Project

General

Profile

Feature #10337

give jobs requiring more nodes a higher priority?

Added by Greg Farnum over 4 years ago. Updated over 4 years ago.

Status:
New
Priority:
High
Assignee:
-
Category:
-
Target version:
-
Start date:
12/16/2014
Due date:
% Done:

0%

Source:
other
Tags:
Backport:
Reviewed:
Affected Versions:

Description

The FS runs have several tests which require 5 machines instead of the default three. These get marked as hung a disproportionate amount of the time because the cluster never has enough machines free that they're allowed to take the locks: the 3-machine tests always get to lock instead.

Either we need a way to order tests which are waiting on locks, or we need to do something like (ugh) increase the reserved number of machines (which teuthology isn't allowed to lock) to a constant value greater than the number of machines used by our largest tests.

History

#1 Updated by Zack Cerza over 4 years ago

  • Tracker changed from Bug to Feature
  • Subject changed from test runs which require more machines than the average tend to starve to give jobs requiring more nodes a higher priority?

#2 Updated by Greg Farnum over 4 years ago

This struck with astounding clarity on http://pulpito.ceph.com/teuthology-2015-01-18_23:10:01-knfs-next-testing-basic-multi/

Six jobs waiting for long enough to get considered hung by the job emailer.

Also available in: Atom PDF