[
https://issues.apache.org/jira/browse/HIVE-25633?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17432417#comment-17432417
]
Eugene Chung commented on HIVE-25633:
-------------------------------------
I missed that in HMSHandler#init, ThreadPool#initialize is called in the
_alwaysThreadsInitialized.compareAndSet(false, true)_ block.
If you think ThreadPool#shutdown is called in HMSHandler#shutdown,
_alwaysThreadsInitialized.set(false)_ should be called.
Or I think ThreadPool#shutdown could be omitted because ThreadPool is created
with daemon flag. It won't block the JVM termination.
> Prevent shutdown of MetaStore scheduled worker ThreadPool
> ---------------------------------------------------------
>
> Key: HIVE-25633
> URL: https://issues.apache.org/jira/browse/HIVE-25633
> Project: Hive
> Issue Type: Bug
> Reporter: Zoltan Haindrich
> Assignee: Zoltan Haindrich
> Priority: Major
> Labels: pull-request-available
> Time Spent: 10m
> Remaining Estimate: 0h
>
> [~lpinter] have noticed that this patch has some sideffect:
> in HIVE-23164 the patch have added a {{ThreadPool#shutdown}} to
> {{HMSHandler#shutdown}} - which could cause trouble in case a {{HMSHandler}}
> is shutdown and a new one is created
> I was looking for cases in which a HMSHandler is created inside the metastore
> (beyond the one HiveMetaStore is using) - and I think tasks like
> PartitionManagementTask which uses Msck which uses HMC to access the
> metastore - and they close the client - which closes the hmshandler client ;
> which will shut down the threadpool
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
