Re: [HTCondor-users] user priority question

[Todd L Miller <tlmiller@xxxxxxxxxxx>]

> I have a pool with 300 cores. User A submits a 500 jobs which request
> single core and runs relatively fast. Then, User B submits 20 jobs which
> request 3 cores. Since only 1 core is requested for user a, does user b
> have to wait till all jobs are done from user a?

	It is certainly possible for the pool to fragment into 300 
single-core slots in that situation, in which case user B is going to be 
waiting for a while.  (It's possible, especially it CLAIM_WORKLIFE is set 
to be relatively short, for three of user A's jobs to finish on the same 
machine and become available as a unit in the next negotiation cycle.)  In 
general, however, this situation is what the condor defrag daemon is for; 
it will ask a certain number of machines to stop accepting new jobs until 
a certain fraction of their resources become available, at which point the 
negotiator -- assuming users A and B had equal priority before this 
example started -- will preferentially hand 3-core slots to B.

	See

https://htcondor.readthedocs.io/en/latest/admin-manual/cm-configuration.html#defragmenting-dynamic-slots

for more information.

- ToddM