Re: Informix crashing UNIX on HP

Silvester_Valiente writes: > I had never heard of this problem before on HP. We have been running > 7.20 - 7.22 > for over a year with as many as 10 x 8 Mb extents at one time (ie > during heavy load). See my previous posting on this. With HP/UX, >3 shared memory segs per process causes performance problems. Depending on your performance expectations, you may not be seeing a "problem". Odds are, though, that if you retune to 1 virt seg, you will see a performance boost in Informix and overall Unix. > Unfortunately there is no messages in the message log. Unix panics and crashes > On the system console we get a > > "Stored message buffer up to panic: > vuseg=810000" > > and the Unix box shutdowns and restarts!! > > The curious thing is that I can allocate Virtual mem using onmode -a > and release > it via onmode -F. But when I monitor the mem segments with onstat -g seg > and I see that the free blocks are almost all exhausted Informix will > at this point > cause the box to reboot. When it reboots, is it always when you have the above 10 virtual segs? Or do you see it with fewer segs? If the former, I'll bet you are hitting shmseg (max # segements, kernel tuneable). OnLine can be somewhat less than graceful at times when it hits memory limitations (e.g. set SHMTOTAL to a real value, then run a query that needs more memory than this: POP goes the weasel!), though I've never seen it bring Unix down with it. It does sound like you are pushing the limits of the system. How much real and virtual (i.e. swap) memory do you have on this box? -- Dave "That's just my opinion - I could be wrong." - Dennis Miller