[
https://issues.apache.org/jira/browse/HBASE-2195?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13096967#comment-13096967
]
Lars Hofhansl commented on HBASE-2195:
--------------------------------------
OK... Right now I made it versioned by observing the current byte cluster id is
always 0, so I was using this as version byte. I'll look into VersionedWritable.
I also have conditioned code that only writes the 16bytes when needed (at the
expense of having to write an extra boolean).
re: Byte ids... The bytes id is just used locally at a replication source. It
has no meaning elsewhere and never ends in the WAL. Otherwise we have to
rethink how one adds and removes sinks at a source (i.e. how do you identify
which sink to remove, unless you like to type in 16 byte ids). Changing this
would also mean that we cannot setup the ZNodes at a replication before the
sink is up - as we cannot get the sinks UUID before it can be reached.
This is actually a fairly clean design (IMHO), as a user at a source can give
the sinks numbers. And these numbers will never clash with the same numbers at
another source.
I'll attach a patch later today, which hopefully makes it clear what I have in
mind.
> 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
