[H-GEN] sig 11s, annoying items.

Bruce Campbell bc at thehub.com.au
Tue Aug 25 13:20:30 EDT 1998


Currently I've got two machines which are terminating random processes
with the dreaded signal 11 (along with a handful of sig 6s and an 8).  One
of the happens to be my main proxy server, the other is zerlargal itself.

Both are running FreeBSD (2.2.7 and 3.0 respectively), have 128Meg of
memory, and normally are under reasonable to heavy load.  Brutha has run
Linux in the distant past, without problems but under a much lighter load
(FreeBSD at the time did not support its particular SCSI chipset).
Zerlargal's current incarnation has not run Linux.

On both machines, I've gone through the sig-11 FAQ
(http://www.bitwizard.nl/sig11/), disabling cache memory, juggling memory,
playing with wait states, changing power supplies etc.  No go.

The ones among you quicker to jump to a conclusion are probably thinking
``aha, they're both the same (dodgy) CPU series''.  Well, you'd be wrong. 
Brutha (the proxy server) is a P133 MMX on an ASUS motherboard (mmm,
onboard SCSI, rather pricey tho), while Zerlargal is a K6/233 on a generic
normal/ATX motherboard.  Both have EDO ram (different brands) rated at
60ns and treated accordingly in the BIOS.  70ns in the BIOS makes no
difference, nor does completely swapping the memory.

Drives?  Rather disparate.  SCSI cards?  Only Adaptec cards here ('cept
for that VL Buslogic one on the backup machine ;) ) so thats a
possibility... 'cept for I've never heard of the adaptecs failing in such
an odd manner.

Normally the machines are at rather seperate physical locations (ie,
completely different power sources) so an odd inconsistency in their
completely different brand power supplies can be ruled out. 

Squid 1.2b24 does reboot both machines when placed under typical network
activity (simulated by a handful of recurring netcats from several
machines to the squid port with random retries, with a few strobes to the
ICP port for good measure), which is a tad annoying for the proxy server. 
Squid 1.2b21 just hangs, sometimes exits (to be restarted by the script
there for just such an occurence), or just forgets to free(). 

I have to admit, its got me at the moment.  Anyone care to suggest
anything that I might be missing here?  I'm going home to catch up on
sleep.  

Speaking of motherboards and Zerlargal, its former troublesome cyrix 5x86
motherboard has had a BIOS flash upgrade sourced for it which apparently
fixes some of the problems I was seeing... now if only I could convince
the thing to actually successfully boot into DOS to apply said upgrade :(

--==--
Bruce.

Squid on RZ58s is scary.  Don't do it.





More information about the General mailing list