CSM Backup Serverfarm for ACS (NOT for HTTP)?

[Ingo de Jager]

Hi,

How can I configure a backup serverfarm, which is at another location with it's own CSM's, for ACS traffic? Redirect won't work - it's only for HTTP.

Thanks,

Ingo

[Gilles Dufour]

the only way is to do client nat.

Let's say you have 2 sites A and B.

on CSM-A you create a server farm that contains a real with CSM-B vip address.

You also do 'nat client' with a pool of ip address from subnet-A.

Traffic coming to CSM-A will be sent to CSM-B with source ip of subnet-A, so the response from CSM-B will come back to CSM-A which can then forwards to client.

since the CSM needs to say both side of the traffic, this is the only solution.

Gilles.

[Ingo de Jager]

Thanks Gilles, I'll give it a try.

Ingo

[Ingo de Jager]

Gilles,

I tried your suggestion and did a quick check by taking my real server out of service. A sh mod csm 4 real det showed that connections are sent to the vip of the second CSM but it also show connections failed.

Below is an extract from my config. It should work but maybe I missed something... I removed some FT configs just to make it read easier.

module ContentSwitchingModule 4

vlan 504 server

ip address 1.1.1.2 255.255.255.0

!

vlan 502 client

ip address 1.1.2.2 255.255.255.0

gateway 1.1.2.1

!

natpool CSM-BACKUP 1.1.2.180 1.1.2.190 netmask 255.255.255.0

!

serverfarm TECHD2_BACKUP

nat server

nat client CSM-BACKUP

failaction purge

real 1.1.3.3

inservice

health retries 30 failed 60

!

serverfarm TECHD2_SRVR00

nat server

no nat client

failaction purge

real 1.1.1.3

inservice

health retries 30 failed 60

!

vserver TYGRRA01-APPCSV

virtual 1.1.2.3 any

serverfarm TECHD2_SRVR00 backup TECHD2_BACKUP sticky

sticky 1

persistent rebalance

inservice

!

[Gilles Dufour · Accepted solution]

this config is good.

You need to check on the other csm if it can ping/reach 1.1.2.180-190.

Capture a sniffer trace and see what is going on.

Thanks,

Gilles.

[Ingo de Jager]

I replicated the entire setup in a lab and it worked like a charm. Now I just need to find the problem in the production environment.

Thanks again.