Re: Informix in "comatose" mode?

Yes. I have experienced the same but running on obsolete hardware (Altos 15000 with 64M ram, 2-486 CPUs), with an obsolete OS (SCO Unix 3.2.4.2), with an early and somewhat buggy version of IDS (7.10UC3and in non-Y2K compliant environment which Informix will not support, I thought I knew why. Informix told me to upgrade. Usually when IDS goes into hang state eventually you'll see a (CKPT REQ) checkpoint requested, at which point it will hang till doomsday. Your only solution: bounce. Sometimes this state is provoked by user threads sitting in a critical section - look for the -X flag when you do an onstat -u. You may have some luck cleaning these up with an onmode -z sessionid. I've tried a few things (Informix support gave me a few suggestions which may help but don't truly solve the problem), one of which was to increase userthreads significantly (from 200 to 400). The problem sometimes seems to be resource starvation at certain times, i.e. when a lot of users (on my system 10-15) are logged in and working - so I increased LRUs, and page cleaners, and lowered LRU MAX and MIN to 3 and 2 respectively, a la Art Kagel (well, not so radically). But then without explanation, sometimes when there is no active user session, the engine goes into hang mode. I know this doesn't help, but it might stoke the fires. Yours, Nick At 10:04 AM 5/20/99 -0400, Alanoly Andrews wrote: > >Hi, > >Occasionally, our Informix Server (7.22 UC2 on AIX 4.1.4) goes into >what I would call "comatose" mode, ie, technically alive but practically > >dead. All "onstat" commands work but after an abnormal delay and >with the initial message of "No response from server process". The >log file does not indicate anything out of the ordinary. >No connections can be made to any db on the server. "onmode -k" >does not work. The only way to handle this seems to be to kill >shared memory (or the master daemon) and do "oninit". >And when the server comes back, everything seems normal. > >Some of you probably may have had the same experience. Any >thoughts on what is causing this and whether it is possible to revive >the server without first killing it? > >Alanoly J. Andrews > > ********************* Nick Nobbe NLS/BPH Library of Congress NNob@loc.gov (202) 707-0548

Nick Nobbe wrote: > Yes. I have experienced the same but running on obsolete hardware (Altos > 15000 with 64M ram, 2-486 CPUs), with an obsolete OS (SCO Unix 3.2.4.2), > with an early and somewhat buggy version of IDS (7.10UC3and in non-Y2K > compliant environment which Informix will not support, I thought I knew > why. Informix told me to upgrade. > > Usually when IDS goes into hang state eventually you'll see a (CKPT REQ) > checkpoint requested, at which point it will hang till doomsday. Your only > solution: bounce. Sometimes this state is provoked by user threads sitting > in a critical section - look for the -X flag when you do an onstat -u. You > may have some luck cleaning these up with an onmode -z sessionid. Have you tried 'onmode -O' (capital O). IDS warns you, but it usually get you thru the CKPT REQ state. > > > I've tried a few things (Informix support gave me a few suggestions which > may help but don't truly solve the problem), one of which was to increase > userthreads significantly (from 200 to 400). The problem sometimes seems to > be resource starvation at certain times, i.e. when a lot of users (on my > system 10-15) are logged in and working - so I increased LRUs, and page > cleaners, and lowered LRU MAX and MIN to 3 and 2 respectively, a la Art > Kagel (well, not so radically). But then without explanation, sometimes > when there is no active user session, the engine goes into hang mode. > > I know this doesn't help, but it might stoke the fires. > > Yours, > Nick > > At 10:04 AM 5/20/99 -0400, Alanoly Andrews wrote: > > > >Hi, > > > >Occasionally, our Informix Server (7.22 UC2 on AIX 4.1.4) goes into > >what I would call "comatose" mode, ie, technically alive but practically > > > >dead. All "onstat" commands work but after an abnormal delay and > >with the initial message of "No response from server process". The > >log file does not indicate anything out of the ordinary. > >No connections can be made to any db on the server. "onmode -k" > >does not work. The only way to handle this seems to be to kill > >shared memory (or the master daemon) and do "oninit". > >And when the server comes back, everything seems normal. > > > >Some of you probably may have had the same experience. Any > >thoughts on what is causing this and whether it is possible to revive > >the server without first killing it? > > > >Alanoly J. Andrews > > > > > ********************* > Nick Nobbe > NLS/BPH > Library of Congress > NNob@loc.gov > (202) 707-0548