[
https://issues.apache.org/jira/browse/HBASE-3351?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jean-Daniel Cryans updated HBASE-3351:
--------------------------------------
Attachment: HBASE-3351_3326.patch
Attaching a patch that covers both this issue and HBASE-3326 (Replication
state's znode should be created else it defaults to false) because they are
very intertwined.
In this patch I fix HBASE-3326 simply by creating the znode and then I add the
better tracking of the state znode which had an impact in more user-side
classes since they had to provide an Abortable in order to be able to start the
tracker in any given situation (before the client wasn't tracking the state
znode). In the end, I think this is much better.
> ReplicationZookeeper goes to ZK every time a znode is modified
> --------------------------------------------------------------
>
> Key: HBASE-3351
> URL: https://issues.apache.org/jira/browse/HBASE-3351
> Project: HBase
> Issue Type: Bug
> Reporter: Jean-Daniel Cryans
> Assignee: Jean-Daniel Cryans
> Fix For: 0.90.0
>
> Attachments: HBASE-3351_3326.patch
>
>
> While debugging other issues, I found that
> ReplicationAdmin.ReplicationStatusTracker is doing a ZK request every time a
> znode changes. Also reading the replication state shouldn't go to ZK if we
> already maintain a local variable that's updated with the tracker.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
