the following problem is to solve: - OpenStack Ocata (Ubuntu 16.04) (deployed with MAAS/Juju) with serveral availability zones / aggregates. Some are for testing, some for production - Test-tenant in its own OpenStack domain and project - Jump server VM in test-tenant (Ubuntu 18.04) with Juju 2.4 - Juju controller (bootstrapped from jump server in OpenStack cloud/region) (NO MAAS in test-tenant !)
Now, we want to deploy a multi-node charm bundle with 'juju deploy' and make sure that every juju-initiated 'machine' would be created only on nova compute nodes in the 'testing' availability zone/aggregate.
Any suggestions or recommendations?
PS: What I did NOT want to achieve with this question: - Hints and tips for 'juju deploy --to 0 xxx' or 'juju add-unit --to 1 xxx' cause the target machines must still exist. I am searching solution for a 'green field' bundle deployment.