Limit concurrent processing canaries

[timsanwald]

Describe the feature

It would be cool if we can restrict the concurrent amount of processing canaries.

What problem are you trying to solve? With a large amount of apps pointing to a "base configmap" e.g. jaeger url, changing values in this configmap triggers all canaries at the same time. The cluster itself though doen't have enough resources to start a second pod of each of the canaries at the same time. This can lead to not starting pods and thus to failing canaries, which have to be triggered then manually.

Proposed solution

A configuration parameter for flagger to process x canaries at the same time.

Any alternatives you've considered?

• Scale the cluster and be prepared for this event -> high costs & unnecessary compute resources
• high timeout for startup probe of canaries -> just masks the problem and more canaries will result in the same error at some point in time