hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jean-Daniel Cryans (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HBASE-2129) Simple Master/Slave replication
Date Fri, 15 Jan 2010 15:54:54 GMT

    [ https://issues.apache.org/jira/browse/HBASE-2129?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12800727#action_12800727
] 

Jean-Daniel Cryans commented on HBASE-2129:
-------------------------------------------

bq. Are there any strategies to parallelize this operation for busy clusters?

Yes, this is one of the things that I scaled down in order to present a patch sooner. Currently
the ReplicationSink cannot get behind edits coming from the master cluster since its blocking
the replication of new entries. The next step is to allow appenders while the file is being
read (logs replayed), then once parallelized (HBASE-2066) puts are available to use them in
that thread.

The plan for the logs themselves in ReplicationSink is to be able to let the thread get behind
on replaying edits, for example because the slave cluster is very busy, and that if the region
server dies to be able to use those logs to replay them somewhere else.

> 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.


Mime
View raw message