Project

General

Profile

Bug #3556

For partitionable slots, VOFE will stop too early

Added by Burt Holzman over 6 years ago. Updated over 6 years ago.

Status:
Assigned
Priority:
Normal
Category:
Frontend
Target version:
Start date:
03/01/2013
Due date:
% Done:

0%

Estimated time:
First Occurred:
Occurs In:
Stakeholders:
Duration:

Description

Partitionable slots always appear to be idle, so we may stop sending jobs to a site too early (it looks like there are many idle glideins, when it just the partitionable slots). A partitionable slot has no more resources to dole out when TotalSlots =?= 1

As a first iteration, we should add a constraint to only count slots as idle if the following expression is true:

((PartitionableSlot =!= True) || (TotalSlots =?= 1))

For a second iteration -- imagine that all the user jobs that the FE can see ask for 8 CPUs. Imagine the partitioanble slot has 4 subslots remaining (TotalSlots == 5). To our system, the partitionable slot isn't idle since we can't match any jobs to it.


Related issues

Related to GlideinWMS - Bug #3438: VOFE won't count partitionable slots correctlyClosed02/06/2013

History

#1 Updated by Burt Holzman over 6 years ago

  • Category changed from Factory to Frontend

#2 Updated by Burt Holzman over 6 years ago

  • Target version changed from v3_1 to v3_x

The first part of this was addressed in #3438.



Also available in: Atom PDF