I have been having trouble with my cache overflowing for a long while, and
have finally tracked it down. By switching from the AIO threaded IO back
to standard/unlinkd, I stay within 3% of my target disk availability. I
used to vary 30%, and often simply run amok. My squid process wouldn't
stop until it had filled the partition (several gigs past it's limit).
Now, this is all gone. Yes, I'm no longer running a multithreaded squid,
but piece of mind is well worth it. I am running on HPUX 11.0, if that
matters. I had the thread limit at 64, and no indication of the queue
filling.
Comments? Has anyone else seen this too?
-- mat.Received on Fri Jan 01 1999 - 19:06:48 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:43:54 MST