Closed smalleni closed 4 months ago
This has been brought up again and we've discussed our approach here. We're nearing the point I think we can start working to implement this. We'll want to have a set of fairly static machines which can be optionally used for this if they are not in a current allocation for starters (and also has a sane, repeatable OSP design in terms of roles.)
The playbooks and Heat templates you and team have created per-machine type will really help here.
Update here: we have 9 x Dell r620's and a new switch in rack b08 to start testing this with.
/me raises hand to consume a brow-beaten openstack and take it another step with openshift.
After a lot of consideration to include the scope of QUADS and our development resources we will not be expanding our capabilities to perform tenant automation/deployment. It's just beyond what we can handle, instead it's better for tenants to utilize the APIv3 that is now available with QUADS 2.0 to perform their own automation.
The most we can extend beyond this is sending webhooks on release of an environment if that's useful too (to perhaps an Ansible Tower/AAC somewhere?) covered in https://github.com/redhat-performance/quads/issues/417 but with APIv3 this really isn't needed either if you just poll the API.
After all the assignments, there could be a few machines idle and it would be awesome if QUADs can grab these machines, build a cloud with it, record machine configs as metadata and push the results to ES using Browbeat.