you might be seeing FD exhaustion..
find the publicly available utility 'lsof' (a must have in
any sysadmin's toolbox, IMHO), and do an 'lsof -p <pid>' on the
squid process, and see how many FD's it's using.
Our squid process exhibits the same problems every week or two,
BUT we're not running port redirectors. We see the same symptoms,
we kill the squid process to restart it, and everything works fine..
Ed
-- On Thu, Aug 07, 1997 at 10:32:39AM +1000, Paul Black said: > I'm wondering if people could help to give me an idea of what is the most > stable squid configuration, ie which base level and which patches will > give the most stable squid configuration. > > Basically the problem I have is this: > > We are running squid 1.1.11 (base version, no patches), it is running > pretty stable, stays up for days and days under moderate load. > Now when I turn the port 80 redirectors on, the load on squid is > increased quite a bit (maybe 200K trans/hour). Now squid is running > quite nicely with a heavier load, but every 5 - 8 hours, something > will happen and users will start to see very long response times (3minutes on > web pages), the access log shows that all traffic is now going direct. > > The fix at this point is killall -1 squid, whereupon squid goes back to > working normally. > > I'm stumped as to even where to start looking for the solution to this problem. > Has anyone come across anything like this? This is why I'm asking about what > the most stable version and set of patches for squid is. > > The software we are using is Linux 2.0.29, DPT Raid 0 SCSI, Libc 5.3.12 > I'd be interested to know what versions of Libc people who are running > as similar (but more stable) configuration are using. > > > Many Thanks... Paul > > > >Received on Thu Aug 07 1997 - 11:49:48 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:36:03 MST