Comment 2 for bug 1708039

One compute node record is created here:

http://logs.openstack.org/54/489754/3/check/gate-grenade-dsvm-neutron-multinode-ubuntu-xenial/7e19ce3/logs/old/screen-n-cpu.txt.gz#_2017-08-01_21_51_12_886

2017-08-01 21:51:12.886 13572 INFO nova.compute.resource_tracker [req-21d845c5-9cc7-4467-be2c-64377f554fea - -] Compute_service record created for ubuntu-xenial-2-node-osic-cloud1-disk-10201108:ubuntu-xenial-2-node-osic-cloud1-disk-10201108

That's 6 minutes before the failures in the conductor logs.

The subnode compute node record is created here:

http://logs.openstack.org/54/489754/3/check/gate-grenade-dsvm-neutron-multinode-ubuntu-xenial/7e19ce3/logs/subnode-2/old/screen-n-cpu.txt.gz#_2017-08-01_21_56_50_543

2017-08-01 21:56:50.543 24035 INFO nova.compute.resource_tracker [req-663ad965-aa47-4bb2-82be-ecc385a23b09 - -] Compute_service record created for ubuntu-xenial-2-node-osic-cloud1-disk-10201108-763019:ubuntu-xenial-2-node-osic-cloud1-disk-10201108-763019

That's about 30 seconds before the failure, and that's the node that is not mapped: ubuntu-xenial-2-node-osic-cloud1-disk-10201108-763019

The weird thing is I don't even see the discover_hosts script being run, but it should happen from this part of the devstack-gate run:

https://github.com/openstack-infra/devstack-gate/blob/94602b3cd50f870df1bac738f596e09c6483dd74/devstack-vm-gate.sh#L707