Open
Description
Not 100% sure yet on how much state should be centralised but even having what was applied by the client would help a lot.
As in, when you'd run skate apply
, whatever you had in your yaml would be also saved to a git repo. Not necessarily what was scheduled, but more the submitted yaml. That would simplify adding nodes (you could see what needs to be rescheduled easily), it would also enable a gitops setup like dokku: a gitops container could run within the skate cluster, and schedule on to skate itself, since it'd be possible to get the global cluster state there (what nodes we have, whats already running).
Metadata
Assignees
Labels
No labels