CSS - all connections dropped when suspending service

[eisenberg]

CSS11501

version: 07.50.3.03

I just suspended two prod services with the understanding that all existing connections would be allowed to timeout. Once I suspended these prod services the connection count (20 total) started dropping slowly until it reached 0 connections. By the time I realized this the damage had been done. Has anyne seen this behavior when suspending services?

thanks

[Jorge Bejarano]

Hello,

You statement is correct, here you have a document which confirm what you said:

http://www.cisco.com/en/US/docs/app_ntwk_services/data_center_app_services/css11500series/v8.10/configuration/content_lb/guide/Services.html#wp1026795

"

When you suspend a service, the CSS rebalances the remaining services using the failover setting.

Suspending a service does not affect existing  content flows, but it prevents additional connections from accessing the  service for its content.

"

Checking some release notes of the version you have, we can notice there are some bugs which might be related with suspension

That should not be an expected behavior

Hope this helps!!!

Jorge

[Jorge Bejarano]

Release notes:

http://www.cisco.com/en/US/docs/app_ntwk_services/data_center_app_services/css11500series/v7.50/release/note/RN750_X.html#wp187493

[eisenberg]

In my case i suspended all services associated to a single content rule. In the past i have suspended a single service leaving at least one service active, all connections on the suspended service were allowed to timeout. Is it possible that this problem occured due to all services associated to the same active content rule having been suspended at the same time? By suspending all services the CSS would not be able to rebalance based on failover settings due to not having any active services.

[rodrguti_2]

Hi,

The problem here is not suspending the services, the problem is that there are no services alive remaining in the content rule, when the content rule is suspended or all the services on that content rule are not available the VIP will go down, which means that the CSS is not able to hold the connections there in the flow table, even if the service is suspended. That is why the behavior is different when you have one service alive, and you suspend the other one.

What you are seeing is expected behavior.

Also another command that you can use in order to speed the user to re generate the connection once the service goes down, is:

flow-srvdown-reset

Under the content rule.

Hope this help.

Rodrigo