ACE - TMG 2010 locks-up when overwhelmed with requests

[edgarfc254]

ACE service module providing loadbalacing to four TMG 2010 boxes. All works well when the Internet is availble. Problem comes when there is an Internet outage for more that 5 minutes, the TMG boxes simple lockup and cannot service any kind of connection.

The situation is brought back to normal by stopping the ACE from servicing traffic and rebooting the TMG boxes when the Internet is back up.

I am looking for ways to limit connections that are sent to the server farm or to stop requests being sent to the server farm when there is an Internet outage. Any ideas?

[Daniel Arrondo Ostiz]

Hi,

I assume that the outage comes from the fact that client connections pile up on the TMG boxes since there is no internet connectivity to serve them. Is this correct?

If so, what you could do is configure a maximum number of connections allowed per server. When that limit is reached, the ACE will stop sending new connections to it. This way, you should be able to stop sending new connections before the server gets overloaded.

For more details on how to configure this, have a look a the link below

http://www.cisco.com/en/US/docs/interfaces_modules/services_modules/ace/vA2_3_0/configuration/slb/guide/rsfarms.html#wp1086777

I hope this helps

Daniel

[edgarfc254]

Your assuption is correct. Will work on configuring the maximum number of connections and monitor the effects.

Thanks Daniel

[Paul Pinto]

Good day gentlemen,

Would it not be an option to implement a health probe to a site or two on the www (eg. http://www.google.com) an apply to each TMG rserver. If there is an Internet outage the probes would fail and the servers would te taken out of service. This is assuming you do not already have this in place.

This in conjunction with the maximum number of connections allowed per server may be an optimal solution?

Just a thought.

Regards

Paul

[Rafael Mendes]

Edgard,

Can post your configuration for this?

I trying configure too...

Tks!