[
https://issues.apache.org/jira/browse/PHOENIX-3970?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16060240#comment-16060240
]
James Taylor commented on PHOENIX-3970:
---------------------------------------
I think there are two separate issues identified here and we're likely all in
agreement:
bq. IMHO we should not use the regular priority in RS-RS communication for any
of our tasks since it always may cause dead lock in some circumstances.
Agreed. RS-RS calls should all be high priority. This is what prevents
deadlocks.
bq. The right(tm) solution is to rebuild indexes asynchronously (either by
returning immediately from the scans after scheduling the work, or my using the
offline index build tools, which since 2.10 can also do partial rebuilds).
Agreed. How about we invoke IndexTool from MetaDataRegionObserver instead of
doing partial rebuild online?
> Ensure that automatic partial index rebuilds are served from the index
> handler pool
> -----------------------------------------------------------------------------------
>
> Key: PHOENIX-3970
> URL: https://issues.apache.org/jira/browse/PHOENIX-3970
> Project: Phoenix
> Issue Type: Bug
> Reporter: Lars Hofhansl
> Assignee: Lars Hofhansl
> Attachments: 3970.txt, 3970-v2.txt
>
>
> This (and other issues) have rendered multiple larger cluster inoperable.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
