[
https://issues.apache.org/jira/browse/HBASE-15970?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16268260#comment-16268260
]
Zheng Hu commented on HBASE-15970:
----------------------------------
bq. Would hbase:replication be used for this tracking ?
I think we save both replication queues and peer states in hbase:replication
meta table. For replication queues, we store the queues with a rowkey:
queueId-<serverName>, For peer states, we can store the peer states with
another rowkey which is a different format, such as peerId-peer-info , or
other format like that , and we store the peer state and peer config in a
difference column: p:s , p:c (p means peer, s means state, c means
configuration ).
Besides, we may need procedures to implement a table based ReplicationPeers
for notifying the RegionServer when add/remove peers or update peer config.
I'd like to prepare the patch for it.
> Move Replication Peers into an HBase table too
> ----------------------------------------------
>
> Key: HBASE-15970
> URL: https://issues.apache.org/jira/browse/HBASE-15970
> Project: HBase
> Issue Type: Sub-task
> Components: Replication
> Reporter: Joseph
> Assignee: Zheng Hu
>
> Currently ReplicationQueuesHBaseTableImpl relies on ReplicationStateZkImpl to
> track information about the available replication peers (used during
> claimQueues). We can also move this into an HBase table instead of relying on
> ZooKeeper
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
