[
https://issues.apache.org/jira/browse/HBASE-20481?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16460966#comment-16460966
]
Duo Zhang commented on HBASE-20481:
-----------------------------------
Sorry, just notice that when calling ReplicationEndpoint.init, we pass a
Context parameter, which contains a ReplicationPeer field, so the current API
is enough to tell the ReplicationEndpoint whether the peer is serial or not...
And for your solution, I think it is too strict. You make everything serial,
but actually, we only need the serial guarantee inside a region, for different
regions we can replicate them parallelly...
> Add a setSerial method to ReplicationEndpoint
> ---------------------------------------------
>
> Key: HBASE-20481
> URL: https://issues.apache.org/jira/browse/HBASE-20481
> Project: HBase
> Issue Type: Sub-task
> Reporter: Duo Zhang
> Assignee: Zheng Hu
> Priority: Major
> Attachments: HBASE-20481.v1.patch
>
>
> When debugging HBASE-20475, [~openinx] found that the
> HBaseInterClusterReplicationEndpoint may send the entries for the same
> regions concurrently, which breaks the serial replication.
> As long as we can have multiple ReplicationEndpoint implementation, just fix
> HBaseInterClusterReplicationEndpoint is not enough, we need to add a
> setSerial method to ReplicationEndpoint, to tell the implementation that you
> should keep the order of the entries from the same region.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
