[
https://issues.apache.org/jira/browse/HBASE-19665?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16392500#comment-16392500
]
Zheng Hu commented on HBASE-19665:
----------------------------------
bq. Why this doesn't store to meta?
The last pushed sequence id is peer level for each region. that's to say,
for each peer, every region will has a sequence id. So better to put this in
hbase:replication table. in other hand, we need to make sure that update wal
position and update last pushed sequence id in a transaction, if store the
this into hbase:meta, we need cross-table transaction which is hard to
implement it in HBase.
bq. Do we need this?
Yes, As the ReplicationQueueStorage#getListOfReplicators said:
{code}
/**
* Get a list of all region servers that have outstanding replication queues.
These servers could
* be alive, dead or from a previous run of the cluster.
* @return a list of server names
*/
List<ServerName> getListOfReplicators() throws ReplicationException;
{code}
So we need a flag (use rs_state:enabled column here ) to indicate that a
region server is a replicator, even if there's no queue and no wals under the
replicator.
bq. If a queue has many wals, how to store them?
The wal family will keep all version, each version present a wal.
> Add table based replication peers/queues storage back
> -----------------------------------------------------
>
> Key: HBASE-19665
> URL: https://issues.apache.org/jira/browse/HBASE-19665
> Project: HBase
> Issue Type: Sub-task
> Components: Replication
> Reporter: Guanghao Zhang
> Assignee: Zheng Hu
> Priority: Major
> Attachments: HBASE-19665.v1.patch, HBASE-19665.v2.patch
>
>
> We removed them after HBASE-19397. So open a issue to track this thing.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
