Re: TCP SACK issue, hung connection, tcpdump included

[Jan Engelhardt <jengelh@computergmbh.de>]

On Jul 29 2007 08:45, Willy Tarreau wrote:
>On Sun, Jul 29, 2007 at 06:59:26AM +0100, Darryl L. Miles wrote:
>> CLIENT = Linux 2.6.20.1-smp [Customer build]
>> SERVER = Linux 2.6.9-55.ELsmp [Red Hat Enterprise Linux AS release 4 
>> (Nahant Update 5)]
>> 
>> The problems start around time index 09:21:39.860302 when the CLIENT issues 
>> a TCP packet with SACK option set (seemingly for a data segment which has 
>> already been seen) from that point on the connection hangs.
>
>Where was the capture taken ? on CLIENT or on SERVER (I suspect client from
>the timers) ? A possible, but very unlikely reason would be an MTU limitation
>somewhere, because the segment which never gets correctly ACKed is also the
>largest one in this trace.[...]

I had something similar (and most strangely enough, it only seemed to happen
when the source and/or destination address was outside a given subnet). It
boiled down to that some switch acted really strange and God knows why. A
tcpdump on *both* sides (server+client) each showed that the *other* peer sent
a FIN/RST. Well in the end I settled with setting sys.net.ipv4.tcp_sack=0,
no idea if the broken switch got replaced in the meantime.
Well, just letting you know that it is not limited to computers.


	Jan
-- 
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/