hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jason Rutherglen (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-2357) Coprocessors: Add read-only region replicas (slaves) for availability and fast region recovery
Date Mon, 13 Jun 2011 19:04:51 GMT

    [ https://issues.apache.org/jira/browse/HBASE-2357?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13048716#comment-13048716

Jason Rutherglen commented on HBASE-2357:

{quote}That is not how HBase RPC works. One connection between the endpoints (in this case
regionserver and regionserver) is established upon demand, reaped when idle too long, and
multiplexed over in the meantime.

Ok, great.  The replication master will need to examine ZK, and find out which slaves to RPC
connect to.  

{quote}And spill the queues under heap pressure presumably. Or give up on a too laggy slave
and have it killed to avoid blowing out heap in the alternative{quote}

Spilling would probably add too much complexity (eg, where would it spill to?).  I think we
need to define how a slave gets too far behind, and then assume it'll need to refresh itself
when it does (on a per-region basis, or does the entire RS need to recover?).

bq. No we need to think of the HMaster as always on the verge of going away to be supplanted
by ZooKeeper mediated distributed actions.

Ok, good to know.  Before implementing we should hammer this part of the design out.

> Coprocessors: Add read-only region replicas (slaves) for availability and fast region
> ----------------------------------------------------------------------------------------------
>                 Key: HBASE-2357
>                 URL: https://issues.apache.org/jira/browse/HBASE-2357
>             Project: HBase
>          Issue Type: Sub-task
>          Components: master, regionserver
>            Reporter: Todd Lipcon
>            Assignee: Andrew Purtell
> I dont plan on working on this in the short term, but the idea is to extend region ownership
to have two modes. Each region has one primary region server and N slave region servers. The
slaves would follow the master (probably by streaming the relevant HLog entries directly from
it) and be able to serve stale reads. The benefit is twofold: (a) provides the ability to
spread read load, (b) enables very fast region failover/rebalance since the memstore is already
nearly up to date on the slave RS.

This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira


View raw message