Re: Swap space on squid boxes

On Tue, 25 Nov 1997, Gregory Maxwell wrote:

> > Depends how the rest of the machine is configured I guess; our top-level
> > proxy is a Sun Ultra E2/2170 with 512Mb RAM and >30Gb of disk (only
> > roughly 17Gb is used for the cache). The machine has two 512Mb swap
> > partitions (on separate disks) giving it a total of 1Gb swap space, or
> > 1.5Gb virtual memory.
[...]
> What you dont include, which would be useful info: How much mem (real +
> swap - os cache) do you have is use?

According to "top":
- "squid" process is 341Mb, 315Mb resident
- memory is 497Mb real, 16Mb free, 355Mb swap, 1080Mb free swap

According to "swap -l":
- swap partition 1 is 1,048,784 blocks, 747,648 blocks free
- swap partition 2 is 1,049,744 blocks, 744,944 blocks free

Anything else? (Actually, while sitting there I saw free real memory drop
to just under 7Mb.)

> Frankly, the only reason for a cache computer to have more then a tiny
> amount of swap is to keep the computer from dieing if things go crazy..

To keep it going at all, to permit for surprise "large objects", to allow
the machine to do "other things" (not necessarily other
services/applications, but other cache-related tasks) .. as noted
previously, since there's no resources taken away from the actual cache
storage area itself there's nothing lost by erring on the safe side in our
case.

I read a lot in this list of people having trouble with their Squids
becoming very slow, crashing when memory is low, &c. - our box doesn't
have these problems as when necessary, Squid (or ftpget, or...) can grow
to be quite large in order to handle in-transit data. *shrug*

As always, YMMV. Ours works. :-)

Cheers..

dave
Received on Tue Nov 25 1997 - 20:42:43 MST