Re: [HTCondor-users] Negotiator only allocating 1 job per machine per cycle

[Todd L Miller <tlmiller@xxxxxxxxxxx>]

> I assume that such definition should be done by adding directives within 
> the condor_config file of the given worker. Could you please provide an 
> example that I could use as a starting point?

	The canonical way to dedicate an entire machine's resources to a 
single partitionable slot is as follows (entries go in the worker 
machine's HTCondor configuration):

	# from https://htcondor.readthedocs.io/en/latest/admin-manual/policy-configuration.html#index-136

	NUM_SLOTS = 1
	NUM_SLOTS_TYPE_1 = 1
	SLOT_TYPE_1 = 100%
	SLOT_TYPE_1_PARTITIONABLE = TRUE

HTCondor has mechanisms to reserve CPUs, memory, and disk for the use of 
the system; look RESERVED_DISK and RESERVED_MEMORY up in the manual (CPUs 
are a little more complicated, IIRC).


	As Christoph said, when you submit jobs with request_cpu, 
request_memory, and request_disk, HTCondor will not start jobs which 
request more resources than are available.  There are various mechanisms 
for enforcing the requested limits, if that becomes necessary.  You can 
also configure HTCondor to reject jobs that don't specify those three 
request values, or to set default values.

- ToddM