RE: 2.3s3 memory leak?

Hi Wayne

Thanks for coming back to me so fast, here are the additional details, does
it add up?

cache_dir ufs /var/spool/squid 100 16 256
This is 100MB but I guess the number of objects varies massively, perhaps if
I used 20 16 256 I might be better and cache mem is only 4 MB

Simon

-----Original Message-----
From: Wayne Smith [mailto:wcsmith@myrealbox.com]
Sent: 13 July 2000 11:59
To: squid-users@ircache.net
Subject: Re: 2.3s3 memory leak?

Simon

You didn't saw how large your cache_dir was. Squid uses 100+ bytes per
object that it's tracking. Number of users isn't as important as number of
objects when it comes to memory useage. Take a look at the number of
objects in your cache. IIRC, the normal formula is 10 megs of memory for
every gig of hard drive space. So if you definied a 4 gig cache_dir, you'd
want at least 40 megs just for squid tracking the objects, plus memory for
squid cache_mem (for hot and intransit objects), plus memory for the OS and
filecaching.

You probably either need more memory, or a smaller cache_dir.

HTH

Wayne

----- Original Message -----
From: "Simon Fawcett" <simon.fawcett@smart.uk.com>
To: "'Duane Wessels'" <wessels@ircache.net>; "'Brandon'"
<dataframe@yahoo.com>
Cc: <squid-users@ircache.net>
Sent: Thursday, July 13, 2000 6:26 AM
Subject: RE: 2.3s3 memory leak?

> I feel that squid 2.3s3 has a memory leak, two people using a new squid 4
a
> week or ten days makes the process grow until it starts swapping to disk
> only 48meg but for two users? on a system with 64meg.
>
> When the process is restarted it is about 8 meg. I have tuned the
> squid.conf to take account of all in the FAQ memory reduction section.
Using
> redhat 6.2 Kernel 2.2.14-12.
>
> Any help would be appreciated
>
> Thanks Simon
Received on Thu Jul 13 2000 - 07:32:36 MDT