Fixed
Pinned fields
Click on the next to a field label to start pinning.
Details
Assignee
Ankush SharmaAnkush SharmaReporter
Ankush SharmaAnkush SharmaIs this a Regression?
UnknownTriage
TriagedDue date
Jun 30, 2023Story Points
0Sprint
NonePriority
CriticalInstabug
Open Instabug
Details
Details
Assignee
Ankush Sharma
Ankush SharmaReporter
Ankush Sharma
Ankush SharmaIs this a Regression?
Unknown
Triage
Triaged
Due date
Jun 30, 2023
Story Points
0
Sprint
None
Priority
Instabug
Open Instabug
PagerDuty
PagerDuty
PagerDuty
Sentry
Sentry
Sentry
Zendesk Support
Zendesk Support
Zendesk Support
Created April 27, 2023 at 3:29 AM
Updated October 11, 2024 at 7:38 AM
Resolved August 4, 2023 at 2:28 PM
Steps To Recreate:
Create a 4 node cluster
Create a magma bucket with (bucket_history_retention_seconds=600,bucket_history_retention_bytes=6000000000)
Create 5000000 items(doc size = 256)
Start new doc ops(update:expiry)
Trigger compaction
SIGKILL memcached once
Observed Memcached crashed in CheckpointManager::expelUnreferencedCheckpointItems (this=0x7f6bcc52de40)
Note:
Though actual test is about crash recovery .Basically keep killing memcached while data loading is going on and between two sigkill test waits for cluster warmup to finish and after warmup finishes test waits for 30 to 60 before next iteration of memcached kill, so total time between two sigkills is = warmup_time+30/60 seconds) , but in the case the crash was observed after first kill itself(since crash was observed memcached was killed just once)
Core Dump was found on node 172.23.121.115
BackTrace:
QE-TEST:
Job: http://qe-jenkins1.sc.couchbase.com/job/test_suite_executor-TAF/24359/consoleFull
Issue
Resolution
In rare cases, after a failover or memcached restart, a replica rollback while under memory pressure might have caused a crash in the Data Service.
Memory pressure recovery logic (Item expelling) is now skipped when replica rollback is in progress.