Shapira, Yoav wrote:
Hi, It seems like connections aren't being closed properly. Some RedHat users have reported a need to set the LD_ASSUME_KERNEL parameter such that NPTL is not used and threading problems go away. I myself don't have experience with that, but if you search this list's archives for LD_ASSUME_KERNEL you'll see what I'm talking about.
Thanks a lot, I'll take a look.
I'm sorry to bother again but I was wondering whether any of you has
some feedback about the following problem.
Don't worry/apologize, you're not bothering anyone. Ask away. If someone doesn't feel like answering, they won't ;)
:-)
Examining the application log I only see a few exceptions caused by an
RDBMS shutdown. I am asking myself if this can be the cause of zombie
threads, but it seems strange to me.
It could be, depending on your driver. If you're using connection pooling such as via DBCP, try adding a validation query to the pool's configuration.
We are using our own Connection Pooling mechanism built upon Jakarta Commons Pool. Basically the jdbc connection is never really released; nevertheless we do have a validation query.
Do you have any ideas? I may increase ThreadPool dimension, but whenthe
block occurred traffic was low and it doesn't seem to justify anychange
on the configuration side. Is it a Tomcat issue? We are planning to
upgrade to 4.1.30 very soon, but could it be worth to go towards 4.1.31
or 5.x.y (not really planned yet)?
The block did not occur because traffic was high: I bet the hung connections accumulated over a long period: possibly even the six months you mentioned.
My guess too.
It's definitely worth your time to test with Tomcat 5.0.28.
:) One day I'll do it, thanks.
Cheers, Giuliano
