Repurposed as:

Backport leader election improvements for schedulers to let production envs using discover_hosts_in_cells_interval instead of installer based handling

The previous version of it was:

Today after scaling out compute nodes in the deployment the human operator needs to manually execute nova-manage cell_v2 discover_hosts to make the new compute usable. We should try to automate this.

For ironic computes this is already automated by the nova-operator.

For EDPM computes nova-operator has no way today to detect when a new EDPM compute is deployed and therefore nova-operator cannot run the discovery. At the moment dataplane-operator knows when an EDPM deployment finished but it does not track if this means a new compute is deployed (it does not even have a concept of an EDPM compute just and EDPM node).