[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Condor-users] Flocking to a NAT



Hello All,

We are deciding how best to connect to separate pools. We are looking at flocking as the first option. I have checked the online docs and haven't turned up much searching.

Our setup is, pool_a is composed of machines on a public net. Pool_b has a central manager that is publicly accessible and it can also access the execute nodes on the NAT. So before we go to far with testing and re-configuring we were wondering if this would work? We are trying to decide if flocking from pool_a to pool_b will work.

It seems that the startd process(pool_a) will contact the central manager in the FLOCK_TO list (pool_b) when needed. But from there it is difficult to tell which processes / machines will communicate with each other? Does the CM (pool_b) verify the request is from a host in the FLOCK_FROM and then negotiate the job on behalf of the remote (pool_a) and then hand it to an execute node (pool_b) in it's pool? Then when the job is complete the execute node (pool_b) would notify it's CM (pool_b) and that CM would in turn notify the startd (pool_a) from the originating submit?

Or once the CM (pool_b) verifies the FLOCK_FROM is valid, it then negotiates with one of it's execute nodes and then tells the originating submit/startd (pool_a) and then from there all communication is between the nodes and not the CM?



TIA
--
Jim Summers
School of Computer Science-University of Oklahoma
-------------------------------------------------