Hello.

The Gang scheduling documentations says

> NOTE: Gang scheduling is performed independently for each partition, so
configuring partitions with overlapping nodes and gang scheduling is
generally not recommended.

but I am not sure what will go wrong.

Specifically I had in mind the following scheme:

SelectType=select/cons_res
SelectTypeParameters=CR_CPU
PreemptType=preempt/partition_prio
PreemptMode=schedule,gang

NodeName=hpc[1-6]  Sockets=4 CoresPerSocket=8 ThreadsPerCore=1
RealMemory=516075
PartitionName=high    Nodes=hpc[1-6]  Priority=2 Shared=FORCE:4 Default=No
  MaxTime=INFINITE State=UP
PartitionName=normal  Nodes=hpc[1-6]  Priority=1 Shared=FORCE:4 Default=yes
 MaxTime=INFINITE State=UP

What I want is basically gang scheduling with temporal complete suspension
of jobs in the normal partition when the corresponding ressource is
required by jobs from the high partition, until these jobs are done.
Particularly I don't want to have more than 4 jobs running/gangin per CPU.
Will this work as expected?

Many thanks
Alex

Reply via email to