Re: FIN_WAIT2 backoff on 1.3.92
BearHeart / Bill Weinman (bearheart@bearnet.com)
Tue, 23 Apr 1996 09:59:35 -0500
At 10:25 am 4/23/96 +0100, Alan Cox spake:
>> >I think this is an old problem, and not due to something recent; I
>> >recall it happening with much earlier kernels.
>>
>> I've been having this problem since about 1.3.72, maybe earlier.
>> But not in 1.2.13.
>>
>> I would love to know what it is, and how to kill the dead
>> sockets.
>
>You can't. Don't use 1.3.x test kernels for busy web sites, they currently have
>bugs that cause sockets to get stuck. I have no idea why but since various
>changes have been made to the TCP layer I'm not actually suprised they are
>still there and in need of fixing.
If nobody tests it on a busy web site, how will we know the
problems exist?
I hope I didn't sound like I was complaining that the bugs are
there--I expect that at this stage of the game. I'm simply poniting
them out and trying to make sure that they are noticed by whomever
is able to fix them. It would be nice if they are not there when
Linus decides to give the HedgeHog a bath, and call 'er 2.0.
That's why I'm running it on my busy web site.
+--------------------------------------------------------------------------+
| http://www.bearnet.com/ || http://www.weinman.com/wew/
| Author of The CGI Book -- http://www.cgibook.com/