staged upgrades and optional downgrade option
Is there any mechanism to stage upgrades in a cluster? Namely, for instance with 11 nodes, upgrading them gradually over the course of several days, maybe in groups of 3 sounds like a good strategy, but how to know when it's safe and when not? Or is it guaranteed to be always safe within a major version?
Secondly, if there is any issue found, how to downgrade or reinstall the affected node to hold off before e.g. resolving the issue?
Currently, when installing from ISO, there is some relatively historical version installed and running upgrade simply brings it to the most recent, so there is no stages like with a kernel, correct?
Would it be easy to e.g. dual-boot 2 separate versions altogether?