I install a production cluster with ocata version for nfv project. During the testing, a compute node(hostname:normal-node2) is reclaimed., and I install a new node with the same hostname. The new nova-compute just has warning in logs, but instances can be created in this compute node with no error.
For example（devstack for master branch,just for check the bug）:
# /usr/bin/nova-compute --config-file /etc/nova/nova-cpu.conf
.............
INFO nova.compute.resource_tracker [None req-4330d5b5-ccb3-41d9-954a-43d42373a325 None None] Final resource view: name=devstack-com phys_ram=15708MB used_ram=512MB phys_disk=492GB used_disk=0GB total_vcpus=8 used_vcpus=0 pci_stats=[]
INFO nova.scheduler.client.report [None req-4330d5b5-ccb3-41d9-954a-43d42373a325 None None] [req-3857fb52-6ca3-44c0-8974-fc30abf038b4] Another thread already created a resource provider with the UUID d494e537-0790-4dfd-9ca2-28905e085f58. Grabbing that record from the placement API.
WARNING nova.scheduler.client.report [None req-4330d5b5-ccb3-41d9-954a-43d42373a325 None None] Unable to refresh my resource provider record
..............
This code just work in two compute nodes with same hostname and restrict the following node. But if the previous node was destroyed, it's useless.
I think it's danger in production environment, with no error in logs can let operators ignore the potential risk. For my project, the new node with wrong resource provider work two weeks before I delete the old record in nova-api database.
Set the status to 'In-Progress' because this report has an assignee.