[jira] [Commented] (HBASE-21539) Should add backoff when replaying failed in SyncReplicationReplayWALProcedure

["Duo Zhang (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HBASE-21539?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16705764#comment-16705764 ] 

Duo Zhang commented on HBASE-21539:
-----------------------------------

[~zghaobac] FYI.

> Should add backoff when replaying failed in SyncReplicationReplayWALProcedure
> -----------------------------------------------------------------------------
>
>                 Key: HBASE-21539
>                 URL: https://issues.apache.org/jira/browse/HBASE-21539
>             Project: HBase
>          Issue Type: Sub-task
>            Reporter: Duo Zhang
>            Priority: Major
>
> I'm still testing serial&sync replication and it is stuck again...
> Still need to find out the root cause but there is another problem, since the replication is stuck, we have lots of wals to replay, and cause too much pressure on the memstore and the region rejects the write requests so the SyncReplicationReplayWALRemoteProcedure fails. But soon we will schedule a new SyncReplicationReplayWALRemoteProcedure without any sleeps, which means we are keep adding pressure to the memstore. The result is very clear, we can not finish the replay, and write too much duplicated data to the region, and can not recover any more...



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)