What version of squid are you using?
If it's 1.1 then you're probably I/O bound and your disks are slowing down or
you're running out of memory. On squid 1.1 this can happen at an HTTP rate
of about 3 HTTP requests per disk spindle eg. 5 disks = 15 HTTP reqs/sec.
I'm talking rough numbers here. As for memory, squid needs to be practically
fully resident. If less than 90% of your squid process is resident (use top)
then you have a memory problem.
> It seems like every evening, after squid running for few days without
> restarting, with the 3GB cache is slowing down. That happens where there are
> about 300 active http sessions going to squid machine. We have 5 machines
> and all of them are showing the same problem.
>
> In the morning, everything works great, (without restarting squid), but of
> course at that time, there are only 50 sessions open.
>
> What I mean slowing down, is when you make request, it takes 5 seconds until
> netscape clears thje window and everything starts downloading slowly.
> Sometimes it will stop in half and wait for 1 minute or so.
>
> Each squid box is P300+ machine with 256MB of ram and 3 GB swap size. It
> runsFreeBSD 2.2.6
>
> In the evening the squid proccess uses about 30-40% of CPU time.
>
> What is wrong? Is that something wrong with the config file? We noticed
> similar behaviour with squid running on BSDI, and Solaris so it is not the
> OS.
>
> What are the limitations of squid?
Received on Mon Aug 24 1998 - 16:28:59 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:41:41 MST