Re: [squid-users] swap.state becoming enormously big.

On Tue, 3 Jul 2001, Igor Karpov wrote:

> Hi!
>
> Got a strange problem. My squid (2.4.stable1) is running under
> FreeBSD-4.3-STABLE.
>
> cache_mem = 48 MB
> maximum_object_size 4096 KB
> cache_dir ufs /usr/local/squid/cache 500 16 256
>
> (I also tried diskd with the same result).
>
> The load is rather low; everything looks OK, but once a week or two the
> swap_state starts rapidly grow, maybe about 1 MB per second. During this
> squid itself occupies the proccessor resources up to the limit. Usually
> it is
>
> PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND
> 2274 nobody 2 0 52380K 51684K poll 2:12 0.10% 0.10% squid
>
> After short time the box gets out of space and crashing.
>
> Any ideas? How to track down this problem?

Sounds like Squid might be trying to remove cache files really
quickly. This might happen if someone requests a really
large file, but your 'maximum_object_size 4096 KB' should also
prevent caching of those large files.

It might also be due to something like a denial of service
attack. Maybe somebody is sending a lot of requests to
your cache? You should be able to tell that by looking
at access.log.
Received on Fri Jul 13 2001 - 11:23:16 MDT