AS7.0.1 distributable war strange behavior

[morphy]

I have a managed bean session scoped in a war which is described to be distributable

it is deployed in the domain configuration using the HA profile for the servers in the server group

that managed bean has a PostConstruct method which i use to bind the principal with a stateful bean (described as clustered) and a PreDestroy method i use to call the remove method of the sfsb

the problem is that after a minute of inactivity, the PreDestroy is called even if the session timeout is configured to be 30 minutes... i tried to play with infinispan configuration but without result...

any idea?

can serialization/deserialization in the cache trigger those lifecycle methods?

bye

later i'll try to remove the sfsb reference even if i think it's not the problem

[morphy]

without documentation... it's almost impossible to understand if i'm missing something or if that's a bug :_(

the problem is in the distributable element... and maybe... cache configuration... for the security domain... i've tried to configure the security domain with cache-type="infinispan" but it doesnt work...

so, is the PreDestroy called for a sort of problemi between cache and security? or the PreDestroy is called by some other reason and the security is a consequence and not a cause?

who knows.... i must revert the webapp to a standalone webapp if i want to go on...

sad...

[pferraro]

FYI - SFSB clustering is not yet implemented - it'll debut in one of the 7.1 pre-releases.

SFSBs in AS 7.0.1 do not yet support passivation, so there shouldn't be anything to trigger SFSB serialization.

Is this SFSB annotated with @StatefulTimeout ?

[morphy]

Hi Paul,

i'm not sure that the sfsb is the problem... i've removed it from the deploy and the problem remains... it seems to be something related to the jboss authentication... my current investigation thread is infinispan... some kind of deserialization taht happens every 60" that loose the authenticated principal when the web app is set as distributable... but i don't have any news right now...

references for 60" are in the cache container for flush-timeout e timeout but i don't master infinispan so it's a hard work by now...