[
https://issues.apache.org/jira/browse/HBASE-15867?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16917256#comment-16917256
]
stack commented on HBASE-15867:
-------------------------------
For history, could be a column family that keeps data for a week or two? In
hbase:meta?
bq. ...what's the real issue we want to solve by moving it from ZK to HBase
table?
We used to have a dictum that had it that you could erase zk content and hbase
would pick up and continue w/o issue. Keeping replication peer data in zk
violates this principal; permanent data belongs in hbase.
We've also had a general tendency in place where we would like to rely less on
zk for services -- especially as it seemed like we'd gone overboard. ZK is an
alien store on the other side of an RPC. All sorts of things can go wrong. We
know the properties of our system better than those of zk and 'storage' is the
name of our game; we shouldn't need to go elsewhere for it.
Thanks.
> Move HBase replication tracking from ZooKeeper to HBase
> -------------------------------------------------------
>
> Key: HBASE-15867
> URL: https://issues.apache.org/jira/browse/HBASE-15867
> Project: HBase
> Issue Type: New Feature
> Components: Replication
> Affects Versions: 2.1.0
> Reporter: Joseph
> Assignee: Zheng Hu
> Priority: Major
> Fix For: 2.3.0
>
>
> Move the WAL file and offset tracking out of ZooKeeper and into an HBase
> table called hbase:replication.
> The largest three new changes will be two classes ReplicationTableBase,
> TableBasedReplicationQueues, and TableBasedReplicationQueuesClient. As of now
> ReplicationPeers and HFileRef's tracking will not be implemented. Subtasks
> have been filed for these two jobs.
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
