[
https://issues.apache.org/jira/browse/SLING-11662?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17628427#comment-17628427
]
Stefan Egli commented on SLING-11662:
-------------------------------------
The problem with the endless loop seems to be due to a breach of contract by
sling.commons.scheduler.QuartzThreadPool:
* [quartz' ThreadPool
javadoc|https://github.com/quartz-scheduler/quartz/blob/v2.3.2/quartz-core/src/main/java/org/quartz/spi/ThreadPool.java#L69-L82]
says that
{quote}The implementation of this method should block until there is at least
one available thread.{quote}
* however
[sling.commons.scheduler.QuartzThreadPool#blockForAvailableThreads|https://github.com/apache/sling-org-apache-sling-commons-scheduler/blob/a9ddf38ea9d9962c8938a381135827072fc9397f/src/main/java/org/apache/sling/commons/scheduler/impl/QuartzThreadPool.java#L80]
does not guarantee {{>0}} - and in particular if "maxPoolSize == queueSize"
then this method will return 0
* that in turn leads quartz to hit the [ironically commented
line|https://github.com/quartz-scheduler/quartz/blob/v2.3.2/quartz-core/src/main/java/org/quartz/core/QuartzSchedulerThread.java#L411-L414]
{code}
} else { // if(availThreadCount > 0)
// should never happen, if
threadPool.blockForAvailableThreads() follows contract
continue; // while (!halted)
}
{code}
so it will just .. continue
* now this game repeats for ever after until .. the CPU becomes too hot due to
constant 100% spinning and it breaks down .. leading to damage in a datacenter
and so on and so forth
Presumably quartz did nothing wrong here - other than perhaps add a
safety/paranoia {{Thread.sleep(1);}} before that {{continue}} to avoid this.
The problem is rather on the sling side.
Question is how to best fix this.. [~cziegeler], [~joerghoh], any suggestions?
> Endless loop in QuartzSchedulerThread.run() with maxPoolSize == queueSize
> -------------------------------------------------------------------------
>
> Key: SLING-11662
> URL: https://issues.apache.org/jira/browse/SLING-11662
> Project: Sling
> Issue Type: Bug
> Components: Commons
> Affects Versions: Commons Scheduler 2.7.12
> Reporter: Stefan Egli
> Priority: Major
> Time Spent: 20m
> Remaining Estimate: 0h
>
> When configuring the ThreadPool with maxPoolSize == queueSize and endless
> loop (can) happen(s) in QuartzSchedulerThread.run() which manifests as
> follows:
> {noformat}
> "MyPool_QuartzSchedulerThread" #123 prio=5 os_prio=0 cpu=5123456.78ms
> elapsed=5163.45s tid=0x000012345678ff00 nid=0x1234 runnable
> [0x000087654321ff00]
> java.lang.Thread.State: RUNNABLE
> at
> org.quartz.core.QuartzSchedulerThread.run(QuartzSchedulerThread.java:413)
> {noformat}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
