At 11:30 AM +0200 9/24/99, Andi Kleen wrote:
>In article <v04205504b410a4c9757c@[130.216.108.150]>,
>[EMAIL PROTECTED] writes:
> > These have been hanging around for hours now.
>
> > ihug# netstat -n | grep 2532
> > tcp 0 59766 130.216.35.104:3128 130.216.35.102:2532 LAST_ACK
>
> > gate.cs# netstat -n | grep 2532
> > tcp 0 0 130.216.35.102.2532 130.216.35.104.3128 FIN_WAIT_2
>
> > ihug is a linux box running redhat, linux kernel linux-2.2.7
> > gate is a Digital alpha running Digital Unix 4.0C .
>
> > Neither end is timing these connections out and removing them from their tables.
>
> > Obviously the process on gate has closed a socket and sent the fin to ihug. ihug
>has sent the ack, closed the socket and sent the fin to gate, moving to the LAST-ACK
>stage. Gate has received the ack to its fin and moved to the FIN_WAIT_2, but never
>received the fin from ihug. Both are now waiting for the other and neither times out.
>
>
>ibug still has a big send queue, so it probably has not even send out the
>FIN yet. Normally Digital Unix should RST when it receives new data on
>a closed socket. If it discards the data without acking this could occur,
>but the sender should time out after some time. The receiver should time
>out too [not specified in the original specification, but most stacks
>start a timer in FIN_WAIT_2 to avoid the protocol bug]
It doesn't look as if they do time out. It has been several days now and they are
still there and the number of these grows slowly over time.
root# netstat -n | grep 2532
tcp 0 59766 130.216.35.104:3128 130.216.35.102:2532 LAST_ACK
root# netstat -n | grep LAST_ACK | wc -l
249
26gate.cs# netstat -n | grep 2532
tcp 0 0 130.216.35.102.2532 130.216.35.104.3128 FIN_WAIT_2
27gate.cs# date
Tue Sep 28 10:05:26 NZST 1999
28gate.cs# netstat -n | grep FIN_WAIT_2 | wc -l
305
>What happens on the wire between the two boxes?
I haven't monitored the traffic between the boxes. There are several GBytes passing by
each day over hundreds of thousands connections. picking the ones out that don't close
properly wouldn't be easy.
>
Rob.
-------------------------------------------------------------------------------
-- Rob Burrowes -- [EMAIL PROTECTED]
-- Computer Services Manager -- Phone:+ 64 + 9 + 373-7999 ext 7972
-- Department of Computer Science, -- Fax: + 64 + 9 + 373-7453
-- University of Auckland, -- Cell Phone: + 64 + 25 + 731-856
-- Private Bag 92019, Auckland, -- Home: + 64 + 9 + 8128-173
-- New Zealand. -- http://www.cs.auckland.ac.nz/~rob/
-------------------------------------------------------------------------------
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
