coreos
plume - Releasing with failing regions
With the current design of plume + os documentation it is not possible to do OS releases without the release image existing in all regions on certain cloud providers. This can cause an issue if the cloud provider is experiencing issues in certain regions/partitions: we cannot release until the issue is fixed and plume is allowed to finish.
This will not be fully solved by code in plume but will likely require some design-flow changes in plume.
A potential idea I've tossed around would be to add a separate step in the plume process that handles cloud specific replication, i.e. a third step (either in the middle or post-release depending on outside changes) that handles replicating the image. It could be either region specific, partition specific, or cloud specific allowing for an easy split of tasks in CI.
