[
https://issues.apache.org/jira/browse/HBASE-21789?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16753229#comment-16753229
]
Duo Zhang commented on HBASE-21789:
-----------------------------------
{quote}
Is there retry under coprocessorService?
{quote}
We have retries for coprocessor call, you can see the
RegionCoprocessorRpcChannel. And there maybe a little problem, that the timeout
config is inherited from Connection, not Table, but this is the problem of our
coprocessor implemention.
{quote}
Should we file another issue to make multi row native rather than CP based?
{quote}
We could but the API needs to be carefully designed. It should only be used by
experts. We have introduced a simple method to do this in our thin hbase
client, and users always put a bunch rows across multiple regions and then ask
why only some of the puts are succeeded...
> Rewrite MetaTableAccessor.multiMutate with Table.coprocessorService
> -------------------------------------------------------------------
>
> Key: HBASE-21789
> URL: https://issues.apache.org/jira/browse/HBASE-21789
> Project: HBase
> Issue Type: Improvement
> Components: Client, Coprocessors
> Reporter: Duo Zhang
> Assignee: Duo Zhang
> Priority: Major
> Fix For: 3.0.0, 2.2.0
>
> Attachments: HBASE-21789-v1.patch, HBASE-21789.patch
>
>
> Now it implements a special RegionServerCallable, which is not necessary. It
> is just a coprocessor call to MultiRowMutationEndpoint, we should use the
> coprocessor related method to do this.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
