[
https://issues.apache.org/jira/browse/HBASE-13367?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14391180#comment-14391180
]
Esteban Gutierrez commented on HBASE-13367:
-------------------------------------------
[~liushaohui]
bq.Yes. We need a external tool to check all edits have been replicated to the
sink cluster.
I don't think it needs to be done via an external tool since it is hard to
guarantee a 0 delta for busy clusters and perhaps every time you run the tool
or depending how long it takes to this tool the table state might have changed.
I think you can use this label to ship the delta between the edit begin shipped
and the latest edit in the source cluster so the sink can track how far is
behind and then via a configuration threshold (time based or transaction based)
transition the replicated table from this dirty state into consistent state.
> Add a replication label to mutations from replication
> -----------------------------------------------------
>
> Key: HBASE-13367
> URL: https://issues.apache.org/jira/browse/HBASE-13367
> Project: HBase
> Issue Type: Improvement
> Reporter: Liu Shaohui
> Assignee: Liu Shaohui
>
> In some scenarios, the regions need to distinguish the mutations from client
> for actual users or from replication of peer cluster.
> - Lower the priority of mutations from replication to improve the latency of
> requests from actuals users
> - Set table in replicated state to keep data consistent. In this state, the
> table will reject mutations from users, but accept mutations from replication
> and read requests from users.
> So we need to add a replication label to mutations from replication.
> Suggestions and discussions are welcomed. Thanks
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
