Zach Seils Wed, 10/31/2007 - 13:03

Mike,


There is currently no way to clear a single connection. Can you provide some additional information about the "hung" connection?


Thanks,

Zach



mike-greene Wed, 10/31/2007 - 13:25

Thanks Zach,


The app (the only app I've had issue with) in question is a home grown product and I'm not quite sure whats causing the WAE to keep connections so long. I've check asymetric routing etc...but traffic is routing correclty. I think it's an app issue but the developers here don't like it when I say that:) My local SE and CCIE worked on this alittle last week in there lab but I had to finally remove the client and server from my WCCP ACL at the central location. Once I did that it removed the hung connections at the remote WAE (it's inline at the site) but when I logged into the 7300 at the central location I noticed they never cleared on that device. It's not building any new connections just hanging onto the old....for instance some have been open for 1000+ hours. I just don't want to force a reboot to get them cleared but it looks like my only option.


Mike

Zach Seils Thu, 11/01/2007 - 07:20

Mike,


TCP keepalives are enabled by default, so a connection can/will stay open if the host systems are responding to the keepalive messages. If the connection is actually closed on the host systems, but still shows up in the WAE, that's a problem. If that's the case, please provide the output from the 'sh tfo conn' command for the connection.


Thanks,

Zach



mike-greene Thu, 11/01/2007 - 14:17

Thanks Zach,

Please see the attachment. This is one of about 1800 stuck connections from the same two servers on this 7300. On September 17th I upgraded all my WAE's to 4.0.13.b23 because of the disk IO issue so they were all rebooted at around noon CDT. I'm not sure if the connection issues were happening before the upgrade or not.


Mike





Zach Seils Mon, 11/05/2007 - 22:31

Mike,


I spoke with Engineering about this, and the feeling is that you are running into defect CSCsk64139. This issue will be resolved in the 4.0.15 maintenance release later this year.


If you can reproduce the problem, we would like to see the output from "debug tfo conn netio" and "debug tfo conn comp-mgr".


Thanks,

Zach



ksumenon Tue, 11/06/2007 - 00:13

Maybe you can try disabling the EPM feature of WAAS and check.


Sunil

mike-greene Tue, 11/06/2007 - 13:30

Hi,

Thanks for the feed back.


Sunil,

If I have the right command it looks to already be disabled..


EPM (MS-PortMapper) adapter is disabled


Zach,

I was able today to reboot my 7326 and clear the connections. I'm able to reproduce the issue but it does not happen very quick. Seems like every ten or fifteen minutes one will hang. This application is polling every five seconds over the WAN...could that be the issue? I'm not sure why the developers have it set to poll that frequent but it generates alot of logs...I'll do my best to get you a hanging connection though.


Mike

Actions

This Discussion