Sergey Soldatov created HDDS-16209:
--------------------------------------
Summary: KeyLifecycleService leaks inFlight entry, permanently
blocking the bucket
Key: HDDS-16209
URL: https://issues.apache.org/jira/browse/HDDS-16209
Project: Apache Ozone
Issue Type: Bug
Components: OM
Affects Versions: 2.3.0
Reporter: Sergey Soldatov
Assignee: Sergey Soldatov
When a KeyLifecycleService task is scheduled for a bucket, the bucket is marked
as "in flight" so it is not scheduled twice. The mark is only cleared when the
task actually runs its work, but if the task starts while the service cannot
run at that moment (the service was just suspended or disabled, or the OM is
not yet leader), the task exits early and the mark is never cleared. After that
the service considers the bucket permanently busy and skips it on every
scheduling round, so lifecycle rules for that bucket silently stop working
until the OM is restarted. This can happen in production after a leader change
or a suspend/resume of the service, and it also causes flaky timeouts in
TestKeyLifecycleService.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
