Comment 1 for bug 1654102

Actually when looking at logstash, the failures are mostly in the non-Ironic jobs which are using keystone v3, and those are failing because we create the compute_nodes record in the database from the resource tracker in the nova-compute service, but we fail to update the compute node record which sets the free_disk_gb value:

http://logs.openstack.org/65/416765/1/check/gate-tempest-dsvm-neutron-identity-v3-only-full-ubuntu-xenial-nv/faf1363/logs/screen-n-cpu.txt.gz#_2017-01-04_23_15_06_674

And it looks like that's happening because of some keystone v3 auth issue when talking to the placement service.

As for the 1/2 date, ~1/3 is when we made the placement-api run in all jobs in devstack:

https://review.openstack.org/#/c/412537/

Which would include the keystone v3 API job.

Why we're hitting the auth issue with keystone v3 I don't know, but it's when we're trying to talk to the placement API for the resource provider for the compute node, which should be handled with a safe_connect decorator...