Details
-
Improvement
-
Status: Resolved
-
Major
-
Resolution: Invalid
-
None
-
None
-
None
Description
Since we want to support the case of simple stack-definition/cluster-definition changes while the cluster is active, we need to store the cluster rev# so that when the agent comes with a heartbeat, the Controller can take smart diffs based on the cluster revision number.
Example: Node1 belongs to
{cluster-ID-10, cluster-rev-100}Now someone changes the cluster's definition to include a new node. This causes the revision# change to cluster-rev-101.
Unless the Controller stored the cluster-rev for Node1, upon receipt of a heartbeat, the Controller wouldn't be able to determine what Node1 would need to run to get to the cluster-rev-101 state (in this case nothing).