I noticed a visible performance regression during deployment of applications, namely web applications, it seems. I packed a simple reproducer EAR application consisting of multiple simple web archives. On the testing machine, on 6.2.0.ER2: deployment time is around 6300 ms on 6.2.0.ER3: deployment time is around 7600 ms This is a regression by 20%. By time needed to deploy, I mean the difference between timestamps of these two logs: 09:26:40,885 INFO [org.jboss.as.server.deployment] (MSC service thread 1-7) JBAS015876: Starting deployment of "webapps.ear" (runtime-name: "webapps.ear") 09:26:47,184 INFO [org.jboss.as.server] (management-handler-thread - 1) JBAS018559: Deployed "webapps.ear" (runtime-name : "webapps.ear") When I deployed other types of applications (webservices, ejb jars,.. even if packed in EAR, too) I did not notice any regression.
Still the same in ER4 and ER5.
This could be caused by a change to unpack the archives in WarStructureDeploymentProcessor (o.jb.as.web.deployment) or in the EAR processor itself. If this can be confirmed (I didn't do it), since this fixed issues, this regression should be considered normal (expanding more has a cost). Other than that, I'm not aware of significant deployment changes.
Is there a regression from 6.1.1 for this test?
Yes, this is a regression compared to 6.1.1.
The regression of the deployment performance for EAP 6.3.0.ER10 against the 6.1.1 is not so significant: 6.1.1.GA - 6.65sec 6.3.0.ER10 - 6.96sec (in average for attached webapps.ear) Due to Comment 3 and discussion with Jan closed as WONTFIX.