CSM- web server failover takes over 3 minutes

[mariolaniel]

we have a csm card in a 6513. A small serverfarm is configured with two real server and the vserver is configured as well, pointing to that server farm. We have a sticky of 5 define with the vserver. When we bring down one of the web server it takes a little over three minutes from the time that the webserver is actually down to failover (for the CSM to say that the server is in failed state). So for at least three minutes the web services are not available. Does anybody knows if there is a way to get around that. Everything commands are set to default.

[Gilles Dufour]

you first need to setup a probe.

ie:

probe myprobe icmp

interval 5

retries 3

Then assign the probe to the serverfarm

serverfarm myservers

probe myprobe

...

In this example, the CSM will ping the real servers every 5 seconds. 3 errors in a row and the server is brought down - so the time to detect a server down will 3 x 5 = 15 seconds.

You can adjust the timers to reduce it if necessary.

Be careful not to reduce it too much, as this will mean lot of probes, so lot of traffic and may be false detection of a server down.

Regards,

Gilles.