On Thu, Dec 20, 2007, Neil Harkins wrote:
> > You need to leave room for the swap state logs
>
> I guess I assumed that squid was factoring in the size
> of its swap.state file. I didn't see any log msgs to the
> effect that it noticed the disk was full this time around.
> Will squid expire more agressively if it notices write failures?
I'm not sure to be honest.
> > or put them somewhere else. That could explain your filling of partition..
>
> I've never seen a config that allows one to relocate swap.state
> outside of the cache_dir. If it is configureable, please enlighten me.
> Not that I think moving it out is a good idea...
http://www.squid-cache.org/Versions/v2/2.6/cfgman/
http://www.squid-cache.org/Versions/v2/3.0/cfgman/
Its one of the _dir commands; swapstate_dir or something?
> Can a limit on the max # of objects be set per cache_dir,
> so that number * the swap.state object struct size can
> be subtracted from the partition size to use as much
> as possible without going over?
The swap.state file is a log, rather than a mapped file; it grows without
bounds until you hit the file size limit on your platform (2gig is easy
to hit if you're not careful and have compiled Squid right :) and don't
run squid -k rotate daily.
> What's "best practices" here?
> Configure it for 95% and reduce
> it by a little each time it fills?
Configure it for less, run squid -k rotate at least daily.
You dont' want to fill a filesystem to 95% anyway; fragmentation will decrease
the already crap performance...
Adrian
-- - Xenion - http://www.xenion.com.au/ - VPS Hosting - Commercial Squid Support -Received on Thu Dec 20 2007 - 18:56:47 MST
This archive was generated by hypermail pre-2.1.9 : Tue Jan 01 2008 - 12:00:02 MST