[jboss-jira] [JBoss JIRA] (JGRP-1912) LockService can fail under heavy load

[Chris Gentile (JIRA)]

    [ https://issues.jboss.org/browse/JGRP-1912?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13042087#comment-13042087 ] 

Chris Gentile commented on JGRP-1912:
-------------------------------------

I relocated the <CENTRAL_LOCK /> tag in my channel config and the problem goes away.

For anyone else that stumbles upon this thread:
The doc says:
   "The locking protocol has to be placed at or towards the top of the stack (close to the channel)."

The stack is lowest-to-highest in channel config files, so "towards the top" is actually at the bottom of the file.
Then it is the last protocol to get added to the stack and is on top.

Bela,
  Perhaps in the documentation you could elaborate on what "close to the channel" actually means.

  Or perhaps an actual example might help.
  The jgroups jar files seem to include several config files already (i.e.: udp.xml).
  Perhaps you could include a locking.xml too?

Thanks,
Chris


> LockService can fail under heavy load
> -------------------------------------
>
>                 Key: JGRP-1912
>                 URL: https://issues.jboss.org/browse/JGRP-1912
>             Project: JGroups
>          Issue Type: Bug
>    Affects Versions: 3.6.1
>         Environment: both windows 8.1 and RHEL
> java 1.7.0_75
>            Reporter: Chris Gentile
>            Assignee: Bela Ban
>             Fix For: 3.6.3
>
>         Attachments: LockServiceTestClient.java, LockServiceTestClient.java, logFile, logFile2, runLockServiceTestClient.sh, udp.xml
>
>
> Once in awhile, the LockService will get stuck acquiring a lock.
> These seems to occur most often when two threads are simultaneously trying to acquire the same lock. (i.e.:  exact same millisecond on a server with multiple processors).



--
This message was sent by Atlassian JIRA
(v6.3.11#6341)