[jboss-jira] [JBoss JIRA] (WFLY-9900) EJB client rarely hangs

[Richard Janík (JIRA)]

    [ https://issues.jboss.org/browse/WFLY-9900?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13538223#comment-13538223 ] 

Richard Janík commented on WFLY-9900:
-------------------------------------

I have a job that can reproduce this now (the jvmkill-dist-sync one, not the other), so I'll have a thread dump soon too. About blocker - I created this as major with the idea that later, when I have a good case and a thread dump, I may want to hike the priority higher, but have no strong opinions on what it should be.
[~rhusar] unfortunately at some point, we taught our automation to kill stuck jobs automatically, without teaching it to gather dumps. :)

> EJB client rarely hangs
> -----------------------
>
>                 Key: WFLY-9900
>                 URL: https://issues.jboss.org/browse/WFLY-9900
>             Project: WildFly
>          Issue Type: Bug
>          Components: Clustering, EJB
>    Affects Versions: 12.0.0.Beta1
>            Reporter: Richard Janík
>            Assignee: David Lloyd
>
> Seen in our clustering tests where remote EJB clients invoke beans on 4 servers in cluster, while each of those is being killed and restarted. This is done in sequence and there's plenty of time in between the kill/start/kill-another (1 minute).
> In two ouf our tests, we saw a small number of clients hang and resist termination at the end. The scenarios are:
> [jvmkill-dist-sync|https://jenkins.hosts.mwqe.eng.bos.redhat.com/hudson/job/perflab_eap-7x-failover-ejb-ejbremote-jvmkill-dist-sync/36/] (36 clients)
> [shutdown-dist-sync|https://jenkins.hosts.mwqe.eng.bos.redhat.com/hudson/job/perflab_eap-7x-failover-ejb-ejbremote-shutdown-dist-sync-3owners/32/] (1 client)
> The hangs are not related to any error or warning message on the server-side. There are ClosedChannelExceptions on the client side, but these seem to be unrelated as the numbers of clients reporting the exceptions are not the same (lower) as the numbers of hanging clients.
> No thread dump as I didn't manage to reproduce yet.
> Might be related to JBEAP-12074 or JBEAP-13169.



--
This message was sent by Atlassian JIRA
(v7.5.0#75005)