[
https://issues.apache.org/jira/browse/HBASE-19525?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16300887#comment-16300887
]
Zheng Hu commented on HBASE-19525:
----------------------------------
[~zghaobac], Seems like RB can not publish the comment after ship it. I
paste the comment here.
bq. Let ReplicationSource own a ReplicationPeerImpl
ReplicationSource should not own ReplicationPeerImpl, should use
ReplicationPeer interface. because there are diff implementations for
ReplicationPeer. Still need an extra method added in ReplicationPeer to read
peer state, such as getPeerState(loadFromBackingStore).
bq. The code may be
replicationSourceManager.getReplicationSource(peerId).refreshPeerState().
Actually the logic are the same, isn't it ？ One thread update the volatile
variable, and another thread read the latest value. I think you mean it's more
readable in that way.
> RS side changes for moving peer modification from zk watcher to procedure
> -------------------------------------------------------------------------
>
> Key: HBASE-19525
> URL: https://issues.apache.org/jira/browse/HBASE-19525
> Project: HBase
> Issue Type: Sub-task
> Components: proc-v2, Replication
> Reporter: Duo Zhang
> Assignee: Zheng Hu
> Fix For: HBASE-19397
>
> Attachments: HBASE-19525.HBASE-19397.v1.patch,
> HBASE-19525.HBASE-19397.v2.patch, HBASE-19525.HBASE-19397.v3.patch,
> HBASE-19525.HBASE-19397.v4.patch, HBASE-19525.HBASE-19397.v5.patch,
> HBASE-19525.HBASE-19397.v6.patch, HBASE-19525.HBASE-19397.v6.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
