Re: Shared memory hogs

>Hi. Ever since upgrading to 7.30.UC6 on my HP9000 (HPUX 10.20) I find that >on occasion one instance will, on occasion, begin allocating shared memory >segments until it runs out of all available memory. At that point I get >the expected "could not allocate" errors in the message log. Nothing, >including applications, general usage, onconfig, OS, or hardware >configuration, has changed except for the version of Informix. (And no, >going back is not an option at this point.) > >After experiencing this on at least one occasion, onmode -F failed to free >them even when there were no users other than Informix on the system. Most >often the command works OK. Tech support tells me that there is nothing in >the problem tracking database, and certainly nothing in the bugs that would >account for this behaviour. > >Has anybody else seen this, or (better still) seen this, found a >resolution, and been able to move forward? This is so intermittant that it >is hard to pin to a causal event, but it is a production database that I >can ill afford to be taking from the users on a regular basis. Also, this >is but one instance (of this version of Informix) of three on the same >computer, and the other two do NOT experience this type of problem. Mmmmm....this _looks_ a lot like the 7.2x problem where certain queries combined with OPTCOMPIND=2 would cause the server to allocate far more virtual memory than it needed, which would never get freed. And that was on HPUX. Is OPTCOMPIND set to 2 on your box, and can you try setting it to 0? Just a guess, but I hope it helps. :-) -- "I'll Be Back" Obnoxio ************************************************ Sane? Hell, if I was sane, why would I be here? Get Your Private, Free Email at http://www.hotmail.com