Nice work Kevan. I got Liferay enterprise to run fine now ;-) Jeff
Kevan Miller wrote: > > On Jul 31, 2006, at 10:45 PM, Matt Hogstrom wrote: > >> Concur...blocked. >> >> Aaron Mulder wrote: >>> Jeff just noted that Liferay seems to take an abnormal amount of >>> memory to run. >>> I just redeployed the console twice and got an OutOfMemoryError. >>> When working on the console for 1.1, I could redeploy dozens of times. >>> I think we've introduced a pretty serious memory problem between 1.1 >>> and now. I know we're close to the deadline, but I think we ought to >>> investigate this before shipping 1.1.1. > > OK. I've found the cause -- it's RC 425160 > (http://svn.apache.org/viewvc?view=rev&revision=425160). Looks like we > were running out of PermGen space. This change roughly double the number > of classes created by G on startup. I could only deploy/undeploy > DayTrader once. After reverting the change, I've run well over 100 > deploy/undeploy cycles. We're leaking two instances of > 'EDU.oswego.cs.dl.util.concurrent.LinkedNode' per deploy/undeploy, but > that's only 32 bytes... > > I haven't confirmed, but I saw evidence that it was at least part of the > DWR leaks that Jeff G reported. > > Here's the Jira associated with the above change -- > http://issues.apache.org/jira/browse/GERONIMO-2204 > > Since the change is optional, I'm reverting the commit. I think we can > safely leave out of 1.1.1. Can discuss how we want to fix, later... > > --kevan > > >
