hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Lars Hofhansl (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-2195) Support cyclic replication
Date Sat, 03 Sep 2011 04:45:09 GMT

    [ https://issues.apache.org/jira/browse/HBASE-2195?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13096578#comment-13096578
] 

Lars Hofhansl commented on HBASE-2195:
--------------------------------------

Amending my solution a bit. The Edits will be written with the full id as described above.
In the near term I think it is better keep the 1 character ID for local identification. 

The reason is:
1. It is nice to be able to identify every slave by a number chosen by the master. (addPeer,
deletePeer, listPeers, etc) - once we support multiple slaves that is.
2. A user at the master does not have to know about cluster UUID.
3. The ZNode handling would not change.

These id is never shared to the slave/sink.
Under the hood the full 16byte cluster id is used in the Edits to avoid replication cycles.

I assume it would be worth going through the trouble of not storing those 16 bytes of UUID
when replication is not enabled(?)


> 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

        

Mime
View raw message