Re: [squid-users] Problem with Restarted Squid Stable 2.6_19 Add

From: Adrian Chadd <adrian@dont-contact.us>
Date: Tue, 22 Apr 2008 10:34:47 +0800

On Mon, Apr 21, 2008, Nicole wrote:

> Hi
> The swaplog files are about 156 megs. Altho I have some servers that have
> swaplogs that are 1.6 gigs but are fine as they the servers have never been
> restarted.
>
> I have never run squid -k rotate. I have another server that just started
> exibiting the same sort of behaviour of slowing down. I tried lowing the
> available disk size to force it trim some files and did a squid -k rotate but
> it was still slow.

Hm. Well, you should run squid -k rotate once a day.

> It's getting to be kind of a drag having to contantly wipe out the cache every
> few months when they get to a larger size. The disks are 146 Gig and are only
> 56% full. I am trying to keep lowering the alloted available cache size to see
> if there is a sweet spot.
>
> How often should squid -k rotate be used. It seems like there are various
> opinions on its usage and frequency.

Are you using AUFS on a recent FreeBSD (FreeBSD > 5.x) ?

I've built a 4 x 18 gig test cache here (not enough RAM atm to run more of the
disks :/) and the rebuild-from-swaplog is quite a bit faster than rebuild-from-
cache. Check cache.log and see if its rebuilding from swaplog, or from cache
(it'll say DIRTY.)

I'd start by looking at iostat to see if Squid is doing a decent amount of IO,
and vmstat / top to see if Squid has hit 100% CPU. The rebuilding logic happens
synchronously - it doesn't use the async io routines to do background processing.
I guess I should make it do that but that'll have to wait.

Adrian

-- 
- Xenion - http://www.xenion.com.au/ - VPS Hosting - Commercial Squid Support -
- $25/pm entry-level VPSes w/ capped bandwidth charges available in WA -
Received on Tue Apr 22 2008 - 13:31:11 MDT

This archive was generated by hypermail 2.2.0 : Thu May 01 2008 - 12:00:04 MDT