serverfarm VIP as rserver of another farm

[kuba.slocinski]

Hi guys,

I'm curious if that's possible. The idea behind is to create "special rserver" which is in fact a webfarm with per-request LB, and can be placed as a backup-rserver when primary one is overloaded.

Let's imageine caching webfarm (A) with URL hash-predictor, then, if one rserver has huge load its probe may switch traffic to backup, failover farm (B) which has per-request basis (let's say round-robin) to spread that (and only that) load. Another rservers from farm (A) are not affected.

Don's see a way to do that in one context. No direct routing between contexts forces to use 6k/other devices to route packets between contexts. Some other ideas to achieve that? Maybe some other ways of solving the pbm?

Regards,

Jakub S.

[Nicolas Fournier]

Hi Jakub,

Instead of using another VIP as rserver of one of your serverfarms (which as you state would need you to first route the traffic out of the ACE before reaching this VIP since we cannot directly send traffic from a VIP to another), I would configure your backup rserver as one of type redirect which would have the client directly send the traffic to the per-request VIP through redirection once the main rserver is overloaded.

Regards,

Nicolas

[kuba.slocinski]

Hello Nicolas,

you pointed the best way indeed, but I thought about doing sth transparent for client. Any other ideas?

Regards,

Jakub

[Nicolas Fournier]

Hi Jakub,

I don't see any as forwarding the traffic from one VIP to another one directly would IMHO create more issue then it would solve but maybe someone else has a better idea 

Regards,

Nicolas