[
https://issues.apache.org/jira/browse/HBASE-20965?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16569054#comment-16569054
]
Yu Li commented on HBASE-20965:
-------------------------------
bq. have you encountered similar issue, as you have a thousands size cluster
We haven't experienced the same issue in our cluster, but I believe it's a real
world issue observed in Xiaomi's production environment. "The jstack shows that
most of handlers response for RSReport" is the evidence :-)
bq. Does it has negative effects on dealing with other rpcs requests?
I think this is a good question and I'd suggest to use {{StealJobQueue}} to
better utilize handlers. [~Yi Mei] [~zghaobac]
> Separate region server report requests to new handlers
> ------------------------------------------------------
>
> Key: HBASE-20965
> URL: https://issues.apache.org/jira/browse/HBASE-20965
> Project: HBase
> Issue Type: Improvement
> Components: Performance
> Reporter: Yi Mei
> Assignee: Yi Mei
> Priority: Major
> Fix For: 3.0.0, 2.2.0
>
> Attachments: HBASE-20965.master.001.patch,
> HBASE-20965.master.002.patch, HBASE-20965.master.003.patch,
> HBASE-20965.master.004.patch
>
>
> In master rpc scheduler, all rpc requests are executed in a thread pool. This
> task separates rs report requests to new handlers.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
