[
https://issues.apache.org/jira/browse/TAJO-525?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13876189#comment-13876189
]
hyoungjunkim edited comment on TAJO-525 at 1/20/14 7:23 AM:
------------------------------------------------------------
Created a review request against branch master in reviewboard
https://reviews.apache.org/r/17114/
This is bug caused by opening too many connection to Worker in the QueryMaster.
So I changed bellow.
- Using RpcConnectionPool, when QueryMaster requests running container to
Worker.
- Configurable number of netty's client
workers("tajo.rpc.client.socket-io-threads"). If this value is zero, The number
of worker per socket connection is Runtime.getRuntime().availableProcessors() *
2 that is netty's default value.
This patch is urgent patch. There is another issues to be resolved like
creation ClientSocketChannelFactory per JVM.
was (Author: hjkim):
Created a review request against branch master in reviewboard
https://reviews.apache.org/r/17114/
> Concurrent queries hang
> -----------------------
>
> Key: TAJO-525
> URL: https://issues.apache.org/jira/browse/TAJO-525
> Project: Tajo
> Issue Type: Bug
> Affects Versions: 0.8-incubating
> Reporter: Min Zhou
> Priority: Critical
> Fix For: 0.8-incubating
>
> Attachments: TAJO-525.patch, pstack
>
>
> When I run a simple query "select count(1) from lineitem;" in one tsql
> session and after that run the 2nd query in another tsql session. The 2nd
> one hung.
> After the 1st one done, the 2nd one still stuck at 0%.
> After that, run the same query in the first tsql session, the 3rd query also
> hung.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
