[
https://issues.apache.org/jira/browse/HBASE-13366?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14388732#comment-14388732
]
Elliott Clark commented on HBASE-13366:
---------------------------------------
When are you seeing requests coming to a read only replica?
* If the requests are coming while a region is getting opened then we'll want
the exceptions to signal wait some time. In fact here it's really really
important not to send a to no retry exception as the read replicas are the best
place to open a failed region. They already have lots of the blocks in cache.
* If the requests are going to a read only replica just normally then there's a
bug and we should fix that.
> Throw DoNotRetryIOException instead of read only IOException
> ------------------------------------------------------------
>
> Key: HBASE-13366
> URL: https://issues.apache.org/jira/browse/HBASE-13366
> Project: HBase
> Issue Type: Improvement
> Reporter: Liu Shaohui
> Assignee: Liu Shaohui
> Priority: Minor
> Fix For: 2.0.0
>
> Attachments: HBASE-13366-v1.diff
>
>
> Currently, the read only region just throws an IOException to the clients who
> send write requests to it. This will cause the clients retry for configured
> times or until operation timeout.
> Changing this exception to DoNotRetryIOException will make the client failed
> fast.
> Suggestions are welcomed~ Thanks
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
