Re: [SQU] Seg faults on Redhat-6.2/Alpha

Hello,
I found that in the kernel i had to setup the proccessor type on linux to be K6 rather than PPRO. and i have 139days uptime now, and no cache downtimes :)
maybe worth thinking about re-compiling your kernel.
thankyou
Jimmy Stewpot

> Hi.

> I'm using 2.3-STABLE3 on Athlon / 512MB RAM, 23GB swap and have the same
> problem. This proxy used by 500-700 clients and we have from 600 to 1100
> requests per minute. Cache dies very often for (once or twice a month)
> with the same message.

> I use thi to configure:
> --prefix=/usr/local/squid --enable-icmp --disable-wccp --enable-snmp
> --enable-heap-replacement --enable-cache-digests --enable-forw-via-db

> More details about configuration are available upon request.

> Thanx for your time.

>> Andrew Gatward wrote:
>>
>> Hopefully someone can help with this... so here goes...
>>
>> I have two AlphaServer DS10s here with the following spec:
>>
>> Alpha EV6 466MHz cpu
>> 1Gb RAM
>> 1 x 9Gb system disk
>> 5 x 9Gb cache disks
>> 3Com 3c905-FX 100-base-FX NIC
>>
>> These systems are running RedHat 6.2 / Alpha.
>>
>> Squid was compiled using the following options:
>>
>> --prefix=/usr/local/squid --enable-dlmalloc --enable-async-io
>> --enable-icmp --disable-wccp --enable-kill-parent-hack --enable-snmp
>> --enable-arp-acl --disable-http-violations --enable-underscores
>>
>> Squid fires up fine... runs for a little while, with the process size
>> growing continually, then eventually I get the following in the log
>> file, and the process restarts:
>>
>> 2000/08/25 08:57:54| NETDB state saved; 0 entries, 4 msec
>> FATAL: Received Segment Violation...dying.
>> 2000/08/25 09:32:47| storeDirWriteCleanLogs: Starting...
>> 2000/08/25 09:32:47| WARNING: Closing open FD 7
>> 2000/08/25 09:32:47| 65536 entries written so far.
>> 2000/08/25 09:32:47| Finished. Wrote 103630 entries.
>> 2000/08/25 09:32:47| Took 0.1 seconds (834044.3 entries/sec).
>> 2000/08/25 09:32:50| Starting Squid Cache version 2.3.STABLE4 for
>> alpha-unknown-
>> linux-gnu...
>>
>> The cache does receive quite a number of hits, as shown from the
>> cachemgr output:
>>
>> Connection information for squid:
>> Number of clients accessing cache: 83
>> Number of HTTP requests received: 7058
>> Number of ICP messages received: 8670
>> Number of ICP messages sent: 8878
>> Number of queued ICP replies: 0
>> Request failure ratio: 0.00%
>> HTTP requests per minute: 663.8
>> ICP messages per minute: 1650.3
>> Select loop called: 101536 times, 6.283 ms avg
>>
>> Before I recompiled squid with the --enable-dlmalloc option, the
>> process would stay running, but the process size would continually
>> grow, eventually exceeding the amout of physical RAM in the machine.
>> Last time it did this, the process size hit 1256MB before grinding to
>> a halt.
>>
>> Does anyone have any ideas as to what I can do with this? I can send
>> copies of squid.conf if necessary.
>>
>> Kind regards, and thanks for an otherwise great product!
>>
>> Andy Gatward
>> Web Solutions
>> Nortel Networks Optical Components Ltd.
>> Phone: +44 (0)1803 667134
>> ESN: 795-7134
>>
>> Company No: 2515751
>> Registered Office: Maidenhead Office Park, Westacott Way, Maidenhead,
>> Berkshire SL6 3QH

> Travis
> --
> *************************************************************
> * Kdyz jsem v tezke situaci, reknu si: "Hlavu vzhuru, mohlo *
> * by to byt horsi!" A opravdu, za chvili to horsi je. *
> *************************************************************

--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html