Cluster Failover with SSO enabled

We have 2 R6 servers in a cluster. We have an Alteon doing SLB (intelligent IP sprayer).

SSO works (we can hold authentication across servers and subdomains). SLB works.

However, when we open a page on server A in a browser… and then kill the HTTP task for that server. Then hit REFRESH. We get Page Cannot Be Displayed. Normally, without SSO, the page would just reload from the other server… this problem only happens with the "

SSO domain" … hence our suspicion that SSO is to blame!!

Is this some confusion of SSO? Has anyone else tried this?

Damian

Subject: Cluster Failover with SSO enabled

Do you have ‘stickiness’ enabled? On some of these load-balancing units, you can force users to go to a certain server to make sure they maintain a login.

Also have your admin check the device itself and make sure that all the appropriate ports are enabled for both of the ‘real’ IP addresses.

Subject: RE: Cluster Failover with SSO enabled

Yes we have session persistence enabled… aka stickiness.

However, the switch should be smart enough to deal with a downed server and reroute appropriately. If not, its a pretty useless piece of $25K hardware.

I’ll do more research here… add it to my list.

Subject: RE: Cluster Failover with SSO enabled

OK, I solved part of this. In case anyone else uses Alteons for HTTP. Do this:

Use minmisses (or hash) metric

Do NOT also use pbind clientip

Using both is what caused my problems. And clientip alone doesnt work across http/https… so just dont use pbind.

I am going to start a new thread relating to SSO and SLB… cuz THAT aint working.