[squid-users] swap.state eating the entire slice

Hi.

I'm using squid caches since long time for now, I have production caches
running 2.7.x, 3.0.x and 3.1.x.
About a year/year and a half ago I started to encounter a problem when
squid eats the entire slice for it's swap.state file.
I still cannot localize this problem, the only thing I know - that this
is somehow connected to squid restarts. For example on servers with long
uptime this doesn't happen at all. But I have a bunch of branch servers,
that are shut down for the nighttime. Almost 2-3 times per month I have
this problem. The cache_dir size on those servers is like 1-2 gigs (I
use squid mostly because of it powerful authorization capabilities), but
the swap.state during some "bad" conditions can eat entire slice, no
matter how big it is - 40 Gigs, 80 Gigs, on one server I saw swap.state
of 120 Gigs. After it eats out the space it crashes.

Does anyone see similar issue ? Is this a configuration problem or a
squid issue ?

Thanks.
Eugene.
Received on Mon Jun 21 2010 - 07:54:37 MDT