Re: [HTCondor-users] Help on FLOCK and CCB

[Ben Cotton <ben.cotton@xxxxxxxxxxxxxxxxxx>]

On Wed, Mar 5, 2014 at 5:58 AM, Batsayan Das <batsayan.das@xxxxxxx> wrote:
> how to define a pool?

There's probably a more rigorous definition, but I'd define a pool as
all of the machines that report to the same central manager.

> Whether pool A and pool B can be a private network?

Definitely. When I was at Purdue, we had flocking set up with several
other institutions where at least some of the machines were on private
addresses.

> how one machine in pool A talks to another machine in pool B?
> in there any routing parameters need to be set for this talking?
>

Left to their own devices, the machines will talk directly to each
other. In the case of A flocking to B, the schedd(s) in A will talk
directly to the central manager in B to negotiate, and then to the
execute nodes in B to run jobs. In cases where that's not possible,
you'll need to configure CCB.

> There may be firewalls. Should we use  Condor Connection Brokering (CCB)
> along with FLOCK ?

Maybe. If pools A and B can communicate directly if the firewalls are
removed, then you might prefer to poke the appropriate holes in your
firewalls and configure HTCondor to only use those ports (see sections
3.3.6 and 3.7 of the HTCondor Manual). If some of the hosts are on
private address space, then you'll need to use CCB. As a matter of
personal preference, I like to keep my central managers and schedds on
public space (firewalled as appropriate) and put the execute nodes
behind CCB for this sort of purpose.


Hope this helps!
BC


-- 
Ben Cotton
main: 888.292.5320

Cycle Computing
Leader in Utility HPC Software

http://www.cyclecomputing.com
twitter: @cyclecomputing