Re: How to tell that my cache is overloaded...

In article <m0whenj-0006dnC@dingo.theplanet.co.uk> you write:

>I'd like a list of things to watch so that I can tell when our cluster of
>cache machines is running out of steam. What points, other than general
>response time, which tends to depend very much one what you are looking
>at, show a cache system that is just getting more hits than it can cope
>with??

I wrote a timer patch (see http://www.IAEhv.nl/users/devet/squid/) which
reports a.o. the time Squid spends waiting for select(2) to return:

97/07/03 22:40:41| time 30016 idle 27174(6233) diskread 217(77) diskwrite 40(227) ...

This means that during the last 30.016 seconds Squid was waiting 27.174
"idle" seconds for select(2) to indicate that new data is available on one
of the filesdescriptors. This means it was not busy at all :-). The
numbers in brackets are the number of calls made, in this case select(2)
calls.

When the reported idle time drops below 1000 ms, you will get very worse
performance. The other things measured (diskread/diskwrite/etc.) can give
some hints about where Squid is spending its time but these measurements
are not yet complete.

Arjan
Received on Thu Jul 03 1997 - 13:58:43 MDT