Description of problem: I have an Icehouse instance. After installation I forgot to configure storage for ephemeral storage on compute nodes. Then I fired up 200 instances with swap spaces. And compute nodes quickly run out of space and nodes got into situation which was hard to fix. I highly recommend to change /usr/lib/python2.7/site-packages/packstack/puppet/templates/nova_sched.pp to "DEFAULT/scheduler_default_filters": value => "RetryFilter,AvailabilityZoneFilter,RamFilter,ComputeFilter,ComputeCapabilitiesFilter,ImagePropertiesFilter,CoreFilter,DiskFilter"; i.e. add DiskFilter. This would make life of OpenStack newbies - like me - much easier.
Fix in https://review.openstack.org/#/c/327800/