We've encountered a case (NDS-103) where two containers have a startup dependency. The current strategy is to use a Pod readinessProbe to look for a specific condition (e.g., waiting for a webapp to be available) that may take time. An external process/script checks for the pod Ready state before starting the dependent container.

We need a long-term strategy for expressing and managing service dependencies.

Outcome:

• Documented approach for handling service dependencies
• Implementation of this approach in existing services (e.g., ELK stack, Dataverse, Clowder)