Is your feature request related to a problem? Please describe.
Right now, when using Restart strategy with Argo Rollouts there is problem that if the reloader controller is failing over leader, or restarting/moving, the config change events will be missed and the restartAt operation will never be fired.
Describe the solution you'd like
How about, similar to annotation implementation, we also submit an annotation to the Argo Rollout object in the same request we issue the spec.restartAt. Adding the annotation will ensure the controller knows if a restart has been issued for that particular config (secrets/cms), and ensure even during restarts, no events will be missed (via the syncAfterRestart)
Additionally, submitting both restartAt and annotation is a single atomic request also minimizes complexity.
Describe alternatives you've considered
EnvVars cause unwanted rollouts.
Additional context
Happy to submit PR if approved.
Is your feature request related to a problem? Please describe.
Right now, when using Restart strategy with Argo Rollouts there is problem that if the reloader controller is failing over leader, or restarting/moving, the config change events will be missed and the restartAt operation will never be fired.
Describe the solution you'd like
How about, similar to annotation implementation, we also submit an annotation to the Argo Rollout object in the same request we issue the
spec.restartAt. Adding the annotation will ensure the controller knows if a restart has been issued for that particular config (secrets/cms), and ensure even during restarts, no events will be missed (via thesyncAfterRestart)Additionally, submitting both restartAt and annotation is a single atomic request also minimizes complexity.
Describe alternatives you've considered
EnvVars cause unwanted rollouts.
Additional context
Happy to submit PR if approved.