Opened 7 years ago

#1398 new defect

RQS for limits not working for parallel jobs

Reported by: Reuti Owned by:
Priority: normal Milestone:
Component: sge Version: 8.0.0d
Severity: minor Keywords:
Cc:

Description

Having a single RQS entry like for a non-consumable:

limit to h_vmem=500M

will work for serial jobs as expected: if h_vmem is requested during job submission, it will be limited to 500M

Parallel jobs requesting h_vmem will never start, although they are both per slot as well as multiplied well inside the limit with the error message: `cannot run because it exceeds limit "/" in rule "fubar/1"'

==

Making h_vmem consumable an attaching it to each exechost will allow parallel jobs to be scheduled again.

Change History (0)

Note: See TracTickets for help on using tickets.