Closed darrylweaver closed 1 year ago
Putting different provider cluster definitions into different namespaces might allow for this separation.
It is not so much different namespaces as different kustomizations for different providers that would help here.
keeping this issue to raise with product at a later date
In the DemoEnvs, we have seen that specially during configuration time of a new provider, there are errors popping up when deploying clusters. You need to go through the motions and hit all the obstacles and configure them out of the way. Each of these errors will halt the flux-system kustomization.
Only command line investigation and manual intervention will bring the flux-system kustomization and thus the capi deployment system back into operation.
This makes CAPI and cluster provisioning brittle. Please discuss and suggest ideas on how to improve this.
One idea might be to run a kustomization for each type of provider. This would possibly contain errors to the provider in question. Liquid Metal would only impact LM provisioning, AKS only Azure and so on.