Re: [squid-users] High CPU usage problem on Squid 2.6 STABLE9

From: Jason Taylor <[email protected]>
Date: Tue, 30 Jan 2007 16:07:28 -0500

Are you logging access_log to syslog?
I used to have "access_log syslog" in my squid.conf and just turned it off.
I saw my iowait drop from 46% to 1% within 10 seconds.

/Jason

Andrew Miehs wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 30/01/2007, at 8:51 PM, Robert wrote:
>> My connection speed is 45 Mbit max, but real traffic is about 30-35
>> Mbit including P2P.
>
> I am not really sure how high the interrupts are allowed to go...
>>
>> 20:33:10 CPU %user %nice %sys %iowait %irq %soft
>> %idle intr/s
>> 20:33:15 all 1,00 0,00 1,00 13,20 1,30 22,95
>> 60,55 9004,01
>> 20:33:15 0 2,20 0,00 1,20 1,00 5,21 89,98
>> 0,40 9004,41
>> 20:33:15 1 1,20 0,00 2,00 34,87 0,00 2,00
>> 60,12 0,00
>> 20:33:15 2 0,20 0,00 0,00 6,81 0,00 0,00
>> 92,99 0,00
>> 20:33:15 3 0,40 0,00 0,60 10,22 0,00 0,00
>> 88,98 0,00
>>
>
> As you see - you have a quite a high IOWait, and high number of interrupts.
> CPU User is NOT the problem. I would take this to mean either disk or
> network issues.
>
>
>>> Is the latency on sites in the cache? or 'new' sites?
>> Hmm i'll see these but eg. openning googll site with squid enabled
>> take 1-3 sec.with squid disabled take max 1 sec
>
>>> Are you running as a 'transperant' proxy? or have the users entered
>>> the proxy details per
>>> hand or per proxy.pak.... ?
>> transparent with iptabes redirect command from port 80 to port 3128.
>
> Do you have this problem as well when you do NOT run in transparent
> mode? (But still via proxy)...
>
>>> I would probably run aufs rather than diskd - but this should not
>>> make a real difference..
>>> You have also only configured 3Gs of cache (3x 1G)...
>>> You may want to try disabling the disk cache altogether and see how
>>> that works - you need to
>>> compile with -enable-storeio=diskd,aufs,null
>>>
>> cache_dir null ? in squid.conf ?
>
> cache_dir null /tmp
>
> this will cause squid not to cache to disk (well not really anyway)
>
>
>>> I would also increase my cache size to perhaps, 2G.. (How much RAM
>>> does the machine have?)
>>>
>> 2 GB RAM
>
> Try increasing your cache size to 1G, and disabling the disk cache.. - -
> see if that improves the
> matter at all.
>
> Stupid question - are you actually gaining anything by caching this
> data? as I can not imagine
> that such a small cache (3G) actually helps in any way. You must have a
> pretty low hit ratio..
>
> Summed up:
>
> 1. Disable transparent caching and try normal browser caching - does
> this help?
>
> if this doesn't change anything then try
>
> 2. change "cache_dir null /tmp" and increase you cache memory to 1G
>
> Cheers
>
> Andrew
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.1 (Darwin)
>
> iD8DBQFFv6d2W126qUNSzvURAuUGAJ45D1BNoI08B8QgjrP/hOprpYDgugCeK9cS
> u6bJPKsTJ7PYyAI0dcpKhO0=
> =5gtW
> -----END PGP SIGNATURE-----
Received on Tue Jan 30 2007 - 14:07:41 MST

This archive was generated by hypermail pre-2.1.9 : Thu Feb 01 2007 - 12:00:01 MST