Re: [squid-users] Performance problems after unclear load situation (2.5.STABLE2, Solaris 8)

From: Henrik Nordstrom <hno@dont-contact.us>
Date: 03 Feb 2003 20:42:39 +0100

What OS?

What Squid version? (STABLE1 is not a Squid version.. 2.5.STABLE1 is)

What tuning have you applied to TCP/IP for high rate proxies? Most OS:es
needs to be slightly tuned in their TCP/IP to handle above ca
100-150req/s.

Are you caching? If so, how is your cache directories configured (number
of drives, drive configuration, filesystem, cache_dir type etc)

Regards
Henrik

mån 2003-02-03 klockan 17.48 skrev Jost Krieger:
> Our cache has been hopping along nicely for quite a long time,
> doing up to 100req./sec. without many problems.
>
> Since a few weeks, however, we repeatedly see the following scenario:
>
> When the general network load goes up (we have reserved soem bandwidth for the
> cache), first the number of requests/sec drops dramtically, then the
> reaction time increases steeply (up to more that a second for a *hit*).
>
> Unfortunately, this state is stable :-) Stopping and restarting squid helps.
>
> We have checked the rest of the load on the machine, but there's enough CPU left,
> there's enough memory left, and it isn't the listen queue either.
>
> I'm running out of ideas where to look. We are already planning a separate cache
> on a Linux system, but I would really like to figure this one out.
>
> Thanks for any help
> Jost
> --
> | Jost.Krieger+sig@ruhr-uni-bochum.de Please help stamp out spam! |
> | Postmaster, JAPH, resident answer machine am RZ der RUB |
> | Pluralitas non est ponenda sine necessitate |
> | William of Ockham (1285-1347/49) |

-- 
Henrik Nordstrom <hno@squid-cache.org>
MARA Systems AB, Sweden
Received on Mon Feb 03 2003 - 12:42:45 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:13:13 MST