[
https://issues.apache.org/jira/browse/HBASE-2129?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12800829#action_12800829
]
Andrew Purtell commented on HBASE-2129:
---------------------------------------
bq. In the case of master/master, you have to manage to not replicate edits
from which you are the origin (I know you know that). I guess you could encode
that information in KV (routing information?).
I was thinking to try overloading HLogKey with source routing information. So
the replicator can skip sending the edit to a replication peer if it is listed
in the source path for the item. Implies globally unique identifiers for peer
clusters but that's not a big deal. Having an admin assign one can be part of
the replication config process. Could be just a byte, treated as unsigned. I do
not expect there will be deployments with > 256 peer HBase clusters for a
while. (Imagine that...)
bq. WRT scoping, I left it out of this particular Jira but it is one of the
next item on my todo list
Ok, cool.
bq. along with live configuration of replication streams and being able to ask
a cluster to replicate edits from time X
Nice!
> Simple Master/Slave replication
> -------------------------------
>
> Key: HBASE-2129
> URL: https://issues.apache.org/jira/browse/HBASE-2129
> Project: Hadoop HBase
> Issue Type: Sub-task
> Reporter: Jean-Daniel Cryans
> Assignee: Jean-Daniel Cryans
> Fix For: 0.21.0
>
> Attachments: HBASE-2129.patch
>
>
> We should first build a very simple replication mechanism to validate our
> assumptions and get a feel of what replication is in this very distributed
> context.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
