Can you duplicate the problem without using a database? We used to think a lot of our crash problems were database related because they only happened on the machine where our DB is used. But after further testing, we were able to cause crashes w/o using the DB.
Do you see the "resuming" message in your server log? If you do, bump up maxconnections until you never see this message in your log and see if it still crashes. I'm not saying this code is broken, but it could have a race condition or something and since it is not exercised very often, most sites would never notice. An alternative would be to lower maxconnections and see if the problem happens more frequently or sooner with your load test. Jim > This sounds like the exact same problem we're getting! At least I know we're > not alone ;) > > More info: our oracle maxidle and maxtimeout are set to 1000000000 already, > and our max stack size is set to 500k, so I doubt those are involved. > > We can replicate the problem with ease through load testing. Thankfully, it > hasn't cropped up in production, yet, but that doesn't mean it won't. > > Anyone else have any ideas? > > Thanks, > Sean > > -----Original Message----- > From: Janine Sisk [mailto:[EMAIL PROTECTED]] > Sent: Friday, September 28, 2001 6:42 AM > Subject: Re: AOLSERVER Digest - 26 Sep 2001 to 27 Sep 2001 (#2001-244) > > > > For mysterious crashes with no explanation, it's probably due to data > > corrupted by blown stacks. > > Kris, > > Would this also apply to situations where nsd is still running, but stops > responding? No zombies or anything, everything appears as normal except > that all threads are frozen; nothing being sent out, and nothing being > written to log files. We have one site which goes into this state on an > unfortunately frequent basis (ie, every coupld of weeks), and I have been > unable to figure out why. > > janine >
