Hi Gabriel,
I do not have experiences with SAP-Portal Loadbalancing but from my experience with loadbalancing I would say that this decision should be done regarding the surrounding environment. If there are proxies around which are used to connected to the loadbalanced VIP I would suggest SSL-Session balancing/stickieness. If clients are directly connecting towards the VIP a client-IP-stickieness might be easiser to troubleshoot.
Regarding the netmask issue:
Unfortunately there is no real documentation (I did not find any so far). I can only tell you the result out of my experience:
A netmask 255.255.255.255 is kind of host stickiness and 255.255.255.0 a stickieness of a /24 network. In other words I think that all Clients with a network of 10.1.1.0/24 are sticked to the decission once taken for one of the hosts in this segment while the sticky timeout did not take place. Maybe one can confirm or correct this.
Regarding the thing with the failure of a real server:
With the corresponding failaction defined in the serverfarm you decide what to do in case of an error of a real server. The stickiness should be corrected correspondig to the results you define with the failaction command.
This is how I dealt with those sticky issue when doing firewall loadbalancing with client-IP-Stickiness.
Hope that helped.
Kind Regards,
Joerg
