Weirdness: HPUX 10.20/IDS 7.30.UC6

Odd thing happened last night. At some point one instance (of several on the machine) decided to stop accepting any connections. There was one user thread returning data and several others (including an oncheck) that had finished and were sitting around waiting for something to do. I mention the oncheck because the pid in onstat -g ses didn't exist; the person who ran oncheck earlier did a ^C to terminate it. onmode -yuck hung, as did onmode -z tid. Anything that required a connection to the database, infact, would hang indefinately. There was nothing in the log file to indicate any problems, and in fact checkpoints were still occuring every 5 minutes, right on schedule. ipcs output looked entirely normal, as did netstat. About the only way to clean this up was to either reboot the computer, or whacking shared memory segments until the engine realized that something was wrong and terminated itself. We chose the reboot because this is production and that was the fastest most reliable way to clean it up. OK, so has anybody else ever seen this kind of problem? Rebooting a computer to solve a problem with connections seems a tad drastic. Thanks, David

I've had the same thing on v7.24 UC6. The fix, arrived at after 5-6 weeks of "analysis" from HP and Informix, was to set RESIDENT=1. Neil David Coburn wrote in message <7gcoj3$2ih$1@news.xmission.com>... > >Odd thing happened last night. At some point one instance (of several on >the machine) decided to stop accepting any connections. There was one user >thread returning data and several others (including an oncheck) that had >finished and were sitting around waiting for something to do. I mention >the oncheck because the pid in onstat -g ses didn't exist; the person who >ran oncheck earlier did a ^C to terminate it. onmode -yuck hung, as did >onmode -z tid. Anything that required a connection to the database, in>fact, would hang indefinately. There was nothing in the log file to >indicate any problems, and in fact checkpoints were still occuring every 5 >minutes, right on schedule. ipcs output looked entirely normal, as did >netstat. > >About the only way to clean this up was to either reboot the computer, or >whacking shared memory segments until the engine realized that something >was wrong and terminated itself. We chose the reboot because this is >production and that was the fastest most reliable way to clean it up. > >OK, so has anybody else ever seen this kind of problem? Rebooting a >computer to solve a problem with connections seems a tad drastic. > >Thanks, > >David > >