This is a continuation of the earlier posts starting with this one: http://ravinote.blogspot.com/2020/09/best-practice-from-networking.html
Operators like update involve scaling down the nodes, then performing the update, and then scaling back up. There is no restriction to reuse logic between operators.
Operators generally work one at a time on the same cluster. This prevents states from being mixed and allowing each reconcile between state and deploys to happen sequentially.
Operators do not have to retain any information between invocations. Anything that needs to be persisted must be part of the state.
There is no limit to the number of operators run during deployment. It is preferred to run them sequentially one after the other. The more granular the operators the better they are for maintenance.
The diagnosability of operators improves with each operator being small and transparent.
The upgrade operator can be rolling upgrades. Then the operator does not require scale up or down.
The operators can invoke methods defined in other operators if they are accessible.
There is a lot in common between installer packages on any other desktop platform and the containerization deployment operators.
The setup of access and access control are part of the operator actions for deployment and they remain some of the core requirements for the deployment of the product.
Networking products keep a journal with the data which helps with packet processing.
Networking products also keep a state machine for the protocol pertaining to the packets.
Networking products allow testing of the packet content, format, sequence, and driver state machine transitions.
No comments:
Post a Comment