[Bug 1073956] [NEW] Scheduler Race Condition at high volume

[Launchpad Bug Tracker]

You have been subscribed to a public bug:

When trying to rapidly schedule hundres of instances the nova scheduler
will not honor cpu_allocation_ratio  and ram_allocation_ratio.

using 'euca-run-instances -n' or a shell loop around 'nova boot'  will
produce this I've seen it with numbers as low as 100 though others
report it taking >300 to reproduce.

On a cloud with a single nova-scheduler host configured with
cpu_allocation_ration=1.0 and targeting a single compute node with 24
VCPUs (named nova-1) this command will schedule 100 instances where only
24 should be allowed:

for i in `seq 1 100`;do nova boot --image someimage --availability-zone
nova:nova-1 --flavor 1 chaff-$i > /dev/null & done

Operating system  is Ubuntu 12.04, using package 2012.2-0ubuntu5~cloud0
from ubuntu-cloud.archive.canonical.com

** Affects: nova
     Importance: High
     Assignee: Vish Ishaya (vishvananda)
         Status: Fix Released

** Affects: nova/folsom
     Importance: High
     Assignee: Vish Ishaya (vishvananda)
         Status: Fix Released

** Affects: nova (Ubuntu)
     Importance: Undecided
         Status: Confirmed

-- 
Scheduler Race Condition at high volume
https://bugs.launchpad.net/bugs/1073956
You received this bug notification because you are a member of Ubuntu Server Team, which is subscribed to nova in Ubuntu.