OS: RHEL 5.3 64 Bit
Squid: squid-2.7.STABLE9
Decided to do a very light load test on a new Squid server (specs above).
I'm running the AUFS file system
cache_dir aufs /usr/local/squid/var/cache 20000 16 256
The test ran for about an hour with about 10 clients loading random web
pages. The CPU/ WAIT IO never got above 2%, so I was surprised to see a
few of these warnings in cache.log:
squidaio_queue_request: WARNING - Queue congestion
Saw about 5 or 6 of these over the hour but they seemed to slow down (or
stop) as time went on. I did some reading and found this:
Queue congestion starts appearing if there are 8 operations queued
and not yet handled by the IO threads. The limit is then doubled
(to 16) before the next warning appears. And so on...
Does this mean the warning limits are doubled each time? If so, that would
seem to just mask the problem, no? I'm not sure if I have a problem or
not.
-Mike
