[
https://issues.apache.org/jira/browse/HBASE-20147?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16404693#comment-16404693
]
Duo Zhang commented on HBASE-20147:
-----------------------------------
Talked with [~openinx] offline. The current approach does not always work. For
an offlined region, or a region crashed before writing anything after we setup
the serial replication peer, then the replication will start from the last
(new) open range directly and will be stuck.
Need to add more stages when modifying a peer. Will be back later.
> Serial replication will be stuck if we create a table with serial replication
> but add it to a peer after there are region moves
> -------------------------------------------------------------------------------------------------------------------------------
>
> Key: HBASE-20147
> URL: https://issues.apache.org/jira/browse/HBASE-20147
> Project: HBase
> Issue Type: Sub-task
> Reporter: Duo Zhang
> Assignee: Duo Zhang
> Priority: Major
> Fix For: 3.0.0
>
> Attachments: HBASE-20147-v1.patch, HBASE-20147-v2.patch,
> HBASE-20147-v3.patch, HBASE-20147.patch, HBASE-20147.patch
>
>
> The start point for serial replication is that, if we are in the first range
> then we are safe to push. And we will record replication barrier when the
> replication scope is set to SERIAL even if the table is not contained in any
> peers. So it could happen that, we record several barriers in the meta
> already and then we add a peer which contains this table. So when
> replicating, we will find that we are not in the first range, and then the
> replication will be stuck.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
