the efs-plugin is frequently failing it's liveness probe on startup. I believe the current configuration in the helm chart is set to check too frequently (period: 2)

it would be simple to change this to allow configuration via helm values. to keep it in line with the existing healthPort I would propose adding the following to the controller and node sections.

  healthInitialDelaySeconds: 10
  healthTimeoutSeconds: 3
  healthPeriodSeconds: 10
  healthFailureThreshold: 5

I can raise a PR.