[
https://issues.apache.org/jira/browse/HBASE-2195?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13096314#comment-13096314
]
Lars Hofhansl commented on HBASE-2195:
--------------------------------------
I have been thinking about this too. Actually I found that the cluster id in
ReplicationSource is always 0 (i.e. it is never set for the Master - and my
tests just worked by accident :( ).
In fact I came to the same conclusion...
I think the main reason that it is nice that the replication members do not
need to be labeled. Knowing their zookeeper ensemble allows us to get the UUID.
I have a patch that uses the UUID about 70% done.
One of the problems is that now one cannot enable replication without the other
side running as it needs to connect to the other zookeeper to get the cluster
id (but I can probably work around that by retrieving the UUID later).
I think getting the attribute based version running is a good start. Then we
can decide whether this is safe enough for 0.92 or we'll do a space optimized
version for 0.94.
> Support cyclic replication
> --------------------------
>
> Key: HBASE-2195
> URL: https://issues.apache.org/jira/browse/HBASE-2195
> Project: HBase
> Issue Type: Sub-task
> Components: replication
> Reporter: Jean-Daniel Cryans
> Attachments: 2195.txt
>
>
> We need to support cyclic replication by using the cluster id of each HlogKey
> and stop replicating when it goes back to the original cluster.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
