Re: server about to crash

Peter T. Breuer (ptb@it.uc3m.es)
Wed, 8 Jul 1998 12:33:38 +0200 (MET DST)


"A month of sundays ago Meelis Roos wrote:"
>
> PTB> But I probably soon will. I've been having problems for months with one
> PTB> of my servers which has been crashing regular as clockwork every five
> PTB> days. I have recently narrowed the suspects down to a kernel or other
> PTB> memory leak of about 10MB a day (64MB machine).
>
> Just a thought:
>
> A month ago, I checked "ipcs" on my 64M machine. It showed 48M (!) on shared

Yes this is a good one. I have been worried about lost and wailing
shared memory segments. Shamefacedly, I must admit I don't know how to
locate or kill them.

> memory segments. I ipcrm'ed them and they haven't come back so I have no idea
> what's causing them. Still waiting :)

ipcrm, eh? Where from?

> --
> Meelis Roos (mroos@tartu.cyber.ee)

The server crashed as I predicted, about 8 hours after I posted. In
contrast to the 2.0.25 kernel, the 2.0.33 kernel behaved well right up
to the end. No exponential slowdown while dying. It died when the
watchdog kicked in for not being able to launch a process at all ...
approximately.

At the time I was recompiling the 2.0.33 kernel under gcc 2.7.2 (not
2.8.0) on a different machine. I brought the server up under
2.0.33-2.7.2 and have been abusing it since. Looks stable to me.

Just completed two nfs mounts of a 4GB system, and tarred the whole lot
into /dev/null. I see a little extra memory use, but about 1-3M, not much
at all. I'll keep on looking.

Peter

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu