[squid-users] FATAL bus error but no core - suggestions?

From: Adam <[email protected]>
Date: Tue, 13 May 2003 12:49:13 -0700

Hello again,

Last week squid crashed around 8am (usage is moderate, not high at that
hour) on my PROD box with the error " FATAL: xmalloc: Unable to allocate
24576 bytes!" Per the FAQ it can either be a lack of swap or "maximum data
segment size." Since ulimit reports unlimited for that, I am assuming that
swap becomes in short supply. There is only 384MB of on disk swap though
with 1GB of RAM, there is usually some left over for swap. cache_mem is
374MB and cache_dir is "aufs 7072 16 256." I do plan on increasing the
size of the ondisk swap filesystem during our next maintenance window (Wed
night next week), so that may also fix this new problem. Is my sizing off.
It seems odd this error has only happened once so the situations where it
completely runs out of swap is rare. Will making ondisk swap 1GB help?

What happened today was squid crashed with the error "FATAL: Received Bus
Error...dying." which, according to the three threads I found on the mailing
list, would indicate a bug (FAQ shows how to analyze the core).
Unfortunately no core was generated. I've set my coredump_dir to /logs
which is owned by squid hence read/writable by squid. Fortunately in both
instances squid was able to restart itself but I am wondering if there is
something else I need to do to enable core creation for squid. I doubt it
but am posting, just in case. I haven't disabled core creation (i.e. using
coreadm or anything). Server is an Ultra 60, 1GB of RAM, Solaris 8, Squid
2.5STABLE2.

thanks,

Adam

today's cache.log extract:
2003/05/13 10:29:11| httpAccept: FD 12: accept failure: (130) Software
caused connection abort
FATAL: Received Bus Error...dying.
2003/05/13 10:29:25| storeDirWriteCleanLogs: Starting...
2003/05/13 10:29:25| WARNING: Closing open FD 12
2003/05/13 10:29:25| 65536 entries written so far.
2003/05/13 10:29:25| 131072 entries written so far.

cache_mem 374 MB
cache_dir aufs /cache 7256 16 256

"top" output a few hours after the crash:
Memory: 1024M real, 461M free, 337M swap in use, 703M swap free

PID USERNAME THR PRI NICE SIZE RES STATE TIME CPU COMMAND
25317 root 20 22 0 315M 313M sleep 30:38 13.48% squid
Received on Tue May 13 2003 - 13:50:07 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:16:38 MST