Re: [squid-users] High load on squid machine after switching from reiserfs to ext3

From: Joe Cooper <joe@dont-contact.us>
Date: Tue, 21 May 2002 13:34:07 -0500

In that case you are at an impasse, I'd say. ext3 doesn't work with
Squid (it may have improved in later versions...try the Red Hat 2.4.18
kernel--I was able to get it built and installed on a 7.2 system with
only two package upgrades to satisfy dependencies).

ReiserFS is trustworthy, in my experience. I have about 40 Squid boxes
in the field, some for two years, all running on ReiserFS. I have only
seen one FS corruption...and it was due to a faulty disk that had to be
replaced. In fact, given the troubles I've had with ext3, ReiserFS is
the only filesystem I trust for Squid. (I have a little test box
locally that started reporting disk full when there is 10GB of free
space, causing Squid to exit...it's running ext3).

But then, as I said, it is wise to recompile with ReiserFS built into
the kernel, rather than as a module. I always have, because there was a
time (maybe even as late as 2.4.9) that ReiserFS would cause problems
when built as a module. Turn off the checks and proc interface, as well.

Just MHO, of course.

Andreas Jung wrote:
> And I don't trust reiserfs...I have seen two boxes with corrupted
> reiserfs partitions...
>
> Andreas
> ----- Original Message -----
> From: "Joe Cooper" <joe@swelltech.com>
> To: "Andreas Jung" <andreas@zope.com>
> Cc: <squid-users@squid-cache.org>
> Sent: Tuesday, May 21, 2002 13:54
> Subject: Re: [squid-users] High load on squid machine after switching from
> reiserfs to ext3
>
>
>
>>Yep. Don't use ext3. ;-)
>>
>>I've been able to crash kernel 2.4.9 (the Red Hat RPM kernels) at will
>>when running a Squid workload on ext3. I never looked very deeply into
>>it, because the performance of ext3 isn't comparable to ReiserFS. It
>>may be worth your time to grab the latest kernel SRPM, and rebuild it
>>with ReiserFS built into the kernel (rather than as a module).
>>
>>I've had no problems with 2.4.9-31, on a quite large number of boxes.
>>
>>Andreas Jung wrote:
>>
>>>We are running Zope behind Squid 2.4Stable6 with squid in acceleration
>>
> mode.
>
>>>The squid box (dual Pentium III 1 GHz, RH 7.2, Linux 2.4.9-21smp, 2GB
>>
> Ram)
>
>>>has during busy hours a normal load of 0.2-0.3 . From time to time
>>>we see spikes over some hours where the load average of the machine
>>>is higher than 1.5 although there are no spikes in the CPU
>>>utilization. Also there is no increase in the number of incoming
>>
> requests.
>
>>>The only change we made was the switch from reiserfs to ext3 for the
>>
> cache
>
>>>partition since reiserfs corrupted its partition some time ago.
>>>
>>>Any insights on this issue?
>>>
>>>Thanks,
>>>Andreas
>>
>>--
>>Joe Cooper <joe@swelltech.com>
>>Web caching appliances and support.
>>http://www.swelltech.com
>>
>>
>
>

-- 
Joe Cooper <joe@swelltech.com>
Web caching appliances and support.
http://www.swelltech.com
Received on Tue May 21 2002 - 12:25:17 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:08:11 MST