CSM - Persistent Rebalance

[duchesne_ced]

Hi,

I've got a question regarding this particular command ('persitent rebalance') in the CSM configuation. Based on the documentation, i had tought that inside one TCP connection between the client and the CSM, all http requests would have been load-balanced among all my reals behind but it's not the case.

Here is a short explanation of my architecture:

I've got clients that connect with HTTP to a reverse proxy. The RP fowards these HTTP requests to a vserver with the command 'persitent rebalance'.

Between the RP and the CSM, the RP keeps the TCP connection opened with an idle timeout of 15 minutes so, in the real life, the TCP connection is never cleared. When i look for sessions opened on the CSM, i've got one connection with my RP and one connection to one real. In other words, HTTP GET/PUT inside the TCP connection are not load-balanced.

Is there any misunderstanding from my side concerning this command 'peristent rebalance'

regards

Cedric

[Gilles Dufour]

Cedric,

the CSM will only keep 1 connection open with the real server.

Did you verify if the connection with the real is changing ?

If it's not changing, you will have to make your vserver an L7 vserver by using a url match all policy.

This will force all the traffic to be inspected.

For L3-4 rules we only look at the SYN and then we do not care about the rest of the traffic.

Regards,

Gilles.

[duchesne_ced]

Hi gilles,

As soon as i clear the connection, i will use another real.

I've defined a cookie insert... and the vserver is referred as L7 (with show ..... vserver name... detail) but still no balancing between my reals.

By the way, i've seen when i use 'cookie insert', whenever the GET has a cookie from CSM or not in the header, the csm will insert it (set-cookie) anyway in the HTTP 200 OK.

regards

Cedric

[Gilles Dufour]

the CSM will always reinsert a cookie, even if no cookie present.

If you're client is receiving a cookie it will use it and the connection will stay stuck to the same real.

Cookie is a special case with 'persistent rebalance'.

If you want to see 'persistent rebalance' I'll suggest the test I mentioned previously.

Or disable cookie on your browser and make a test with cookie-insert.

Finally, capture a sniffer trace of the csm etherchannel if you want to verify that there is no cookie and that the connection is rebalanced.

Gilles.