[
https://issues.apache.org/jira/browse/IGNITE-26051?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18011901#comment-18011901
]
Kirill Tkalenko commented on IGNITE-26051:
------------------------------------------
Looks good.
> With enabled colocation, write intent switch might fail due to table
> destruction
> --------------------------------------------------------------------------------
>
> Key: IGNITE-26051
> URL: https://issues.apache.org/jira/browse/IGNITE-26051
> Project: Ignite
> Issue Type: Bug
> Reporter: Roman Puchkovskiy
> Assignee: Roman Puchkovskiy
> Priority: Major
> Labels: MakeTeamcityGreenAgain, ignite-3
> Time Spent: 20m
> Remaining Estimate: 0h
>
> ItPartitionDestructionTest#
> partitionIsDestroyedOnTableDestructionOnNodeRecoveryWithColocation()
> demonstrates the problem: it is enough to remove the muting annotation from
> the test and then run it. The test passes, but in the log the following
> appears:
>
> {noformat}
> [2025-07-28T15:30:28,532][ERROR][%ipdt_pidotdonrwc_3344%JRaft-FSMCaller-Disruptor_stripe_0-0][FailureManager]
> Critical system error detected. Will be handled accordingly to configured
> handler [hnd=NoOpFailureHandler [super=AbstractFailureHandler
> [ignoredFailureTypes=UnmodifiableSet [SYSTEM_WORKER_BLOCKED,
> SYSTEM_CRITICAL_OPERATION_TIMEOUT]]], failureCtx=CRITICAL_ERROR]
> org.apache.ignite.internal.failure.StackTraceCapturingException: Unknown error
> at
> org.apache.ignite.internal.failure.FailureManager.process(FailureManager.java:191)
> at
> org.apache.ignite.internal.failure.FailureManager.process(FailureManager.java:168)
> at
> org.apache.ignite.internal.raft.server.impl.JraftServerImpl$DelegatingStateMachine.onApply(JraftServerImpl.java:866)
> at
> org.apache.ignite.raft.jraft.core.FSMCallerImpl.doApplyTasks(FSMCallerImpl.java:578)
> at
> org.apache.ignite.raft.jraft.core.FSMCallerImpl.doCommitted(FSMCallerImpl.java:544)
> at
> org.apache.ignite.raft.jraft.core.FSMCallerImpl.runApplyTask(FSMCallerImpl.java:462)
> at
> org.apache.ignite.raft.jraft.core.FSMCallerImpl$ApplyTaskHandler.onEvent(FSMCallerImpl.java:131)
> at
> org.apache.ignite.raft.jraft.core.FSMCallerImpl$ApplyTaskHandler.onEvent(FSMCallerImpl.java:125)
> at
> org.apache.ignite.raft.jraft.disruptor.StripedDisruptor$StripeEntryHandler.onEvent(StripedDisruptor.java:330)
> at
> org.apache.ignite.raft.jraft.disruptor.StripedDisruptor$StripeEntryHandler.onEvent(StripedDisruptor.java:287)
> at
> com.lmax.disruptor.BatchEventProcessor.processEvents(BatchEventProcessor.java:167)
> at com.lmax.disruptor.BatchEventProcessor.run(BatchEventProcessor.java:122)
> at java.base/java.lang.Thread.run(Thread.java:833)
> Caused by: java.lang.AssertionError: No RAFT table processor found by table
> ID 21
> at
> org.apache.ignite.internal.partition.replicator.raft.handlers.WriteIntentSwitchCommandHandler.raftTableProcessor(WriteIntentSwitchCommandHandler.java:70)
> at
> org.apache.ignite.internal.partition.replicator.raft.handlers.WriteIntentSwitchCommandHandler.handleInternally(WriteIntentSwitchCommandHandler.java:58)
> at
> org.apache.ignite.internal.partition.replicator.raft.handlers.WriteIntentSwitchCommandHandler.handleInternally(WriteIntentSwitchCommandHandler.java:33)
> at
> org.apache.ignite.internal.partition.replicator.raft.handlers.AbstractCommandHandler.handle(AbstractCommandHandler.java:45)
> at
> org.apache.ignite.internal.partition.replicator.raft.ZonePartitionRaftListener.processWriteCommand(ZonePartitionRaftListener.java:228)
> at
> org.apache.ignite.internal.partition.replicator.raft.ZonePartitionRaftListener.lambda$onWrite$1(ZonePartitionRaftListener.java:162)
> at java.base/java.util.Iterator.forEachRemaining(Iterator.java:133)
> at
> org.apache.ignite.internal.partition.replicator.raft.ZonePartitionRaftListener.onWrite(ZonePartitionRaftListener.java:160)
> at
> org.apache.ignite.internal.raft.server.impl.JraftServerImpl$DelegatingStateMachine.onApply(JraftServerImpl.java:845)
> ... 10 more{noformat}
>
> There appears to a race between table raft processor removal (due to table
> destruction) and a write intent switch command application (that tries to
> invoke that processor).
>
> It seems that this can be easily fixed by ignoring tables for which there is
> no processor. On the one hand, this seems fine as the table is already
> destroyed together with all its data and not-yet-switched write intents; on
> the other hand, this will lead to partial application of a write intent, so
> it should be carefully considered if it IS ok to do so.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
