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

From: "Mike Diggins" <diggins@McMaster.CA>
> I run
> Squid 2.4S7 on Solaris 8. I did open a bug report but, so far, I have been
> unable to convince Squid (or Solaris) to dump Core during a crash. I don't
> think my problem is SWAP as I have lots of memory and lot's of Swap space
> free. I'm just testing Squid 2.5S2 and was hoping this would resolve my
> crashes but based on your report I'm not so sure.

Not sure what squid port you are using but we use one lower than 1024 so are
unable to start squid as anything other than root. What I did do was use
coreadm to enable global and per-process setid cores. By default (at least
on my prod and dev Sol8 squid boxes), these were both disabled. Only
"per-process core dumps" were/are enabled but those won't work. Some good
explanations of what does and does not core are on google. Here is the best
explanation I found for Sun's newish coring policies:
http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&oe=UTF-8&selm=916ahn%2476
0%241%40engnews2.Eng.Sun.COM

> Henrik, are you saying that if I shutdown Squid and restart it under the
> Squid userid that it would dump core during a segmentation/bus error?

Wouldn't dare speak for Henrik :) but my understanding is that if you use a
port above 1024 then yeah you can do that and since "per process core dumps"
are already enabled by default, it should dump core. You can use coreadm to
specify where it should dump core but should probably use the squid conf
parameter coredump_dir - make sure it is to a directory that is both larger
than your max process size and to which squid has write perms.

Let the list know in the end whatever you discover as someone else may have
the same problem (e.g. me :).

thanks,

Adam
Received on Wed May 14 2003 - 11:50:03 MDT