Re: Web App Redeploy Memory Leak

Scott Ferguson <[email protected]> Wed, 31 May 2006 15:29:53 -0700
Newsgroups gmane.text.xml.resin.user
Message-ID <[email protected]>
On May 30, 2006, at 1:17 AM, sks wrote:

> Some of you may remember my on-going memory leak problem when re- 
> deploying
> my web app.
>
> Scott said it might have been related to a JVM bug regarding primitive
> arrays and JVMTI but this bug has now been fixed in 1.5.07 and  
> still it was
> happening.
> http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=5093520

I'll check this.

> I have found the source of the bug anyway and it is in Resin (or I  
> suppose Java's) DataSource code.
>
> I call getConnection obviously to get a connection.
>
> Now if I remove the DataSource code and instead use a  
> java.sql.DriverManager lookup each time it works perfectly and all  
> objects are garbage collected. I have not had to start Resin for 3  
> days on my development machine (when its normally every hour or so).

But that's the original issue.  Resin's <database> is doing  
introspection on the underlying driver (can you let us know which  
driver class you're using?).  If the introspection has anything to do  
with a byte[] field or method, the JVM will get stuck.

If you avoid the DataSource entirely, Resin won't introspect any  
classes and therefore won't trigger the bug.

-- Scott
>
>
>
>