Readiness and liveness probe values were introduced in Amsterdam and Beijing. The size and complexity of ONAP has increased dramatically since then, and as such the current probe values are not necessarily valid any more. In addition, applications now depend more and more on "shared" components like Cassandra and Mariadb Clusters, AAF, DMaaP, Logging, etc.

To prevent continued tweaking of probe values and the unnecassry crash/backoff/restart cycles of applications, we need to re-evaluate the dependency mechanism being used.

Proposals include:

1. Start all "shared" applications first as a separate prerequisite action 'e.g. helm deploy -f onap-common.yaml'.
2. Using new k8s capabilities to improve startup ordering: https://kubernetes.io/docs/concepts/configuration/pod-priority-preemption/
3. Use helm hook or new init container, to block initialization of applications that are await shared components to start. Delays starting readiness/liveness probes until unblocked.