Late or no RST on a TCP connection

From: Niclas Sodergard (nickus@gmail.com)
Date: Sun Feb 17 2008 - 07:54:50 EST


Hi Sun Managers,

I have a problem that I can't seem to understand. We have a V490
running Solaris 10u3. We have 10 zones on the server and we each zone
has 6 IP addresses each. We are running a custom java app in each of
these zones. This system worked fine until a couple of days ago and it
has been in use for almost a year.

When you do a "telnet 192.168.1.2 29" you expected to get a
"Connection refused." Basically, when we send a SYN to a port where
this is no listener we expect go get back a RST. This works for a
couple of hours but after a while the RST gets delayed, it takes about
4-40 seconds before they come back. After a while we don't get them
back and telnet says there is no response after approx. 4 minutes. We
have a high amount of connection probes to ports without any listener
but it is my understanding that this shouldn't cause a problem. The
problem also persist if I do this from localhost so it is not a
network or firewall issue. If we snoop the interface we can in fact
see that there is no RST sent back (or very late).

I've tried to tune various ndd variables (e.g tcp_time_wait_interval)
but to no avail. I don't expect to get a solid answer from you sun
managers without more information but I don't seem to find much. *stat
(including netstat) shows no drops, no illegal packets. There is not
that much load on the machine either. Where could I start to look? Has
anyone experienced a similar problem? I've been beating my head
against the wall for four days and I have really no clue how to
continue.

TIA,
Niclas

-- 
Have a look at my blog for sysadmins!
http://aspiringsysadmin.com
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers


This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 23:42:47 EDT