Why do we have to separate Session Owner and Session Setup Device in an A/A-Cluster

[sylvia]

Hello,

in the documentation it is said: "The separation of session owner and session setup devices is necessary to avoid race conditions that can occur in asymmetrically routed environments.”

But nothing more in detail.

I absolutely understand, what the session owner and what the session setup device is doing and I understand the configuration.

But I do not understand the need for separation. Why isn't it possible that the session owner is doing the setup as well?

Can somebody explain?

Many thanks,

Sylvia

[mikand]

I dont know if im right on this but I can imagine that session owner is the statetable itself which even in A/A configurations actually is working like A/P. That is one of the boxes owns the session but mirror this to the other box (in case of failure).

While session setup is the physical box which the packet actually arrived to (which we dont know which box it will be in A/A setup).

That is if box1 is session owner (mirroring to box2) and a packet that arrives to box1 is then processed by the session owner (who happens to be at box1) and then forwarded if policy matches. If a packet arrives to box2 this box will forward this to box1 to notify the session owner (and setup a session) but then it will forward the packet on its own once the session is setup and it has a "cache" of it in its own memory.

Or if its the other way around (session setup is the statetable while session owner is which physical box the packet arrived to) :smileysilly:

Because of this session owner and session setup is two different processes?