> > that means that the average object size lost is 29.6 bytes per request.
That makes sense to me, since I'm also experiencing a memory leak on our 2.2
p3 Squid cache running on 2.0.36/transparent proxy, but NOT on our other
Squid cache...
(Our transparent proxy/redirector only has 32 meg of memory, of which Squid
NORMALLY uses 12 meg for the first few hours (only 1 meg of cache on hard
drive, mem_pools set to 1 meg, etc) but leaks... when it uses approximately
27 meg total it starts to swap out and lag really really bad... could it be
the handling of Host: headers and something not getting freed there? (I'm
thinking average size of the Host: header would be around 27 bytes or so)
I've had to temporarily disable our cache server because I was having to
restart it about once a day to free the memory! (And I have tried your two
memory leak patches--no luck) The leak was also in 2.2 patch 2, but was not
in 2.1.patch2... our main cache is still running 2.2.patch2 and has yet to
run out of memory, but it doesn't do the transparent redirecting.
--Jeff Barrow, Internet Connections, Inc.
Received on Mon Jun 28 1999 - 17:05:13 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:47:02 MST