Mark Elsen wrote:
Hello,
One of our proxies has a problem which causes the response time to explode.
We've been unable to find a cause for this behaviour, but I want to implement
a workaround: when the median response time grows over 1 second (normal
behaviour is a median response time of about 100/110ms) I want to
automatically restart the proxy.
- Wouldn´t that explode the issue you are trying to tackle ?
No. As I said, median time is normal about 110ms, and the highest (apart from
the strange irregular peaks) I've seen is about 150ms. Proxy averages 200-250
requests/sec, peaking to 300 on the busiest time. Median time of 1000ms or
higher indicates a problem, and so far, restarting the proxy solved this problem.
The provider of the upstream proxy does monitoring of the parent and of Squid
from an internal client, and the parent proxy (a loadbalanced Finjan cluster)
has no problems at all when Squid is acting weird.
- Better is too solve the issue ;
We're replacing the proxy with a loadbalanced cluster in a few months, so
restarting when the problem occurs works for me.
1) - SQUID version.
Currently 2.5.10.
2) - Post access log entry, for long taking URL-s
Don't have them at hand, but -all- files take extremely long, even 2 byte
files. The only files served at normal speed are files delivered from cache.
And error log shows nothing, apart from the usual 'illegal hostname with a _'.
3) - Does SQUID at all times have sufficient memory ?
Yes, machine has 2G memory, Squid uses about 600MB, and it's the only service
running on this machine.
Joost
--
Du hast mein Herz zerrissen, meine Seele geraubt
Das es so enden würde hätt` ich nie geglaubt [Aus der Ruinen -]
Ohne Rücksicht auf Verluste, hast Du meine Welt zerstört [L'Âme Immortelle]
Eine Welt, die vor kurzem nur uns beiden hat gehört
