quality@glassfish.java.net

v3 not deploying huge EAR anymore

From: matthias.fraass_at_tricoder.net <matthias.fraass_at_gmail.com>
Date: Thu, 29 Oct 2009 15:03:29 +0100

Hi,

I'm perplexed:
An EAR that deploys fine under GF v2 doesn't deploy at all under v3 (b70).
It sits there consuming CPU time and timeouts after 10 minutes:

>asadmin deploy whow-thats-a-huge.ear
I/O Error: Read timed out
Command deploy failed.

I have set javax.enterprise.system.tools.deployment=FINE but get no
server.log entries, nothing.
It seems that the deployment is still going on after the timeout, though.
But I didn't have the patience to wait for more than 20minutes after that.

Originally it was a 100+MB EAR containing, amongst other JARs, an ejb-jar
with 45.000+ classes.

Please don't ask if the design is alright :). We have plans to do refatoring
but we have to do things one by one.
- It's my task to migrate that behemoth from Weblogic to Glassfish.
- It's deploying fine under a stock GF v2.1, so it should work under v3 or
at least give some hints what went wrong

I've stripped down the EAR so that it only contains the ejb-jar with a
minmal ejb-jar.xml (stripped from 5MB to only 1 SessionBean) - that's still
60MB.
Still, it won't deploy unless I remove most of the classes from that
ejb-jar, leaving only about 1000 classes. And for every 10 classes I add,
the deployment takes 15s longer!

Again: v2 chews on this without a grumble and deploys it within 2-3 minutes!

-> As I've heard that v3 contains major rewrites - is there something wrong
with the scan processes during deployment?

Sorry for these unclear statements, but this is a showstopper :/.

I appreciate any ideas on this very much.

Cheers,

Matthias

--
http://twitter.com/MatthiasFraass