[jira] [Updated] (STORM-165) Adding more supervisor causes topology (LinearDRPC) to fail

["Rick Kellogg (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/STORM-165?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Rick Kellogg updated STORM-165:
-------------------------------
    Component/s: storm-core

> Adding more supervisor causes topology (LinearDRPC) to fail
> -----------------------------------------------------------
>
>                 Key: STORM-165
>                 URL: https://issues.apache.org/jira/browse/STORM-165
>             Project: Apache Storm
>          Issue Type: Bug
>          Components: storm-core
>            Reporter: James Xu
>            Priority: Minor
>
> https://github.com/nathanmarz/storm/issues/292
> Following my previous email, i just want to add this to the issue log so everyone is aware. One new thing I found is that it's not associated with a "magic number" of supervisor. I have it running at 3 but not at 4. I think it has to do with this specific error:
> 2012-08-14 11:29:33 worker [WARN] Received invalid messages for unknown tasks. Dropping... 
> 2012-08-14 11:29:33 worker [WARN] Received invalid messages for unknown tasks. Dropping... 
> 2012-08-14 11:29:33 worker [WARN] Received invalid messages for unknown tasks. Dropping...
> Seems the fourth one gets messages for tasks it does not know about- possibly out of sync.
> Steps to reproduce:
> 1. Bring one supervisor up 2 mvn -f m2-pom.xml compile exec:java -Dexec.classpathScope=compile -Dexec.mainClass=storm.starter.DRPCTestClient -Dexec.args="2000"
> DRPRC CLIENT
> SIZE 2000
> Result = []
> db time 3268 ms
> [INFO] ------------
> 2. Repeat adding supervisor and rerun the client.
> #DRPRC CLIENT #
> SIZE 2000
> [WARNING] 
> java.lang.reflect.InvocationTargetException
> at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
> ..
> ----------
> amuraru: @nathanmarz What kind of key is used to uniquely identify the supervisor in nimbus? I saw similar errors when running multiple supervisors on different machines but with the same hostname (e.g. localhost.localdomain) listed in storm ui. COuld be this an issue?
> ----------
> nathanmarz: Supervisors generate a unique id for themselves and store it in their local dir. It's fine to run many supervisors on the same host as long as they don't share a local dir.
> ----------
> willfcareer: Hi, there can't be more supervisor in trident drpc ?



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)