hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Nicolas Liochon (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-9139) Independent timeout configuration for rpc channel between cluster nodes
Date Tue, 06 Aug 2013 16:18:48 GMT

    [ https://issues.apache.org/jira/browse/HBASE-9139?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13730898#comment-13730898
] 

Nicolas Liochon commented on HBASE-9139:
----------------------------------------

I think it could become difficult to manage: them we would have half-internal client like
map reduce and so one.
So to me, if you want different settings just use different configuration files.

The real differences is in the operations imho. Some operations take no time and can be retried.
So having the same timeout for such operations and for operations that can take longer (for
example because they write in hdfs) is not good. Internally, in HBase code, we should use
a different setting for such operations. 


                
> Independent timeout configuration for rpc channel between cluster nodes
> -----------------------------------------------------------------------
>
>                 Key: HBASE-9139
>                 URL: https://issues.apache.org/jira/browse/HBASE-9139
>             Project: HBase
>          Issue Type: Improvement
>          Components: IPC/RPC, regionserver
>    Affects Versions: 0.94.10, 0.96.0
>            Reporter: Julian Zhou
>            Priority: Minor
>             Fix For: 0.94.11, 0.96.0
>
>
> Default of "hbase.rpc.timeout" is 60000 ms (1 min). User sometimes
> increase them to a bigger value such as 600000 ms (10 mins) for many
> concurrent loading application from client. Some user share the same
> hbase-site.xml for both client and server. HRegionServer
> #tryRegionServerReport via rpc channel to report to live master, but
> there was a window for master failover scenario. That region server
> attempting to connect to master, which was just killed, backup master
> took the active role immediately and put to /hbase/master, but region
> server was still waiting for the rpc timeout from connecting to the dead
> master. If "hbase.rpc.timeout" is too long, this master failover process
> will be long due to long rpc timeout from dead master.
> If so, could we separate with 2 options, "hbase.rpc.timeout" is still
> for hbase client, while "hbase.rpc.internal.timeout" was for this
> regionserver/master rpc channel, which could be set shorted value
> without affect real client rpc timeout value?

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message