giraph-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Maja Kabiljo (JIRA)" <>
Subject [jira] [Commented] (GIRAPH-386) ClassCastException when giraph.SplitMasterWorker=false
Date Thu, 25 Oct 2012 00:46:12 GMT


Maja Kabiljo commented on GIRAPH-386:

I see, this is because we added mapping from taskid to InetSocketAddress to NettyClient. And
in mode with master and worker on the same node, we have two addresses corresponding to the
same taskid. Let me think what's the best way to fix this. Just today we were talking about
having our own ids for workers because of some other optimization, maybe that can also be
used to fix this problem.
> ClassCastException when giraph.SplitMasterWorker=false
> ------------------------------------------------------
>                 Key: GIRAPH-386
>                 URL:
>             Project: Giraph
>          Issue Type: Bug
>            Reporter: Jaeho Shin
>            Assignee: Maja Kabiljo
> Using -Dgiraph.SplitMasterWorker=false with a recent trunk (r1401165) caused the machine
who is playing the master role (showing itself as {{ALL}} from the task tracker page) to throw
ClassCastException (SendVertexRequest -> MasterRequest) from MasterRequestServerHandler
class.  I'm trying to use as many machines as possible for actual computation (can't afford
to waste one for master+ZK).  This worked fine with r1388628 (roughly a month ago), so a recent
change must've broken something.  Here's the relevant log I captured:
> {code}
> 2012-10-24 23:08:02,152 WARN org.apache.giraph.comm.netty.handler.RequestServerHandler:
exceptionCaught: Channel failed with remote address /10.x.y.z:41780
> java.lang.ClassCastException: org.apache.giraph.comm.requests.SendVertexRequest cannot
be cast to org.apache.giraph.comm.requests.MasterRequest
> 	at org.apache.giraph.comm.netty.handler.MasterRequestServerHandler.processRequest(
> 	at org.apache.giraph.comm.netty.handler.RequestServerHandler.messageReceived(
> 	at
> 	at org.jboss.netty.handler.codec.oneone.OneToOneDecoder.handleUpstream(
> 	at org.jboss.netty.handler.execution.ChannelUpstreamEventRunnable.doRun(
> 	at
> 	at java.util.concurrent.ThreadPoolExecutor$Worker.runTask(
> 	at java.util.concurrent.ThreadPoolExecutor$
> 	at
> {code}

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:

View raw message