[
https://issues.apache.org/jira/browse/HBASE-1160?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12667879#action_12667879
]
ryan rawson commented on HBASE-1160:
------------------------------------
We'd probably want to avoid spanning a cluster across datacenter boundaries...
it's too easy to get into irreconcilable network splits.
Replication as a site-to-site backup is a great approach. Multi-master
replication would allow concurrent read/write when the link between the sites
is down, but the internet connection is still available (very common scenario
surprisingly).
I don't think we need a statistical 'eventual consistency' multi-write (Eg:
paxos), but instead a straight forward replication stream (a la redo log
replication from Oracle, mysql, etc) would be pretty good. Tools to monitor
that replication stream (replicas are X seconds behind kind of thing) would
also be necessary.
> Multi-site replication support
> ------------------------------
>
> Key: HBASE-1160
> URL: https://issues.apache.org/jira/browse/HBASE-1160
> Project: Hadoop HBase
> Issue Type: New Feature
> Reporter: Jean-Daniel Cryans
>
> HBase should support multi-site replication in order to provide higher
> availability. A first version with eventual consistency could be easy to
> implement.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
