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

Re: [Condor-users] Jobs will not run in another pool

I had two clusters, Scout and Mastodon, with no firewalls in between. Jobs without specific requirements would freely flock back and forth between them. If I specified a requirement of InMastodon from within the Mastodon cluster, the job would run on one of the Mastodon nodes, and likewise with Scout. If I submitted a job with requirement of InMastodon from within the Scout cluster, the job would not flock to Mastodon.

I've changed our setup greatly since then, so I can't test that this problem still exists, but It caused me lots of trouble at the time for various reasons.

- dave

Kewley, J (John) wrote:
I've tried the same sort of thing before with no success, and I know there was no communications issue.

Another thing to check is that you have set
FLOCK_FROM at one end and FLOCK_TO at the other.

I don't know how the various HOST_ALLOW settings interact with flocking,
maybe they could interfere.

It seems that flocking negotiations do not work in this way.

Don't work in what way?

If you specify requirements that can be met only by machines in the other pool, Condor never seems to match them.

I am fairly certain that I have been able to do this without problem in the past,
but haven't the ability to test it just now.

A year or so back we setup secure flocking between different institutions in UK.
Since our application was to build lumps of s/w on different architectures, our
jobs had to move to different pools which satisfied those archtectures to match.
Only one pool had Solaris nodes for instance.



Condor-users mailing list