Re: [squid-users] poor performance

From: Matus UHLAR - fantomas <[email protected]>
Date: Fri, 31 Aug 2007 15:26:20 +0200

> On 30-Aug-07 My Secret NSA Wiretap Overheard Lutieri G. Saying :
> > /dev/da0s1a on / (ufs, local)
> > devfs on /dev (devfs, local)
> > /dev/da0s1e on /tmp (ufs, local, soft-updates)
> > /dev/da0s1f on /usr (ufs, local, soft-updates)
> > /dev/da0s1d on /var (ufs, local, soft-updates)
> > /dev/da0s2a on /cache (ufs, local, soft-updates)
> > devfs on /var/chroot/named/dev (devfs, local)

On 30.08.07 12:20, Nicole wrote:
> I am curious why you opted for a seperate slice for the /cache?
> I found that to be slower than just having a seperate partition, as odd as
> that seems.

pardon, he has separate _partition_ - 1 and 2 are partitions, a to f are
slices. However, the performance should very similar, if not the same.
How did you find out the difference?

For proxy cache, separate disk should be used.

> > cache_dir diskd /usr/local/squid/cache/cache1 5120 16 256 Q1=128 Q2=100
> > cache_dir diskd /usr/local/squid/cache/cache2 5120 16 256 Q1=128 Q2=100
>
> I'm confused. Above your showed a df with a /cache yet here you are
> aloting space on /usr?

that might be a reason why is his system that slow ;)

Also, hawing two cache directories on one disk is usually pointless and may
slow system a bit.

And I think that on system with threads the aufs should perform better...

> > cache_mem 1536 MB

Having this much of cache_mem usually wastes memory and doesn't leave it
free for other usage. Note that memory cache is only used for objects
fetched from network. I'd decrease it to 512MB

> > maximum_object_size 64 MB
> > minimum_object_size 0 KB
> > maximum_object_size_in_memory 2560 Kb
>
> 64 Mb is a pretty large objest size.

Not if he uses LFUDA replacement policy and decent cache size. While he has
10G, that will a waste...

> So your df above was from the old box? Why not show the new unit?
> As someone else suggested. Having one disk be, say as big as needed for OS
> and programs and the rest be cache and then the 2nd disk be all cache.

not if those SAS disks are mirrored. I have one (yet) such proxy, system and
other data are mirrored, but I use rest of space proxy data as two separate
cache dirs on two separate disks.

> > After 10 minutes squidclient mgr:info return me 70 clients and the
> > speed connection goes very low for the users.
>
> Perhaps try a netstat to see if you have too many users hanging on. Based on
> the amount of dropped connections that may be likely.

turning half_closed_clients off might also help.
 

-- 
Matus UHLAR - fantomas, [email protected] ; http://www.fantomas.sk/
Warning: I wish NOT to receive e-mail advertising to this address.
Varovanie: na tuto adresu chcem NEDOSTAVAT akukolvek reklamnu postu.
2B|!2B, that's a question!
Received on Fri Aug 31 2007 - 07:26:24 MDT

This archive was generated by hypermail pre-2.1.9 : Sat Sep 01 2007 - 12:00:04 MDT