node_t *node, int base_health); int pe__node_health(pe_node_t *node); static inline enum pcmk__health_strategy pe__health_strategy(pe_working_set_t *data_set) { return pcmk__parse_health_strategy(pe_pref(data_set->config_hash, PCMK__OPT_NODE_HEALTH_STRATEGY)); } static inline int pe__health_score(const char *option, pe_working_set_t *data_set) { return char2score(pe_pref(data_set->config_hash, option)); } #endif diff --git a/lib/pengine/unpack.c b/lib/pengine/unpack.c index 426022013d..6c3174472d 100644 --- a/lib/pengine/unpack.c +++ b/lib/pengine/unpack.c @@ -1,4141 +1,4211 @@ /* * Copyright 2004-2022 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 CRM_TRACE_INIT_DATA(pe_status); /* This uses pcmk__set_flags_as()/pcmk__clear_flags_as() directly rather than * use pe__set_working_set_flags()/pe__clear_working_set_flags() so that the * flag is stringified more readably in log messages. */ #define set_config_flag(data_set, option, flag) do { \ const char *scf_value = pe_pref((data_set)->config_hash, (option)); \ if (scf_value != NULL) { \ if (crm_is_true(scf_value)) { \ (data_set)->flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, "Working set", \ crm_system_name, (data_set)->flags, \ (flag), #flag); \ } else { \ (data_set)->flags = pcmk__clear_flags_as(__func__, __LINE__,\ LOG_TRACE, "Working set", \ crm_system_name, (data_set)->flags, \ (flag), #flag); \ } \ } \ } while(0) static void unpack_rsc_op(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op, xmlNode **last_failure, enum action_fail_response *failed, pe_working_set_t *data_set); static void determine_remote_online_status(pe_working_set_t *data_set, pe_node_t *this_node); static void add_node_attrs(xmlNode *attrs, pe_node_t *node, bool overwrite, pe_working_set_t *data_set); static void determine_online_status(xmlNode *node_state, pe_node_t *this_node, pe_working_set_t *data_set); static void unpack_node_lrm(pe_node_t *node, xmlNode *xml, pe_working_set_t *data_set); // Bitmask for warnings we only want to print once uint32_t pe_wo = 0; static gboolean is_dangling_guest_node(pe_node_t *node) { /* we are looking for a remote-node that was supposed to be mapped to a * container resource, but all traces of that container have disappeared * from both the config and the status section. */ if (pe__is_guest_or_remote_node(node) && node->details->remote_rsc && node->details->remote_rsc->container == NULL && pcmk_is_set(node->details->remote_rsc->flags, pe_rsc_orphan_container_filler)) { return TRUE; } return FALSE; } /*! * \brief Schedule a fence action for a node * * \param[in,out] data_set Current working set of cluster * \param[in,out] node Node to fence * \param[in] reason Text description of why fencing is needed * \param[in] priority_delay Whether to consider `priority-fencing-delay` */ void pe_fence_node(pe_working_set_t * data_set, pe_node_t * node, const char *reason, bool priority_delay) { CRM_CHECK(node, return); /* A guest node is fenced by marking its container as failed */ if (pe__is_guest_node(node)) { pe_resource_t *rsc = node->details->remote_rsc->container; if (!pcmk_is_set(rsc->flags, pe_rsc_failed)) { if (!pcmk_is_set(rsc->flags, pe_rsc_managed)) { crm_notice("Not fencing guest node %s " "(otherwise would because %s): " "its guest resource %s is unmanaged", node->details->uname, reason, rsc->id); } else { crm_warn("Guest node %s will be fenced " "(by recovering its guest resource %s): %s", node->details->uname, rsc->id, reason); /* We don't mark the node as unclean because that would prevent the * node from running resources. We want to allow it to run resources * in this transition if the recovery succeeds. */ node->details->remote_requires_reset = TRUE; pe__set_resource_flags(rsc, pe_rsc_failed|pe_rsc_stop); } } } else if (is_dangling_guest_node(node)) { crm_info("Cleaning up dangling connection for guest node %s: " "fencing was already done because %s, " "and guest resource no longer exists", node->details->uname, reason); pe__set_resource_flags(node->details->remote_rsc, pe_rsc_failed|pe_rsc_stop); } else if (pe__is_remote_node(node)) { pe_resource_t *rsc = node->details->remote_rsc; if ((rsc != NULL) && !pcmk_is_set(rsc->flags, pe_rsc_managed)) { crm_notice("Not fencing remote node %s " "(otherwise would because %s): connection is unmanaged", node->details->uname, reason); } else if(node->details->remote_requires_reset == FALSE) { node->details->remote_requires_reset = TRUE; crm_warn("Remote node %s %s: %s", node->details->uname, pe_can_fence(data_set, node)? "will be fenced" : "is unclean", reason); } node->details->unclean = TRUE; // No need to apply `priority-fencing-delay` for remote nodes pe_fence_op(node, NULL, TRUE, reason, FALSE, data_set); } else if (node->details->unclean) { crm_trace("Cluster node %s %s because %s", node->details->uname, pe_can_fence(data_set, node)? "would also be fenced" : "also is unclean", reason); } else { crm_warn("Cluster node %s %s: %s", node->details->uname, pe_can_fence(data_set, node)? "will be fenced" : "is unclean", reason); node->details->unclean = TRUE; pe_fence_op(node, NULL, TRUE, reason, priority_delay, data_set); } } // @TODO xpaths can't handle templates, rules, or id-refs // nvpair with provides or requires set to unfencing #define XPATH_UNFENCING_NVPAIR XML_CIB_TAG_NVPAIR \ "[(@" XML_NVPAIR_ATTR_NAME "='" PCMK_STONITH_PROVIDES "'" \ "or @" XML_NVPAIR_ATTR_NAME "='" XML_RSC_ATTR_REQUIRES "') " \ "and @" XML_NVPAIR_ATTR_VALUE "='unfencing']" // unfencing in rsc_defaults or any resource #define XPATH_ENABLE_UNFENCING \ "/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_RESOURCES \ "//" XML_TAG_META_SETS "/" XPATH_UNFENCING_NVPAIR \ "|/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_RSCCONFIG \ "/" XML_TAG_META_SETS "/" XPATH_UNFENCING_NVPAIR static void set_if_xpath(uint64_t flag, const char *xpath, pe_working_set_t *data_set) { xmlXPathObjectPtr result = NULL; if (!pcmk_is_set(data_set->flags, flag)) { result = xpath_search(data_set->input, xpath); if (result && (numXpathResults(result) > 0)) { pe__set_working_set_flags(data_set, flag); } freeXpathObject(result); } } gboolean unpack_config(xmlNode * config, pe_working_set_t * data_set) { const char *value = NULL; GHashTable *config_hash = pcmk__strkey_table(free, free); pe_rule_eval_data_t rule_data = { .node_hash = NULL, .role = RSC_ROLE_UNKNOWN, .now = data_set->now, .match_data = NULL, .rsc_data = NULL, .op_data = NULL }; data_set->config_hash = config_hash; pe__unpack_dataset_nvpairs(config, XML_CIB_TAG_PROPSET, &rule_data, config_hash, CIB_OPTIONS_FIRST, FALSE, data_set); verify_pe_options(data_set->config_hash); set_config_flag(data_set, "enable-startup-probes", pe_flag_startup_probes); if (!pcmk_is_set(data_set->flags, pe_flag_startup_probes)) { crm_info("Startup probes: disabled (dangerous)"); } value = pe_pref(data_set->config_hash, XML_ATTR_HAVE_WATCHDOG); if (value && crm_is_true(value)) { crm_info("Watchdog-based self-fencing will be performed via SBD if " "fencing is required and stonith-watchdog-timeout is nonzero"); pe__set_working_set_flags(data_set, pe_flag_have_stonith_resource); } /* Set certain flags via xpath here, so they can be used before the relevant * configuration sections are unpacked. */ set_if_xpath(pe_flag_enable_unfencing, XPATH_ENABLE_UNFENCING, data_set); value = pe_pref(data_set->config_hash, "stonith-timeout"); data_set->stonith_timeout = (int) crm_parse_interval_spec(value); crm_debug("STONITH timeout: %d", data_set->stonith_timeout); set_config_flag(data_set, "stonith-enabled", pe_flag_stonith_enabled); crm_debug("STONITH of failed nodes is %s", pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)? "enabled" : "disabled"); data_set->stonith_action = pe_pref(data_set->config_hash, "stonith-action"); if (!strcmp(data_set->stonith_action, "poweroff")) { pe_warn_once(pe_wo_poweroff, "Support for stonith-action of 'poweroff' is deprecated " "and will be removed in a future release (use 'off' instead)"); data_set->stonith_action = "off"; } crm_trace("STONITH will %s nodes", data_set->stonith_action); set_config_flag(data_set, "concurrent-fencing", pe_flag_concurrent_fencing); crm_debug("Concurrent fencing is %s", pcmk_is_set(data_set->flags, pe_flag_concurrent_fencing)? "enabled" : "disabled"); value = pe_pref(data_set->config_hash, XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY); if (value) { data_set->priority_fencing_delay = crm_parse_interval_spec(value) / 1000; crm_trace("Priority fencing delay is %ds", data_set->priority_fencing_delay); } set_config_flag(data_set, "stop-all-resources", pe_flag_stop_everything); crm_debug("Stop all active resources: %s", pcmk__btoa(pcmk_is_set(data_set->flags, pe_flag_stop_everything))); set_config_flag(data_set, "symmetric-cluster", pe_flag_symmetric_cluster); if (pcmk_is_set(data_set->flags, pe_flag_symmetric_cluster)) { crm_debug("Cluster is symmetric" " - resources can run anywhere by default"); } value = pe_pref(data_set->config_hash, "no-quorum-policy"); if (pcmk__str_eq(value, "ignore", pcmk__str_casei)) { data_set->no_quorum_policy = no_quorum_ignore; } else if (pcmk__str_eq(value, "freeze", pcmk__str_casei)) { data_set->no_quorum_policy = no_quorum_freeze; } else if (pcmk__str_eq(value, "demote", pcmk__str_casei)) { data_set->no_quorum_policy = no_quorum_demote; } else if (pcmk__str_eq(value, "suicide", pcmk__str_casei)) { if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { int do_panic = 0; crm_element_value_int(data_set->input, XML_ATTR_QUORUM_PANIC, &do_panic); if (do_panic || pcmk_is_set(data_set->flags, pe_flag_have_quorum)) { data_set->no_quorum_policy = no_quorum_suicide; } else { crm_notice("Resetting no-quorum-policy to 'stop': cluster has never had quorum"); data_set->no_quorum_policy = no_quorum_stop; } } else { pcmk__config_err("Resetting no-quorum-policy to 'stop' because " "fencing is disabled"); data_set->no_quorum_policy = no_quorum_stop; } } else { data_set->no_quorum_policy = no_quorum_stop; } switch (data_set->no_quorum_policy) { case no_quorum_freeze: crm_debug("On loss of quorum: Freeze resources"); break; case no_quorum_stop: crm_debug("On loss of quorum: Stop ALL resources"); break; case no_quorum_demote: crm_debug("On loss of quorum: " "Demote promotable resources and stop other resources"); break; case no_quorum_suicide: crm_notice("On loss of quorum: Fence all remaining nodes"); break; case no_quorum_ignore: crm_notice("On loss of quorum: Ignore"); break; } set_config_flag(data_set, "stop-orphan-resources", pe_flag_stop_rsc_orphans); crm_trace("Orphan resources are %s", pcmk_is_set(data_set->flags, pe_flag_stop_rsc_orphans)? "stopped" : "ignored"); set_config_flag(data_set, "stop-orphan-actions", pe_flag_stop_action_orphans); crm_trace("Orphan resource actions are %s", pcmk_is_set(data_set->flags, pe_flag_stop_action_orphans)? "stopped" : "ignored"); value = pe_pref(data_set->config_hash, "remove-after-stop"); if (value != NULL) { if (crm_is_true(value)) { pe__set_working_set_flags(data_set, pe_flag_remove_after_stop); #ifndef PCMK__COMPAT_2_0 pe_warn_once(pe_wo_remove_after, "Support for the remove-after-stop cluster property is" " deprecated and will be removed in a future release"); #endif } else { pe__clear_working_set_flags(data_set, pe_flag_remove_after_stop); } } set_config_flag(data_set, "maintenance-mode", pe_flag_maintenance_mode); crm_trace("Maintenance mode: %s", pcmk__btoa(pcmk_is_set(data_set->flags, pe_flag_maintenance_mode))); set_config_flag(data_set, "start-failure-is-fatal", pe_flag_start_failure_fatal); crm_trace("Start failures are %s", pcmk_is_set(data_set->flags, pe_flag_start_failure_fatal)? "always fatal" : "handled by failcount"); if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { set_config_flag(data_set, "startup-fencing", pe_flag_startup_fencing); } if (pcmk_is_set(data_set->flags, pe_flag_startup_fencing)) { crm_trace("Unseen nodes will be fenced"); } else { pe_warn_once(pe_wo_blind, "Blind faith: not fencing unseen nodes"); } pe__unpack_node_health_scores(data_set); data_set->placement_strategy = pe_pref(data_set->config_hash, "placement-strategy"); crm_trace("Placement strategy: %s", data_set->placement_strategy); set_config_flag(data_set, "shutdown-lock", pe_flag_shutdown_lock); crm_trace("Resources will%s be locked to cleanly shut down nodes", (pcmk_is_set(data_set->flags, pe_flag_shutdown_lock)? "" : " not")); if (pcmk_is_set(data_set->flags, pe_flag_shutdown_lock)) { value = pe_pref(data_set->config_hash, XML_CONFIG_ATTR_SHUTDOWN_LOCK_LIMIT); data_set->shutdown_lock = crm_parse_interval_spec(value) / 1000; crm_trace("Shutdown locks expire after %us", data_set->shutdown_lock); } return TRUE; } pe_node_t * pe_create_node(const char *id, const char *uname, const char *type, const char *score, pe_working_set_t * data_set) { pe_node_t *new_node = NULL; if (pe_find_node(data_set->nodes, uname) != NULL) { pcmk__config_warn("More than one node entry has name '%s'", uname); } new_node = calloc(1, sizeof(pe_node_t)); if (new_node == NULL) { return NULL; } new_node->weight = char2score(score); new_node->fixed = FALSE; new_node->details = calloc(1, sizeof(struct pe_node_shared_s)); if (new_node->details == NULL) { free(new_node); return NULL; } crm_trace("Creating node for entry %s/%s", uname, id); new_node->details->id = id; new_node->details->uname = uname; new_node->details->online = FALSE; new_node->details->shutdown = FALSE; new_node->details->rsc_discovery_enabled = TRUE; new_node->details->running_rsc = NULL; new_node->details->data_set = data_set; if (pcmk__str_eq(type, "member", pcmk__str_null_matches | pcmk__str_casei)) { new_node->details->type = node_member; } else if (pcmk__str_eq(type, "remote", pcmk__str_casei)) { new_node->details->type = node_remote; pe__set_working_set_flags(data_set, pe_flag_have_remote_nodes); } else { /* @COMPAT 'ping' is the default for backward compatibility, but it * should be changed to 'member' at a compatibility break */ if (!pcmk__str_eq(type, "ping", pcmk__str_casei)) { pcmk__config_warn("Node %s has unrecognized type '%s', " "assuming 'ping'", crm_str(uname), type); } pe_warn_once(pe_wo_ping_node, "Support for nodes of type 'ping' (such as %s) is " "deprecated and will be removed in a future release", crm_str(uname)); new_node->details->type = node_ping; } new_node->details->attrs = pcmk__strkey_table(free, free); if (pe__is_guest_or_remote_node(new_node)) { g_hash_table_insert(new_node->details->attrs, strdup(CRM_ATTR_KIND), strdup("remote")); } else { g_hash_table_insert(new_node->details->attrs, strdup(CRM_ATTR_KIND), strdup("cluster")); } new_node->details->utilization = pcmk__strkey_table(free, free); new_node->details->digest_cache = pcmk__strkey_table(free, pe__free_digests); data_set->nodes = g_list_insert_sorted(data_set->nodes, new_node, sort_node_uname); return new_node; } static const char * expand_remote_rsc_meta(xmlNode *xml_obj, xmlNode *parent, pe_working_set_t *data) { xmlNode *attr_set = NULL; xmlNode *attr = NULL; const char *container_id = ID(xml_obj); const char *remote_name = NULL; const char *remote_server = NULL; const char *remote_port = NULL; const char *connect_timeout = "60s"; const char *remote_allow_migrate=NULL; const char *is_managed = NULL; for (attr_set = pcmk__xe_first_child(xml_obj); attr_set != NULL; attr_set = pcmk__xe_next(attr_set)) { if (!pcmk__str_eq((const char *)attr_set->name, XML_TAG_META_SETS, pcmk__str_casei)) { continue; } for (attr = pcmk__xe_first_child(attr_set); attr != NULL; attr = pcmk__xe_next(attr)) { const char *value = crm_element_value(attr, XML_NVPAIR_ATTR_VALUE); const char *name = crm_element_value(attr, XML_NVPAIR_ATTR_NAME); if (pcmk__str_eq(name, XML_RSC_ATTR_REMOTE_NODE, pcmk__str_casei)) { remote_name = value; } else if (pcmk__str_eq(name, "remote-addr", pcmk__str_casei)) { remote_server = value; } else if (pcmk__str_eq(name, "remote-port", pcmk__str_casei)) { remote_port = value; } else if (pcmk__str_eq(name, "remote-connect-timeout", pcmk__str_casei)) { connect_timeout = value; } else if (pcmk__str_eq(name, "remote-allow-migrate", pcmk__str_casei)) { remote_allow_migrate=value; } else if (pcmk__str_eq(name, XML_RSC_ATTR_MANAGED, pcmk__str_casei)) { is_managed = value; } } } if (remote_name == NULL) { return NULL; } if (pe_find_resource(data->resources, remote_name) != NULL) { return NULL; } pe_create_remote_xml(parent, remote_name, container_id, remote_allow_migrate, is_managed, connect_timeout, remote_server, remote_port); return remote_name; } static void handle_startup_fencing(pe_working_set_t *data_set, pe_node_t *new_node) { if ((new_node->details->type == node_remote) && (new_node->details->remote_rsc == NULL)) { /* Ignore fencing for remote nodes that don't have a connection resource * associated with them. This happens when remote node entries get left * in the nodes section after the connection resource is removed. */ return; } if (pcmk_is_set(data_set->flags, pe_flag_startup_fencing)) { // All nodes are unclean until we've seen their status entry new_node->details->unclean = TRUE; } else { // Blind faith ... new_node->details->unclean = FALSE; } /* We need to be able to determine if a node's status section * exists or not separate from whether the node is unclean. */ new_node->details->unseen = TRUE; } gboolean unpack_nodes(xmlNode * xml_nodes, pe_working_set_t * data_set) { xmlNode *xml_obj = NULL; pe_node_t *new_node = NULL; const char *id = NULL; const char *uname = NULL; const char *type = NULL; const char *score = NULL; pe_rule_eval_data_t rule_data = { .node_hash = NULL, .role = RSC_ROLE_UNKNOWN, .now = data_set->now, .match_data = NULL, .rsc_data = NULL, .op_data = NULL }; for (xml_obj = pcmk__xe_first_child(xml_nodes); xml_obj != NULL; xml_obj = pcmk__xe_next(xml_obj)) { if (pcmk__str_eq((const char *)xml_obj->name, XML_CIB_TAG_NODE, pcmk__str_none)) { new_node = NULL; id = crm_element_value(xml_obj, XML_ATTR_ID); uname = crm_element_value(xml_obj, XML_ATTR_UNAME); type = crm_element_value(xml_obj, XML_ATTR_TYPE); score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE); crm_trace("Processing node %s/%s", uname, id); if (id == NULL) { pcmk__config_err("Ignoring <" XML_CIB_TAG_NODE "> entry in configuration without id"); continue; } new_node = pe_create_node(id, uname, type, score, data_set); if (new_node == NULL) { return FALSE; } /* if(data_set->have_quorum == FALSE */ /* && data_set->no_quorum_policy == no_quorum_stop) { */ /* /\* start shutting resources down *\/ */ /* new_node->weight = -INFINITY; */ /* } */ handle_startup_fencing(data_set, new_node); add_node_attrs(xml_obj, new_node, FALSE, data_set); pe__unpack_dataset_nvpairs(xml_obj, XML_TAG_UTILIZATION, &rule_data, new_node->details->utilization, NULL, FALSE, data_set); crm_trace("Done with node %s", crm_element_value(xml_obj, XML_ATTR_UNAME)); } } if (data_set->localhost && pe_find_node(data_set->nodes, data_set->localhost) == NULL) { crm_info("Creating a fake local node"); pe_create_node(data_set->localhost, data_set->localhost, NULL, 0, data_set); } return TRUE; } static void setup_container(pe_resource_t * rsc, pe_working_set_t * data_set) { const char *container_id = NULL; if (rsc->children) { g_list_foreach(rsc->children, (GFunc) setup_container, data_set); return; } container_id = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_CONTAINER); if (container_id && !pcmk__str_eq(container_id, rsc->id, pcmk__str_casei)) { pe_resource_t *container = pe_find_resource(data_set->resources, container_id); if (container) { rsc->container = container; pe__set_resource_flags(container, pe_rsc_is_container); container->fillers = g_list_append(container->fillers, rsc); pe_rsc_trace(rsc, "Resource %s's container is %s", rsc->id, container_id); } else { pe_err("Resource %s: Unknown resource container (%s)", rsc->id, container_id); } } } gboolean unpack_remote_nodes(xmlNode * xml_resources, pe_working_set_t * data_set) { xmlNode *xml_obj = NULL; /* Create remote nodes and guest nodes from the resource configuration * before unpacking resources. */ for (xml_obj = pcmk__xe_first_child(xml_resources); xml_obj != NULL; xml_obj = pcmk__xe_next(xml_obj)) { const char *new_node_id = NULL; /* Check for remote nodes, which are defined by ocf:pacemaker:remote * primitives. */ if (xml_contains_remote_node(xml_obj)) { new_node_id = ID(xml_obj); /* The "pe_find_node" check is here to make sure we don't iterate over * an expanded node that has already been added to the node list. */ if (new_node_id && pe_find_node(data_set->nodes, new_node_id) == NULL) { crm_trace("Found remote node %s defined by resource %s", new_node_id, ID(xml_obj)); pe_create_node(new_node_id, new_node_id, "remote", NULL, data_set); } continue; } /* Check for guest nodes, which are defined by special meta-attributes * of a primitive of any type (for example, VirtualDomain or Xen). */ if (pcmk__str_eq((const char *)xml_obj->name, XML_CIB_TAG_RESOURCE, pcmk__str_none)) { /* This will add an ocf:pacemaker:remote primitive to the * configuration for the guest node's connection, to be unpacked * later. */ new_node_id = expand_remote_rsc_meta(xml_obj, xml_resources, data_set); if (new_node_id && pe_find_node(data_set->nodes, new_node_id) == NULL) { crm_trace("Found guest node %s in resource %s", new_node_id, ID(xml_obj)); pe_create_node(new_node_id, new_node_id, "remote", NULL, data_set); } continue; } /* Check for guest nodes inside a group. Clones are currently not * supported as guest nodes. */ if (pcmk__str_eq((const char *)xml_obj->name, XML_CIB_TAG_GROUP, pcmk__str_none)) { xmlNode *xml_obj2 = NULL; for (xml_obj2 = pcmk__xe_first_child(xml_obj); xml_obj2 != NULL; xml_obj2 = pcmk__xe_next(xml_obj2)) { new_node_id = expand_remote_rsc_meta(xml_obj2, xml_resources, data_set); if (new_node_id && pe_find_node(data_set->nodes, new_node_id) == NULL) { crm_trace("Found guest node %s in resource %s inside group %s", new_node_id, ID(xml_obj2), ID(xml_obj)); pe_create_node(new_node_id, new_node_id, "remote", NULL, data_set); } } } } return TRUE; } /* Call this after all the nodes and resources have been * unpacked, but before the status section is read. * * A remote node's online status is reflected by the state * of the remote node's connection resource. We need to link * the remote node to this connection resource so we can have * easy access to the connection resource during the scheduler calculations. */ static void link_rsc2remotenode(pe_working_set_t *data_set, pe_resource_t *new_rsc) { pe_node_t *remote_node = NULL; if (new_rsc->is_remote_node == FALSE) { return; } if (pcmk_is_set(data_set->flags, pe_flag_quick_location)) { /* remote_nodes and remote_resources are not linked in quick location calculations */ return; } remote_node = pe_find_node(data_set->nodes, new_rsc->id); CRM_CHECK(remote_node != NULL, return); pe_rsc_trace(new_rsc, "Linking remote connection resource %s to node %s", new_rsc->id, remote_node->details->uname); remote_node->details->remote_rsc = new_rsc; if (new_rsc->container == NULL) { /* Handle start-up fencing for remote nodes (as opposed to guest nodes) * the same as is done for cluster nodes. */ handle_startup_fencing(data_set, remote_node); } else { /* pe_create_node() marks the new node as "remote" or "cluster"; now * that we know the node is a guest node, update it correctly. */ g_hash_table_replace(remote_node->details->attrs, strdup(CRM_ATTR_KIND), strdup("container")); } } static void destroy_tag(gpointer data) { pe_tag_t *tag = data; if (tag) { free(tag->id); g_list_free_full(tag->refs, free); free(tag); } } /*! * \internal * \brief Parse configuration XML for resource information * * \param[in] xml_resources Top of resource configuration XML * \param[in,out] data_set Where to put resource information * * \return TRUE * * \note unpack_remote_nodes() MUST be called before this, so that the nodes can * be used when common_unpack() calls resource_location() */ gboolean unpack_resources(xmlNode * xml_resources, pe_working_set_t * data_set) { xmlNode *xml_obj = NULL; GList *gIter = NULL; data_set->template_rsc_sets = pcmk__strkey_table(free, destroy_tag); for (xml_obj = pcmk__xe_first_child(xml_resources); xml_obj != NULL; xml_obj = pcmk__xe_next(xml_obj)) { pe_resource_t *new_rsc = NULL; if (pcmk__str_eq((const char *)xml_obj->name, XML_CIB_TAG_RSC_TEMPLATE, pcmk__str_none)) { const char *template_id = ID(xml_obj); if (template_id && g_hash_table_lookup_extended(data_set->template_rsc_sets, template_id, NULL, NULL) == FALSE) { /* Record the template's ID for the knowledge of its existence anyway. */ g_hash_table_insert(data_set->template_rsc_sets, strdup(template_id), NULL); } continue; } crm_trace("Beginning unpack... <%s id=%s... >", crm_element_name(xml_obj), ID(xml_obj)); if (common_unpack(xml_obj, &new_rsc, NULL, data_set) && (new_rsc != NULL)) { data_set->resources = g_list_append(data_set->resources, new_rsc); pe_rsc_trace(new_rsc, "Added resource %s", new_rsc->id); } else { pcmk__config_err("Ignoring <%s> resource '%s' " "because configuration is invalid", crm_element_name(xml_obj), crm_str(ID(xml_obj))); if (new_rsc != NULL && new_rsc->fns != NULL) { new_rsc->fns->free(new_rsc); } } } for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) { pe_resource_t *rsc = (pe_resource_t *) gIter->data; setup_container(rsc, data_set); link_rsc2remotenode(data_set, rsc); } data_set->resources = g_list_sort(data_set->resources, sort_rsc_priority); if (pcmk_is_set(data_set->flags, pe_flag_quick_location)) { /* Ignore */ } else if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled) && !pcmk_is_set(data_set->flags, pe_flag_have_stonith_resource)) { pcmk__config_err("Resource start-up disabled since no STONITH resources have been defined"); pcmk__config_err("Either configure some or disable STONITH with the stonith-enabled option"); pcmk__config_err("NOTE: Clusters with shared data need STONITH to ensure data integrity"); } return TRUE; } gboolean unpack_tags(xmlNode * xml_tags, pe_working_set_t * data_set) { xmlNode *xml_tag = NULL; data_set->tags = pcmk__strkey_table(free, destroy_tag); for (xml_tag = pcmk__xe_first_child(xml_tags); xml_tag != NULL; xml_tag = pcmk__xe_next(xml_tag)) { xmlNode *xml_obj_ref = NULL; const char *tag_id = ID(xml_tag); if (!pcmk__str_eq((const char *)xml_tag->name, XML_CIB_TAG_TAG, pcmk__str_none)) { continue; } if (tag_id == NULL) { pcmk__config_err("Ignoring <%s> without " XML_ATTR_ID, crm_element_name(xml_tag)); continue; } for (xml_obj_ref = pcmk__xe_first_child(xml_tag); xml_obj_ref != NULL; xml_obj_ref = pcmk__xe_next(xml_obj_ref)) { const char *obj_ref = ID(xml_obj_ref); if (!pcmk__str_eq((const char *)xml_obj_ref->name, XML_CIB_TAG_OBJ_REF, pcmk__str_none)) { continue; } if (obj_ref == NULL) { pcmk__config_err("Ignoring <%s> for tag '%s' without " XML_ATTR_ID, crm_element_name(xml_obj_ref), tag_id); continue; } if (add_tag_ref(data_set->tags, tag_id, obj_ref) == FALSE) { return FALSE; } } } return TRUE; } /* The ticket state section: * "/cib/status/tickets/ticket_state" */ static gboolean unpack_ticket_state(xmlNode * xml_ticket, pe_working_set_t * data_set) { const char *ticket_id = NULL; const char *granted = NULL; const char *last_granted = NULL; const char *standby = NULL; xmlAttrPtr xIter = NULL; pe_ticket_t *ticket = NULL; ticket_id = ID(xml_ticket); if (pcmk__str_empty(ticket_id)) { return FALSE; } crm_trace("Processing ticket state for %s", ticket_id); ticket = g_hash_table_lookup(data_set->tickets, ticket_id); if (ticket == NULL) { ticket = ticket_new(ticket_id, data_set); if (ticket == NULL) { return FALSE; } } for (xIter = xml_ticket->properties; xIter; xIter = xIter->next) { const char *prop_name = (const char *)xIter->name; const char *prop_value = crm_element_value(xml_ticket, prop_name); if (pcmk__str_eq(prop_name, XML_ATTR_ID, pcmk__str_none)) { continue; } g_hash_table_replace(ticket->state, strdup(prop_name), strdup(prop_value)); } granted = g_hash_table_lookup(ticket->state, "granted"); if (granted && crm_is_true(granted)) { ticket->granted = TRUE; crm_info("We have ticket '%s'", ticket->id); } else { ticket->granted = FALSE; crm_info("We do not have ticket '%s'", ticket->id); } last_granted = g_hash_table_lookup(ticket->state, "last-granted"); if (last_granted) { long long last_granted_ll; pcmk__scan_ll(last_granted, &last_granted_ll, 0LL); ticket->last_granted = (time_t) last_granted_ll; } standby = g_hash_table_lookup(ticket->state, "standby"); if (standby && crm_is_true(standby)) { ticket->standby = TRUE; if (ticket->granted) { crm_info("Granted ticket '%s' is in standby-mode", ticket->id); } } else { ticket->standby = FALSE; } crm_trace("Done with ticket state for %s", ticket_id); return TRUE; } static gboolean unpack_tickets_state(xmlNode * xml_tickets, pe_working_set_t * data_set) { xmlNode *xml_obj = NULL; for (xml_obj = pcmk__xe_first_child(xml_tickets); xml_obj != NULL; xml_obj = pcmk__xe_next(xml_obj)) { if (!pcmk__str_eq((const char *)xml_obj->name, XML_CIB_TAG_TICKET_STATE, pcmk__str_none)) { continue; } unpack_ticket_state(xml_obj, data_set); } return TRUE; } static void unpack_handle_remote_attrs(pe_node_t *this_node, xmlNode *state, pe_working_set_t * data_set) { const char *resource_discovery_enabled = NULL; xmlNode *attrs = NULL; pe_resource_t *rsc = NULL; if (!pcmk__str_eq((const char *)state->name, XML_CIB_TAG_STATE, pcmk__str_none)) { return; } if ((this_node == NULL) || !pe__is_guest_or_remote_node(this_node)) { return; } crm_trace("Processing remote node id=%s, uname=%s", this_node->details->id, this_node->details->uname); pcmk__scan_min_int(crm_element_value(state, XML_NODE_IS_MAINTENANCE), &(this_node->details->remote_maintenance), 0); rsc = this_node->details->remote_rsc; if (this_node->details->remote_requires_reset == FALSE) { this_node->details->unclean = FALSE; this_node->details->unseen = FALSE; } attrs = find_xml_node(state, XML_TAG_TRANSIENT_NODEATTRS, FALSE); add_node_attrs(attrs, this_node, TRUE, data_set); if (pe__shutdown_requested(this_node)) { crm_info("Node %s is shutting down", this_node->details->uname); this_node->details->shutdown = TRUE; } if (crm_is_true(pe_node_attribute_raw(this_node, "standby"))) { crm_info("Node %s is in standby-mode", this_node->details->uname); this_node->details->standby = TRUE; } if (crm_is_true(pe_node_attribute_raw(this_node, "maintenance")) || ((rsc != NULL) && !pcmk_is_set(rsc->flags, pe_rsc_managed))) { crm_info("Node %s is in maintenance-mode", this_node->details->uname); this_node->details->maintenance = TRUE; } resource_discovery_enabled = pe_node_attribute_raw(this_node, XML_NODE_ATTR_RSC_DISCOVERY); if (resource_discovery_enabled && !crm_is_true(resource_discovery_enabled)) { if (pe__is_remote_node(this_node) && !pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { crm_warn("Ignoring %s attribute on remote node %s because stonith is disabled", XML_NODE_ATTR_RSC_DISCOVERY, this_node->details->uname); } else { /* This is either a remote node with fencing enabled, or a guest * node. We don't care whether fencing is enabled when fencing guest * nodes, because they are "fenced" by recovering their containing * resource. */ crm_info("Node %s has resource discovery disabled", this_node->details->uname); this_node->details->rsc_discovery_enabled = FALSE; } } } /*! * \internal * \brief Unpack a cluster node's transient attributes * * \param[in] state CIB node state XML * \param[in] node Cluster node whose attributes are being unpacked * \param[in] data_set Cluster working set */ static void unpack_transient_attributes(xmlNode *state, pe_node_t *node, pe_working_set_t *data_set) { const char *discovery = NULL; xmlNode *attrs = find_xml_node(state, XML_TAG_TRANSIENT_NODEATTRS, FALSE); add_node_attrs(attrs, node, TRUE, data_set); if (crm_is_true(pe_node_attribute_raw(node, "standby"))) { crm_info("Node %s is in standby-mode", node->details->uname); node->details->standby = TRUE; } if (crm_is_true(pe_node_attribute_raw(node, "maintenance"))) { crm_info("Node %s is in maintenance-mode", node->details->uname); node->details->maintenance = TRUE; } discovery = pe_node_attribute_raw(node, XML_NODE_ATTR_RSC_DISCOVERY); if ((discovery != NULL) && !crm_is_true(discovery)) { crm_warn("Ignoring %s attribute for node %s because disabling " "resource discovery is not allowed for cluster nodes", XML_NODE_ATTR_RSC_DISCOVERY, node->details->uname); } } /*! * \internal * \brief Unpack a node state entry (first pass) * * Unpack one node state entry from status. This unpacks information from the * node_state element itself and node attributes inside it, but not the * resource history inside it. Multiple passes through the status are needed to * fully unpack everything. * * \param[in] state CIB node state XML * \param[in] data_set Cluster working set */ static void unpack_node_state(xmlNode *state, pe_working_set_t *data_set) { const char *id = NULL; const char *uname = NULL; pe_node_t *this_node = NULL; id = crm_element_value(state, XML_ATTR_ID); if (id == NULL) { crm_warn("Ignoring malformed " XML_CIB_TAG_STATE " entry without " XML_ATTR_ID); return; } uname = crm_element_value(state, XML_ATTR_UNAME); if (uname == NULL) { crm_warn("Ignoring malformed " XML_CIB_TAG_STATE " entry without " XML_ATTR_UNAME); return; } this_node = pe_find_node_any(data_set->nodes, id, uname); if (this_node == NULL) { pcmk__config_warn("Ignoring recorded node state for '%s' because " "it is no longer in the configuration", uname); return; } if (pe__is_guest_or_remote_node(this_node)) { /* We can't determine the online status of Pacemaker Remote nodes until * after all resource history has been unpacked. In this first pass, we * do need to mark whether the node has been fenced, as this plays a * role during unpacking cluster node resource state. */ pcmk__scan_min_int(crm_element_value(state, XML_NODE_IS_FENCED), &(this_node->details->remote_was_fenced), 0); return; } unpack_transient_attributes(state, this_node, data_set); /* Provisionally mark this cluster node as clean. We have at least seen it * in the current cluster's lifetime. */ this_node->details->unclean = FALSE; this_node->details->unseen = FALSE; crm_trace("Determining online status of cluster node %s (id %s)", this_node->details->uname, id); determine_online_status(state, this_node, data_set); if (!pcmk_is_set(data_set->flags, pe_flag_have_quorum) && this_node->details->online && (data_set->no_quorum_policy == no_quorum_suicide)) { /* Everything else should flow from this automatically * (at least until the scheduler becomes able to migrate off * healthy resources) */ pe_fence_node(data_set, this_node, "cluster does not have quorum", FALSE); } } /*! * \internal * \brief Unpack nodes' resource history as much as possible * * Unpack as many nodes' resource history as possible in one pass through the * status. We need to process Pacemaker Remote nodes' connections/containers * before unpacking their history; the connection/container history will be * in another node's history, so it might take multiple passes to unpack * everything. * * \param[in] status CIB XML status section * \param[in] fence If true, treat any not-yet-unpacked nodes as unseen * \param[in] data_set Cluster working set * * \return Standard Pacemaker return code (specifically pcmk_rc_ok if done, * or EAGAIN if more unpacking remains to be done) */ static int unpack_node_history(xmlNode *status, bool fence, pe_working_set_t *data_set) { int rc = pcmk_rc_ok; // Loop through all node_state entries in CIB status for (xmlNode *state = first_named_child(status, XML_CIB_TAG_STATE); state != NULL; state = crm_next_same_xml(state)) { const char *id = ID(state); const char *uname = crm_element_value(state, XML_ATTR_UNAME); pe_node_t *this_node = NULL; if ((id == NULL) || (uname == NULL)) { // Warning already logged in first pass through status section crm_trace("Not unpacking resource history from malformed " XML_CIB_TAG_STATE " without id and/or uname"); continue; } this_node = pe_find_node_any(data_set->nodes, id, uname); if (this_node == NULL) { // Warning already logged in first pass through status section crm_trace("Not unpacking resource history for node %s because " "no longer in configuration", id); continue; } if (this_node->details->unpacked) { crm_trace("Not unpacking resource history for node %s because " "already unpacked", id); continue; } if (fence) { // We're processing all remaining nodes } else if (pe__is_guest_node(this_node)) { /* We can unpack a guest node's history only after we've unpacked * other resource history to the point that we know that the node's * connection and containing resource are both up. */ pe_resource_t *rsc = this_node->details->remote_rsc; if ((rsc == NULL) || (rsc->role != RSC_ROLE_STARTED) || (rsc->container->role != RSC_ROLE_STARTED)) { crm_trace("Not unpacking resource history for guest node %s " "because container and connection are not known to " "be up", id); continue; } } else if (pe__is_remote_node(this_node)) { /* We can unpack a remote node's history only after we've unpacked * other resource history to the point that we know that the node's * connection is up, with the exception of when shutdown locks are * in use. */ pe_resource_t *rsc = this_node->details->remote_rsc; if ((rsc == NULL) || (!pcmk_is_set(data_set->flags, pe_flag_shutdown_lock) && (rsc->role != RSC_ROLE_STARTED))) { crm_trace("Not unpacking resource history for remote node %s " "because connection is not known to be up", id); continue; } /* If fencing and shutdown locks are disabled and we're not processing * unseen nodes, then we don't want to unpack offline nodes until online * nodes have been unpacked. This allows us to number active clone * instances first. */ } else if (!pcmk_any_flags_set(data_set->flags, pe_flag_stonith_enabled |pe_flag_shutdown_lock) && !this_node->details->online) { crm_trace("Not unpacking resource history for offline " "cluster node %s", id); continue; } if (pe__is_guest_or_remote_node(this_node)) { determine_remote_online_status(data_set, this_node); unpack_handle_remote_attrs(this_node, state, data_set); } crm_trace("Unpacking resource history for %snode %s", (fence? "unseen " : ""), id); this_node->details->unpacked = TRUE; unpack_node_lrm(this_node, state, data_set); rc = EAGAIN; // Other node histories might depend on this one } return rc; } /* remove nodes that are down, stopping */ /* create positive rsc_to_node constraints between resources and the nodes they are running on */ /* anything else? */ gboolean unpack_status(xmlNode * status, pe_working_set_t * data_set) { xmlNode *state = NULL; crm_trace("Beginning unpack"); if (data_set->tickets == NULL) { data_set->tickets = pcmk__strkey_table(free, destroy_ticket); } for (state = pcmk__xe_first_child(status); state != NULL; state = pcmk__xe_next(state)) { if (pcmk__str_eq((const char *)state->name, XML_CIB_TAG_TICKETS, pcmk__str_none)) { unpack_tickets_state((xmlNode *) state, data_set); } else if (pcmk__str_eq((const char *)state->name, XML_CIB_TAG_STATE, pcmk__str_none)) { unpack_node_state(state, data_set); } } while (unpack_node_history(status, FALSE, data_set) == EAGAIN) { crm_trace("Another pass through node resource histories is needed"); } // Now catch any nodes we didn't see unpack_node_history(status, pcmk_is_set(data_set->flags, pe_flag_stonith_enabled), data_set); /* Now that we know where resources are, we can schedule stops of containers * with failed bundle connections */ if (data_set->stop_needed != NULL) { for (GList *item = data_set->stop_needed; item; item = item->next) { pe_resource_t *container = item->data; pe_node_t *node = pe__current_node(container); if (node) { stop_action(container, node, FALSE); } } g_list_free(data_set->stop_needed); data_set->stop_needed = NULL; } /* Now that we know status of all Pacemaker Remote connections and nodes, * we can stop connections for node shutdowns, and check the online status * of remote/guest nodes that didn't have any node history to unpack. */ for (GList *gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { pe_node_t *this_node = gIter->data; if (!pe__is_guest_or_remote_node(this_node)) { continue; } if (this_node->details->shutdown && (this_node->details->remote_rsc != NULL)) { pe__set_next_role(this_node->details->remote_rsc, RSC_ROLE_STOPPED, "remote shutdown"); } if (!this_node->details->unpacked) { determine_remote_online_status(data_set, this_node); } } return TRUE; } static gboolean determine_online_status_no_fencing(pe_working_set_t * data_set, xmlNode * node_state, pe_node_t * this_node) { gboolean online = FALSE; const char *join = crm_element_value(node_state, XML_NODE_JOIN_STATE); const char *is_peer = crm_element_value(node_state, XML_NODE_IS_PEER); const char *in_cluster = crm_element_value(node_state, XML_NODE_IN_CLUSTER); const char *exp_state = crm_element_value(node_state, XML_NODE_EXPECTED); if (!crm_is_true(in_cluster)) { crm_trace("Node is down: in_cluster=%s", crm_str(in_cluster)); } else if (pcmk__str_eq(is_peer, ONLINESTATUS, pcmk__str_casei)) { if (pcmk__str_eq(join, CRMD_JOINSTATE_MEMBER, pcmk__str_casei)) { online = TRUE; } else { crm_debug("Node is not ready to run resources: %s", join); } } else if (this_node->details->expected_up == FALSE) { crm_trace("Controller is down: in_cluster=%s", crm_str(in_cluster)); crm_trace("\tis_peer=%s, join=%s, expected=%s", crm_str(is_peer), crm_str(join), crm_str(exp_state)); } else { /* mark it unclean */ pe_fence_node(data_set, this_node, "peer is unexpectedly down", FALSE); crm_info("\tin_cluster=%s, is_peer=%s, join=%s, expected=%s", crm_str(in_cluster), crm_str(is_peer), crm_str(join), crm_str(exp_state)); } return online; } static gboolean determine_online_status_fencing(pe_working_set_t * data_set, xmlNode * node_state, pe_node_t * this_node) { gboolean online = FALSE; gboolean do_terminate = FALSE; bool crmd_online = FALSE; const char *join = crm_element_value(node_state, XML_NODE_JOIN_STATE); const char *is_peer = crm_element_value(node_state, XML_NODE_IS_PEER); const char *in_cluster = crm_element_value(node_state, XML_NODE_IN_CLUSTER); const char *exp_state = crm_element_value(node_state, XML_NODE_EXPECTED); const char *terminate = pe_node_attribute_raw(this_node, "terminate"); /* - XML_NODE_IN_CLUSTER ::= true|false - XML_NODE_IS_PEER ::= online|offline - XML_NODE_JOIN_STATE ::= member|down|pending|banned - XML_NODE_EXPECTED ::= member|down */ if (crm_is_true(terminate)) { do_terminate = TRUE; } else if (terminate != NULL && strlen(terminate) > 0) { /* could be a time() value */ char t = terminate[0]; if (t != '0' && isdigit(t)) { do_terminate = TRUE; } } crm_trace("%s: in_cluster=%s, is_peer=%s, join=%s, expected=%s, term=%d", this_node->details->uname, crm_str(in_cluster), crm_str(is_peer), crm_str(join), crm_str(exp_state), do_terminate); online = crm_is_true(in_cluster); crmd_online = pcmk__str_eq(is_peer, ONLINESTATUS, pcmk__str_casei); if (exp_state == NULL) { exp_state = CRMD_JOINSTATE_DOWN; } if (this_node->details->shutdown) { crm_debug("%s is shutting down", this_node->details->uname); /* Slightly different criteria since we can't shut down a dead peer */ online = crmd_online; } else if (in_cluster == NULL) { pe_fence_node(data_set, this_node, "peer has not been seen by the cluster", FALSE); } else if (pcmk__str_eq(join, CRMD_JOINSTATE_NACK, pcmk__str_casei)) { pe_fence_node(data_set, this_node, "peer failed Pacemaker membership criteria", FALSE); } else if (do_terminate == FALSE && pcmk__str_eq(exp_state, CRMD_JOINSTATE_DOWN, pcmk__str_casei)) { if (crm_is_true(in_cluster) || crmd_online) { crm_info("- Node %s is not ready to run resources", this_node->details->uname); this_node->details->standby = TRUE; this_node->details->pending = TRUE; } else { crm_trace("%s is down or still coming up", this_node->details->uname); } } else if (do_terminate && pcmk__str_eq(join, CRMD_JOINSTATE_DOWN, pcmk__str_casei) && crm_is_true(in_cluster) == FALSE && !crmd_online) { crm_info("Node %s was just shot", this_node->details->uname); online = FALSE; } else if (crm_is_true(in_cluster) == FALSE) { // Consider `priority-fencing-delay` for lost nodes pe_fence_node(data_set, this_node, "peer is no longer part of the cluster", TRUE); } else if (!crmd_online) { pe_fence_node(data_set, this_node, "peer process is no longer available", FALSE); /* Everything is running at this point, now check join state */ } else if (do_terminate) { pe_fence_node(data_set, this_node, "termination was requested", FALSE); } else if (pcmk__str_eq(join, CRMD_JOINSTATE_MEMBER, pcmk__str_casei)) { crm_info("Node %s is active", this_node->details->uname); } else if (pcmk__strcase_any_of(join, CRMD_JOINSTATE_PENDING, CRMD_JOINSTATE_DOWN, NULL)) { crm_info("Node %s is not ready to run resources", this_node->details->uname); this_node->details->standby = TRUE; this_node->details->pending = TRUE; } else { pe_fence_node(data_set, this_node, "peer was in an unknown state", FALSE); crm_warn("%s: in-cluster=%s, is-peer=%s, join=%s, expected=%s, term=%d, shutdown=%d", this_node->details->uname, crm_str(in_cluster), crm_str(is_peer), crm_str(join), crm_str(exp_state), do_terminate, this_node->details->shutdown); } return online; } static void determine_remote_online_status(pe_working_set_t * data_set, pe_node_t * this_node) { pe_resource_t *rsc = this_node->details->remote_rsc; pe_resource_t *container = NULL; pe_node_t *host = NULL; /* If there is a node state entry for a (former) Pacemaker Remote node * but no resource creating that node, the node's connection resource will * be NULL. Consider it an offline remote node in that case. */ if (rsc == NULL) { this_node->details->online = FALSE; goto remote_online_done; } container = rsc->container; if (container && pcmk__list_of_1(rsc->running_on)) { host = rsc->running_on->data; } /* If the resource is currently started, mark it online. */ if (rsc->role == RSC_ROLE_STARTED) { crm_trace("%s node %s presumed ONLINE because connection resource is started", (container? "Guest" : "Remote"), this_node->details->id); this_node->details->online = TRUE; } /* consider this node shutting down if transitioning start->stop */ if (rsc->role == RSC_ROLE_STARTED && rsc->next_role == RSC_ROLE_STOPPED) { crm_trace("%s node %s shutting down because connection resource is stopping", (container? "Guest" : "Remote"), this_node->details->id); this_node->details->shutdown = TRUE; } /* Now check all the failure conditions. */ if(container && pcmk_is_set(container->flags, pe_rsc_failed)) { crm_trace("Guest node %s UNCLEAN because guest resource failed", this_node->details->id); this_node->details->online = FALSE; this_node->details->remote_requires_reset = TRUE; } else if (pcmk_is_set(rsc->flags, pe_rsc_failed)) { crm_trace("%s node %s OFFLINE because connection resource failed", (container? "Guest" : "Remote"), this_node->details->id); this_node->details->online = FALSE; } else if (rsc->role == RSC_ROLE_STOPPED || (container && container->role == RSC_ROLE_STOPPED)) { crm_trace("%s node %s OFFLINE because its resource is stopped", (container? "Guest" : "Remote"), this_node->details->id); this_node->details->online = FALSE; this_node->details->remote_requires_reset = FALSE; } else if (host && (host->details->online == FALSE) && host->details->unclean) { crm_trace("Guest node %s UNCLEAN because host is unclean", this_node->details->id); this_node->details->online = FALSE; this_node->details->remote_requires_reset = TRUE; } remote_online_done: crm_trace("Remote node %s online=%s", this_node->details->id, this_node->details->online ? "TRUE" : "FALSE"); } static void determine_online_status(xmlNode * node_state, pe_node_t * this_node, pe_working_set_t * data_set) { gboolean online = FALSE; const char *exp_state = crm_element_value(node_state, XML_NODE_EXPECTED); CRM_CHECK(this_node != NULL, return); this_node->details->shutdown = FALSE; this_node->details->expected_up = FALSE; if (pe__shutdown_requested(this_node)) { this_node->details->shutdown = TRUE; } else if (pcmk__str_eq(exp_state, CRMD_JOINSTATE_MEMBER, pcmk__str_casei)) { this_node->details->expected_up = TRUE; } if (this_node->details->type == node_ping) { this_node->details->unclean = FALSE; online = FALSE; /* As far as resource management is concerned, * the node is safely offline. * Anyone caught abusing this logic will be shot */ } else if (!pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { online = determine_online_status_no_fencing(data_set, node_state, this_node); } else { online = determine_online_status_fencing(data_set, node_state, this_node); } if (online) { this_node->details->online = TRUE; } else { /* remove node from contention */ this_node->fixed = TRUE; this_node->weight = -INFINITY; } if (online && this_node->details->shutdown) { /* don't run resources here */ this_node->fixed = TRUE; this_node->weight = -INFINITY; } if (this_node->details->type == node_ping) { crm_info("Node %s is not a Pacemaker node", this_node->details->uname); } else if (this_node->details->unclean) { pe_proc_warn("Node %s is unclean", this_node->details->uname); } else if (this_node->details->online) { crm_info("Node %s is %s", this_node->details->uname, this_node->details->shutdown ? "shutting down" : this_node->details->pending ? "pending" : this_node->details->standby ? "standby" : this_node->details->maintenance ? "maintenance" : "online"); } else { crm_trace("Node %s is offline", this_node->details->uname); } } /*! * \internal * \brief Find the end of a resource's name, excluding any clone suffix * * \param[in] id Resource ID to check * * \return Pointer to last character of resource's base name */ const char * pe_base_name_end(const char *id) { if (!pcmk__str_empty(id)) { const char *end = id + strlen(id) - 1; for (const char *s = end; s > id; --s) { switch (*s) { case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': break; case ':': return (s == end)? s : (s - 1); default: return end; } } return end; } return NULL; } /*! * \internal * \brief Get a resource name excluding any clone suffix * * \param[in] last_rsc_id Resource ID to check * * \return Pointer to newly allocated string with resource's base name * \note It is the caller's responsibility to free() the result. * This asserts on error, so callers can assume result is not NULL. */ char * clone_strip(const char *last_rsc_id) { const char *end = pe_base_name_end(last_rsc_id); char *basename = NULL; CRM_ASSERT(end); basename = strndup(last_rsc_id, end - last_rsc_id + 1); CRM_ASSERT(basename); return basename; } /*! * \internal * \brief Get the name of the first instance of a cloned resource * * \param[in] last_rsc_id Resource ID to check * * \return Pointer to newly allocated string with resource's base name plus :0 * \note It is the caller's responsibility to free() the result. * This asserts on error, so callers can assume result is not NULL. */ char * clone_zero(const char *last_rsc_id) { const char *end = pe_base_name_end(last_rsc_id); size_t base_name_len = end - last_rsc_id + 1; char *zero = NULL; CRM_ASSERT(end); zero = calloc(base_name_len + 3, sizeof(char)); CRM_ASSERT(zero); memcpy(zero, last_rsc_id, base_name_len); zero[base_name_len] = ':'; zero[base_name_len + 1] = '0'; return zero; } static pe_resource_t * create_fake_resource(const char *rsc_id, xmlNode * rsc_entry, pe_working_set_t * data_set) { pe_resource_t *rsc = NULL; xmlNode *xml_rsc = create_xml_node(NULL, XML_CIB_TAG_RESOURCE); copy_in_properties(xml_rsc, rsc_entry); crm_xml_add(xml_rsc, XML_ATTR_ID, rsc_id); crm_log_xml_debug(xml_rsc, "Orphan resource"); if (!common_unpack(xml_rsc, &rsc, NULL, data_set)) { return NULL; } if (xml_contains_remote_node(xml_rsc)) { pe_node_t *node; crm_debug("Detected orphaned remote node %s", rsc_id); node = pe_find_node(data_set->nodes, rsc_id); if (node == NULL) { node = pe_create_node(rsc_id, rsc_id, "remote", NULL, data_set); } link_rsc2remotenode(data_set, rsc); if (node) { crm_trace("Setting node %s as shutting down due to orphaned connection resource", rsc_id); node->details->shutdown = TRUE; } } if (crm_element_value(rsc_entry, XML_RSC_ATTR_CONTAINER)) { /* This orphaned rsc needs to be mapped to a container. */ crm_trace("Detected orphaned container filler %s", rsc_id); pe__set_resource_flags(rsc, pe_rsc_orphan_container_filler); } pe__set_resource_flags(rsc, pe_rsc_orphan); data_set->resources = g_list_append(data_set->resources, rsc); return rsc; } /*! * \internal * \brief Create orphan instance for anonymous clone resource history */ static pe_resource_t * create_anonymous_orphan(pe_resource_t *parent, const char *rsc_id, pe_node_t *node, pe_working_set_t *data_set) { pe_resource_t *top = pe__create_clone_child(parent, data_set); // find_rsc() because we might be a cloned group pe_resource_t *orphan = top->fns->find_rsc(top, rsc_id, NULL, pe_find_clone); pe_rsc_debug(parent, "Created orphan %s for %s: %s on %s", top->id, parent->id, rsc_id, node->details->uname); return orphan; } /*! * \internal * \brief Check a node for an instance of an anonymous clone * * Return a child instance of the specified anonymous clone, in order of * preference: (1) the instance running on the specified node, if any; * (2) an inactive instance (i.e. within the total of clone-max instances); * (3) a newly created orphan (i.e. clone-max instances are already active). * * \param[in] data_set Cluster information * \param[in] node Node on which to check for instance * \param[in] parent Clone to check * \param[in] rsc_id Name of cloned resource in history (without instance) */ static pe_resource_t * find_anonymous_clone(pe_working_set_t * data_set, pe_node_t * node, pe_resource_t * parent, const char *rsc_id) { GList *rIter = NULL; pe_resource_t *rsc = NULL; pe_resource_t *inactive_instance = NULL; gboolean skip_inactive = FALSE; CRM_ASSERT(parent != NULL); CRM_ASSERT(pe_rsc_is_clone(parent)); CRM_ASSERT(!pcmk_is_set(parent->flags, pe_rsc_unique)); // Check for active (or partially active, for cloned groups) instance pe_rsc_trace(parent, "Looking for %s on %s in %s", rsc_id, node->details->uname, parent->id); for (rIter = parent->children; rsc == NULL && rIter; rIter = rIter->next) { GList *locations = NULL; pe_resource_t *child = rIter->data; /* Check whether this instance is already known to be active or pending * anywhere, at this stage of unpacking. Because this function is called * for a resource before the resource's individual operation history * entries are unpacked, locations will generally not contain the * desired node. * * However, there are three exceptions: * (1) when child is a cloned group and we have already unpacked the * history of another member of the group on the same node; * (2) when we've already unpacked the history of another numbered * instance on the same node (which can happen if globally-unique * was flipped from true to false); and * (3) when we re-run calculations on the same data set as part of a * simulation. */ child->fns->location(child, &locations, 2); if (locations) { /* We should never associate the same numbered anonymous clone * instance with multiple nodes, and clone instances can't migrate, * so there must be only one location, regardless of history. */ CRM_LOG_ASSERT(locations->next == NULL); if (((pe_node_t *)locations->data)->details == node->details) { /* This child instance is active on the requested node, so check * for a corresponding configured resource. We use find_rsc() * instead of child because child may be a cloned group, and we * need the particular member corresponding to rsc_id. * * If the history entry is orphaned, rsc will be NULL. */ rsc = parent->fns->find_rsc(child, rsc_id, NULL, pe_find_clone); if (rsc) { /* If there are multiple instance history entries for an * anonymous clone in a single node's history (which can * happen if globally-unique is switched from true to * false), we want to consider the instances beyond the * first as orphans, even if there are inactive instance * numbers available. */ if (rsc->running_on) { crm_notice("Active (now-)anonymous clone %s has " "multiple (orphan) instance histories on %s", parent->id, node->details->uname); skip_inactive = TRUE; rsc = NULL; } else { pe_rsc_trace(parent, "Resource %s, active", rsc->id); } } } g_list_free(locations); } else { pe_rsc_trace(parent, "Resource %s, skip inactive", child->id); if (!skip_inactive && !inactive_instance && !pcmk_is_set(child->flags, pe_rsc_block)) { // Remember one inactive instance in case we don't find active inactive_instance = parent->fns->find_rsc(child, rsc_id, NULL, pe_find_clone); /* ... but don't use it if it was already associated with a * pending action on another node */ if (inactive_instance && inactive_instance->pending_node && (inactive_instance->pending_node->details != node->details)) { inactive_instance = NULL; } } } } if ((rsc == NULL) && !skip_inactive && (inactive_instance != NULL)) { pe_rsc_trace(parent, "Resource %s, empty slot", inactive_instance->id); rsc = inactive_instance; } /* If the resource has "requires" set to "quorum" or "nothing", and we don't * have a clone instance for every node, we don't want to consume a valid * instance number for unclean nodes. Such instances may appear to be active * according to the history, but should be considered inactive, so we can * start an instance elsewhere. Treat such instances as orphans. * * An exception is instances running on guest nodes -- since guest node * "fencing" is actually just a resource stop, requires shouldn't apply. * * @TODO Ideally, we'd use an inactive instance number if it is not needed * for any clean instances. However, we don't know that at this point. */ if ((rsc != NULL) && !pcmk_is_set(rsc->flags, pe_rsc_needs_fencing) && (!node->details->online || node->details->unclean) && !pe__is_guest_node(node) && !pe__is_universal_clone(parent, data_set)) { rsc = NULL; } if (rsc == NULL) { rsc = create_anonymous_orphan(parent, rsc_id, node, data_set); pe_rsc_trace(parent, "Resource %s, orphan", rsc->id); } return rsc; } static pe_resource_t * unpack_find_resource(pe_working_set_t * data_set, pe_node_t * node, const char *rsc_id, xmlNode * rsc_entry) { pe_resource_t *rsc = NULL; pe_resource_t *parent = NULL; crm_trace("looking for %s", rsc_id); rsc = pe_find_resource(data_set->resources, rsc_id); if (rsc == NULL) { /* If we didn't find the resource by its name in the operation history, * check it again as a clone instance. Even when clone-max=0, we create * a single :0 orphan to match against here. */ char *clone0_id = clone_zero(rsc_id); pe_resource_t *clone0 = pe_find_resource(data_set->resources, clone0_id); if (clone0 && !pcmk_is_set(clone0->flags, pe_rsc_unique)) { rsc = clone0; parent = uber_parent(clone0); crm_trace("%s found as %s (%s)", rsc_id, clone0_id, parent->id); } else { crm_trace("%s is not known as %s either (orphan)", rsc_id, clone0_id); } free(clone0_id); } else if (rsc->variant > pe_native) { crm_trace("Resource history for %s is orphaned because it is no longer primitive", rsc_id); return NULL; } else { parent = uber_parent(rsc); } if (pe_rsc_is_anon_clone(parent)) { if (pe_rsc_is_bundled(parent)) { rsc = pe__find_bundle_replica(parent->parent, node); } else { char *base = clone_strip(rsc_id); rsc = find_anonymous_clone(data_set, node, parent, base); free(base); CRM_ASSERT(rsc != NULL); } } if (rsc && !pcmk__str_eq(rsc_id, rsc->id, pcmk__str_casei) && !pcmk__str_eq(rsc_id, rsc->clone_name, pcmk__str_casei)) { pcmk__str_update(&rsc->clone_name, rsc_id); pe_rsc_debug(rsc, "Internally renamed %s on %s to %s%s", rsc_id, node->details->uname, rsc->id, (pcmk_is_set(rsc->flags, pe_rsc_orphan)? " (ORPHAN)" : "")); } return rsc; } static pe_resource_t * process_orphan_resource(xmlNode * rsc_entry, pe_node_t * node, pe_working_set_t * data_set) { pe_resource_t *rsc = NULL; const char *rsc_id = crm_element_value(rsc_entry, XML_ATTR_ID); crm_debug("Detected orphan resource %s on %s", rsc_id, node->details->uname); rsc = create_fake_resource(rsc_id, rsc_entry, data_set); if (rsc == NULL) { return NULL; } if (!pcmk_is_set(data_set->flags, pe_flag_stop_rsc_orphans)) { pe__clear_resource_flags(rsc, pe_rsc_managed); } else { CRM_CHECK(rsc != NULL, return NULL); pe_rsc_trace(rsc, "Added orphan %s", rsc->id); resource_location(rsc, NULL, -INFINITY, "__orphan_do_not_run__", data_set); } return rsc; } static void process_rsc_state(pe_resource_t * rsc, pe_node_t * node, enum action_fail_response on_fail, xmlNode * migrate_op, pe_working_set_t * data_set) { pe_node_t *tmpnode = NULL; char *reason = NULL; enum action_fail_response save_on_fail = action_fail_ignore; CRM_ASSERT(rsc); pe_rsc_trace(rsc, "Resource %s is %s on %s: on_fail=%s", rsc->id, role2text(rsc->role), node->details->uname, fail2text(on_fail)); /* process current state */ if (rsc->role != RSC_ROLE_UNKNOWN) { pe_resource_t *iter = rsc; while (iter) { if (g_hash_table_lookup(iter->known_on, node->details->id) == NULL) { pe_node_t *n = pe__copy_node(node); pe_rsc_trace(rsc, "%s%s%s known on %s", rsc->id, ((rsc->clone_name == NULL)? "" : " also known as "), ((rsc->clone_name == NULL)? "" : rsc->clone_name), n->details->uname); g_hash_table_insert(iter->known_on, (gpointer) n->details->id, n); } if (pcmk_is_set(iter->flags, pe_rsc_unique)) { break; } iter = iter->parent; } } /* If a managed resource is believed to be running, but node is down ... */ if (rsc->role > RSC_ROLE_STOPPED && node->details->online == FALSE && node->details->maintenance == FALSE && pcmk_is_set(rsc->flags, pe_rsc_managed)) { gboolean should_fence = FALSE; /* If this is a guest node, fence it (regardless of whether fencing is * enabled, because guest node fencing is done by recovery of the * container resource rather than by the fencer). Mark the resource * we're processing as failed. When the guest comes back up, its * operation history in the CIB will be cleared, freeing the affected * resource to run again once we are sure we know its state. */ if (pe__is_guest_node(node)) { pe__set_resource_flags(rsc, pe_rsc_failed|pe_rsc_stop); should_fence = TRUE; } else if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { if (pe__is_remote_node(node) && node->details->remote_rsc && !pcmk_is_set(node->details->remote_rsc->flags, pe_rsc_failed)) { /* Setting unseen means that fencing of the remote node will * occur only if the connection resource is not going to start * somewhere. This allows connection resources on a failed * cluster node to move to another node without requiring the * remote nodes to be fenced as well. */ node->details->unseen = TRUE; reason = crm_strdup_printf("%s is active there (fencing will be" " revoked if remote connection can " "be re-established elsewhere)", rsc->id); } should_fence = TRUE; } if (should_fence) { if (reason == NULL) { reason = crm_strdup_printf("%s is thought to be active there", rsc->id); } pe_fence_node(data_set, node, reason, FALSE); } free(reason); } /* In order to calculate priority_fencing_delay correctly, save the failure information and pass it to native_add_running(). */ save_on_fail = on_fail; if (node->details->unclean) { /* No extra processing needed * Also allows resources to be started again after a node is shot */ on_fail = action_fail_ignore; } switch (on_fail) { case action_fail_ignore: /* nothing to do */ break; case action_fail_demote: pe__set_resource_flags(rsc, pe_rsc_failed); demote_action(rsc, node, FALSE); break; case action_fail_fence: /* treat it as if it is still running * but also mark the node as unclean */ reason = crm_strdup_printf("%s failed there", rsc->id); pe_fence_node(data_set, node, reason, FALSE); free(reason); break; case action_fail_standby: node->details->standby = TRUE; node->details->standby_onfail = TRUE; break; case action_fail_block: /* is_managed == FALSE will prevent any * actions being sent for the resource */ pe__clear_resource_flags(rsc, pe_rsc_managed); pe__set_resource_flags(rsc, pe_rsc_block); break; case action_fail_migrate: /* make sure it comes up somewhere else * or not at all */ resource_location(rsc, node, -INFINITY, "__action_migration_auto__", data_set); break; case action_fail_stop: pe__set_next_role(rsc, RSC_ROLE_STOPPED, "on-fail=stop"); break; case action_fail_recover: if (rsc->role != RSC_ROLE_STOPPED && rsc->role != RSC_ROLE_UNKNOWN) { pe__set_resource_flags(rsc, pe_rsc_failed|pe_rsc_stop); stop_action(rsc, node, FALSE); } break; case action_fail_restart_container: pe__set_resource_flags(rsc, pe_rsc_failed|pe_rsc_stop); if (rsc->container && pe_rsc_is_bundled(rsc)) { /* A bundle's remote connection can run on a different node than * the bundle's container. We don't necessarily know where the * container is running yet, so remember it and add a stop * action for it later. */ data_set->stop_needed = g_list_prepend(data_set->stop_needed, rsc->container); } else if (rsc->container) { stop_action(rsc->container, node, FALSE); } else if (rsc->role != RSC_ROLE_STOPPED && rsc->role != RSC_ROLE_UNKNOWN) { stop_action(rsc, node, FALSE); } break; case action_fail_reset_remote: pe__set_resource_flags(rsc, pe_rsc_failed|pe_rsc_stop); if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { tmpnode = NULL; if (rsc->is_remote_node) { tmpnode = pe_find_node(data_set->nodes, rsc->id); } if (tmpnode && pe__is_remote_node(tmpnode) && tmpnode->details->remote_was_fenced == 0) { /* The remote connection resource failed in a way that * should result in fencing the remote node. */ pe_fence_node(data_set, tmpnode, "remote connection is unrecoverable", FALSE); } } /* require the stop action regardless if fencing is occurring or not. */ if (rsc->role > RSC_ROLE_STOPPED) { stop_action(rsc, node, FALSE); } /* if reconnect delay is in use, prevent the connection from exiting the * "STOPPED" role until the failure is cleared by the delay timeout. */ if (rsc->remote_reconnect_ms) { pe__set_next_role(rsc, RSC_ROLE_STOPPED, "remote reset"); } break; } /* ensure a remote-node connection failure forces an unclean remote-node * to be fenced. By setting unseen = FALSE, the remote-node failure will * result in a fencing operation regardless if we're going to attempt to * reconnect to the remote-node in this transition or not. */ if (pcmk_is_set(rsc->flags, pe_rsc_failed) && rsc->is_remote_node) { tmpnode = pe_find_node(data_set->nodes, rsc->id); if (tmpnode && tmpnode->details->unclean) { tmpnode->details->unseen = FALSE; } } if (rsc->role != RSC_ROLE_STOPPED && rsc->role != RSC_ROLE_UNKNOWN) { if (pcmk_is_set(rsc->flags, pe_rsc_orphan)) { if (pcmk_is_set(rsc->flags, pe_rsc_managed)) { pcmk__config_warn("Detected active orphan %s running on %s", rsc->id, node->details->uname); } else { pcmk__config_warn("Resource '%s' must be stopped manually on " "%s because cluster is configured not to " "stop active orphans", rsc->id, node->details->uname); } } native_add_running(rsc, node, data_set, (save_on_fail != action_fail_ignore)); switch (on_fail) { case action_fail_ignore: break; case action_fail_demote: case action_fail_block: pe__set_resource_flags(rsc, pe_rsc_failed); break; default: pe__set_resource_flags(rsc, pe_rsc_failed|pe_rsc_stop); break; } } else if (rsc->clone_name && strchr(rsc->clone_name, ':') != NULL) { /* Only do this for older status sections that included instance numbers * Otherwise stopped instances will appear as orphans */ pe_rsc_trace(rsc, "Resetting clone_name %s for %s (stopped)", rsc->clone_name, rsc->id); free(rsc->clone_name); rsc->clone_name = NULL; } else { GList *possible_matches = pe__resource_actions(rsc, node, RSC_STOP, FALSE); GList *gIter = possible_matches; for (; gIter != NULL; gIter = gIter->next) { pe_action_t *stop = (pe_action_t *) gIter->data; pe__set_action_flags(stop, pe_action_optional); } g_list_free(possible_matches); } } /* create active recurring operations as optional */ static void process_recurring(pe_node_t * node, pe_resource_t * rsc, int start_index, int stop_index, GList *sorted_op_list, pe_working_set_t * data_set) { int counter = -1; const char *task = NULL; const char *status = NULL; GList *gIter = sorted_op_list; CRM_ASSERT(rsc); pe_rsc_trace(rsc, "%s: Start index %d, stop index = %d", rsc->id, start_index, stop_index); for (; gIter != NULL; gIter = gIter->next) { xmlNode *rsc_op = (xmlNode *) gIter->data; guint interval_ms = 0; char *key = NULL; const char *id = ID(rsc_op); counter++; if (node->details->online == FALSE) { pe_rsc_trace(rsc, "Skipping %s/%s: node is offline", rsc->id, node->details->uname); break; /* Need to check if there's a monitor for role="Stopped" */ } else if (start_index < stop_index && counter <= stop_index) { pe_rsc_trace(rsc, "Skipping %s/%s: resource is not active", id, node->details->uname); continue; } else if (counter < start_index) { pe_rsc_trace(rsc, "Skipping %s/%s: old %d", id, node->details->uname, counter); continue; } crm_element_value_ms(rsc_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); if (interval_ms == 0) { pe_rsc_trace(rsc, "Skipping %s/%s: non-recurring", id, node->details->uname); continue; } status = crm_element_value(rsc_op, XML_LRM_ATTR_OPSTATUS); if (pcmk__str_eq(status, "-1", pcmk__str_casei)) { pe_rsc_trace(rsc, "Skipping %s/%s: status", id, node->details->uname); continue; } task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); /* create the action */ key = pcmk__op_key(rsc->id, task, interval_ms); pe_rsc_trace(rsc, "Creating %s/%s", key, node->details->uname); custom_action(rsc, key, task, node, TRUE, TRUE, data_set); } } void calculate_active_ops(GList *sorted_op_list, int *start_index, int *stop_index) { int counter = -1; int implied_monitor_start = -1; int implied_clone_start = -1; const char *task = NULL; const char *status = NULL; GList *gIter = sorted_op_list; *stop_index = -1; *start_index = -1; for (; gIter != NULL; gIter = gIter->next) { xmlNode *rsc_op = (xmlNode *) gIter->data; counter++; task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); status = crm_element_value(rsc_op, XML_LRM_ATTR_OPSTATUS); if (pcmk__str_eq(task, CRMD_ACTION_STOP, pcmk__str_casei) && pcmk__str_eq(status, "0", pcmk__str_casei)) { *stop_index = counter; } else if (pcmk__strcase_any_of(task, CRMD_ACTION_START, CRMD_ACTION_MIGRATED, NULL)) { *start_index = counter; } else if ((implied_monitor_start <= *stop_index) && pcmk__str_eq(task, CRMD_ACTION_STATUS, pcmk__str_casei)) { const char *rc = crm_element_value(rsc_op, XML_LRM_ATTR_RC); if (pcmk__strcase_any_of(rc, "0", "8", NULL)) { implied_monitor_start = counter; } } else if (pcmk__strcase_any_of(task, CRMD_ACTION_PROMOTE, CRMD_ACTION_DEMOTE, NULL)) { implied_clone_start = counter; } } if (*start_index == -1) { if (implied_clone_start != -1) { *start_index = implied_clone_start; } else if (implied_monitor_start != -1) { *start_index = implied_monitor_start; } } } // If resource history entry has shutdown lock, remember lock node and time static void unpack_shutdown_lock(xmlNode *rsc_entry, pe_resource_t *rsc, pe_node_t *node, pe_working_set_t *data_set) { time_t lock_time = 0; // When lock started (i.e. node shutdown time) if ((crm_element_value_epoch(rsc_entry, XML_CONFIG_ATTR_SHUTDOWN_LOCK, &lock_time) == pcmk_ok) && (lock_time != 0)) { if ((data_set->shutdown_lock > 0) && (get_effective_time(data_set) > (lock_time + data_set->shutdown_lock))) { pe_rsc_info(rsc, "Shutdown lock for %s on %s expired", rsc->id, node->details->uname); pe__clear_resource_history(rsc, node, data_set); } else { rsc->lock_node = node; rsc->lock_time = lock_time; } } } /*! * \internal * \brief Unpack one lrm_resource entry from a node's CIB status * * \param[in] node Node whose status is being unpacked * \param[in] rsc_entry lrm_resource XML being unpacked * \param[in] data_set Cluster working set * * \return Resource corresponding to the entry, or NULL if no operation history */ static pe_resource_t * unpack_lrm_resource(pe_node_t *node, xmlNode *lrm_resource, pe_working_set_t *data_set) { GList *gIter = NULL; int stop_index = -1; int start_index = -1; enum rsc_role_e req_role = RSC_ROLE_UNKNOWN; const char *task = NULL; const char *rsc_id = ID(lrm_resource); pe_resource_t *rsc = NULL; GList *op_list = NULL; GList *sorted_op_list = NULL; xmlNode *migrate_op = NULL; xmlNode *rsc_op = NULL; xmlNode *last_failure = NULL; enum action_fail_response on_fail = action_fail_ignore; enum rsc_role_e saved_role = RSC_ROLE_UNKNOWN; if (rsc_id == NULL) { crm_warn("Ignoring malformed " XML_LRM_TAG_RESOURCE " entry without id"); return NULL; } crm_trace("Unpacking " XML_LRM_TAG_RESOURCE " for %s on %s", rsc_id, node->details->uname); // Build a list of individual lrm_rsc_op entries, so we can sort them for (rsc_op = first_named_child(lrm_resource, XML_LRM_TAG_RSC_OP); rsc_op != NULL; rsc_op = crm_next_same_xml(rsc_op)) { op_list = g_list_prepend(op_list, rsc_op); } if (!pcmk_is_set(data_set->flags, pe_flag_shutdown_lock)) { if (op_list == NULL) { // If there are no operations, there is nothing to do return NULL; } } /* find the resource */ rsc = unpack_find_resource(data_set, node, rsc_id, lrm_resource); if (rsc == NULL) { if (op_list == NULL) { // If there are no operations, there is nothing to do return NULL; } else { rsc = process_orphan_resource(lrm_resource, node, data_set); } } CRM_ASSERT(rsc != NULL); // Check whether the resource is "shutdown-locked" to this node if (pcmk_is_set(data_set->flags, pe_flag_shutdown_lock)) { unpack_shutdown_lock(lrm_resource, rsc, node, data_set); } /* process operations */ saved_role = rsc->role; rsc->role = RSC_ROLE_UNKNOWN; sorted_op_list = g_list_sort(op_list, sort_op_by_callid); for (gIter = sorted_op_list; gIter != NULL; gIter = gIter->next) { xmlNode *rsc_op = (xmlNode *) gIter->data; task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); if (pcmk__str_eq(task, CRMD_ACTION_MIGRATED, pcmk__str_casei)) { migrate_op = rsc_op; } unpack_rsc_op(rsc, node, rsc_op, &last_failure, &on_fail, data_set); } /* create active recurring operations as optional */ calculate_active_ops(sorted_op_list, &start_index, &stop_index); process_recurring(node, rsc, start_index, stop_index, sorted_op_list, data_set); /* no need to free the contents */ g_list_free(sorted_op_list); process_rsc_state(rsc, node, on_fail, migrate_op, data_set); if (get_target_role(rsc, &req_role)) { if (rsc->next_role == RSC_ROLE_UNKNOWN || req_role < rsc->next_role) { pe__set_next_role(rsc, req_role, XML_RSC_ATTR_TARGET_ROLE); } else if (req_role > rsc->next_role) { pe_rsc_info(rsc, "%s: Not overwriting calculated next role %s" " with requested next role %s", rsc->id, role2text(rsc->next_role), role2text(req_role)); } } if (saved_role > rsc->role) { rsc->role = saved_role; } return rsc; } static void handle_orphaned_container_fillers(xmlNode * lrm_rsc_list, pe_working_set_t * data_set) { xmlNode *rsc_entry = NULL; for (rsc_entry = pcmk__xe_first_child(lrm_rsc_list); rsc_entry != NULL; rsc_entry = pcmk__xe_next(rsc_entry)) { pe_resource_t *rsc; pe_resource_t *container; const char *rsc_id; const char *container_id; if (!pcmk__str_eq((const char *)rsc_entry->name, XML_LRM_TAG_RESOURCE, pcmk__str_casei)) { continue; } container_id = crm_element_value(rsc_entry, XML_RSC_ATTR_CONTAINER); rsc_id = crm_element_value(rsc_entry, XML_ATTR_ID); if (container_id == NULL || rsc_id == NULL) { continue; } container = pe_find_resource(data_set->resources, container_id); if (container == NULL) { continue; } rsc = pe_find_resource(data_set->resources, rsc_id); if (rsc == NULL || !pcmk_is_set(rsc->flags, pe_rsc_orphan_container_filler) || rsc->container != NULL) { continue; } pe_rsc_trace(rsc, "Mapped container of orphaned resource %s to %s", rsc->id, container_id); rsc->container = container; container->fillers = g_list_append(container->fillers, rsc); } } /*! * \internal * \brief Unpack one node's lrm status section * * \param[in] node Node whose status is being unpacked * \param[in] xml CIB node state XML * \param[in] data_set Cluster working set */ static void unpack_node_lrm(pe_node_t *node, xmlNode *xml, pe_working_set_t *data_set) { bool found_orphaned_container_filler = false; // Drill down to lrm_resources section xml = find_xml_node(xml, XML_CIB_TAG_LRM, FALSE); if (xml == NULL) { return; } xml = find_xml_node(xml, XML_LRM_TAG_RESOURCES, FALSE); if (xml == NULL) { return; } // Unpack each lrm_resource entry for (xmlNode *rsc_entry = first_named_child(xml, XML_LRM_TAG_RESOURCE); rsc_entry != NULL; rsc_entry = crm_next_same_xml(rsc_entry)) { pe_resource_t *rsc = unpack_lrm_resource(node, rsc_entry, data_set); if ((rsc != NULL) && pcmk_is_set(rsc->flags, pe_rsc_orphan_container_filler)) { found_orphaned_container_filler = true; } } /* Now that all resource state has been unpacked for this node, map any * orphaned container fillers to their container resource. */ if (found_orphaned_container_filler) { handle_orphaned_container_fillers(xml, data_set); } } static void set_active(pe_resource_t * rsc) { pe_resource_t *top = uber_parent(rsc); if (top && pcmk_is_set(top->flags, pe_rsc_promotable)) { rsc->role = RSC_ROLE_UNPROMOTED; } else { rsc->role = RSC_ROLE_STARTED; } } static void set_node_score(gpointer key, gpointer value, gpointer user_data) { pe_node_t *node = value; int *score = user_data; node->weight = *score; } #define STATUS_PATH_MAX 1024 static xmlNode * find_lrm_op(const char *resource, const char *op, const char *node, const char *source, - bool success_only, pe_working_set_t *data_set) + int target_rc, pe_working_set_t *data_set) { int offset = 0; char xpath[STATUS_PATH_MAX]; xmlNode *xml = NULL; offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "//node_state[@uname='%s']", node); offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "//" XML_LRM_TAG_RESOURCE "[@id='%s']", resource); /* Need to check against transition_magic too? */ if (source && pcmk__str_eq(op, CRMD_ACTION_MIGRATE, pcmk__str_casei)) { offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "/" XML_LRM_TAG_RSC_OP "[@operation='%s' and @migrate_target='%s']", op, source); } else if (source && pcmk__str_eq(op, CRMD_ACTION_MIGRATED, pcmk__str_casei)) { offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "/" XML_LRM_TAG_RSC_OP "[@operation='%s' and @migrate_source='%s']", op, source); } else { offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "/" XML_LRM_TAG_RSC_OP "[@operation='%s']", op); } CRM_LOG_ASSERT(offset > 0); xml = get_xpath_object(xpath, data_set->input, LOG_DEBUG); - if (xml && success_only) { + if (xml && target_rc >= 0) { int rc = PCMK_OCF_UNKNOWN_ERROR; int status = PCMK_EXEC_ERROR; crm_element_value_int(xml, XML_LRM_ATTR_RC, &rc); crm_element_value_int(xml, XML_LRM_ATTR_OPSTATUS, &status); - if ((rc != PCMK_OCF_OK) || (status != PCMK_EXEC_DONE)) { + if ((rc != target_rc) || (status != PCMK_EXEC_DONE)) { return NULL; } } return xml; } +static xmlNode * +find_lrm_resource(const char *rsc_id, const char *node_name, + pe_working_set_t *data_set) +{ + int offset = 0; + char xpath[STATUS_PATH_MAX]; + xmlNode *xml = NULL; + + offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, + "//node_state[@uname='%s']", node_name); + offset += + snprintf(xpath + offset, STATUS_PATH_MAX - offset, + "//" XML_LRM_TAG_RESOURCE "[@id='%s']", rsc_id); + + CRM_LOG_ASSERT(offset > 0); + xml = get_xpath_object(xpath, data_set->input, LOG_DEBUG); + + return xml; +} + +static bool +unknown_on_node(const char *rsc_id, const char *node_name, + pe_working_set_t *data_set) +{ + xmlNode *lrm_resource = NULL; + + lrm_resource = find_lrm_resource(rsc_id, node_name, data_set); + + /* If the resource has no lrm_rsc_op history on the node, that means its + * state is unknown there. + */ + return (lrm_resource == NULL + || first_named_child(lrm_resource, XML_LRM_TAG_RSC_OP) == NULL); +} + +/*! + * \brief Check whether a probe/monitor indicating the resource was not running + * on a node happened after some event + * + * \param[in] rsc_id Resource being checked + * \param[in] node_name Node being checked + * \param[in] xml_op Event that monitor is being compared to + * \param[in] data_set Cluster working set + * + * \return true if such a monitor happened after event, false otherwise + */ +static bool +monitor_not_running_after(const char *rsc_id, const char *node_name, + xmlNode *xml_op, pe_working_set_t *data_set) +{ + /* Any probe/monitor operation on the node indicating it was not running + * there + */ + xmlNode *monitor = find_lrm_op(rsc_id, CRMD_ACTION_STATUS, node_name, + NULL, PCMK_OCF_NOT_RUNNING, data_set); + + return (monitor && pe__is_newer_op(monitor, xml_op) > 0); +} + static int pe__call_id(xmlNode *op_xml) { int id = 0; if (op_xml) { crm_element_value_int(op_xml, XML_LRM_ATTR_CALLID, &id); } return id; } /*! * \brief Check whether a stop happened on the same node after some event * * \param[in] rsc Resource being checked * \param[in] node Node being checked * \param[in] xml_op Event that stop is being compared to * \param[in] data_set Cluster working set * * \return TRUE if stop happened after event, FALSE otherwise * * \note This is really unnecessary, but kept as a safety mechanism. We * currently don't save more than one successful event in history, so this * only matters when processing really old CIB files that we don't * technically support anymore, or as preparation for logging an extended * history in the future. */ static bool stop_happened_after(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op, pe_working_set_t *data_set) { xmlNode *stop_op = find_lrm_op(rsc->id, CRMD_ACTION_STOP, - node->details->uname, NULL, TRUE, data_set); + node->details->uname, NULL, PCMK_OCF_OK, data_set); return (stop_op && (pe__call_id(stop_op) > pe__call_id(xml_op))); } static void unpack_migrate_to_success(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op, pe_working_set_t *data_set) { /* A successful migration sequence is: * migrate_to on source node * migrate_from on target node * stop on source node * * If a migrate_to is followed by a stop, the entire migration (successful * or failed) is complete, and we don't care what happened on the target. * * If no migrate_from has happened, the migration is considered to be * "partial". If the migrate_from failed, make sure the resource gets * stopped on both source and target (if up). * * If the migrate_to and migrate_from both succeeded (which also implies the * resource is no longer running on the source), but there is no stop, the * migration is considered to be "dangling". Schedule a stop on the source * in this case. */ int from_rc = 0; int from_status = 0; pe_node_t *target_node = NULL; pe_node_t *source_node = NULL; xmlNode *migrate_from = NULL; const char *source = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_SOURCE); const char *target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET); // Sanity check CRM_CHECK(source && target && !strcmp(source, node->details->uname), return); if (stop_happened_after(rsc, node, xml_op, data_set)) { return; } // Clones are not allowed to migrate, so role can't be promoted rsc->role = RSC_ROLE_STARTED; target_node = pe_find_node(data_set->nodes, target); source_node = pe_find_node(data_set->nodes, source); // Check whether there was a migrate_from action on the target migrate_from = find_lrm_op(rsc->id, CRMD_ACTION_MIGRATED, target, - source, FALSE, data_set); + source, -1, data_set); if (migrate_from) { crm_element_value_int(migrate_from, XML_LRM_ATTR_RC, &from_rc); crm_element_value_int(migrate_from, XML_LRM_ATTR_OPSTATUS, &from_status); pe_rsc_trace(rsc, "%s op on %s exited with status=%d, rc=%d", ID(migrate_from), target, from_status, from_rc); } if (migrate_from && from_rc == PCMK_OCF_OK && (from_status == PCMK_EXEC_DONE)) { /* The migrate_to and migrate_from both succeeded, so mark the migration * as "dangling". This will be used to schedule a stop action on the * source without affecting the target. */ pe_rsc_trace(rsc, "Detected dangling migration op: %s on %s", ID(xml_op), source); rsc->role = RSC_ROLE_STOPPED; rsc->dangling_migrations = g_list_prepend(rsc->dangling_migrations, node); } else if (migrate_from && (from_status != PCMK_EXEC_PENDING)) { // Failed if (target_node && target_node->details->online) { pe_rsc_trace(rsc, "Marking active on %s %p %d", target, target_node, target_node->details->online); native_add_running(rsc, target_node, data_set, TRUE); } } else { // Pending, or complete but erased if (target_node && target_node->details->online) { pe_rsc_trace(rsc, "Marking active on %s %p %d", target, target_node, target_node->details->online); native_add_running(rsc, target_node, data_set, FALSE); if (source_node && source_node->details->online) { /* This is a partial migration: the migrate_to completed * successfully on the source, but the migrate_from has not * completed. Remember the source and target; if the newly * chosen target remains the same when we schedule actions * later, we may continue with the migration. */ rsc->partial_migration_target = target_node; rsc->partial_migration_source = source_node; } } else { /* Consider it failed here - forces a restart, prevents migration */ pe__set_resource_flags(rsc, pe_rsc_failed|pe_rsc_stop); pe__clear_resource_flags(rsc, pe_rsc_allow_migrate); } } } // Is there an action_name in node_name's rsc history newer than call_id? static bool newer_op(pe_resource_t *rsc, const char *action_name, const char *node_name, int call_id, pe_working_set_t *data_set) { - xmlNode *action = find_lrm_op(rsc->id, action_name, node_name, NULL, TRUE, + xmlNode *action = find_lrm_op(rsc->id, action_name, node_name, NULL, PCMK_OCF_OK, data_set); return pe__call_id(action) > call_id; } static void unpack_migrate_to_failure(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op, pe_working_set_t *data_set) { int target_stop_id = 0; int target_migrate_from_id = 0; xmlNode *target_stop = NULL; xmlNode *target_migrate_from = NULL; const char *source = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_SOURCE); const char *target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET); // Sanity check CRM_CHECK(source && target && !strcmp(source, node->details->uname), return); /* If a migration failed, we have to assume the resource is active. Clones * are not allowed to migrate, so role can't be promoted. */ rsc->role = RSC_ROLE_STARTED; // Check for stop on the target target_stop = find_lrm_op(rsc->id, CRMD_ACTION_STOP, target, NULL, - TRUE, data_set); + PCMK_OCF_OK, data_set); target_stop_id = pe__call_id(target_stop); // Check for migrate_from on the target target_migrate_from = find_lrm_op(rsc->id, CRMD_ACTION_MIGRATED, target, - source, TRUE, data_set); + source, PCMK_OCF_OK, data_set); target_migrate_from_id = pe__call_id(target_migrate_from); - if ((target_stop == NULL) || (target_stop_id < target_migrate_from_id)) { + if (/* If the resource state is unknown on the target, it will likely be + * probed there. + * Don't just consider it running there. We will get back here anyway in + * case the probe detects it's running there. + */ + !unknown_on_node(rsc->id, target, data_set) + /* If there's any probe/monitor operation on the target newer than this + * failed migrate_to indicating it was not running there, this migrate_to + * failure no longer matters for the target. + */ + && !monitor_not_running_after(rsc->id, target, xml_op, data_set) + && ((target_stop == NULL) || (target_stop_id < target_migrate_from_id))) { /* There was no stop on the target, or a stop that happened before a * migrate_from, so assume the resource is still active on the target * (if it is up). */ pe_node_t *target_node = pe_find_node(data_set->nodes, target); pe_rsc_trace(rsc, "stop (%d) + migrate_from (%d)", target_stop_id, target_migrate_from_id); if (target_node && target_node->details->online) { native_add_running(rsc, target_node, data_set, FALSE); } } else if (target_migrate_from == NULL) { /* We know there was a stop on the target, but there may not have been a * migrate_from (the stop could have happened before migrate_from was * scheduled or attempted). * * That means this could be a "dangling" migration. But first, check * whether there is a newer successful stop, start, or migrate_from on * the source node -- it's possible the failed migration was followed by * a successful stop, full restart, or migration in the reverse * direction, in which case we don't want to force a stop. */ int source_migrate_to_id = pe__call_id(xml_op); if (newer_op(rsc, CRMD_ACTION_MIGRATED, source, source_migrate_to_id, data_set) || newer_op(rsc, CRMD_ACTION_START, source, source_migrate_to_id, data_set) || newer_op(rsc, CRMD_ACTION_STOP, source, source_migrate_to_id, data_set)) { return; } // Mark node as having dangling migration so we can force a stop later rsc->dangling_migrations = g_list_prepend(rsc->dangling_migrations, node); } } static void unpack_migrate_from_failure(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op, pe_working_set_t *data_set) { xmlNode *source_stop = NULL; xmlNode *source_migrate_to = NULL; const char *source = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_SOURCE); const char *target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET); // Sanity check CRM_CHECK(source && target && !strcmp(target, node->details->uname), return); /* If a migration failed, we have to assume the resource is active. Clones * are not allowed to migrate, so role can't be promoted. */ rsc->role = RSC_ROLE_STARTED; // Check for a stop on the source source_stop = find_lrm_op(rsc->id, CRMD_ACTION_STOP, source, NULL, - TRUE, data_set); + PCMK_OCF_OK, data_set); // Check for a migrate_to on the source source_migrate_to = find_lrm_op(rsc->id, CRMD_ACTION_MIGRATE, - source, target, TRUE, data_set); + source, target, PCMK_OCF_OK, data_set); if ((source_stop == NULL) || (pe__call_id(source_stop) < pe__call_id(source_migrate_to))) { /* There was no stop on the source, or a stop that happened before * migrate_to, so assume the resource is still active on the source (if * it is up). */ pe_node_t *source_node = pe_find_node(data_set->nodes, source); if (source_node && source_node->details->online) { native_add_running(rsc, source_node, data_set, TRUE); } } } static void record_failed_op(xmlNode *op, const pe_node_t *node, const pe_resource_t *rsc, pe_working_set_t *data_set) { xmlNode *xIter = NULL; const char *op_key = crm_element_value(op, XML_LRM_ATTR_TASK_KEY); if (node->details->online == FALSE) { return; } for (xIter = data_set->failed->children; xIter; xIter = xIter->next) { const char *key = crm_element_value(xIter, XML_LRM_ATTR_TASK_KEY); const char *uname = crm_element_value(xIter, XML_ATTR_UNAME); if(pcmk__str_eq(op_key, key, pcmk__str_casei) && pcmk__str_eq(uname, node->details->uname, pcmk__str_casei)) { crm_trace("Skipping duplicate entry %s on %s", op_key, node->details->uname); return; } } crm_trace("Adding entry %s on %s", op_key, node->details->uname); crm_xml_add(op, XML_ATTR_UNAME, node->details->uname); crm_xml_add(op, XML_LRM_ATTR_RSCID, rsc->id); add_node_copy(data_set->failed, op); } static const char *get_op_key(xmlNode *xml_op) { const char *key = crm_element_value(xml_op, XML_LRM_ATTR_TASK_KEY); if(key == NULL) { key = ID(xml_op); } return key; } static const char * last_change_str(xmlNode *xml_op) { time_t when; const char *when_s = NULL; if (crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &when) == pcmk_ok) { when_s = pcmk__epoch2str(&when); if (when_s) { // Skip day of week to make message shorter when_s = strchr(when_s, ' '); if (when_s) { ++when_s; } } } return ((when_s && *when_s)? when_s : "unknown time"); } /*! * \internal * \brief Compare two on-fail values * * \param[in] first One on-fail value to compare * \param[in] second The other on-fail value to compare * * \return A negative number if second is more severe than first, zero if they * are equal, or a positive number if first is more severe than second. * \note This is only needed until the action_fail_response values can be * renumbered at the next API compatibility break. */ static int cmp_on_fail(enum action_fail_response first, enum action_fail_response second) { switch (first) { case action_fail_demote: switch (second) { case action_fail_ignore: return 1; case action_fail_demote: return 0; default: return -1; } break; case action_fail_reset_remote: switch (second) { case action_fail_ignore: case action_fail_demote: case action_fail_recover: return 1; case action_fail_reset_remote: return 0; default: return -1; } break; case action_fail_restart_container: switch (second) { case action_fail_ignore: case action_fail_demote: case action_fail_recover: case action_fail_reset_remote: return 1; case action_fail_restart_container: return 0; default: return -1; } break; default: break; } switch (second) { case action_fail_demote: return (first == action_fail_ignore)? -1 : 1; case action_fail_reset_remote: switch (first) { case action_fail_ignore: case action_fail_demote: case action_fail_recover: return -1; default: return 1; } break; case action_fail_restart_container: switch (first) { case action_fail_ignore: case action_fail_demote: case action_fail_recover: case action_fail_reset_remote: return -1; default: return 1; } break; default: break; } return first - second; } static void unpack_rsc_op_failure(pe_resource_t * rsc, pe_node_t * node, int rc, xmlNode * xml_op, xmlNode ** last_failure, enum action_fail_response * on_fail, pe_working_set_t * data_set) { bool is_probe = false; pe_action_t *action = NULL; const char *key = get_op_key(xml_op); const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); const char *exit_reason = crm_element_value(xml_op, XML_LRM_ATTR_EXIT_REASON); CRM_ASSERT(rsc); CRM_CHECK(task != NULL, return); *last_failure = xml_op; is_probe = pcmk_xe_is_probe(xml_op); if (exit_reason == NULL) { exit_reason = ""; } if (!pcmk_is_set(data_set->flags, pe_flag_symmetric_cluster) && (rc == PCMK_OCF_NOT_INSTALLED)) { crm_trace("Unexpected result (%s%s%s) was recorded for " "%s of %s on %s at %s " CRM_XS " rc=%d id=%s", services_ocf_exitcode_str(rc), (*exit_reason? ": " : ""), exit_reason, (is_probe? "probe" : task), rsc->id, node->details->uname, last_change_str(xml_op), rc, ID(xml_op)); } else { crm_warn("Unexpected result (%s%s%s) was recorded for " "%s of %s on %s at %s " CRM_XS " rc=%d id=%s", services_ocf_exitcode_str(rc), (*exit_reason? ": " : ""), exit_reason, (is_probe? "probe" : task), rsc->id, node->details->uname, last_change_str(xml_op), rc, ID(xml_op)); if (is_probe && (rc != PCMK_OCF_OK) && (rc != PCMK_OCF_NOT_RUNNING) && (rc != PCMK_OCF_RUNNING_PROMOTED)) { /* A failed (not just unexpected) probe result could mean the user * didn't know resources will be probed even where they can't run. */ crm_notice("If it is not possible for %s to run on %s, see " "the resource-discovery option for location constraints", rsc->id, node->details->uname); } record_failed_op(xml_op, node, rsc, data_set); } action = custom_action(rsc, strdup(key), task, NULL, TRUE, FALSE, data_set); if (cmp_on_fail(*on_fail, action->on_fail) < 0) { pe_rsc_trace(rsc, "on-fail %s -> %s for %s (%s)", fail2text(*on_fail), fail2text(action->on_fail), action->uuid, key); *on_fail = action->on_fail; } if (!strcmp(task, CRMD_ACTION_STOP)) { resource_location(rsc, node, -INFINITY, "__stop_fail__", data_set); } else if (!strcmp(task, CRMD_ACTION_MIGRATE)) { unpack_migrate_to_failure(rsc, node, xml_op, data_set); } else if (!strcmp(task, CRMD_ACTION_MIGRATED)) { unpack_migrate_from_failure(rsc, node, xml_op, data_set); } else if (!strcmp(task, CRMD_ACTION_PROMOTE)) { rsc->role = RSC_ROLE_PROMOTED; } else if (!strcmp(task, CRMD_ACTION_DEMOTE)) { if (action->on_fail == action_fail_block) { rsc->role = RSC_ROLE_PROMOTED; pe__set_next_role(rsc, RSC_ROLE_STOPPED, "demote with on-fail=block"); } else if(rc == PCMK_OCF_NOT_RUNNING) { rsc->role = RSC_ROLE_STOPPED; } else { /* Staying in the promoted role would put the scheduler and * controller into a loop. Setting the role to unpromoted is not * dangerous because the resource will be stopped as part of * recovery, and any promotion will be ordered after that stop. */ rsc->role = RSC_ROLE_UNPROMOTED; } } if(is_probe && rc == PCMK_OCF_NOT_INSTALLED) { /* leave stopped */ pe_rsc_trace(rsc, "Leaving %s stopped", rsc->id); rsc->role = RSC_ROLE_STOPPED; } else if (rsc->role < RSC_ROLE_STARTED) { pe_rsc_trace(rsc, "Setting %s active", rsc->id); set_active(rsc); } pe_rsc_trace(rsc, "Resource %s: role=%s, unclean=%s, on_fail=%s, fail_role=%s", rsc->id, role2text(rsc->role), pcmk__btoa(node->details->unclean), fail2text(action->on_fail), role2text(action->fail_role)); if (action->fail_role != RSC_ROLE_STARTED && rsc->next_role < action->fail_role) { pe__set_next_role(rsc, action->fail_role, "failure"); } if (action->fail_role == RSC_ROLE_STOPPED) { int score = -INFINITY; pe_resource_t *fail_rsc = rsc; if (fail_rsc->parent) { pe_resource_t *parent = uber_parent(fail_rsc); if (pe_rsc_is_clone(parent) && !pcmk_is_set(parent->flags, pe_rsc_unique)) { /* For clone resources, if a child fails on an operation * with on-fail = stop, all the resources fail. Do this by preventing * the parent from coming up again. */ fail_rsc = parent; } } crm_notice("%s will not be started under current conditions", fail_rsc->id); /* make sure it doesn't come up again */ if (fail_rsc->allowed_nodes != NULL) { g_hash_table_destroy(fail_rsc->allowed_nodes); } fail_rsc->allowed_nodes = pe__node_list2table(data_set->nodes); g_hash_table_foreach(fail_rsc->allowed_nodes, set_node_score, &score); } pe_free_action(action); } /*! * \internal * \brief Remap informational monitor results and operation status * * For the monitor results, certain OCF codes are for providing extended information * to the user about services that aren't yet failed but not entirely healthy either. * These must be treated as the "normal" result by Pacemaker. * * For operation status, the action result can be used to determine an appropriate * status for the purposes of responding to the action. The status provided by the * executor is not directly usable since the executor does not know what was expected. * * \param[in] xml_op Operation history entry XML from CIB status * \param[in,out] rsc Resource that operation history entry is for * \param[in] node Node where operation was executed * \param[in] data_set Current cluster working set * \param[in,out] on_fail What should be done about the result * \param[in] target_rc Expected return code of operation * \param[in,out] rc Actual return code of operation * \param[in,out] status Operation execution status * * \note If the result is remapped and the node is not shutting down or failed, * the operation will be recorded in the data set's list of failed operations * to highlight it for the user. * * \note This may update the resource's current and next role. */ static void remap_operation(xmlNode *xml_op, pe_resource_t *rsc, pe_node_t *node, pe_working_set_t *data_set, enum action_fail_response *on_fail, int target_rc, int *rc, int *status) { bool is_probe = false; const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); const char *key = get_op_key(xml_op); const char *exit_reason = crm_element_value(xml_op, XML_LRM_ATTR_EXIT_REASON); if (pcmk__str_eq(task, CRMD_ACTION_STATUS, pcmk__str_none)) { int remapped_rc = pcmk__effective_rc(*rc); if (*rc != remapped_rc) { crm_trace("Remapping monitor result %d to %d", *rc, remapped_rc); if (!node->details->shutdown || node->details->online) { record_failed_op(xml_op, node, rsc, data_set); } *rc = remapped_rc; } } if (!pe_rsc_is_bundled(rsc) && pcmk_xe_mask_probe_failure(xml_op)) { *status = PCMK_EXEC_DONE; *rc = PCMK_OCF_NOT_RUNNING; } /* If the executor reported an operation status of anything but done or * error, consider that final. But for done or error, we know better whether * it should be treated as a failure or not, because we know the expected * result. */ if (*status != PCMK_EXEC_DONE && *status != PCMK_EXEC_ERROR) { return; } CRM_ASSERT(rsc); CRM_CHECK(task != NULL, *status = PCMK_EXEC_ERROR; return); *status = PCMK_EXEC_DONE; if (exit_reason == NULL) { exit_reason = ""; } is_probe = pcmk_xe_is_probe(xml_op); if (is_probe) { task = "probe"; } if (target_rc < 0) { /* Pre-1.0 Pacemaker versions, and Pacemaker 1.1.6 or earlier with * Heartbeat 2.0.7 or earlier as the cluster layer, did not include the * target_rc in the transition key, which (along with the similar case * of a corrupted transition key in the CIB) will be reported to this * function as -1. Pacemaker 2.0+ does not support rolling upgrades from * those versions or processing of saved CIB files from those versions, * so we do not need to care much about this case. */ *status = PCMK_EXEC_ERROR; crm_warn("Expected result not found for %s on %s (corrupt or obsolete CIB?)", key, node->details->uname); } else if (target_rc != *rc) { *status = PCMK_EXEC_ERROR; pe_rsc_debug(rsc, "%s on %s: expected %d (%s), got %d (%s%s%s)", key, node->details->uname, target_rc, services_ocf_exitcode_str(target_rc), *rc, services_ocf_exitcode_str(*rc), (*exit_reason? ": " : ""), exit_reason); } switch (*rc) { case PCMK_OCF_OK: if (is_probe && (target_rc == PCMK_OCF_NOT_RUNNING)) { *status = PCMK_EXEC_DONE; pe_rsc_info(rsc, "Probe found %s active on %s at %s", rsc->id, node->details->uname, last_change_str(xml_op)); } break; case PCMK_OCF_NOT_RUNNING: if (is_probe || (target_rc == *rc) || !pcmk_is_set(rsc->flags, pe_rsc_managed)) { *status = PCMK_EXEC_DONE; rsc->role = RSC_ROLE_STOPPED; /* clear any previous failure actions */ *on_fail = action_fail_ignore; pe__set_next_role(rsc, RSC_ROLE_UNKNOWN, "not running"); } break; case PCMK_OCF_RUNNING_PROMOTED: if (is_probe && (*rc != target_rc)) { *status = PCMK_EXEC_DONE; pe_rsc_info(rsc, "Probe found %s active and promoted on %s at %s", rsc->id, node->details->uname, last_change_str(xml_op)); } rsc->role = RSC_ROLE_PROMOTED; break; case PCMK_OCF_DEGRADED_PROMOTED: case PCMK_OCF_FAILED_PROMOTED: rsc->role = RSC_ROLE_PROMOTED; *status = PCMK_EXEC_ERROR; break; case PCMK_OCF_NOT_CONFIGURED: *status = PCMK_EXEC_ERROR_FATAL; break; case PCMK_OCF_UNIMPLEMENT_FEATURE: { guint interval_ms = 0; crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); if (interval_ms > 0) { *status = PCMK_EXEC_NOT_SUPPORTED; break; } // fall through } case PCMK_OCF_NOT_INSTALLED: case PCMK_OCF_INVALID_PARAM: case PCMK_OCF_INSUFFICIENT_PRIV: if (!pe_can_fence(data_set, node) && !strcmp(task, CRMD_ACTION_STOP)) { /* If a stop fails and we can't fence, there's nothing else we can do */ pe_proc_err("No further recovery can be attempted for %s " "because %s on %s failed (%s%s%s) at %s " CRM_XS " rc=%d id=%s", rsc->id, task, node->details->uname, services_ocf_exitcode_str(*rc), (*exit_reason? ": " : ""), exit_reason, last_change_str(xml_op), *rc, ID(xml_op)); pe__clear_resource_flags(rsc, pe_rsc_managed); pe__set_resource_flags(rsc, pe_rsc_block); } *status = PCMK_EXEC_ERROR_HARD; break; default: if (*status == PCMK_EXEC_DONE) { crm_info("Treating unknown exit status %d from %s of %s " "on %s at %s as failure", *rc, task, rsc->id, node->details->uname, last_change_str(xml_op)); *status = PCMK_EXEC_ERROR; } break; } pe_rsc_trace(rsc, "Remapped %s status to '%s'", key, pcmk_exec_status_str(*status)); } // return TRUE if start or monitor last failure but parameters changed static bool should_clear_for_param_change(xmlNode *xml_op, const char *task, pe_resource_t *rsc, pe_node_t *node, pe_working_set_t *data_set) { if (!strcmp(task, "start") || !strcmp(task, "monitor")) { if (pe__bundle_needs_remote_name(rsc, data_set)) { /* We haven't allocated resources yet, so we can't reliably * substitute addr parameters for the REMOTE_CONTAINER_HACK. * When that's needed, defer the check until later. */ pe__add_param_check(xml_op, rsc, node, pe_check_last_failure, data_set); } else { op_digest_cache_t *digest_data = NULL; digest_data = rsc_action_digest_cmp(rsc, xml_op, node, data_set); switch (digest_data->rc) { case RSC_DIGEST_UNKNOWN: crm_trace("Resource %s history entry %s on %s" " has no digest to compare", rsc->id, get_op_key(xml_op), node->details->id); break; case RSC_DIGEST_MATCH: break; default: return TRUE; } } } return FALSE; } // Order action after fencing of remote node, given connection rsc static void order_after_remote_fencing(pe_action_t *action, pe_resource_t *remote_conn, pe_working_set_t *data_set) { pe_node_t *remote_node = pe_find_node(data_set->nodes, remote_conn->id); if (remote_node) { pe_action_t *fence = pe_fence_op(remote_node, NULL, TRUE, NULL, FALSE, data_set); order_actions(fence, action, pe_order_implies_then); } } static bool should_ignore_failure_timeout(pe_resource_t *rsc, xmlNode *xml_op, const char *task, guint interval_ms, bool is_last_failure, pe_working_set_t *data_set) { /* Clearing failures of recurring monitors has special concerns. The * executor reports only changes in the monitor result, so if the * monitor is still active and still getting the same failure result, * that will go undetected after the failure is cleared. * * Also, the operation history will have the time when the recurring * monitor result changed to the given code, not the time when the * result last happened. * * @TODO We probably should clear such failures only when the failure * timeout has passed since the last occurrence of the failed result. * However we don't record that information. We could maybe approximate * that by clearing only if there is a more recent successful monitor or * stop result, but we don't even have that information at this point * since we are still unpacking the resource's operation history. * * This is especially important for remote connection resources with a * reconnect interval, so in that case, we skip clearing failures * if the remote node hasn't been fenced. */ if (rsc->remote_reconnect_ms && pcmk_is_set(data_set->flags, pe_flag_stonith_enabled) && (interval_ms != 0) && pcmk__str_eq(task, CRMD_ACTION_STATUS, pcmk__str_casei)) { pe_node_t *remote_node = pe_find_node(data_set->nodes, rsc->id); if (remote_node && !remote_node->details->remote_was_fenced) { if (is_last_failure) { crm_info("Waiting to clear monitor failure for remote node %s" " until fencing has occurred", rsc->id); } return TRUE; } } return FALSE; } /*! * \internal * \brief Check operation age and schedule failure clearing when appropriate * * This function has two distinct purposes. The first is to check whether an * operation history entry is expired (i.e. the resource has a failure timeout, * the entry is older than the timeout, and the resource either has no fail * count or its fail count is entirely older than the timeout). The second is to * schedule fail count clearing when appropriate (i.e. the operation is expired * and either the resource has an expired fail count or the operation is a * last_failure for a remote connection resource with a reconnect interval, * or the operation is a last_failure for a start or monitor operation and the * resource's parameters have changed since the operation). * * \param[in] rsc Resource that operation happened to * \param[in] node Node that operation happened on * \param[in] rc Actual result of operation * \param[in] xml_op Operation history entry XML * \param[in] data_set Current working set * * \return TRUE if operation history entry is expired, FALSE otherwise */ static bool check_operation_expiry(pe_resource_t *rsc, pe_node_t *node, int rc, xmlNode *xml_op, pe_working_set_t *data_set) { bool expired = FALSE; bool is_last_failure = pcmk__ends_with(ID(xml_op), "_last_failure_0"); time_t last_run = 0; guint interval_ms = 0; int unexpired_fail_count = 0; const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); const char *clear_reason = NULL; crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); if ((rsc->failure_timeout > 0) && (crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &last_run) == 0)) { // Resource has a failure-timeout, and history entry has a timestamp time_t now = get_effective_time(data_set); time_t last_failure = 0; // Is this particular operation history older than the failure timeout? if ((now >= (last_run + rsc->failure_timeout)) && !should_ignore_failure_timeout(rsc, xml_op, task, interval_ms, is_last_failure, data_set)) { expired = TRUE; } // Does the resource as a whole have an unexpired fail count? unexpired_fail_count = pe_get_failcount(node, rsc, &last_failure, pe_fc_effective, xml_op, data_set); // Update scheduler recheck time according to *last* failure crm_trace("%s@%lld is %sexpired @%lld with unexpired_failures=%d timeout=%ds" " last-failure@%lld", ID(xml_op), (long long) last_run, (expired? "" : "not "), (long long) now, unexpired_fail_count, rsc->failure_timeout, (long long) last_failure); last_failure += rsc->failure_timeout + 1; if (unexpired_fail_count && (now < last_failure)) { pe__update_recheck_time(last_failure, data_set); } } if (expired) { if (pe_get_failcount(node, rsc, NULL, pe_fc_default, xml_op, data_set)) { // There is a fail count ignoring timeout if (unexpired_fail_count == 0) { // There is no fail count considering timeout clear_reason = "it expired"; } else { /* This operation is old, but there is an unexpired fail count. * In a properly functioning cluster, this should only be * possible if this operation is not a failure (otherwise the * fail count should be expired too), so this is really just a * failsafe. */ expired = FALSE; } } else if (is_last_failure && rsc->remote_reconnect_ms) { /* Clear any expired last failure when reconnect interval is set, * even if there is no fail count. */ clear_reason = "reconnect interval is set"; } } if (!expired && is_last_failure && should_clear_for_param_change(xml_op, task, rsc, node, data_set)) { clear_reason = "resource parameters have changed"; } if (clear_reason != NULL) { // Schedule clearing of the fail count pe_action_t *clear_op = pe__clear_failcount(rsc, node, clear_reason, data_set); if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled) && rsc->remote_reconnect_ms) { /* If we're clearing a remote connection due to a reconnect * interval, we want to wait until any scheduled fencing * completes. * * We could limit this to remote_node->details->unclean, but at * this point, that's always true (it won't be reliable until * after unpack_node_history() is done). */ crm_info("Clearing %s failure will wait until any scheduled " "fencing of %s completes", task, rsc->id); order_after_remote_fencing(clear_op, rsc, data_set); } } if (expired && (interval_ms == 0) && pcmk__str_eq(task, CRMD_ACTION_STATUS, pcmk__str_casei)) { switch(rc) { case PCMK_OCF_OK: case PCMK_OCF_NOT_RUNNING: case PCMK_OCF_RUNNING_PROMOTED: case PCMK_OCF_DEGRADED: case PCMK_OCF_DEGRADED_PROMOTED: // Don't expire probes that return these values expired = FALSE; break; } } return expired; } int pe__target_rc_from_xml(xmlNode *xml_op) { int target_rc = 0; const char *key = crm_element_value(xml_op, XML_ATTR_TRANSITION_KEY); if (key == NULL) { return -1; } decode_transition_key(key, NULL, NULL, NULL, &target_rc); return target_rc; } static enum action_fail_response get_action_on_fail(pe_resource_t *rsc, const char *key, const char *task, pe_working_set_t * data_set) { enum action_fail_response result = action_fail_recover; pe_action_t *action = custom_action(rsc, strdup(key), task, NULL, TRUE, FALSE, data_set); result = action->on_fail; pe_free_action(action); return result; } static void update_resource_state(pe_resource_t * rsc, pe_node_t * node, xmlNode * xml_op, const char * task, int rc, xmlNode * last_failure, enum action_fail_response * on_fail, pe_working_set_t * data_set) { gboolean clear_past_failure = FALSE; CRM_ASSERT(rsc); CRM_ASSERT(xml_op); if (rc == PCMK_OCF_NOT_INSTALLED || (!pe_rsc_is_bundled(rsc) && pcmk_xe_mask_probe_failure(xml_op))) { rsc->role = RSC_ROLE_STOPPED; } else if (rc == PCMK_OCF_NOT_RUNNING) { clear_past_failure = TRUE; } else if (pcmk__str_eq(task, CRMD_ACTION_STATUS, pcmk__str_casei)) { if (last_failure) { const char *op_key = get_op_key(xml_op); const char *last_failure_key = get_op_key(last_failure); if (pcmk__str_eq(op_key, last_failure_key, pcmk__str_casei)) { clear_past_failure = TRUE; } } if (rsc->role < RSC_ROLE_STARTED) { set_active(rsc); } } else if (pcmk__str_eq(task, CRMD_ACTION_START, pcmk__str_casei)) { rsc->role = RSC_ROLE_STARTED; clear_past_failure = TRUE; } else if (pcmk__str_eq(task, CRMD_ACTION_STOP, pcmk__str_casei)) { rsc->role = RSC_ROLE_STOPPED; clear_past_failure = TRUE; } else if (pcmk__str_eq(task, CRMD_ACTION_PROMOTE, pcmk__str_casei)) { rsc->role = RSC_ROLE_PROMOTED; clear_past_failure = TRUE; } else if (pcmk__str_eq(task, CRMD_ACTION_DEMOTE, pcmk__str_casei)) { if (*on_fail == action_fail_demote) { // Demote clears an error only if on-fail=demote clear_past_failure = TRUE; } rsc->role = RSC_ROLE_UNPROMOTED; } else if (pcmk__str_eq(task, CRMD_ACTION_MIGRATED, pcmk__str_casei)) { rsc->role = RSC_ROLE_STARTED; clear_past_failure = TRUE; } else if (pcmk__str_eq(task, CRMD_ACTION_MIGRATE, pcmk__str_casei)) { unpack_migrate_to_success(rsc, node, xml_op, data_set); } else if (rsc->role < RSC_ROLE_STARTED) { pe_rsc_trace(rsc, "%s active on %s", rsc->id, node->details->uname); set_active(rsc); } /* clear any previous failure actions */ if (clear_past_failure) { switch (*on_fail) { case action_fail_stop: case action_fail_fence: case action_fail_migrate: case action_fail_standby: pe_rsc_trace(rsc, "%s.%s is not cleared by a completed stop", rsc->id, fail2text(*on_fail)); break; case action_fail_block: case action_fail_ignore: case action_fail_demote: case action_fail_recover: case action_fail_restart_container: *on_fail = action_fail_ignore; pe__set_next_role(rsc, RSC_ROLE_UNKNOWN, "clear past failures"); break; case action_fail_reset_remote: if (rsc->remote_reconnect_ms == 0) { /* With no reconnect interval, the connection is allowed to * start again after the remote node is fenced and * completely stopped. (With a reconnect interval, we wait * for the failure to be cleared entirely before attempting * to reconnect.) */ *on_fail = action_fail_ignore; pe__set_next_role(rsc, RSC_ROLE_UNKNOWN, "clear past failures and reset remote"); } break; } } } static void unpack_rsc_op(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op, xmlNode **last_failure, enum action_fail_response *on_fail, pe_working_set_t *data_set) { int rc = 0; int old_rc = 0; int task_id = 0; int target_rc = 0; int old_target_rc = 0; int status = PCMK_EXEC_UNKNOWN; guint interval_ms = 0; const char *task = NULL; const char *task_key = NULL; const char *exit_reason = NULL; bool expired = false; pe_resource_t *parent = rsc; enum action_fail_response failure_strategy = action_fail_recover; bool maskable_probe_failure = false; CRM_CHECK(rsc && node && xml_op, return); target_rc = pe__target_rc_from_xml(xml_op); task_key = get_op_key(xml_op); task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); exit_reason = crm_element_value(xml_op, XML_LRM_ATTR_EXIT_REASON); if (exit_reason == NULL) { exit_reason = ""; } crm_element_value_int(xml_op, XML_LRM_ATTR_RC, &rc); crm_element_value_int(xml_op, XML_LRM_ATTR_CALLID, &task_id); crm_element_value_int(xml_op, XML_LRM_ATTR_OPSTATUS, &status); crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); CRM_CHECK(task != NULL, return); CRM_CHECK((status >= PCMK_EXEC_PENDING) && (status <= PCMK_EXEC_MAX), return); if (!strcmp(task, CRMD_ACTION_NOTIFY) || !strcmp(task, CRMD_ACTION_METADATA)) { /* safe to ignore these */ return; } if (!pcmk_is_set(rsc->flags, pe_rsc_unique)) { parent = uber_parent(rsc); } pe_rsc_trace(rsc, "Unpacking task %s/%s (call_id=%d, status=%d, rc=%d) on %s (role=%s)", task_key, task, task_id, status, rc, node->details->uname, role2text(rsc->role)); if (node->details->unclean) { pe_rsc_trace(rsc, "Node %s (where %s is running) is unclean." " Further action depends on the value of the stop's on-fail attribute", node->details->uname, rsc->id); } /* It should be possible to call remap_operation() first then call * check_operation_expiry() only if rc != target_rc, because there should * never be a fail count without at least one unexpected result in the * resource history. That would be more efficient by avoiding having to call * check_operation_expiry() for expected results. * * However, we do have such configurations in the scheduler regression * tests, even if it shouldn't be possible with the current code. It's * probably a good idea anyway, but that would require updating the test * inputs to something currently possible. */ if ((status != PCMK_EXEC_NOT_INSTALLED) && check_operation_expiry(rsc, node, rc, xml_op, data_set)) { expired = true; } old_rc = rc; old_target_rc = target_rc; remap_operation(xml_op, rsc, node, data_set, on_fail, target_rc, &rc, &status); maskable_probe_failure = !pe_rsc_is_bundled(rsc) && pcmk_xe_mask_probe_failure(xml_op); if (expired && maskable_probe_failure && old_rc != old_target_rc) { if (rsc->role <= RSC_ROLE_STOPPED) { rsc->role = RSC_ROLE_UNKNOWN; } goto done; } else if (expired && (rc != target_rc)) { const char *magic = crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC); if (interval_ms == 0) { crm_notice("Ignoring expired %s failure on %s " CRM_XS " actual=%d expected=%d magic=%s", task_key, node->details->uname, rc, target_rc, magic); goto done; } else if(node->details->online && node->details->unclean == FALSE) { /* Reschedule the recurring monitor. schedule_cancel() won't work at * this stage, so as a hacky workaround, forcibly change the restart * digest so pcmk__check_action_config() does what we want later. * * @TODO We should skip this if there is a newer successful monitor. * Also, this causes rescheduling only if the history entry * has an op-digest (which the expire-non-blocked-failure * scheduler regression test doesn't, but that may not be a * realistic scenario in production). */ crm_notice("Rescheduling %s after failure expired on %s " CRM_XS " actual=%d expected=%d magic=%s", task_key, node->details->uname, rc, target_rc, magic); crm_xml_add(xml_op, XML_LRM_ATTR_RESTART_DIGEST, "calculated-failure-timeout"); goto done; } } if (maskable_probe_failure) { crm_notice("Treating probe result '%s' for %s on %s as 'not running'", services_ocf_exitcode_str(old_rc), rsc->id, node->details->uname); update_resource_state(rsc, node, xml_op, task, target_rc, *last_failure, on_fail, data_set); crm_xml_add(xml_op, XML_ATTR_UNAME, node->details->uname); record_failed_op(xml_op, node, rsc, data_set); resource_location(parent, node, -INFINITY, "masked-probe-failure", data_set); goto done; } switch (status) { case PCMK_EXEC_CANCELLED: // Should never happen pe_err("Resource history contains cancellation '%s' " "(%s of %s on %s at %s)", ID(xml_op), task, rsc->id, node->details->uname, last_change_str(xml_op)); goto done; case PCMK_EXEC_PENDING: if (!strcmp(task, CRMD_ACTION_START)) { pe__set_resource_flags(rsc, pe_rsc_start_pending); set_active(rsc); } else if (!strcmp(task, CRMD_ACTION_PROMOTE)) { rsc->role = RSC_ROLE_PROMOTED; } else if (!strcmp(task, CRMD_ACTION_MIGRATE) && node->details->unclean) { /* If a pending migrate_to action is out on a unclean node, * we have to force the stop action on the target. */ const char *migrate_target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET); pe_node_t *target = pe_find_node(data_set->nodes, migrate_target); if (target) { stop_action(rsc, target, FALSE); } } if (rsc->pending_task == NULL) { if ((interval_ms != 0) || strcmp(task, CRMD_ACTION_STATUS)) { rsc->pending_task = strdup(task); rsc->pending_node = node; } else { /* Pending probes are not printed, even if pending * operations are requested. If someone ever requests that * behavior, enable the below and the corresponding part of * native.c:native_pending_task(). */ #if 0 rsc->pending_task = strdup("probe"); rsc->pending_node = node; #endif } } goto done; case PCMK_EXEC_DONE: pe_rsc_trace(rsc, "%s of %s on %s completed at %s " CRM_XS " id=%s", task, rsc->id, node->details->uname, last_change_str(xml_op), ID(xml_op)); update_resource_state(rsc, node, xml_op, task, rc, *last_failure, on_fail, data_set); goto done; case PCMK_EXEC_NOT_INSTALLED: failure_strategy = get_action_on_fail(rsc, task_key, task, data_set); if (failure_strategy == action_fail_ignore) { crm_warn("Cannot ignore failed %s of %s on %s: " "Resource agent doesn't exist " CRM_XS " status=%d rc=%d id=%s", task, rsc->id, node->details->uname, status, rc, ID(xml_op)); /* Also for printing it as "FAILED" by marking it as pe_rsc_failed later */ *on_fail = action_fail_migrate; } resource_location(parent, node, -INFINITY, "hard-error", data_set); unpack_rsc_op_failure(rsc, node, rc, xml_op, last_failure, on_fail, data_set); goto done; case PCMK_EXEC_NOT_CONNECTED: if (pe__is_guest_or_remote_node(node) && pcmk_is_set(node->details->remote_rsc->flags, pe_rsc_managed)) { /* We should never get into a situation where a managed remote * connection resource is considered OK but a resource action * behind the connection gets a "not connected" status. But as a * fail-safe in case a bug or unusual circumstances do lead to * that, ensure the remote connection is considered failed. */ pe__set_resource_flags(node->details->remote_rsc, pe_rsc_failed|pe_rsc_stop); } break; // Not done, do error handling case PCMK_EXEC_ERROR: case PCMK_EXEC_ERROR_HARD: case PCMK_EXEC_ERROR_FATAL: case PCMK_EXEC_TIMEOUT: case PCMK_EXEC_NOT_SUPPORTED: case PCMK_EXEC_INVALID: break; // Not done, do error handling case PCMK_EXEC_NO_FENCE_DEVICE: case PCMK_EXEC_NO_SECRETS: status = PCMK_EXEC_ERROR_HARD; break; // Not done, do error handling } failure_strategy = get_action_on_fail(rsc, task_key, task, data_set); if ((failure_strategy == action_fail_ignore) || (failure_strategy == action_fail_restart_container && !strcmp(task, CRMD_ACTION_STOP))) { crm_warn("Pretending failed %s (%s%s%s) of %s on %s at %s " "succeeded " CRM_XS " rc=%d id=%s", task, services_ocf_exitcode_str(rc), (*exit_reason? ": " : ""), exit_reason, rsc->id, node->details->uname, last_change_str(xml_op), rc, ID(xml_op)); update_resource_state(rsc, node, xml_op, task, target_rc, *last_failure, on_fail, data_set); crm_xml_add(xml_op, XML_ATTR_UNAME, node->details->uname); pe__set_resource_flags(rsc, pe_rsc_failure_ignored); record_failed_op(xml_op, node, rsc, data_set); if ((failure_strategy == action_fail_restart_container) && cmp_on_fail(*on_fail, action_fail_recover) <= 0) { *on_fail = failure_strategy; } } else { unpack_rsc_op_failure(rsc, node, rc, xml_op, last_failure, on_fail, data_set); if (status == PCMK_EXEC_ERROR_HARD) { do_crm_log(rc != PCMK_OCF_NOT_INSTALLED?LOG_ERR:LOG_NOTICE, "Preventing %s from restarting on %s because " "of hard failure (%s%s%s)" CRM_XS " rc=%d id=%s", parent->id, node->details->uname, services_ocf_exitcode_str(rc), (*exit_reason? ": " : ""), exit_reason, rc, ID(xml_op)); resource_location(parent, node, -INFINITY, "hard-error", data_set); } else if (status == PCMK_EXEC_ERROR_FATAL) { crm_err("Preventing %s from restarting anywhere because " "of fatal failure (%s%s%s) " CRM_XS " rc=%d id=%s", parent->id, services_ocf_exitcode_str(rc), (*exit_reason? ": " : ""), exit_reason, rc, ID(xml_op)); resource_location(parent, NULL, -INFINITY, "fatal-error", data_set); } } done: pe_rsc_trace(rsc, "Resource %s after %s: role=%s, next=%s", rsc->id, task, role2text(rsc->role), role2text(rsc->next_role)); } static void add_node_attrs(xmlNode *xml_obj, pe_node_t *node, bool overwrite, pe_working_set_t *data_set) { const char *cluster_name = NULL; pe_rule_eval_data_t rule_data = { .node_hash = NULL, .role = RSC_ROLE_UNKNOWN, .now = data_set->now, .match_data = NULL, .rsc_data = NULL, .op_data = NULL }; g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_UNAME), strdup(node->details->uname)); g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_ID), strdup(node->details->id)); if (pcmk__str_eq(node->details->id, data_set->dc_uuid, pcmk__str_casei)) { data_set->dc_node = node; node->details->is_dc = TRUE; g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_IS_DC), strdup(XML_BOOLEAN_TRUE)); } else { g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_IS_DC), strdup(XML_BOOLEAN_FALSE)); } cluster_name = g_hash_table_lookup(data_set->config_hash, "cluster-name"); if (cluster_name) { g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_CLUSTER_NAME), strdup(cluster_name)); } pe__unpack_dataset_nvpairs(xml_obj, XML_TAG_ATTR_SETS, &rule_data, node->details->attrs, NULL, overwrite, data_set); if (pe_node_attribute_raw(node, CRM_ATTR_SITE_NAME) == NULL) { const char *site_name = pe_node_attribute_raw(node, "site-name"); if (site_name) { g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_SITE_NAME), strdup(site_name)); } else if (cluster_name) { /* Default to cluster-name if unset */ g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_SITE_NAME), strdup(cluster_name)); } } } static GList * extract_operations(const char *node, const char *rsc, xmlNode * rsc_entry, gboolean active_filter) { int counter = -1; int stop_index = -1; int start_index = -1; xmlNode *rsc_op = NULL; GList *gIter = NULL; GList *op_list = NULL; GList *sorted_op_list = NULL; /* extract operations */ op_list = NULL; sorted_op_list = NULL; for (rsc_op = pcmk__xe_first_child(rsc_entry); rsc_op != NULL; rsc_op = pcmk__xe_next(rsc_op)) { if (pcmk__str_eq((const char *)rsc_op->name, XML_LRM_TAG_RSC_OP, pcmk__str_none)) { crm_xml_add(rsc_op, "resource", rsc); crm_xml_add(rsc_op, XML_ATTR_UNAME, node); op_list = g_list_prepend(op_list, rsc_op); } } if (op_list == NULL) { /* if there are no operations, there is nothing to do */ return NULL; } sorted_op_list = g_list_sort(op_list, sort_op_by_callid); /* create active recurring operations as optional */ if (active_filter == FALSE) { return sorted_op_list; } op_list = NULL; calculate_active_ops(sorted_op_list, &start_index, &stop_index); for (gIter = sorted_op_list; gIter != NULL; gIter = gIter->next) { xmlNode *rsc_op = (xmlNode *) gIter->data; counter++; if (start_index < stop_index) { crm_trace("Skipping %s: not active", ID(rsc_entry)); break; } else if (counter < start_index) { crm_trace("Skipping %s: old", ID(rsc_op)); continue; } op_list = g_list_append(op_list, rsc_op); } g_list_free(sorted_op_list); return op_list; } GList * find_operations(const char *rsc, const char *node, gboolean active_filter, pe_working_set_t * data_set) { GList *output = NULL; GList *intermediate = NULL; xmlNode *tmp = NULL; xmlNode *status = find_xml_node(data_set->input, XML_CIB_TAG_STATUS, TRUE); pe_node_t *this_node = NULL; xmlNode *node_state = NULL; for (node_state = pcmk__xe_first_child(status); node_state != NULL; node_state = pcmk__xe_next(node_state)) { if (pcmk__str_eq((const char *)node_state->name, XML_CIB_TAG_STATE, pcmk__str_none)) { const char *uname = crm_element_value(node_state, XML_ATTR_UNAME); if (node != NULL && !pcmk__str_eq(uname, node, pcmk__str_casei)) { continue; } this_node = pe_find_node(data_set->nodes, uname); if(this_node == NULL) { CRM_LOG_ASSERT(this_node != NULL); continue; } else if (pe__is_guest_or_remote_node(this_node)) { determine_remote_online_status(data_set, this_node); } else { determine_online_status(node_state, this_node, data_set); } if (this_node->details->online || pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { /* offline nodes run no resources... * unless stonith is enabled in which case we need to * make sure rsc start events happen after the stonith */ xmlNode *lrm_rsc = NULL; tmp = find_xml_node(node_state, XML_CIB_TAG_LRM, FALSE); tmp = find_xml_node(tmp, XML_LRM_TAG_RESOURCES, FALSE); for (lrm_rsc = pcmk__xe_first_child(tmp); lrm_rsc != NULL; lrm_rsc = pcmk__xe_next(lrm_rsc)) { if (pcmk__str_eq((const char *)lrm_rsc->name, XML_LRM_TAG_RESOURCE, pcmk__str_none)) { const char *rsc_id = crm_element_value(lrm_rsc, XML_ATTR_ID); if (rsc != NULL && !pcmk__str_eq(rsc_id, rsc, pcmk__str_casei)) { continue; } intermediate = extract_operations(uname, rsc_id, lrm_rsc, active_filter); output = g_list_concat(output, intermediate); } } } } } return output; } diff --git a/lib/pengine/utils.c b/lib/pengine/utils.c index c0ddc7d23a..32c43bce8c 100644 --- a/lib/pengine/utils.c +++ b/lib/pengine/utils.c @@ -1,2616 +1,2626 @@ /* * Copyright 2004-2022 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 "pe_status_private.h" extern bool pcmk__is_daemon; void print_str_str(gpointer key, gpointer value, gpointer user_data); gboolean ghash_free_str_str(gpointer key, gpointer value, gpointer user_data); static void unpack_operation(pe_action_t * action, xmlNode * xml_obj, pe_resource_t * container, pe_working_set_t * data_set, guint interval_ms); static xmlNode *find_rsc_op_entry_helper(pe_resource_t * rsc, const char *key, gboolean include_disabled); #if ENABLE_VERSIONED_ATTRS pe_rsc_action_details_t * pe_rsc_action_details(pe_action_t *action) { pe_rsc_action_details_t *details; CRM_CHECK(action != NULL, return NULL); if (action->action_details == NULL) { action->action_details = calloc(1, sizeof(pe_rsc_action_details_t)); CRM_CHECK(action->action_details != NULL, return NULL); } details = (pe_rsc_action_details_t *) action->action_details; if (details->versioned_parameters == NULL) { details->versioned_parameters = create_xml_node(NULL, XML_TAG_OP_VER_ATTRS); } if (details->versioned_meta == NULL) { details->versioned_meta = create_xml_node(NULL, XML_TAG_OP_VER_META); } return details; } static void pe_free_rsc_action_details(pe_action_t *action) { pe_rsc_action_details_t *details; if ((action == NULL) || (action->action_details == NULL)) { return; } details = (pe_rsc_action_details_t *) action->action_details; if (details->versioned_parameters) { free_xml(details->versioned_parameters); } if (details->versioned_meta) { free_xml(details->versioned_meta); } action->action_details = NULL; } #endif /*! * \internal * \brief Check whether we can fence a particular node * * \param[in] data_set Working set for cluster * \param[in] node Name of node to check * * \return true if node can be fenced, false otherwise */ bool pe_can_fence(pe_working_set_t *data_set, pe_node_t *node) { if (pe__is_guest_node(node)) { /* Guest nodes are fenced by stopping their container resource. We can * do that if the container's host is either online or fenceable. */ pe_resource_t *rsc = node->details->remote_rsc->container; for (GList *n = rsc->running_on; n != NULL; n = n->next) { pe_node_t *container_node = n->data; if (!container_node->details->online && !pe_can_fence(data_set, container_node)) { return false; } } return true; } else if (!pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { return false; /* Turned off */ } else if (!pcmk_is_set(data_set->flags, pe_flag_have_stonith_resource)) { return false; /* No devices */ } else if (pcmk_is_set(data_set->flags, pe_flag_have_quorum)) { return true; } else if (data_set->no_quorum_policy == no_quorum_ignore) { return true; } else if(node == NULL) { return false; } else if(node->details->online) { crm_notice("We can fence %s without quorum because they're in our membership", node->details->uname); return true; } crm_trace("Cannot fence %s", node->details->uname); return false; } /*! * \internal * \brief Copy a node object * * \param[in] this_node Node object to copy * * \return Newly allocated shallow copy of this_node * \note This function asserts on errors and is guaranteed to return non-NULL. */ pe_node_t * pe__copy_node(const pe_node_t *this_node) { pe_node_t *new_node = NULL; CRM_ASSERT(this_node != NULL); new_node = calloc(1, sizeof(pe_node_t)); CRM_ASSERT(new_node != NULL); new_node->rsc_discover_mode = this_node->rsc_discover_mode; new_node->weight = this_node->weight; new_node->fixed = this_node->fixed; new_node->details = this_node->details; return new_node; } /* any node in list1 or list2 and not in the other gets a score of -INFINITY */ void node_list_exclude(GHashTable * hash, GList *list, gboolean merge_scores) { GHashTable *result = hash; pe_node_t *other_node = NULL; GList *gIter = list; GHashTableIter iter; pe_node_t *node = NULL; g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { other_node = pe_find_node_id(list, node->details->id); if (other_node == NULL) { node->weight = -INFINITY; } else if (merge_scores) { node->weight = pcmk__add_scores(node->weight, other_node->weight); } } for (; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; other_node = pe_hash_table_lookup(result, node->details->id); if (other_node == NULL) { pe_node_t *new_node = pe__copy_node(node); new_node->weight = -INFINITY; g_hash_table_insert(result, (gpointer) new_node->details->id, new_node); } } } /*! * \internal * \brief Create a node hash table from a node list * * \param[in] list Node list * * \return Hash table equivalent of node list */ GHashTable * pe__node_list2table(GList *list) { GHashTable *result = NULL; result = pcmk__strkey_table(NULL, free); for (GList *gIter = list; gIter != NULL; gIter = gIter->next) { pe_node_t *new_node = pe__copy_node((pe_node_t *) gIter->data); g_hash_table_insert(result, (gpointer) new_node->details->id, new_node); } return result; } gint sort_node_uname(gconstpointer a, gconstpointer b) { return pcmk__numeric_strcasecmp(((const pe_node_t *) a)->details->uname, ((const pe_node_t *) b)->details->uname); } /*! * \internal * \brief Output node weights to stdout * * \param[in] rsc Use allowed nodes for this resource * \param[in] comment Text description to prefix lines with * \param[in] nodes If rsc is not specified, use these nodes */ static void pe__output_node_weights(pe_resource_t *rsc, const char *comment, GHashTable *nodes, pe_working_set_t *data_set) { pcmk__output_t *out = data_set->priv; char score[128]; // Stack-allocated since this is called frequently // Sort the nodes so the output is consistent for regression tests GList *list = g_list_sort(g_hash_table_get_values(nodes), sort_node_uname); for (GList *gIter = list; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; score2char_stack(node->weight, score, sizeof(score)); out->message(out, "node-weight", rsc, comment, node->details->uname, score); } g_list_free(list); } /*! * \internal * \brief Log node weights at trace level * * \param[in] file Caller's filename * \param[in] function Caller's function name * \param[in] line Caller's line number * \param[in] rsc Use allowed nodes for this resource * \param[in] comment Text description to prefix lines with * \param[in] nodes If rsc is not specified, use these nodes */ static void pe__log_node_weights(const char *file, const char *function, int line, pe_resource_t *rsc, const char *comment, GHashTable *nodes) { GHashTableIter iter; pe_node_t *node = NULL; char score[128]; // Stack-allocated since this is called frequently // Don't waste time if we're not tracing at this point pcmk__log_else(LOG_TRACE, return); g_hash_table_iter_init(&iter, nodes); while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) { score2char_stack(node->weight, score, sizeof(score)); if (rsc) { qb_log_from_external_source(function, file, "%s: %s allocation score on %s: %s", LOG_TRACE, line, 0, comment, rsc->id, node->details->uname, score); } else { qb_log_from_external_source(function, file, "%s: %s = %s", LOG_TRACE, line, 0, comment, node->details->uname, score); } } } /*! * \internal * \brief Log or output node weights * * \param[in] file Caller's filename * \param[in] function Caller's function name * \param[in] line Caller's line number * \param[in] to_log Log if true, otherwise output * \param[in] rsc Use allowed nodes for this resource * \param[in] comment Text description to prefix lines with * \param[in] nodes Use these nodes */ void pe__show_node_weights_as(const char *file, const char *function, int line, bool to_log, pe_resource_t *rsc, const char *comment, GHashTable *nodes, pe_working_set_t *data_set) { if (rsc != NULL && pcmk_is_set(rsc->flags, pe_rsc_orphan)) { // Don't show allocation scores for orphans return; } if (nodes == NULL) { // Nothing to show return; } if (to_log) { pe__log_node_weights(file, function, line, rsc, comment, nodes); } else { pe__output_node_weights(rsc, comment, nodes, data_set); } // If this resource has children, repeat recursively for each if (rsc && rsc->children) { for (GList *gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *) gIter->data; pe__show_node_weights_as(file, function, line, to_log, child, comment, child->allowed_nodes, data_set); } } } gint sort_rsc_index(gconstpointer a, gconstpointer b) { const pe_resource_t *resource1 = (const pe_resource_t *)a; const pe_resource_t *resource2 = (const pe_resource_t *)b; if (a == NULL && b == NULL) { return 0; } if (a == NULL) { return 1; } if (b == NULL) { return -1; } if (resource1->sort_index > resource2->sort_index) { return -1; } if (resource1->sort_index < resource2->sort_index) { return 1; } return 0; } gint sort_rsc_priority(gconstpointer a, gconstpointer b) { const pe_resource_t *resource1 = (const pe_resource_t *)a; const pe_resource_t *resource2 = (const pe_resource_t *)b; if (a == NULL && b == NULL) { return 0; } if (a == NULL) { return 1; } if (b == NULL) { return -1; } if (resource1->priority > resource2->priority) { return -1; } if (resource1->priority < resource2->priority) { return 1; } return 0; } static enum pe_quorum_policy effective_quorum_policy(pe_resource_t *rsc, pe_working_set_t *data_set) { enum pe_quorum_policy policy = data_set->no_quorum_policy; if (pcmk_is_set(data_set->flags, pe_flag_have_quorum)) { policy = no_quorum_ignore; } else if (data_set->no_quorum_policy == no_quorum_demote) { switch (rsc->role) { case RSC_ROLE_PROMOTED: case RSC_ROLE_UNPROMOTED: if (rsc->next_role > RSC_ROLE_UNPROMOTED) { pe__set_next_role(rsc, RSC_ROLE_UNPROMOTED, "no-quorum-policy=demote"); } policy = no_quorum_ignore; break; default: policy = no_quorum_stop; break; } } return policy; } static void add_singleton(pe_working_set_t *data_set, pe_action_t *action) { if (data_set->singletons == NULL) { data_set->singletons = pcmk__strkey_table(NULL, NULL); } g_hash_table_insert(data_set->singletons, action->uuid, action); } static pe_action_t * lookup_singleton(pe_working_set_t *data_set, const char *action_uuid) { if (data_set->singletons == NULL) { return NULL; } return g_hash_table_lookup(data_set->singletons, action_uuid); } /*! * \internal * \brief Find an existing action that matches arguments * * \param[in] key Action key to match * \param[in] rsc Resource to match (if any) * \param[in] node Node to match (if any) * \param[in] data_set Cluster working set * * \return Existing action that matches arguments (or NULL if none) */ static pe_action_t * find_existing_action(const char *key, pe_resource_t *rsc, pe_node_t *node, pe_working_set_t *data_set) { GList *matches = NULL; pe_action_t *action = NULL; /* When rsc is NULL, it would be quicker to check data_set->singletons, * but checking all data_set->actions takes the node into account. */ matches = find_actions(((rsc == NULL)? data_set->actions : rsc->actions), key, node); if (matches == NULL) { return NULL; } CRM_LOG_ASSERT(!pcmk__list_of_multiple(matches)); action = matches->data; g_list_free(matches); return action; } /*! * \internal * \brief Create a new action object * * \param[in] key Action key * \param[in] task Action name * \param[in] rsc Resource that action is for (if any) * \param[in] node Node that action is on (if any) * \param[in] optional Whether action should be considered optional * \param[in] for_graph Whether action should be recorded in transition graph * \param[in] data_set Cluster working set * * \return Newly allocated action * \note This function takes ownership of \p key. It is the caller's * responsibility to free the return value with pe_free_action(). */ static pe_action_t * new_action(char *key, const char *task, pe_resource_t *rsc, pe_node_t *node, bool optional, bool for_graph, pe_working_set_t *data_set) { pe_action_t *action = calloc(1, sizeof(pe_action_t)); CRM_ASSERT(action != NULL); action->rsc = rsc; action->task = strdup(task); CRM_ASSERT(action->task != NULL); action->uuid = key; action->extra = pcmk__strkey_table(free, free); action->meta = pcmk__strkey_table(free, free); if (node) { action->node = pe__copy_node(node); } if (pcmk__str_eq(task, CRM_OP_LRM_DELETE, pcmk__str_casei)) { // Resource history deletion for a node can be done on the DC pe__set_action_flags(action, pe_action_dc); } pe__set_action_flags(action, pe_action_runnable); if (optional) { pe__set_action_flags(action, pe_action_optional); } else { pe__clear_action_flags(action, pe_action_optional); } if (rsc != NULL) { guint interval_ms = 0; action->op_entry = find_rsc_op_entry_helper(rsc, key, TRUE); parse_op_key(key, NULL, NULL, &interval_ms); unpack_operation(action, action->op_entry, rsc->container, data_set, interval_ms); } if (for_graph) { pe_rsc_trace(rsc, "Created %s action %d (%s): %s for %s on %s", (optional? "optional" : "required"), data_set->action_id, key, task, ((rsc == NULL)? "no resource" : rsc->id), ((node == NULL)? "no node" : node->details->uname)); action->id = data_set->action_id++; data_set->actions = g_list_prepend(data_set->actions, action); if (rsc == NULL) { add_singleton(data_set, action); } else { rsc->actions = g_list_prepend(rsc->actions, action); } } return action; } /*! * \internal * \brief Evaluate node attribute values for an action * * \param[in] action Action to unpack attributes for * \param[in] data_set Cluster working set */ static void unpack_action_node_attributes(pe_action_t *action, pe_working_set_t *data_set) { if (!pcmk_is_set(action->flags, pe_action_have_node_attrs) && (action->op_entry != NULL)) { pe_rule_eval_data_t rule_data = { .node_hash = action->node->details->attrs, .role = RSC_ROLE_UNKNOWN, .now = data_set->now, .match_data = NULL, .rsc_data = NULL, .op_data = NULL }; pe__set_action_flags(action, pe_action_have_node_attrs); pe__unpack_dataset_nvpairs(action->op_entry, XML_TAG_ATTR_SETS, &rule_data, action->extra, NULL, FALSE, data_set); } } /*! * \internal * \brief Update an action's optional flag * * \param[in] action Action to update * \param[in] optional Requested optional status */ static void update_action_optional(pe_action_t *action, gboolean optional) { // Force a non-recurring action to be optional if its resource is unmanaged if ((action->rsc != NULL) && (action->node != NULL) && !pcmk_is_set(action->flags, pe_action_pseudo) && !pcmk_is_set(action->rsc->flags, pe_rsc_managed) && (g_hash_table_lookup(action->meta, XML_LRM_ATTR_INTERVAL_MS) == NULL)) { pe_rsc_debug(action->rsc, "%s on %s is optional (%s is unmanaged)", action->uuid, action->node->details->uname, action->rsc->id); pe__set_action_flags(action, pe_action_optional); // We shouldn't clear runnable here because ... something // Otherwise require the action if requested } else if (!optional) { pe__clear_action_flags(action, pe_action_optional); } } /*! * \internal * \brief Update a resource action's runnable flag * * \param[in] action Action to update * \param[in] for_graph Whether action should be recorded in transition graph * \param[in] data_set Cluster working set * * \note This may also schedule fencing if a stop is unrunnable. */ static void update_resource_action_runnable(pe_action_t *action, bool for_graph, pe_working_set_t *data_set) { if (pcmk_is_set(action->flags, pe_action_pseudo)) { return; } if (action->node == NULL) { pe_rsc_trace(action->rsc, "%s is unrunnable (unallocated)", action->uuid); pe__clear_action_flags(action, pe_action_runnable); } else if (!pcmk_is_set(action->flags, pe_action_dc) && !(action->node->details->online) && (!pe__is_guest_node(action->node) || action->node->details->remote_requires_reset)) { pe__clear_action_flags(action, pe_action_runnable); do_crm_log((for_graph? LOG_WARNING: LOG_TRACE), "%s on %s is unrunnable (node is offline)", action->uuid, action->node->details->uname); if (pcmk_is_set(action->rsc->flags, pe_rsc_managed) && for_graph && pcmk__str_eq(action->task, CRMD_ACTION_STOP, pcmk__str_casei) && !(action->node->details->unclean)) { pe_fence_node(data_set, action->node, "stop is unrunnable", false); } } else if (!pcmk_is_set(action->flags, pe_action_dc) && action->node->details->pending) { pe__clear_action_flags(action, pe_action_runnable); do_crm_log((for_graph? LOG_WARNING: LOG_TRACE), "Action %s on %s is unrunnable (node is pending)", action->uuid, action->node->details->uname); } else if (action->needs == rsc_req_nothing) { pe_action_set_reason(action, NULL, TRUE); if (pe__is_guest_node(action->node) && !pe_can_fence(data_set, action->node)) { /* An action that requires nothing usually does not require any * fencing in order to be runnable. However, there is an exception: * such an action cannot be completed if it is on a guest node whose * host is unclean and cannot be fenced. */ pe_rsc_debug(action->rsc, "%s on %s is unrunnable " "(node's host cannot be fenced)", action->uuid, action->node->details->uname); pe__clear_action_flags(action, pe_action_runnable); } else { pe_rsc_trace(action->rsc, "%s on %s does not require fencing or quorum", action->uuid, action->node->details->uname); pe__set_action_flags(action, pe_action_runnable); } } else { switch (effective_quorum_policy(action->rsc, data_set)) { case no_quorum_stop: pe_rsc_debug(action->rsc, "%s on %s is unrunnable (no quorum)", action->uuid, action->node->details->uname); pe__clear_action_flags(action, pe_action_runnable); pe_action_set_reason(action, "no quorum", true); break; case no_quorum_freeze: if (!action->rsc->fns->active(action->rsc, TRUE) || (action->rsc->next_role > action->rsc->role)) { pe_rsc_debug(action->rsc, "%s on %s is unrunnable (no quorum)", action->uuid, action->node->details->uname); pe__clear_action_flags(action, pe_action_runnable); pe_action_set_reason(action, "quorum freeze", true); } break; default: //pe_action_set_reason(action, NULL, TRUE); pe__set_action_flags(action, pe_action_runnable); break; } } } /*! * \internal * \brief Update a resource object's flags for a new action on it * * \param[in] rsc Resource that action is for (if any) * \param[in] action New action */ static void update_resource_flags_for_action(pe_resource_t *rsc, pe_action_t *action) { /* @COMPAT pe_rsc_starting and pe_rsc_stopping are not actually used * within Pacemaker, and should be deprecated and eventually removed */ if (pcmk__str_eq(action->task, CRMD_ACTION_STOP, pcmk__str_casei)) { pe__set_resource_flags(rsc, pe_rsc_stopping); } else if (pcmk__str_eq(action->task, CRMD_ACTION_START, pcmk__str_casei)) { if (pcmk_is_set(action->flags, pe_action_runnable)) { pe__set_resource_flags(rsc, pe_rsc_starting); } else { pe__clear_resource_flags(rsc, pe_rsc_starting); } } } /*! * \brief Create or update an action object * * \param[in] rsc Resource that action is for (if any) * \param[in] key Action key (must be non-NULL) * \param[in] task Action name (must be non-NULL) * \param[in] on_node Node that action is on (if any) * \param[in] optional Whether action should be considered optional * \param[in] save_action Whether action should be recorded in transition graph * \param[in] data_set Cluster working set * * \return Action object corresponding to arguments * \note This function takes ownership of (and might free) \p key. If * \p save_action is true, \p data_set will own the returned action, * otherwise it is the caller's responsibility to free the return value * with pe_free_action(). */ pe_action_t * custom_action(pe_resource_t *rsc, char *key, const char *task, pe_node_t *on_node, gboolean optional, gboolean save_action, pe_working_set_t *data_set) { pe_action_t *action = NULL; CRM_ASSERT((key != NULL) && (task != NULL) && (data_set != NULL)); if (save_action) { action = find_existing_action(key, rsc, on_node, data_set); } if (action == NULL) { action = new_action(key, task, rsc, on_node, optional, save_action, data_set); } else { free(key); } update_action_optional(action, optional); if (rsc != NULL) { if (action->node != NULL) { unpack_action_node_attributes(action, data_set); } update_resource_action_runnable(action, save_action, data_set); if (save_action) { update_resource_flags_for_action(rsc, action); } } return action; } static bool valid_stop_on_fail(const char *value) { return !pcmk__strcase_any_of(value, "standby", "demote", "stop", NULL); } static const char * unpack_operation_on_fail(pe_action_t * action) { const char *name = NULL; const char *role = NULL; const char *on_fail = NULL; const char *interval_spec = NULL; const char *value = g_hash_table_lookup(action->meta, XML_OP_ATTR_ON_FAIL); if (pcmk__str_eq(action->task, CRMD_ACTION_STOP, pcmk__str_casei) && !valid_stop_on_fail(value)) { pcmk__config_err("Resetting '" XML_OP_ATTR_ON_FAIL "' for %s stop " "action to default value because '%s' is not " "allowed for stop", action->rsc->id, value); return NULL; } else if (pcmk__str_eq(action->task, CRMD_ACTION_DEMOTE, pcmk__str_casei) && !value) { // demote on_fail defaults to monitor value for promoted role if present xmlNode *operation = NULL; CRM_CHECK(action->rsc != NULL, return NULL); for (operation = pcmk__xe_first_child(action->rsc->ops_xml); (operation != NULL) && (value == NULL); operation = pcmk__xe_next(operation)) { bool enabled = false; if (!pcmk__str_eq((const char *)operation->name, "op", pcmk__str_none)) { continue; } name = crm_element_value(operation, "name"); role = crm_element_value(operation, "role"); on_fail = crm_element_value(operation, XML_OP_ATTR_ON_FAIL); interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL); if (!on_fail) { continue; } else if (pcmk__xe_get_bool_attr(operation, "enabled", &enabled) == pcmk_rc_ok && !enabled) { continue; } else if (!pcmk__str_eq(name, "monitor", pcmk__str_casei) || !pcmk__strcase_any_of(role, RSC_ROLE_PROMOTED_S, RSC_ROLE_PROMOTED_LEGACY_S, NULL)) { continue; } else if (crm_parse_interval_spec(interval_spec) == 0) { continue; } else if (pcmk__str_eq(on_fail, "demote", pcmk__str_casei)) { continue; } value = on_fail; } } else if (pcmk__str_eq(action->task, CRM_OP_LRM_DELETE, pcmk__str_casei)) { value = "ignore"; } else if (pcmk__str_eq(value, "demote", pcmk__str_casei)) { name = crm_element_value(action->op_entry, "name"); role = crm_element_value(action->op_entry, "role"); interval_spec = crm_element_value(action->op_entry, XML_LRM_ATTR_INTERVAL); if (!pcmk__str_eq(name, CRMD_ACTION_PROMOTE, pcmk__str_casei) && (!pcmk__str_eq(name, CRMD_ACTION_STATUS, pcmk__str_casei) || !pcmk__strcase_any_of(role, RSC_ROLE_PROMOTED_S, RSC_ROLE_PROMOTED_LEGACY_S, NULL) || (crm_parse_interval_spec(interval_spec) == 0))) { pcmk__config_err("Resetting '" XML_OP_ATTR_ON_FAIL "' for %s %s " "action to default value because 'demote' is not " "allowed for it", action->rsc->id, name); return NULL; } } return value; } static xmlNode * find_min_interval_mon(pe_resource_t * rsc, gboolean include_disabled) { guint interval_ms = 0; guint min_interval_ms = G_MAXUINT; const char *name = NULL; const char *interval_spec = NULL; xmlNode *op = NULL; xmlNode *operation = NULL; for (operation = pcmk__xe_first_child(rsc->ops_xml); operation != NULL; operation = pcmk__xe_next(operation)) { if (pcmk__str_eq((const char *)operation->name, "op", pcmk__str_none)) { bool enabled = false; name = crm_element_value(operation, "name"); interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL); if (!include_disabled && pcmk__xe_get_bool_attr(operation, "enabled", &enabled) == pcmk_rc_ok && !enabled) { continue; } if (!pcmk__str_eq(name, RSC_STATUS, pcmk__str_casei)) { continue; } interval_ms = crm_parse_interval_spec(interval_spec); if (interval_ms && (interval_ms < min_interval_ms)) { min_interval_ms = interval_ms; op = operation; } } } return op; } static int unpack_start_delay(const char *value, GHashTable *meta) { int start_delay = 0; if (value != NULL) { start_delay = crm_get_msec(value); if (start_delay < 0) { start_delay = 0; } if (meta) { g_hash_table_replace(meta, strdup(XML_OP_ATTR_START_DELAY), pcmk__itoa(start_delay)); } } return start_delay; } // true if value contains valid, non-NULL interval origin for recurring op static bool unpack_interval_origin(const char *value, xmlNode *xml_obj, guint interval_ms, crm_time_t *now, long long *start_delay) { long long result = 0; guint interval_sec = interval_ms / 1000; crm_time_t *origin = NULL; // Ignore unspecified values and non-recurring operations if ((value == NULL) || (interval_ms == 0) || (now == NULL)) { return false; } // Parse interval origin from text origin = crm_time_new(value); if (origin == NULL) { pcmk__config_err("Ignoring '" XML_OP_ATTR_ORIGIN "' for operation " "'%s' because '%s' is not valid", (ID(xml_obj)? ID(xml_obj) : "(missing ID)"), value); return false; } // Get seconds since origin (negative if origin is in the future) result = crm_time_get_seconds(now) - crm_time_get_seconds(origin); crm_time_free(origin); // Calculate seconds from closest interval to now result = result % interval_sec; // Calculate seconds remaining until next interval result = ((result <= 0)? 0 : interval_sec) - result; crm_info("Calculated a start delay of %llds for operation '%s'", result, (ID(xml_obj)? ID(xml_obj) : "(unspecified)")); if (start_delay != NULL) { *start_delay = result * 1000; // milliseconds } return true; } static int unpack_timeout(const char *value) { int timeout_ms = crm_get_msec(value); if (timeout_ms < 0) { timeout_ms = crm_get_msec(CRM_DEFAULT_OP_TIMEOUT_S); } return timeout_ms; } int pe_get_configured_timeout(pe_resource_t *rsc, const char *action, pe_working_set_t *data_set) { xmlNode *child = NULL; GHashTable *action_meta = NULL; const char *timeout_spec = NULL; int timeout_ms = 0; pe_rule_eval_data_t rule_data = { .node_hash = NULL, .role = RSC_ROLE_UNKNOWN, .now = data_set->now, .match_data = NULL, .rsc_data = NULL, .op_data = NULL }; for (child = first_named_child(rsc->ops_xml, XML_ATTR_OP); child != NULL; child = crm_next_same_xml(child)) { if (pcmk__str_eq(action, crm_element_value(child, XML_NVPAIR_ATTR_NAME), pcmk__str_casei)) { timeout_spec = crm_element_value(child, XML_ATTR_TIMEOUT); break; } } if (timeout_spec == NULL && data_set->op_defaults) { action_meta = pcmk__strkey_table(free, free); pe__unpack_dataset_nvpairs(data_set->op_defaults, XML_TAG_META_SETS, &rule_data, action_meta, NULL, FALSE, data_set); timeout_spec = g_hash_table_lookup(action_meta, XML_ATTR_TIMEOUT); } // @TODO check meta-attributes (including versioned meta-attributes) // @TODO maybe use min-interval monitor timeout as default for monitors timeout_ms = crm_get_msec(timeout_spec); if (timeout_ms < 0) { timeout_ms = crm_get_msec(CRM_DEFAULT_OP_TIMEOUT_S); } if (action_meta != NULL) { g_hash_table_destroy(action_meta); } return timeout_ms; } #if ENABLE_VERSIONED_ATTRS static void unpack_versioned_meta(xmlNode *versioned_meta, xmlNode *xml_obj, guint interval_ms, crm_time_t *now) { xmlNode *attrs = NULL; xmlNode *attr = NULL; for (attrs = pcmk__xe_first_child(versioned_meta); attrs != NULL; attrs = pcmk__xe_next(attrs)) { for (attr = pcmk__xe_first_child(attrs); attr != NULL; attr = pcmk__xe_next(attr)) { const char *name = crm_element_value(attr, XML_NVPAIR_ATTR_NAME); const char *value = crm_element_value(attr, XML_NVPAIR_ATTR_VALUE); if (pcmk__str_eq(name, XML_OP_ATTR_START_DELAY, pcmk__str_casei)) { int start_delay = unpack_start_delay(value, NULL); crm_xml_add_int(attr, XML_NVPAIR_ATTR_VALUE, start_delay); } else if (pcmk__str_eq(name, XML_OP_ATTR_ORIGIN, pcmk__str_casei)) { long long start_delay = 0; if (unpack_interval_origin(value, xml_obj, interval_ms, now, &start_delay)) { crm_xml_add(attr, XML_NVPAIR_ATTR_NAME, XML_OP_ATTR_START_DELAY); crm_xml_add_ll(attr, XML_NVPAIR_ATTR_VALUE, start_delay); } } else if (pcmk__str_eq(name, XML_ATTR_TIMEOUT, pcmk__str_casei)) { int timeout_ms = unpack_timeout(value); crm_xml_add_int(attr, XML_NVPAIR_ATTR_VALUE, timeout_ms); } } } } #endif /*! * \brief Unpack operation XML into an action structure * * Unpack an operation's meta-attributes (normalizing the interval, timeout, * and start delay values as integer milliseconds), requirements, and * failure policy. * * \param[in,out] action Action to unpack into * \param[in] xml_obj Operation XML (or NULL if all defaults) * \param[in] container Resource that contains affected resource, if any * \param[in] data_set Cluster state * \param[in] interval_ms How frequently to perform the operation */ static void unpack_operation(pe_action_t * action, xmlNode * xml_obj, pe_resource_t * container, pe_working_set_t * data_set, guint interval_ms) { int timeout_ms = 0; const char *value = NULL; bool is_probe = false; #if ENABLE_VERSIONED_ATTRS pe_rsc_action_details_t *rsc_details = NULL; #endif pe_rsc_eval_data_t rsc_rule_data = { .standard = crm_element_value(action->rsc->xml, XML_AGENT_ATTR_CLASS), .provider = crm_element_value(action->rsc->xml, XML_AGENT_ATTR_PROVIDER), .agent = crm_element_value(action->rsc->xml, XML_EXPR_ATTR_TYPE) }; pe_op_eval_data_t op_rule_data = { .op_name = action->task, .interval = interval_ms }; pe_rule_eval_data_t rule_data = { .node_hash = NULL, .role = RSC_ROLE_UNKNOWN, .now = data_set->now, .match_data = NULL, .rsc_data = &rsc_rule_data, .op_data = &op_rule_data }; CRM_CHECK(action && action->rsc, return); is_probe = pcmk_is_probe(action->task, interval_ms); // Cluster-wide pe__unpack_dataset_nvpairs(data_set->op_defaults, XML_TAG_META_SETS, &rule_data, action->meta, NULL, FALSE, data_set); // Determine probe default timeout differently if (is_probe) { xmlNode *min_interval_mon = find_min_interval_mon(action->rsc, FALSE); if (min_interval_mon) { value = crm_element_value(min_interval_mon, XML_ATTR_TIMEOUT); if (value) { crm_trace("\t%s: Setting default timeout to minimum-interval " "monitor's timeout '%s'", action->uuid, value); g_hash_table_replace(action->meta, strdup(XML_ATTR_TIMEOUT), strdup(value)); } } } if (xml_obj) { xmlAttrPtr xIter = NULL; // take precedence over defaults pe__unpack_dataset_nvpairs(xml_obj, XML_TAG_META_SETS, &rule_data, action->meta, NULL, TRUE, data_set); #if ENABLE_VERSIONED_ATTRS rsc_details = pe_rsc_action_details(action); /* Non-versioned attributes also unpack XML_TAG_ATTR_SETS, but that * capability is deprecated, so we don't need to extend that support to * versioned attributes. */ pe_eval_versioned_attributes(data_set->input, xml_obj, XML_TAG_META_SETS, &rule_data, rsc_details->versioned_meta, NULL); #endif /* Anything set as an XML property has highest precedence. * This ensures we use the name and interval from the tag. */ for (xIter = xml_obj->properties; xIter; xIter = xIter->next) { const char *prop_name = (const char *)xIter->name; const char *prop_value = crm_element_value(xml_obj, prop_name); g_hash_table_replace(action->meta, strdup(prop_name), strdup(prop_value)); } } g_hash_table_remove(action->meta, "id"); // Normalize interval to milliseconds if (interval_ms > 0) { g_hash_table_replace(action->meta, strdup(XML_LRM_ATTR_INTERVAL), crm_strdup_printf("%u", interval_ms)); } else { g_hash_table_remove(action->meta, XML_LRM_ATTR_INTERVAL); } /* * Timeout order of precedence: * 1. pcmk_monitor_timeout (if rsc has pcmk_ra_cap_fence_params * and task is start or a probe; pcmk_monitor_timeout works * by default for a recurring monitor) * 2. explicit op timeout on the primitive * 3. default op timeout * a. if probe, then min-interval monitor's timeout * b. else, in XML_CIB_TAG_OPCONFIG * 4. CRM_DEFAULT_OP_TIMEOUT_S * * #1 overrides general rule of XML property having highest * precedence. */ if (pcmk_is_set(pcmk_get_ra_caps(rsc_rule_data.standard), pcmk_ra_cap_fence_params) && (pcmk__str_eq(action->task, RSC_START, pcmk__str_casei) || is_probe)) { GHashTable *params = pe_rsc_params(action->rsc, action->node, data_set); value = g_hash_table_lookup(params, "pcmk_monitor_timeout"); if (value) { crm_trace("\t%s: Setting timeout to pcmk_monitor_timeout '%s', " "overriding default", action->uuid, value); g_hash_table_replace(action->meta, strdup(XML_ATTR_TIMEOUT), strdup(value)); } } // Normalize timeout to positive milliseconds value = g_hash_table_lookup(action->meta, XML_ATTR_TIMEOUT); timeout_ms = unpack_timeout(value); g_hash_table_replace(action->meta, strdup(XML_ATTR_TIMEOUT), pcmk__itoa(timeout_ms)); if (!pcmk__strcase_any_of(action->task, RSC_START, RSC_PROMOTE, NULL)) { action->needs = rsc_req_nothing; value = "nothing (not start or promote)"; } else if (pcmk_is_set(action->rsc->flags, pe_rsc_needs_fencing)) { action->needs = rsc_req_stonith; value = "fencing"; } else if (pcmk_is_set(action->rsc->flags, pe_rsc_needs_quorum)) { action->needs = rsc_req_quorum; value = "quorum"; } else { action->needs = rsc_req_nothing; value = "nothing"; } pe_rsc_trace(action->rsc, "%s requires %s", action->uuid, value); value = unpack_operation_on_fail(action); if (value == NULL) { } else if (pcmk__str_eq(value, "block", pcmk__str_casei)) { action->on_fail = action_fail_block; g_hash_table_insert(action->meta, strdup(XML_OP_ATTR_ON_FAIL), strdup("block")); value = "block"; // The above could destroy the original string } else if (pcmk__str_eq(value, "fence", pcmk__str_casei)) { action->on_fail = action_fail_fence; value = "node fencing"; if (!pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { pcmk__config_err("Resetting '" XML_OP_ATTR_ON_FAIL "' for " "operation '%s' to 'stop' because 'fence' is not " "valid when fencing is disabled", action->uuid); action->on_fail = action_fail_stop; action->fail_role = RSC_ROLE_STOPPED; value = "stop resource"; } } else if (pcmk__str_eq(value, "standby", pcmk__str_casei)) { action->on_fail = action_fail_standby; value = "node standby"; } else if (pcmk__strcase_any_of(value, "ignore", "nothing", NULL)) { action->on_fail = action_fail_ignore; value = "ignore"; } else if (pcmk__str_eq(value, "migrate", pcmk__str_casei)) { action->on_fail = action_fail_migrate; value = "force migration"; } else if (pcmk__str_eq(value, "stop", pcmk__str_casei)) { action->on_fail = action_fail_stop; action->fail_role = RSC_ROLE_STOPPED; value = "stop resource"; } else if (pcmk__str_eq(value, "restart", pcmk__str_casei)) { action->on_fail = action_fail_recover; value = "restart (and possibly migrate)"; } else if (pcmk__str_eq(value, "restart-container", pcmk__str_casei)) { if (container) { action->on_fail = action_fail_restart_container; value = "restart container (and possibly migrate)"; } else { value = NULL; } } else if (pcmk__str_eq(value, "demote", pcmk__str_casei)) { action->on_fail = action_fail_demote; value = "demote instance"; } else { pe_err("Resource %s: Unknown failure type (%s)", action->rsc->id, value); value = NULL; } /* defaults */ if (value == NULL && container) { action->on_fail = action_fail_restart_container; value = "restart container (and possibly migrate) (default)"; /* For remote nodes, ensure that any failure that results in dropping an * active connection to the node results in fencing of the node. * * There are only two action failures that don't result in fencing. * 1. probes - probe failures are expected. * 2. start - a start failure indicates that an active connection does not already * exist. The user can set op on-fail=fence if they really want to fence start * failures. */ } else if (((value == NULL) || !pcmk_is_set(action->rsc->flags, pe_rsc_managed)) && pe__resource_is_remote_conn(action->rsc, data_set) && !(pcmk__str_eq(action->task, CRMD_ACTION_STATUS, pcmk__str_casei) && (interval_ms == 0)) && !pcmk__str_eq(action->task, CRMD_ACTION_START, pcmk__str_casei)) { if (!pcmk_is_set(action->rsc->flags, pe_rsc_managed)) { action->on_fail = action_fail_stop; action->fail_role = RSC_ROLE_STOPPED; value = "stop unmanaged remote node (enforcing default)"; } else { if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { value = "fence remote node (default)"; } else { value = "recover remote node connection (default)"; } if (action->rsc->remote_reconnect_ms) { action->fail_role = RSC_ROLE_STOPPED; } action->on_fail = action_fail_reset_remote; } } else if (value == NULL && pcmk__str_eq(action->task, CRMD_ACTION_STOP, pcmk__str_casei)) { if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { action->on_fail = action_fail_fence; value = "resource fence (default)"; } else { action->on_fail = action_fail_block; value = "resource block (default)"; } } else if (value == NULL) { action->on_fail = action_fail_recover; value = "restart (and possibly migrate) (default)"; } pe_rsc_trace(action->rsc, "%s failure handling: %s", action->uuid, value); value = NULL; if (xml_obj != NULL) { value = g_hash_table_lookup(action->meta, "role_after_failure"); if (value) { pe_warn_once(pe_wo_role_after, "Support for role_after_failure is deprecated and will be removed in a future release"); } } if (value != NULL && action->fail_role == RSC_ROLE_UNKNOWN) { action->fail_role = text2role(value); } /* defaults */ if (action->fail_role == RSC_ROLE_UNKNOWN) { if (pcmk__str_eq(action->task, CRMD_ACTION_PROMOTE, pcmk__str_casei)) { action->fail_role = RSC_ROLE_UNPROMOTED; } else { action->fail_role = RSC_ROLE_STARTED; } } pe_rsc_trace(action->rsc, "%s failure results in: %s", action->uuid, role2text(action->fail_role)); value = g_hash_table_lookup(action->meta, XML_OP_ATTR_START_DELAY); if (value) { unpack_start_delay(value, action->meta); } else { long long start_delay = 0; value = g_hash_table_lookup(action->meta, XML_OP_ATTR_ORIGIN); if (unpack_interval_origin(value, xml_obj, interval_ms, data_set->now, &start_delay)) { g_hash_table_replace(action->meta, strdup(XML_OP_ATTR_START_DELAY), crm_strdup_printf("%lld", start_delay)); } } #if ENABLE_VERSIONED_ATTRS unpack_versioned_meta(rsc_details->versioned_meta, xml_obj, interval_ms, data_set->now); #endif } static xmlNode * find_rsc_op_entry_helper(pe_resource_t * rsc, const char *key, gboolean include_disabled) { guint interval_ms = 0; gboolean do_retry = TRUE; char *local_key = NULL; const char *name = NULL; const char *interval_spec = NULL; char *match_key = NULL; xmlNode *op = NULL; xmlNode *operation = NULL; retry: for (operation = pcmk__xe_first_child(rsc->ops_xml); operation != NULL; operation = pcmk__xe_next(operation)) { if (pcmk__str_eq((const char *)operation->name, "op", pcmk__str_none)) { bool enabled = false; name = crm_element_value(operation, "name"); interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL); if (!include_disabled && pcmk__xe_get_bool_attr(operation, "enabled", &enabled) == pcmk_rc_ok && !enabled) { continue; } interval_ms = crm_parse_interval_spec(interval_spec); match_key = pcmk__op_key(rsc->id, name, interval_ms); if (pcmk__str_eq(key, match_key, pcmk__str_casei)) { op = operation; } free(match_key); if (rsc->clone_name) { match_key = pcmk__op_key(rsc->clone_name, name, interval_ms); if (pcmk__str_eq(key, match_key, pcmk__str_casei)) { op = operation; } free(match_key); } if (op != NULL) { free(local_key); return op; } } } free(local_key); if (do_retry == FALSE) { return NULL; } do_retry = FALSE; if (strstr(key, CRMD_ACTION_MIGRATE) || strstr(key, CRMD_ACTION_MIGRATED)) { local_key = pcmk__op_key(rsc->id, "migrate", 0); key = local_key; goto retry; } else if (strstr(key, "_notify_")) { local_key = pcmk__op_key(rsc->id, "notify", 0); key = local_key; goto retry; } return NULL; } xmlNode * find_rsc_op_entry(pe_resource_t * rsc, const char *key) { return find_rsc_op_entry_helper(rsc, key, FALSE); } /* * Used by the HashTable for-loop */ void print_str_str(gpointer key, gpointer value, gpointer user_data) { crm_trace("%s%s %s ==> %s", user_data == NULL ? "" : (char *)user_data, user_data == NULL ? "" : ": ", (char *)key, (char *)value); } void pe_free_action(pe_action_t * action) { if (action == NULL) { return; } g_list_free_full(action->actions_before, free); /* pe_action_wrapper_t* */ g_list_free_full(action->actions_after, free); /* pe_action_wrapper_t* */ if (action->extra) { g_hash_table_destroy(action->extra); } if (action->meta) { g_hash_table_destroy(action->meta); } #if ENABLE_VERSIONED_ATTRS if (action->rsc) { pe_free_rsc_action_details(action); } #endif free(action->cancel_task); free(action->reason); free(action->task); free(action->uuid); free(action->node); free(action); } GList * find_recurring_actions(GList *input, pe_node_t * not_on_node) { const char *value = NULL; GList *result = NULL; GList *gIter = input; CRM_CHECK(input != NULL, return NULL); for (; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; value = g_hash_table_lookup(action->meta, XML_LRM_ATTR_INTERVAL_MS); if (value == NULL) { /* skip */ } else if (pcmk__str_eq(value, "0", pcmk__str_casei)) { /* skip */ } else if (pcmk__str_eq(CRMD_ACTION_CANCEL, action->task, pcmk__str_casei)) { /* skip */ } else if (not_on_node == NULL) { crm_trace("(null) Found: %s", action->uuid); result = g_list_prepend(result, action); } else if (action->node == NULL) { /* skip */ } else if (action->node->details != not_on_node->details) { crm_trace("Found: %s", action->uuid); result = g_list_prepend(result, action); } } return result; } enum action_tasks get_complex_task(pe_resource_t * rsc, const char *name, gboolean allow_non_atomic) { enum action_tasks task = text2task(name); if (rsc == NULL) { return task; } else if (allow_non_atomic == FALSE || rsc->variant == pe_native) { switch (task) { case stopped_rsc: case started_rsc: case action_demoted: case action_promoted: crm_trace("Folding %s back into its atomic counterpart for %s", name, rsc->id); return task - 1; default: break; } } return task; } pe_action_t * find_first_action(GList *input, const char *uuid, const char *task, pe_node_t * on_node) { GList *gIter = NULL; CRM_CHECK(uuid || task, return NULL); for (gIter = input; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; if (uuid != NULL && !pcmk__str_eq(uuid, action->uuid, pcmk__str_casei)) { continue; } else if (task != NULL && !pcmk__str_eq(task, action->task, pcmk__str_casei)) { continue; } else if (on_node == NULL) { return action; } else if (action->node == NULL) { continue; } else if (on_node->details == action->node->details) { return action; } } return NULL; } GList * find_actions(GList *input, const char *key, const pe_node_t *on_node) { GList *gIter = input; GList *result = NULL; CRM_CHECK(key != NULL, return NULL); for (; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; if (!pcmk__str_eq(key, action->uuid, pcmk__str_casei)) { continue; } else if (on_node == NULL) { crm_trace("Action %s matches (ignoring node)", key); result = g_list_prepend(result, action); } else if (action->node == NULL) { crm_trace("Action %s matches (unallocated, assigning to %s)", key, on_node->details->uname); action->node = pe__copy_node(on_node); result = g_list_prepend(result, action); } else if (on_node->details == action->node->details) { crm_trace("Action %s on %s matches", key, on_node->details->uname); result = g_list_prepend(result, action); } } return result; } GList * find_actions_exact(GList *input, const char *key, const pe_node_t *on_node) { GList *result = NULL; CRM_CHECK(key != NULL, return NULL); if (on_node == NULL) { return NULL; } for (GList *gIter = input; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; if ((action->node != NULL) && pcmk__str_eq(key, action->uuid, pcmk__str_casei) && pcmk__str_eq(on_node->details->id, action->node->details->id, pcmk__str_casei)) { crm_trace("Action %s on %s matches", key, on_node->details->uname); result = g_list_prepend(result, action); } } return result; } /*! * \brief Find all actions of given type for a resource * * \param[in] rsc Resource to search * \param[in] node Find only actions scheduled on this node * \param[in] task Action name to search for * \param[in] require_node If TRUE, NULL node or action node will not match * * \return List of actions found (or NULL if none) * \note If node is not NULL and require_node is FALSE, matching actions * without a node will be assigned to node. */ GList * pe__resource_actions(const pe_resource_t *rsc, const pe_node_t *node, const char *task, bool require_node) { GList *result = NULL; char *key = pcmk__op_key(rsc->id, task, 0); if (require_node) { result = find_actions_exact(rsc->actions, key, node); } else { result = find_actions(rsc->actions, key, node); } free(key); return result; } static void resource_node_score(pe_resource_t * rsc, pe_node_t * node, int score, const char *tag) { pe_node_t *match = NULL; if ((rsc->exclusive_discover || (node->rsc_discover_mode == pe_discover_never)) && pcmk__str_eq(tag, "symmetric_default", pcmk__str_casei)) { /* This string comparision may be fragile, but exclusive resources and * exclusive nodes should not have the symmetric_default constraint * applied to them. */ return; } else if (rsc->children) { GList *gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; resource_node_score(child_rsc, node, score, tag); } } pe_rsc_trace(rsc, "Setting %s for %s on %s: %d", tag, rsc->id, node->details->uname, score); match = pe_hash_table_lookup(rsc->allowed_nodes, node->details->id); if (match == NULL) { match = pe__copy_node(node); g_hash_table_insert(rsc->allowed_nodes, (gpointer) match->details->id, match); } match->weight = pcmk__add_scores(match->weight, score); } void resource_location(pe_resource_t * rsc, pe_node_t * node, int score, const char *tag, pe_working_set_t * data_set) { if (node != NULL) { resource_node_score(rsc, node, score, tag); } else if (data_set != NULL) { GList *gIter = data_set->nodes; for (; gIter != NULL; gIter = gIter->next) { pe_node_t *node_iter = (pe_node_t *) gIter->data; resource_node_score(rsc, node_iter, score, tag); } } else { GHashTableIter iter; pe_node_t *node_iter = NULL; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node_iter)) { resource_node_score(rsc, node_iter, score, tag); } } if (node == NULL && score == -INFINITY) { if (rsc->allocated_to) { crm_info("Deallocating %s from %s", rsc->id, rsc->allocated_to->details->uname); free(rsc->allocated_to); rsc->allocated_to = NULL; } } } #define sort_return(an_int, why) do { \ free(a_uuid); \ free(b_uuid); \ crm_trace("%s (%d) %c %s (%d) : %s", \ a_xml_id, a_call_id, an_int>0?'>':an_int<0?'<':'=', \ b_xml_id, b_call_id, why); \ return an_int; \ } while(0) -gint -sort_op_by_callid(gconstpointer a, gconstpointer b) +int +pe__is_newer_op(const xmlNode *xml_a, const xmlNode *xml_b) { int a_call_id = -1; int b_call_id = -1; char *a_uuid = NULL; char *b_uuid = NULL; - const xmlNode *xml_a = a; - const xmlNode *xml_b = b; - const char *a_xml_id = crm_element_value(xml_a, XML_ATTR_ID); const char *b_xml_id = crm_element_value(xml_b, XML_ATTR_ID); - if (pcmk__str_eq(a_xml_id, b_xml_id, pcmk__str_casei)) { + const char *a_node = crm_element_value(xml_a, XML_LRM_ATTR_TARGET); + const char *b_node = crm_element_value(xml_b, XML_LRM_ATTR_TARGET); + bool same_node = pcmk__str_eq(a_node, b_node, pcmk__str_casei); + + if (same_node && pcmk__str_eq(a_xml_id, b_xml_id, pcmk__str_none)) { /* We have duplicate lrm_rsc_op entries in the status * section which is unlikely to be a good thing * - we can handle it easily enough, but we need to get * to the bottom of why it's happening. */ pe_err("Duplicate lrm_rsc_op entries named %s", a_xml_id); sort_return(0, "duplicate"); } crm_element_value_int(xml_a, XML_LRM_ATTR_CALLID, &a_call_id); crm_element_value_int(xml_b, XML_LRM_ATTR_CALLID, &b_call_id); if (a_call_id == -1 && b_call_id == -1) { /* both are pending ops so it doesn't matter since * stops are never pending */ sort_return(0, "pending"); - } else if (a_call_id >= 0 && a_call_id < b_call_id) { + } else if (same_node && a_call_id >= 0 && a_call_id < b_call_id) { sort_return(-1, "call id"); - } else if (b_call_id >= 0 && a_call_id > b_call_id) { + } else if (same_node && b_call_id >= 0 && a_call_id > b_call_id) { sort_return(1, "call id"); - } else if (b_call_id >= 0 && a_call_id == b_call_id) { + } else if (a_call_id >= 0 && b_call_id >= 0 + && (!same_node || a_call_id == b_call_id)) { /* * The op and last_failed_op are the same * Order on last-rc-change */ time_t last_a = -1; time_t last_b = -1; crm_element_value_epoch(xml_a, XML_RSC_OP_LAST_CHANGE, &last_a); crm_element_value_epoch(xml_b, XML_RSC_OP_LAST_CHANGE, &last_b); crm_trace("rc-change: %lld vs %lld", (long long) last_a, (long long) last_b); if (last_a >= 0 && last_a < last_b) { sort_return(-1, "rc-change"); } else if (last_b >= 0 && last_a > last_b) { sort_return(1, "rc-change"); } sort_return(0, "rc-change"); } else { /* One of the inputs is a pending operation * Attempt to use XML_ATTR_TRANSITION_MAGIC to determine its age relative to the other */ int a_id = -1; int b_id = -1; const char *a_magic = crm_element_value(xml_a, XML_ATTR_TRANSITION_MAGIC); const char *b_magic = crm_element_value(xml_b, XML_ATTR_TRANSITION_MAGIC); CRM_CHECK(a_magic != NULL && b_magic != NULL, sort_return(0, "No magic")); if (!decode_transition_magic(a_magic, &a_uuid, &a_id, NULL, NULL, NULL, NULL)) { sort_return(0, "bad magic a"); } if (!decode_transition_magic(b_magic, &b_uuid, &b_id, NULL, NULL, NULL, NULL)) { sort_return(0, "bad magic b"); } /* try to determine the relative age of the operation... * some pending operations (e.g. a start) may have been superseded * by a subsequent stop * * [a|b]_id == -1 means it's a shutdown operation and _always_ comes last */ if (!pcmk__str_eq(a_uuid, b_uuid, pcmk__str_casei) || a_id == b_id) { /* * some of the logic in here may be redundant... * * if the UUID from the TE doesn't match then one better * be a pending operation. * pending operations don't survive between elections and joins * because we query the LRM directly */ if (b_call_id == -1) { sort_return(-1, "transition + call"); } else if (a_call_id == -1) { sort_return(1, "transition + call"); } } else if ((a_id >= 0 && a_id < b_id) || b_id == -1) { sort_return(-1, "transition"); } else if ((b_id >= 0 && a_id > b_id) || a_id == -1) { sort_return(1, "transition"); } } /* we should never end up here */ CRM_CHECK(FALSE, sort_return(0, "default")); +} + +gint +sort_op_by_callid(gconstpointer a, gconstpointer b) +{ + const xmlNode *xml_a = a; + const xmlNode *xml_b = b; + return pe__is_newer_op(xml_a, xml_b); } time_t get_effective_time(pe_working_set_t * data_set) { if(data_set) { if (data_set->now == NULL) { crm_trace("Recording a new 'now'"); data_set->now = crm_time_new(NULL); } return crm_time_get_seconds_since_epoch(data_set->now); } crm_trace("Defaulting to 'now'"); return time(NULL); } gboolean get_target_role(pe_resource_t * rsc, enum rsc_role_e * role) { enum rsc_role_e local_role = RSC_ROLE_UNKNOWN; const char *value = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE); CRM_CHECK(role != NULL, return FALSE); if (pcmk__str_eq(value, "started", pcmk__str_null_matches | pcmk__str_casei) || pcmk__str_eq("default", value, pcmk__str_casei)) { return FALSE; } local_role = text2role(value); if (local_role == RSC_ROLE_UNKNOWN) { pcmk__config_err("Ignoring '" XML_RSC_ATTR_TARGET_ROLE "' for %s " "because '%s' is not valid", rsc->id, value); return FALSE; } else if (local_role > RSC_ROLE_STARTED) { if (pcmk_is_set(uber_parent(rsc)->flags, pe_rsc_promotable)) { if (local_role > RSC_ROLE_UNPROMOTED) { /* This is what we'd do anyway, just leave the default to avoid messing up the placement algorithm */ return FALSE; } } else { pcmk__config_err("Ignoring '" XML_RSC_ATTR_TARGET_ROLE "' for %s " "because '%s' only makes sense for promotable " "clones", rsc->id, value); return FALSE; } } *role = local_role; return TRUE; } gboolean order_actions(pe_action_t * lh_action, pe_action_t * rh_action, enum pe_ordering order) { GList *gIter = NULL; pe_action_wrapper_t *wrapper = NULL; GList *list = NULL; if (order == pe_order_none) { return FALSE; } if (lh_action == NULL || rh_action == NULL) { return FALSE; } crm_trace("Creating action wrappers for ordering: %s then %s", lh_action->uuid, rh_action->uuid); /* Ensure we never create a dependency on ourselves... it's happened */ CRM_ASSERT(lh_action != rh_action); /* Filter dups, otherwise update_action_states() has too much work to do */ gIter = lh_action->actions_after; for (; gIter != NULL; gIter = gIter->next) { pe_action_wrapper_t *after = (pe_action_wrapper_t *) gIter->data; if (after->action == rh_action && (after->type & order)) { return FALSE; } } wrapper = calloc(1, sizeof(pe_action_wrapper_t)); wrapper->action = rh_action; wrapper->type = order; list = lh_action->actions_after; list = g_list_prepend(list, wrapper); lh_action->actions_after = list; wrapper = calloc(1, sizeof(pe_action_wrapper_t)); wrapper->action = lh_action; wrapper->type = order; list = rh_action->actions_before; list = g_list_prepend(list, wrapper); rh_action->actions_before = list; return TRUE; } pe_action_t * get_pseudo_op(const char *name, pe_working_set_t * data_set) { pe_action_t *op = lookup_singleton(data_set, name); if (op == NULL) { op = custom_action(NULL, strdup(name), name, NULL, TRUE, TRUE, data_set); pe__set_action_flags(op, pe_action_pseudo|pe_action_runnable); } return op; } void destroy_ticket(gpointer data) { pe_ticket_t *ticket = data; if (ticket->state) { g_hash_table_destroy(ticket->state); } free(ticket->id); free(ticket); } pe_ticket_t * ticket_new(const char *ticket_id, pe_working_set_t * data_set) { pe_ticket_t *ticket = NULL; if (pcmk__str_empty(ticket_id)) { return NULL; } if (data_set->tickets == NULL) { data_set->tickets = pcmk__strkey_table(free, destroy_ticket); } ticket = g_hash_table_lookup(data_set->tickets, ticket_id); if (ticket == NULL) { ticket = calloc(1, sizeof(pe_ticket_t)); if (ticket == NULL) { crm_err("Cannot allocate ticket '%s'", ticket_id); return NULL; } crm_trace("Creaing ticket entry for %s", ticket_id); ticket->id = strdup(ticket_id); ticket->granted = FALSE; ticket->last_granted = -1; ticket->standby = FALSE; ticket->state = pcmk__strkey_table(free, free); g_hash_table_insert(data_set->tickets, strdup(ticket->id), ticket); } return ticket; } const char *rsc_printable_id(pe_resource_t *rsc) { if (!pcmk_is_set(rsc->flags, pe_rsc_unique)) { return ID(rsc->xml); } return rsc->id; } void pe__clear_resource_flags_recursive(pe_resource_t *rsc, uint64_t flags) { pe__clear_resource_flags(rsc, flags); for (GList *gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe__clear_resource_flags_recursive((pe_resource_t *) gIter->data, flags); } } void pe__clear_resource_flags_on_all(pe_working_set_t *data_set, uint64_t flag) { for (GList *lpc = data_set->resources; lpc != NULL; lpc = lpc->next) { pe_resource_t *r = (pe_resource_t *) lpc->data; pe__clear_resource_flags_recursive(r, flag); } } void pe__set_resource_flags_recursive(pe_resource_t *rsc, uint64_t flags) { pe__set_resource_flags(rsc, flags); for (GList *gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe__set_resource_flags_recursive((pe_resource_t *) gIter->data, flags); } } static GList * find_unfencing_devices(GList *candidates, GList *matches) { for (GList *gIter = candidates; gIter != NULL; gIter = gIter->next) { pe_resource_t *candidate = gIter->data; const char *provides = g_hash_table_lookup(candidate->meta, PCMK_STONITH_PROVIDES); const char *requires = g_hash_table_lookup(candidate->meta, XML_RSC_ATTR_REQUIRES); if(candidate->children) { matches = find_unfencing_devices(candidate->children, matches); } else if (!pcmk_is_set(candidate->flags, pe_rsc_fence_device)) { continue; } else if (pcmk__str_eq(provides, "unfencing", pcmk__str_casei) || pcmk__str_eq(requires, "unfencing", pcmk__str_casei)) { matches = g_list_prepend(matches, candidate); } } return matches; } static int node_priority_fencing_delay(pe_node_t * node, pe_working_set_t * data_set) { int member_count = 0; int online_count = 0; int top_priority = 0; int lowest_priority = 0; GList *gIter = NULL; // `priority-fencing-delay` is disabled if (data_set->priority_fencing_delay <= 0) { return 0; } /* No need to request a delay if the fencing target is not a normal cluster * member, for example if it's a remote node or a guest node. */ if (node->details->type != node_member) { return 0; } // No need to request a delay if the fencing target is in our partition if (node->details->online) { return 0; } for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { pe_node_t *n = gIter->data; if (n->details->type != node_member) { continue; } member_count ++; if (n->details->online) { online_count++; } if (member_count == 1 || n->details->priority > top_priority) { top_priority = n->details->priority; } if (member_count == 1 || n->details->priority < lowest_priority) { lowest_priority = n->details->priority; } } // No need to delay if we have more than half of the cluster members if (online_count > member_count / 2) { return 0; } /* All the nodes have equal priority. * Any configured corresponding `pcmk_delay_base/max` will be applied. */ if (lowest_priority == top_priority) { return 0; } if (node->details->priority < top_priority) { return 0; } return data_set->priority_fencing_delay; } pe_action_t * pe_fence_op(pe_node_t * node, const char *op, bool optional, const char *reason, bool priority_delay, pe_working_set_t * data_set) { char *op_key = NULL; pe_action_t *stonith_op = NULL; if(op == NULL) { op = data_set->stonith_action; } op_key = crm_strdup_printf("%s-%s-%s", CRM_OP_FENCE, node->details->uname, op); stonith_op = lookup_singleton(data_set, op_key); if(stonith_op == NULL) { stonith_op = custom_action(NULL, op_key, CRM_OP_FENCE, node, TRUE, TRUE, data_set); add_hash_param(stonith_op->meta, XML_LRM_ATTR_TARGET, node->details->uname); add_hash_param(stonith_op->meta, XML_LRM_ATTR_TARGET_UUID, node->details->id); add_hash_param(stonith_op->meta, "stonith_action", op); if (pe__is_guest_or_remote_node(node) && pcmk_is_set(data_set->flags, pe_flag_enable_unfencing)) { /* Extra work to detect device changes on remotes * * We may do this for all nodes in the future, but for now * the pcmk__check_action_config() based stuff works fine. */ long max = 1024; long digests_all_offset = 0; long digests_secure_offset = 0; char *digests_all = calloc(max, sizeof(char)); char *digests_secure = calloc(max, sizeof(char)); GList *matches = find_unfencing_devices(data_set->resources, NULL); for (GList *gIter = matches; gIter != NULL; gIter = gIter->next) { pe_resource_t *match = gIter->data; const char *agent = g_hash_table_lookup(match->meta, XML_ATTR_TYPE); op_digest_cache_t *data = NULL; data = pe__compare_fencing_digest(match, agent, node, data_set); if(data->rc == RSC_DIGEST_ALL) { optional = FALSE; crm_notice("Unfencing %s (remote): because the definition of %s changed", node->details->uname, match->id); if (!pcmk__is_daemon && data_set->priv != NULL) { pcmk__output_t *out = data_set->priv; out->info(out, "notice: Unfencing %s (remote): because the definition of %s changed", node->details->uname, match->id); } } digests_all_offset += snprintf( digests_all+digests_all_offset, max-digests_all_offset, "%s:%s:%s,", match->id, agent, data->digest_all_calc); digests_secure_offset += snprintf( digests_secure+digests_secure_offset, max-digests_secure_offset, "%s:%s:%s,", match->id, agent, data->digest_secure_calc); } g_hash_table_insert(stonith_op->meta, strdup(XML_OP_ATTR_DIGESTS_ALL), digests_all); g_hash_table_insert(stonith_op->meta, strdup(XML_OP_ATTR_DIGESTS_SECURE), digests_secure); } } else { free(op_key); } if (data_set->priority_fencing_delay > 0 /* It's a suitable case where `priority-fencing-delay` applies. * At least add `priority-fencing-delay` field as an indicator. */ && (priority_delay /* The priority delay needs to be recalculated if this function has * been called by schedule_fencing_and_shutdowns() after node * priority has already been calculated by native_add_running(). */ || g_hash_table_lookup(stonith_op->meta, XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY) != NULL)) { /* Add `priority-fencing-delay` to the fencing op even if it's 0 for * the targeting node. So that it takes precedence over any possible * `pcmk_delay_base/max`. */ char *delay_s = pcmk__itoa(node_priority_fencing_delay(node, data_set)); g_hash_table_insert(stonith_op->meta, strdup(XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY), delay_s); } if(optional == FALSE && pe_can_fence(data_set, node)) { pe__clear_action_flags(stonith_op, pe_action_optional); pe_action_set_reason(stonith_op, reason, false); } else if(reason && stonith_op->reason == NULL) { stonith_op->reason = strdup(reason); } return stonith_op; } void trigger_unfencing( pe_resource_t * rsc, pe_node_t *node, const char *reason, pe_action_t *dependency, pe_working_set_t * data_set) { if (!pcmk_is_set(data_set->flags, pe_flag_enable_unfencing)) { /* No resources require it */ return; } else if ((rsc != NULL) && !pcmk_is_set(rsc->flags, pe_rsc_fence_device)) { /* Wasn't a stonith device */ return; } else if(node && node->details->online && node->details->unclean == FALSE && node->details->shutdown == FALSE) { pe_action_t *unfence = pe_fence_op(node, "on", FALSE, reason, FALSE, data_set); if(dependency) { order_actions(unfence, dependency, pe_order_optional); } } else if(rsc) { GHashTableIter iter; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if(node->details->online && node->details->unclean == FALSE && node->details->shutdown == FALSE) { trigger_unfencing(rsc, node, reason, dependency, data_set); } } } } gboolean add_tag_ref(GHashTable * tags, const char * tag_name, const char * obj_ref) { pe_tag_t *tag = NULL; GList *gIter = NULL; gboolean is_existing = FALSE; CRM_CHECK(tags && tag_name && obj_ref, return FALSE); tag = g_hash_table_lookup(tags, tag_name); if (tag == NULL) { tag = calloc(1, sizeof(pe_tag_t)); if (tag == NULL) { return FALSE; } tag->id = strdup(tag_name); tag->refs = NULL; g_hash_table_insert(tags, strdup(tag_name), tag); } for (gIter = tag->refs; gIter != NULL; gIter = gIter->next) { const char *existing_ref = (const char *) gIter->data; if (pcmk__str_eq(existing_ref, obj_ref, pcmk__str_none)){ is_existing = TRUE; break; } } if (is_existing == FALSE) { tag->refs = g_list_append(tag->refs, strdup(obj_ref)); crm_trace("Added: tag=%s ref=%s", tag->id, obj_ref); } return TRUE; } /*! * \internal * \brief Create an action reason string based on the action itself * * \param[in] action Action to create reason string for * \param[in] flag Action flag that was cleared * * \return Newly allocated string suitable for use as action reason * \note It is the caller's responsibility to free() the result. */ char * pe__action2reason(pe_action_t *action, enum pe_action_flags flag) { const char *change = NULL; switch (flag) { case pe_action_runnable: case pe_action_migrate_runnable: change = "unrunnable"; break; case pe_action_optional: change = "required"; break; default: // Bug: caller passed unsupported flag CRM_CHECK(change != NULL, change = ""); break; } return crm_strdup_printf("%s%s%s %s", change, (action->rsc == NULL)? "" : " ", (action->rsc == NULL)? "" : action->rsc->id, action->task); } void pe_action_set_reason(pe_action_t *action, const char *reason, bool overwrite) { if (action->reason != NULL && overwrite) { pe_rsc_trace(action->rsc, "Changing %s reason from '%s' to '%s'", action->uuid, action->reason, crm_str(reason)); } else if (action->reason == NULL) { pe_rsc_trace(action->rsc, "Set %s reason to '%s'", action->uuid, crm_str(reason)); } else { // crm_assert(action->reason != NULL && !overwrite); return; } pcmk__str_update(&action->reason, reason); } /*! * \internal * \brief Check whether shutdown has been requested for a node * * \param[in] node Node to check * * \return TRUE if node has shutdown attribute set and nonzero, FALSE otherwise * \note This differs from simply using node->details->shutdown in that it can * be used before that has been determined (and in fact to determine it), * and it can also be used to distinguish requested shutdown from implicit * shutdown of remote nodes by virtue of their connection stopping. */ bool pe__shutdown_requested(pe_node_t *node) { const char *shutdown = pe_node_attribute_raw(node, XML_CIB_ATTR_SHUTDOWN); return !pcmk__str_eq(shutdown, "0", pcmk__str_null_matches); } /*! * \internal * \brief Update a data set's "recheck by" time * * \param[in] recheck Epoch time when recheck should happen * \param[in,out] data_set Current working set */ void pe__update_recheck_time(time_t recheck, pe_working_set_t *data_set) { if ((recheck > get_effective_time(data_set)) && ((data_set->recheck_by == 0) || (data_set->recheck_by > recheck))) { data_set->recheck_by = recheck; } } /*! * \internal * \brief Wrapper for pe_unpack_nvpairs() using a cluster working set */ void pe__unpack_dataset_nvpairs(xmlNode *xml_obj, const char *set_name, pe_rule_eval_data_t *rule_data, GHashTable *hash, const char *always_first, gboolean overwrite, pe_working_set_t *data_set) { crm_time_t *next_change = crm_time_new_undefined(); pe_eval_nvpairs(data_set->input, xml_obj, set_name, rule_data, hash, always_first, overwrite, next_change); if (crm_time_is_defined(next_change)) { time_t recheck = (time_t) crm_time_get_seconds_since_epoch(next_change); pe__update_recheck_time(recheck, data_set); } crm_time_free(next_change); } bool pe__resource_is_disabled(pe_resource_t *rsc) { const char *target_role = NULL; CRM_CHECK(rsc != NULL, return false); target_role = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE); if (target_role) { enum rsc_role_e target_role_e = text2role(target_role); if ((target_role_e == RSC_ROLE_STOPPED) || ((target_role_e == RSC_ROLE_UNPROMOTED) && pcmk_is_set(uber_parent(rsc)->flags, pe_rsc_promotable))) { return true; } } return false; } /*! * \internal * \brief Create an action to clear a resource's history from CIB * * \param[in] rsc Resource to clear * \param[in] node Node to clear history on * * \return New action to clear resource history */ pe_action_t * pe__clear_resource_history(pe_resource_t *rsc, pe_node_t *node, pe_working_set_t *data_set) { char *key = NULL; CRM_ASSERT(rsc && node); key = pcmk__op_key(rsc->id, CRM_OP_LRM_DELETE, 0); return custom_action(rsc, key, CRM_OP_LRM_DELETE, node, FALSE, TRUE, data_set); } bool pe__rsc_running_on_any(pe_resource_t *rsc, GList *node_list) { for (GList *ele = rsc->running_on; ele; ele = ele->next) { pe_node_t *node = (pe_node_t *) ele->data; if (pcmk__str_in_list(node->details->uname, node_list, pcmk__str_star_matches|pcmk__str_casei)) { return true; } } return false; } bool pcmk__rsc_filtered_by_node(pe_resource_t *rsc, GList *only_node) { return (rsc->fns->active(rsc, FALSE) && !pe__rsc_running_on_any(rsc, only_node)); } GList * pe__filter_rsc_list(GList *rscs, GList *filter) { GList *retval = NULL; for (GList *gIter = rscs; gIter; gIter = gIter->next) { pe_resource_t *rsc = (pe_resource_t *) gIter->data; /* I think the second condition is safe here for all callers of this * function. If not, it needs to move into pe__node_text. */ if (pcmk__str_in_list(rsc_printable_id(rsc), filter, pcmk__str_star_matches) || (rsc->parent && pcmk__str_in_list(rsc_printable_id(rsc->parent), filter, pcmk__str_star_matches))) { retval = g_list_prepend(retval, rsc); } } return retval; } GList * pe__build_node_name_list(pe_working_set_t *data_set, const char *s) { GList *nodes = NULL; if (pcmk__str_eq(s, "*", pcmk__str_null_matches)) { /* Nothing was given so return a list of all node names. Or, '*' was * given. This would normally fall into the pe__unames_with_tag branch * where it will return an empty list. Catch it here instead. */ nodes = g_list_prepend(nodes, strdup("*")); } else { pe_node_t *node = pe_find_node(data_set->nodes, s); if (node) { /* The given string was a valid uname for a node. Return a * singleton list containing just that uname. */ nodes = g_list_prepend(nodes, strdup(s)); } else { /* The given string was not a valid uname. It's either a tag or * it's a typo or something. In the first case, we'll return a * list of all the unames of the nodes with the given tag. In the * second case, we'll return a NULL pointer and nothing will * get displayed. */ nodes = pe__unames_with_tag(data_set, s); } } return nodes; } GList * pe__build_rsc_list(pe_working_set_t *data_set, const char *s) { GList *resources = NULL; if (pcmk__str_eq(s, "*", pcmk__str_null_matches)) { resources = g_list_prepend(resources, strdup("*")); } else { pe_resource_t *rsc = pe_find_resource_with_flags(data_set->resources, s, pe_find_renamed|pe_find_any); if (rsc) { /* A colon in the name we were given means we're being asked to filter * on a specific instance of a cloned resource. Put that exact string * into the filter list. Otherwise, use the printable ID of whatever * resource was found that matches what was asked for. */ if (strstr(s, ":") != NULL) { resources = g_list_prepend(resources, strdup(rsc->id)); } else { resources = g_list_prepend(resources, strdup(rsc_printable_id(rsc))); } } else { /* The given string was not a valid resource name. It's either * a tag or it's a typo or something. See build_uname_list for * more detail. */ resources = pe__rscs_with_tag(data_set, s); } } return resources; } xmlNode * pe__failed_probe_for_rsc(pe_resource_t *rsc, const char *name) { pe_resource_t *parent = uber_parent(rsc); const char *rsc_id = rsc->id; if (rsc->variant == pe_clone) { rsc_id = pe__clone_child_id(rsc); } else if (parent->variant == pe_clone) { rsc_id = pe__clone_child_id(parent); } for (xmlNode *xml_op = pcmk__xml_first_child(rsc->cluster->failed); xml_op != NULL; xml_op = pcmk__xml_next(xml_op)) { const char *value = NULL; char *op_id = NULL; /* This resource operation is not a failed probe. */ if (!pcmk_xe_mask_probe_failure(xml_op)) { continue; } /* This resource operation was not run on the given node. Note that if name is * NULL, this will always succeed. */ value = crm_element_value(xml_op, XML_LRM_ATTR_TARGET); if (value == NULL || !pcmk__str_eq(value, name, pcmk__str_casei|pcmk__str_null_matches)) { continue; } /* This resource operation has no operation_key. */ value = crm_element_value(xml_op, XML_LRM_ATTR_TASK_KEY); if (!parse_op_key(value ? value : ID(xml_op), &op_id, NULL, NULL)) { continue; } /* This resource operation's ID does not match the rsc_id we are looking for. */ if (!pcmk__str_eq(op_id, rsc_id, pcmk__str_none)) { free(op_id); continue; } free(op_id); return xml_op; } return NULL; }