[
https://issues.apache.org/jira/browse/HBASE-16868?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16217984#comment-16217984
]
Duo Zhang commented on HBASE-16868:
-----------------------------------
So in general, if I create a peer with no namespace/table, we will set
replicate_all flag automatically? And then if you add namespace/table to this
peer, the operation will fail as you need to manually set the replicate_all
flag to false first? Then what is the behavior for an empty peer with a false
replicate_all flag? And is it allowed to set replicate_all flag to true for a
peer which already has namespace/table?
The code is fine, and I see you add UTs for the new feature, good. But I think
you need to fix the rubocop and ruby-lint violations. The only concern for me
is that we need to give a very clear behavior, and also the best practice on
how to config a peer.
Thanks.
> Add a replicate_all flag to avoid misuse the namespaces and table-cfs config
> of replication peer
> ------------------------------------------------------------------------------------------------
>
> Key: HBASE-16868
> URL: https://issues.apache.org/jira/browse/HBASE-16868
> Project: HBase
> Issue Type: Improvement
> Components: Replication
> Affects Versions: 2.0.0, 3.0.0
> Reporter: Guanghao Zhang
> Assignee: Guanghao Zhang
> Priority: Critical
> Fix For: 2.0.0-beta-1
>
> Attachments: HBASE-16868.master.001.patch,
> HBASE-16868.master.002.patch, HBASE-16868.master.003.patch,
> HBASE-16868.master.004.patch, HBASE-16868.master.005.patch,
> HBASE-16868.master.006.patch, HBASE-16868.master.007.patch
>
>
> First add a new peer by shell cmd.
> {code}
> add_peer '1', CLUSTER_KEY => "server1.cie.com:2181:/hbase".
> {code}
> If we don't set namespaces and table cfs in peer config. It means replicate
> all tables to the peer cluster.
> Then append a table to the peer config.
> {code}
> append_peer_tableCFs '1', {"table1" => []}
> {code}
> Then this peer will only replicate table1 to the peer cluster. It changes to
> replicate only one table from replicate all tables in the cluster. It is very
> easy to misuse in production cluster. So we should avoid appending table to a
> peer which replicates all table.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
