-
Spike
-
Resolution: Done
-
Normal
-
None
-
None
Background
This might be by exposing a route, or by having it actually listen there.
My only concern with exposing a route would be that I don't know if there are platforms/deployments where our OpenShift router/haproxy doesn't exist.
This is different than the MCS though because we want to have it exposed both to the cluster AND to the host network.
As long as the MCO is still writing files before we rebase, I think we can still cheat with the cluster pull secrets, but if that ever goes away we're going to have to figure out how to get the initial pull secret on the node so it can scale up.
TODO(jkyros): This card needs more background
Goal
- The node that is bootstrapping will have a way to auth (pull secret, probably) to the internal registry, even though the new node is outside the cluster
- The internal registry is available on the host network such that nodes that are bootstraping/scaling up can pull their images from it
Done When
- A custom OSImageURL referencing the internal registry can be assigned to a pool, and that pool can be scaled up successfully while using it