There's a report[1] that in OSP12 we can no longer scale out compute nodes without also running "nova-manage cell_v2 discover_hosts" on the controllers (or at least a node running nova api). This is only an issue when scaling out computes while using the blacklist feature so that no changes are pushed to controllers. Using the blacklist enables scaling computes without changing any other nodes in your deployment. I (believe) it's been traditionally well understood in OpenStack that you can scale out compute nodes without having to also modify/update your control plane services. But, this no longer seems possible if you must also run discover_hosts on a node where nova api is running. [1] http://post-office.corp.redhat.com/archives/rhos-tech/2018-March/msg00363.html
this may be as simple as cron'ing our custom discover_hosts script or not documenting that NovaSchedulerDiscoverHostsInCellsInterval must be set for this use case to work. Could also consider enabling NovaSchedulerDiscoverHostsInCellsInterval by default as currently it's disabled unless deploying Ironic in the overcloud.
*** Bug 1644418 has been marked as a duplicate of this bug. ***
verified on compose 2019-04-18.2
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. https://access.redhat.com/errata/RHBA-2019:0939