Description of problem: GroupAffinityFilter doesn't work as expected. There's an upstream bug https://bugs.launchpad.net/nova/+bug/1218878 tracking the issue about GroupAffinityFilter and GroupAntiAffinityFilter. Commit for the same is proposed in https://github.com/openstack/nova/commit/4233e16ba2f5bb85e93494b6eeb9ffff31dbdf30 Version-Release number of selected component (if applicable): RHOS 3.0 How reproducible: Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info:
There is already a backport of the patch for grizzly upstream, if it gets in we'll get it in the next rebase.
How to reproduce: 1. Deploy openstack with two compute nodes (compute1 and compute2). 2. Set the key 'scheduler_default_filters' in /etc/nova/nova.conf to have the following filters: * AvailabilityZoneFilter,RamFilter,ComputeFilter,GroupAntiAffinityFilter 3. Restart the scheduler: * service openstack-nova-scheduler restart 4. Start two instances, one on each compute node: * nova boot --image cirros --flavor m1.tiny --availability-zone nova:compute1 vm-compute1 * nova boot --image cirros --flavor m1.tiny --availability-zone nova:compute2 vm-compute2 5. Start another instance in a group: * nova boot --image cirros --flavor m1.tiny --hint group=foo vm-compute 6. Without the fix the instance will move to the ERROR state and a WARNING will appear in scheduler.log.
Xavier, how is group "foo" actually defined (and the policy set)? I thought the API for this didn't get accepted yet?
In 3.0 only the GroupAntiAffinityFilter exists.
(In reply to Stephen Gordon from comment #4) > Xavier, how is group "foo" actually defined (and the policy set)? I thought > the API for this didn't get accepted yet? Until the blueprint linked from bug 978507 is completed (expected for Icehouse), the group's policy will be defined by the group filter the scheduler is using. In Havana both affinity and anti affinity filters are available, while in Grizzly only the anti affinity exists.
verified according to https://bugzilla.redhat.com/show_bug.cgi?id=1014948#c2: I have now successfully created 2 instances in group, 2 in each one of the computes. No errors found in nova (scheduler and compute)
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. http://rhn.redhat.com/errata/RHSA-2014-0112.html