[infinispan-dev] Hanging state transfer

[Sanne Grinovero]

I've resolved it, thanks all - especially Erik and Dan for their suggestions.

The problem was a hidden exception being swallowed somewhere in the
deployment stack: a "not serializable" exception which pointed to a
classloader issue during bootup, as my custom Externalizer wasn't
registered.

After fixing my classloader issue I'll see if I can figure were things
are being silently swallowed. Would be nice if someone in Infinispan
core could make sure that the state-transfer isn't going to spin
forever in such a case? I'd rather have the sending node at least
reply with some kind of abort code.

Thanks,
Sanne

On 21 April 2015 at 17:23, Erik Salter <an1310 at hotmail.com> wrote:
> Hi Sanne,
>
> This looks a lot like ISPN-5106
> <https://issues.jboss.org/browse/ISPN-5106>  If it is, I work around it in
> my production environment by not having my nodes wait on the initial state
> transfer.
>
> Erik
>
> On 4/21/15, 9:25 AM, "Sanne Grinovero" <sanne at infinispan.org> wrote:
>
>>Hi all,
>>any idea or suggestion please? Thread dumps are attached.
>>
>>It's running two WildFly 9 instances, in "standalone" mode so I don't
>>expect other JGroups channels to be open.
>>I deploy an Hibernate Search application in them, and this simply
>>starts a CacheManager. That hangs on start().
>>
>>If I run the exact same test, with exact same Infinispan version
>>(7.1.1.Final) on WildFly 8.2 + the as modules (embedded) distributed
>>by the Infinispan release it works just fine.
>>
>>Wildfly 9 includes this Infinispan version, so the difference is that
>>now I'm running the modules which are shipped with WildFly 9 directly:
>>this should work for end users w/o additional downloads.
>>
>>Don't take the current master of WF9 as reference: there are some
>>(other) problems with these modules, so I'm fixing them.. the other
>>problems were classloader / Search / Hibernate related and I have
>>patches for those, this run is using them.
>>But this one I don't understand? Any idea of what to look for? It's
>>probably caused by some extension point too much, or too few, but it's
>>not giving a hint.
>>
>>Another notable difference, is that Infinispan 7.1.1.Final shipped
>>with JGroups 3.6.1.Final - so that's what I'm using in the version
>>which works - while WildFly 9 is using JGroups 3.6.2.Final.
>>I've tried to patch WildFly9 to use the older JGroups but it didn't
>>change the result.
>>
>>Thanks,
>>Sanne
>>_______________________________________________
>>infinispan-dev mailing list
>>infinispan-dev at lists.jboss.org
>>https://lists.jboss.org/mailman/listinfo/infinispan-dev
>
>
> _______________________________________________
> infinispan-dev mailing list
> infinispan-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/infinispan-dev