db-derby-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Knut Anders Hatlen <Knut.Hat...@sun.com>
Subject Re: AntiGC Not Shutting Down
Date Mon, 24 Oct 2005 14:03:03 GMT
Eric Lenio <eric@lenio.net> writes:

> On Mon, Oct 24, 2005 at 10:54:37AM +0200, Knut Anders Hatlen wrote:
>> Eric Lenio <eric@lenio.net> writes:
>> 
>> > I have a similar setup as you, and previously I had witnessed similar
>> > behavior with AntiGC.  I recommend this link if you haven't already seen
>> > it:
>> >
>> > http://www-128.ibm.com/developerworks/db2/library/techarticle/dm-0508bader/
>> >
>> > Be especially sure to put derby.jar in the correct path.  In my case the
>> > AntiGC thread issue went away when derby.jar was placed in
>> > tomcat/common/lib (like the above article states in the "casual
>> > scenario" section).  The embedded and network drivers worked perfectly
>> > after I did this.  Hope this helps.
>> 
>> Hi Eric,
>> 
>> Are you running the official Apache Derby 10.1.1.0 or the patched
>> version I sent you?
>> 
>
> Hi Knut, yes I am indeed using your patched version which has run very
> stable for me.  I've been stress testing my Tomcat application for
> several weeks now without having to restart Tomcat or Derby even once,
> whereas before I had to restart several times per hour during heavy
> development.

Then I guess the original poster has to use the development version or
patch and build his own 10.1 jar in order to solve the problem. 

> I would also add that there *may* be another thread issue with
> DRDAConnThread, I've been monitoring Tomcat with Jconsole for a while
> now and these types of threads seem to be growing (I've got 11 right
> now, but haven't touched the database in 24 hours).  I have no idea
> about what DRDAConnThread even does, so maybe what I am witnessing is
> normal.

The network server has one DRDAConnThread for each client which is
connected. DRDAConnThreads can be reused, but DERBY-206 states that
"[unless] derby.drda.maxThreads parameter is set the number of
DRDAConnThreads will be equal to the maximum number of simultaneous
connections ever and there is no way to free the memory without
shutting down and restarting the server."

As long as the network server is running, seeing a non-decreasing
number of DRDAConnThreads is normal, but if you see DRDAConnThreads
even after you have unloaded the Derby driver, I would say it's a bug.

-- 
Knut Anders


Mime
View raw message