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

Re: [Condor-users] flocking questions

[BTW your original message used characters from a non-ascii charset, for
mailing list, it is usually

-----Original Message-----
> From: condor-users-bounces@xxxxxxxxxxx
> On Behalf Of Yu Fu
> Sent: Friday, August 10, 2007 4:47 PM
> To: Condor-Users Mail List
> Subject: [Condor-users] flocking questions

> Hi there,

Hi, BTW your original message contained characters from character sets
that my
mail reader couldn't cope with. I don't know what this group's policy
but personally I prefer ascii text for such newsgroups.

> I have a question about flocking: in order to make condor flocking
work on all submitting
> nodes in a cluster, do I need to define the FLOCK... related
parameters on all of the
> submitting nodes or I just need to do that on the central condor
management host?

FLOCK_TO and FLOCK_FROM need setting up on the participating central
(one on each for one way, both for 2 way). You will also need to ensure
that all 
potential execute nodes can see all potential submit nodes (and vice
versa), i.e.
no NATs or firewalls getting in the way. You may also have to adjust
HOSTALLOW settings.

> Another question is about condor_q: I noticed that each submitting
node has its own
> queue when listed with condor_q, the condor_q command only display the
jobs submitted from
> the local host. How can I make it centralized, that is, a condor_q on
any submitting node
> and and the central management host will give all of the queued jobs
from all submitting
> nodes?

Each submitting node has its own queue.
condor_q -global
lists jobs on all queues
(including, if I remember rightly, queues on submit nodes on a flocked

I hope this helps a little