[
https://issues.apache.org/jira/browse/TIKA-4844?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Tim Allison updated TIKA-4844:
------------------------------
Description: In shared server mode, a client whose in-flight parse was
killed by another client's restart files its crash report after that restart
completes – ensureRunning holds the lock across the whole fork, so siblings
cannot report any earlier. pendingRestart carries no process identity, so the
report is applied to the healthy replacement, which is then destroyed and
re-forked. One process death produces two restarts and a second round of
destroyed in-flight work. Under sustained concurrent load it sustains itself at
one spurious restart per round. Affects 4.0.0 and earlier; shared mode only,
which is not the default.
> Shared pipes server: stale crash reports trigger spurious restarts of a
> healthy worker
> --------------------------------------------------------------------------------------
>
> Key: TIKA-4844
> URL: https://issues.apache.org/jira/browse/TIKA-4844
> Project: Tika
> Issue Type: Task
> Reporter: Tim Allison
> Priority: Major
>
> In shared server mode, a client whose in-flight parse was killed by another
> client's restart files its crash report after that restart completes –
> ensureRunning holds the lock across the whole fork, so siblings cannot report
> any earlier. pendingRestart carries no process identity, so the report is
> applied to the healthy replacement, which is then destroyed and re-forked.
> One process death produces two restarts and a second round of destroyed
> in-flight work. Under sustained concurrent load it sustains itself at one
> spurious restart per round. Affects 4.0.0 and earlier; shared mode only,
> which is not the default.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
