Since I set:

vm.filemin=2
vm.filemax=4

in my /etc/sysctl.conf file, my two boxes hardly swap.  All other values set to
defaults.


On Fri, 22 Oct 2004 13:18:52 +0300 (EEST)
Arto Selonen <arto@selonen.org> wrote:

> Hi!
> 
> On October 7th, I sent the following to current-users:
> 
> 	http://mail-index.netbsd.org/current-users/2004/10/07/0014.html
> 
> Since I got no replies to that one, I'm making another effort here.
> 
> Basically, what I am trying to achieve is to keep 'squid' in memory.
> I've set the following limits (after having sent the above email):
> 
> 	vm.anonmin = 65
> 	vm.execmin = 2
> 	vm.filemin = 10
> 	vm.anonmax = 80
> 	vm.execmax = 5
> 	vm.filemax = 15
> 
> And I've checked the proc.<pid-of-squid>.rlimit values for the running
> process. I've also set reasonable values to those in /etc/login.conf
> (I haven't yet modified the squid startup script to use ulimit).
> 
> However, no matter what I do, as soon as RSS of squid grows to
> ~330-350 MB, it starts to throw pages to swap (ie. swap usage starts
> to grow and RSS of squid shrinks). At the same time, file cache is
> kept at ~350-400MB range. For a 1 GB system that really does not use
> memory for anything else besides squid, this is not bad, but I would
> like to at least feel like *I* am the one controlling the balance
> between memory and disk caching (from squid's point of view).
> Squid also seems to take a performance it once the swap issue surfaces
> (from looking at the access statistics; kind of makes sense if swap is
> used).
> 
> Why can't I shoot myself in the foot, and control the VM usage?
> 
> 
> <RANT>
> In case this could be a bug of any sort, here are some thoughts
> surrounding the issue. After having set the vm.anon{min,max}
> and friends, I thought everything was set. After seeing squid bounce
> back from a magical ~350MB limit several times, I learned that even
> though root session has ulimits set to max/unlimited, memorylocked
> was set to ~350MB (or 35% of 1GB). I thought that I had finally found
> the reason, and set new, higher limits, and started squid again.
> I made sure that sysctl reported those limits for the running process.
> Now that the process is again hitting an invisible limit of ~340MB
> I am starting to think that either there is a bug, or I'm still missing
> some pieces of this puzzle. Is there no way for a process to grow
> to ~500MB size in a 1GB system where the OS is the only other major memory
> consumer?
> 
> And no, I don't claim to understand what I've done so far, though
> I think I've grasped some of the most trivial explanations from
> that 'Bad response' thread on current-users. I thought I knew
> how the system would behave depending on the vm.{anon,exec,file}{min,max}
> values, but my system seems to be making fun of my expectations. :)
> Who is to blame, man or machine?
> </RANT>
> 
> Any comments or suggestions are welcome...
> 
> 
> Artsi
> 
> PS. In case it makes any difference, I'm running -current from ~20041012
> -- 
> #######======------  http://www.selonen.org/arto/  --------========########
> Everstinkuja 5 B 35                               Don't mind doing it.
> FIN-02600 Espoo        arto@selonen.org         Don't mind not doing it.
> Finland              tel +358 50 560 4826     Don't know anything about it.
> 


-- 
Julio M. Merino Vidal <jmmv@menta.net>
http://www.livejournal.com/users/jmmv/
The NetBSD Project - http://www.NetBSD.org/