Re: [squid-users] Any tips on analysing squid crashes?

On 12/11/10 04:02, Declan White wrote:
> I have a squid that falls from the sky once a week. It is running with -C to provide coredumps, however, the coredumps contain only:
>
> Solaris 2.9 pstack output:
> core 'core.squid' of 14292: (squid) -YC -f /etc/squid3.conf
> ffffffff7dba8ddc _kill (ffffffff7e10dac8, ffffffff7e10cc38, 1127bc, 10604c, 2, ffffffff7e10dac8) + 8
> ffffffff7e0064d4 __1cH__CimplRdefault_terminate6F_v_ (ffffffff7e10dac8, ffffffff7e10cc38, 190298, 10604c, 104e40, ffffffff7e0064d0) + 4
> ffffffff7e0062b4 __1cH__CimplMex_terminate6F_v_ (ffffffff7e10de40, 0, 0, ffffffff7e10de40, ffffffff7e10c978, 1) + 24
> ffffffff7e007078 _ex_rethrow_body (ffffffff7e10de40, 1, 0, 105984, ffffffff7e4c0608, 100272ba8) + 88
> ffffffff7e006fc8 __1cG__CrunKex_rethrow6F_v_ (100514d50, 1000c9640, 100514d50, ffffffff7e10de40, ffffffff7e4c8528, 1) + 40
> 000000010014d0b4 $XDGBaHss9zzMWWh.__1cNSquidMainSafe6Fippc_i_ (4, ffffffff7ffffc18, ffffffff7ffffc40, 0, ffffffff7e10dec8, 100514d50) + 94
> 00000001000714dc _start (0, 0, 0, 0, 0, 0) + 17c
>
> This appears to be saying that a 'catch' handler in SquidMainSafe was
> triggered, yet nothing appeared in the errorlog, and nothing in the stack
> hints as to why it died.

One of those hex values should be the exception details. I'm not sure
how "core" utility display them though.

SquidMain also dumps to stderr any details about the original exception
it can. That may be in your syslog at the time of death.

>
> Also does anyone have any rough estimate of CPU usage of squid3 vs squid2?
> It looks like squid3 takes more CPU than squid2.
>
> Declan

Exact squid release version and details of this to squid-dev please.
Last few cache.log entries before it restarted or stopped completely?

Amos

-- 
Please be using
   Current Stable Squid 2.7.STABLE9 or 3.1.9
   Beta testers wanted for 3.2.0.3