Re: memory usage squid 1.1.3

From: Ed Knowles <ed@dont-contact.us>
Date: Tue, 21 Jan 1997 10:05:13 -0400

G'day Dirk!

Dirk Lutzebaeck wrote:
>
> Duane Wessels writes:
> >
> > The reported "Process Size" is determined from the getrusage(2)
> > system call. Some systems may report it as kbytes, some as pages.
> > We're probably multiplying the ru_maxrss value by getpagesize()
> > when we should not be. Check your man page for getrusage(2)
> > and let me know if it reports kbytes or pages.
>
> Is this problem resolved?

The 'problem' can only be resolved by people seeing that Maximum Resident Size
looks odd and then reporting if ru_maxrss is in bytes or pages. At the moment
_SQUID_SGI_, _SQUID_OSF_ and BSD4_4 are treated as kbytes, everything else as
pages.

> But when you look at the process size squid eats up 7.7 MB (which is
> ever increasing in the long run). It never should be more than
> cache_mem if I understand it correctly:

No. The cache_mem setting is used to control 'Pool for in-memory object' ...
you also have to allow for all the other bits and pieces. See section 3.17 of
http://squid.nlanr.net/Squid/FAQ.html for more info.

> This is what the Linux getrusage(2) man page says:

You need to go down another page where it explains the different fields:

The fields are interpreted as follows:

     ru_utime the total amount of time spent executing in user mode.

     ru_stime the total amount of time spent in the system executing on
                    behalf of the process(es).

     ru_maxrss the maximum resident set size utilized (in kilobytes).

What does your ru_maxrss say?

Later
Ed

-- 
Ed Knowles aka Jasper				   Phone : +61 2 9385 4962
E-mail: ed@fatboy.geog.unsw.edu.au	           Fax   : +61 2 9313 7878
            What I lack in morals I make up for in principles.
Received on Mon Jan 20 1997 - 15:18:12 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:34:07 MST