Testing the health of a rollout

[dudicoco]

Hi,

Tools such as flagger and argo rollouts allow you to validate each step of a rollout/canary by performing smoke tests or checking metrics such as number of errors and latency.

Are there any plans on supporting this within kruise rollouts? What is the point of a progressive rollouts without an automatic way of validating the progress?

[wparr-circle]

Was looking for the same. It would be good if on the website https://openkruise.io/rollouts/introduction#kruise-rollouts-vs-other-components clearly documented that Metric Analysis is not currently supported (I created https://github.com/openkruise/openkruise.io/pull/200 to clarify).

It looks like there is a skeleton built for metric analysis during a canary upgrade, however this is currently implemented as: https://github.com/openkruise/rollouts/blob/master/pkg/controller/rollout/rollout_canary.go#L288

func (m *canaryReleaseManager) doCanaryMetricsAnalysis(c *RolloutContext) (bool, error) {
	// todo
	return true, nil
}

[wparr-circle]

Duplicates https://github.com/openkruise/rollouts/issues/16