Re: [squid-users] squid dies with signal 6

>I had some problems with exits like this, it was due to bad hardware - a
>SCSI
>disk which was dying. However, this may not be the case for you.
>
>The reason you have had few replies is because it is genuinely difficult to
>pin down what this might be. Have you looked at memory usage of the cache
>before it dies, numbers of requests, cache.log, access.log... the processes
>running... and of course, you haven't told us what your setup is like:
>aufs,
>diskd, ufs? how much cache_mem and disk space for cache_dir, how much
>physical
>memory you have, etc?

Hello,

I was previously on RH6.2 Linux (with the kernel.org source tree, not the
redhat one) with 1GB RAM, 3 18G 10K SCSI disks on 3 10GB cache partitions,
aufs, running Squid2.4STABLE4, on a PIII-800. I am also running wccp, 80Mb
cache_mem.

I have since rebuilt the machine on a RH7.2 P4-1.8Ghz, 2GB RAM and it did
not die on signal 6 for 6 days. The cache dirs filled up and about 48 hours
after being at 97% usage, I got three signal 6 aborts in about 5 minutes,
still with no cache.log entries as to why. I was hoping my GLIBC was to
blame, however I suppose it is not since these crashes have continued on the
new distro.

I am still using the ip_wccp.o module on kernel 2.4.9-31 (redhat proprietary
source tree); could this be a potential reason? Other than the fact it is a
"unsigned module", the kernel does not complain about it and it compiles and
inserts correctly.

Thanks,

Ron.

_________________________________________________________________
Chat with friends online, try MSN Messenger: http://messenger.msn.com
Received on Mon Mar 18 2002 - 08:53:47 MST