[CASSANDRA-16585] Periodic failures in *RepairCoordinator*Test caused by race condition with nodetool repair

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Fixed
• Fix Version/s: 4.0-rc1, 4.0
• Component/s: CI, Consistency/Repair, Test/dtest/java
• Bug Category: Correctness - Transient Incorrect Response
• Severity: Normal
• Complexity: Normal
• Discovered By: Unit Test
• Platform: All
• Since Version: 4.0-alpha1

Source Control Link

https://github.com/apache/cassandra/commit/5cbdb2e58e870535af61204898a1e2bbf6cb5f64

Test and Documentation Plan

added tests

• Assignee: David Capwell
• Reporter: David Capwell

Authors

David Capwell

Reviewers

Marcus Eriksson

• Created: 09/Apr/21 22:31
• Updated: 16/Mar/22 15:41
• Resolved: 13/Apr/21 18:48

Description

Periodic failures in *RepairCoordinator*Test cause errors such as

FullRepairCoordinatorNeighbourDownTest#validationParticipentCrashesAndComesBack[DATACENTER_AWARE/true]

nodetool command [repair, distributed_test_keyspace, validationparticipentcrashesandcomesback_full_datacenter_aware_true, --dc-parallel, --full] Error message 'Some repair failed' does not contain any of [/127.0.0.2:7012 died]
stdout:
[2021-04-07 22:45:24,887] Starting repair command #10 (f129cb60-97f2-11eb-9316-794aa6ab8411), repairing keyspace distributed_test_keyspace with repair options (parallelism: dc_parallel, primary range: false, incremental: false, job threads: 1, ColumnFamilies: [validationparticipentcrashesandcomesback_full_datacenter_aware_true], dataCenters: [], hosts: [], previewKind: NONE, # of ranges: 2, pull repair: false, force repair: false, optimise streams: false, ignore unreplicated keyspaces: false)
[2021-04-07 22:45:32,864] Repair command #10 failed with error Repair session f1342ba0-97f2-11eb-9316-794aa6ab8411 for range [(-1,9223372036854775805], (9223372036854775805,-1]] failed with error Endpoint /127.0.0.2:7012 died
[2021-04-07 22:45:32,887] After waiting for poll interval of 1 seconds queried for parent session status and discovered repair failed.
[2021-04-07 22:45:32,887] Repair command #10 finished with error
[2021-04-07 22:45:32,887] Some repair failed
[2021-04-07 22:45:32,888] Repair command #10 finished with error

stderr:
error: Some repair failed
-- StackTrace --
java.io.IOException: Some repair failed
at org.apache.cassandra.tools.RepairRunner.queryForCompletedRepair(RepairRunner.java:167)
at org.apache.cassandra.tools.RepairRunner.run(RepairRunner.java:72)
at org.apache.cassandra.tools.NodeProbe.repairAsync(NodeProbe.java:431)
at org.apache.cassandra.tools.nodetool.Repair.execute(Repair.java:171)
at org.apache.cassandra.tools.NodeTool$NodeToolCmd.runInternal(NodeTool.java:358)
at org.apache.cassandra.tools.NodeTool$NodeToolCmd.run(NodeTool.java:343)
at org.apache.cassandra.tools.NodeTool.execute(NodeTool.java:246)
at org.apache.cassandra.distributed.impl.Instance$DTestNodeTool.execute(Instance.java:836)
at org.apache.cassandra.distributed.impl.Instance.lambda$nodetoolResult$38(Instance.java:746)
at java.base/java.util.concurrent.FutureTask.run(FutureTask.java:264)
at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128)
at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628)
at io.netty.util.concurrent.FastThreadLocalRunnable.run(FastThreadLocalRunnable.java:30)
at java.base/java.lang.Thread.run(Thread.java:834)


Notifications:
Notification{type=START, src=repair:10, message=Starting repair command #10 (f129cb60-97f2-11eb-9316-794aa6ab8411), repairing keyspace distributed_test_keyspace with repair options (parallelism: dc_parallel, primary range: false, incremental: false, job threads: 1, ColumnFamilies: [validationparticipentcrashesandcomesback_full_datacenter_aware_true], dataCenters: [], hosts: [], previewKind: NONE, # of ranges: 2, pull repair: false, force repair: false, optimise streams: false, ignore unreplicated keyspaces: false)}
Notification{type=ERROR, src=repair:10, message=Repair command #10 failed with error Repair session f1342ba0-97f2-11eb-9316-794aa6ab8411 for range [(-1,9223372036854775805], (9223372036854775805,-1]] failed with error Endpoint /127.0.0.2:7012 died}
Notification{type=COMPLETE, src=repair:10, message=Repair command #10 finished with error}
Error:
java.io.IOException: Some repair failed
at org.apache.cassandra.tools.RepairRunner.queryForCompletedRepair(RepairRunner.java:167)
at org.apache.cassandra.tools.RepairRunner.run(RepairRunner.java:72)
at org.apache.cassandra.tools.NodeProbe.repairAsync(NodeProbe.java:431)
at org.apache.cassandra.tools.nodetool.Repair.execute(Repair.java:171)
at org.apache.cassandra.tools.NodeTool$NodeToolCmd.runInternal(NodeTool.java:358)
at org.apache.cassandra.tools.NodeTool$NodeToolCmd.run(NodeTool.java:343)
at org.apache.cassandra.tools.NodeTool.execute(NodeTool.java:246)
at org.apache.cassandra.distributed.impl.Instance$DTestNodeTool.execute(Instance.java:836)
at org.apache.cassandra.distributed.impl.Instance.lambda$nodetoolResult$38(Instance.java:746)
at java.base/java.util.concurrent.FutureTask.run(FutureTask.java:264)
at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128)
at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628)
at io.netty.util.concurrent.FastThreadLocalRunnable.run(FastThreadLocalRunnable.java:30)
at java.base/java.lang.Thread.run(Thread.java:834)

Seems there is a race condition in nodetool repair where we query the error state before we get the notification, then we throw a generic error rather than the specific error.

Issue Links

• links to: Circle CI: trunk
• links to: GH PR: trunk