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

Re: [HTCondor-users] HTCondor flock and Negotiator interval



On 8/29/19 12:21 PM, Vikrant Aggarwal wrote:
Hello Experts,

I want to confirm my understanding about the behavior of requirements and flock_to list:

D is the primary cluster of submitter node which shows up in "condor_host". B, C and A are the pools to which submitter node can flock.

FLOCK_TO = B, C, A

Requirements = ( (MachineOwner=="A") Â|| Â(MachineOwner=="B") Â|| Â(MachineOwner=="C") Â|| (MachineOwner=="D")

My understanding:

- Irrespective of the order it always tries first to find slot in primarly pool in this case D.


This is correct.


- If any pool mentioned in flock_to list but not in requirement that will not be considered.


Once condor starts trying to flock, it just does normal matchmaking -- if the job's requirements expression matches the slot's (and vice-versa) in the remote pool, that machine is eligible to run the job.


If my above understanding is correct then how can I force the submit box to try running the job on pool D for 30 mins and if the slots are not available after 30mins then move to the other pools mentioned in requirements.


Probably the easiest way to do this is explicitly in the job's requirements:

Requirements = (Collector_host_string = "D") || ((time() - QDate) > 1800)

This method may create a lot of autoclusters if you have a large number of jobs, which can slow down negotiation.

-greg