Re: Linux 2.0.29-ISS and squid = too many fds open?

Hi,

during the last days I investigated this problem a little bit and I think
it's caused by the increased FD lifetime patch (BTW, Duane could you change
the diff-format to unified? easier to read) from 1.1.8 to 1.1.9.

It happens under the following conditions:
- Client requests data
- Client hits stop button before the transfer has started (finished? I'm
not sure about that), ERR_CLIENT_ABORT is logged
- the FD now has a big lifetime of around 200 minutes (?!) until it gets
closed, thus the increasing number of sockets in CLOSE state

I think 1.1.9 misses to catch one specific situation which causes this. Due
to the long timeout it's easy to hit the max FD limit also on a small
cache, it depends on the access patterns of the users.

I hope this info helps the guys that are familiar with the code a little bit.

Ciao,
Franz.

--
--------------------------------------------------------------------------
URLs	<mailto:Franz.Sirl@munich.netsurf.de>
	<http://homepages.munich.netsurf.de/Franz.Sirl>
--------------------------------------------------------------------------