Re: [squid-users] Intermittent High Utilization

From: Hendrik Voigtländer <hendrik@dont-contact.us>
Date: Thu, 03 Jun 2004 19:05:36 +0200

> I wasn't aware MRTG could monitor squid.. I'll have to look into that.
mrtg can monitor probably a mouse pad, if it speaks snmp :-) You can use
any snmp-capable monitoring tool (cacti, cricket, whatever). I use mrtg
  because we use it since ages and I just integrated squid into it.
However, we are going to switch over to cricket soon. Seem to be easier
to configure and maintain if you have multiple similar targets.

>
> I have built and reinstalled from source using aufs. I still get the
> utilization spikes, but it seems to have settled a little now it is when
> there are a pretty significant amount of users accessing the proxy
> concurrently. It still has the same behavior though. X amount of time
> very low utilization (2-10 percent cpu) X/2 amoutn of time high
> utilization 99 percent solid. I have tried switching to heap
> replacement policiies. If this was a 2.6 kernel issue wouldn't other
> people also see it?
>
Yes, but probably it is an issue only with your particular build of the
SuSE-kernel. Need an example? SMP-Kernel of SuSE 8.1 Pro loads the
network module just fine but is unable to send a single bit over the
wire. Happened to me on two different platforms (SMicro Server with Dual
CPU, IBM Desktop with HT).

> I have not tried the 2.4.x kernel but we did have this same setup in
> place and without the utilization spikes on Redhat 9 (2.4 kernel) and
> Mandrake 9.1 (2.4 kernel).
>
Well, why not give it a shot then? If you try an RPM, it is easier than
to compile squid :-)

Good luck.

Regards, Hendrik
Received on Thu Jun 03 2004 - 11:05:56 MDT

This archive was generated by hypermail pre-2.1.9 : Thu Jul 01 2004 - 12:00:02 MDT