[
https://issues.apache.org/jira/browse/SLING-4981?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Julian Sedding resolved SLING-4981.
-----------------------------------
Resolution: Fixed
Assignee: Julian Sedding
Fix Version/s: Commons Scheduler 2.4.12
Fixed in [r1701254|http://svn.apache.org/r1701254] by removing the
{{synchronized}} keyword from all callbacks of the {{ServiceTrackerCustomizer}}.
In local tests I could re-create the deadlock from the attached dump by
repeatedly restarting a bundle that registered several scheduled tasks on
startup and subsequently unregistered some of them again.
With my changes I was not able to reproduce.
> Deadlock involving Scheduler Whiteboard
> ---------------------------------------
>
> Key: SLING-4981
> URL: https://issues.apache.org/jira/browse/SLING-4981
> Project: Sling
> Issue Type: Bug
> Components: Commons
> Affects Versions: Commons Scheduler 2.4.6
> Reporter: Julian Sedding
> Assignee: Julian Sedding
> Fix For: Commons Scheduler 2.4.12
>
> Attachments: scheduler-scr-deadlock.log
>
>
> I repeatedly observed a deadlock situation in an AEM 6.1 installation
> (involving custom code).
> The result, amongst others, is a blocked SCR Actor thread.
> Hypothesis: I suspect that this can occur in situations where a scheduled Job
> (A) has a reference to another SCR component (B) that is lazily initialized
> (immediate=false) and itself registered via the whiteboard. If A is disabled
> while B is being added by the Scheduler, this triggers deregistration of B as
> well.
> [~cziegeler] could you please take a look at the attached thread-dump?
> Several subsequent thread-dumps showed no progress. I would appreciate your
> opinion on this.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
