Flagd is commonly managed using a container orchestration tool such as Kubernetes. In these environments, it's important to understand the status of the container to determine if it's capable of performing flag evaluations. Liveness and readiness probes are commonly used to periodically check the status of the container and take corrective measures when an unhealthy state is detected.
Proposal
Add a liveness probe compatible http or gRPC endpoint to flagd. This endpoint should return a success status when the process has started and all sync providers have loaded. Once flagd has this endpoint, the OpenFeature Operator can then be configured to support liveness and readiness probes when flagd is injected into a workload.
Questions
Should probes be enabled by default in flagd?
Should the OFO be configured to set probes by default?
Overview
Flagd is commonly managed using a container orchestration tool such as Kubernetes. In these environments, it's important to understand the status of the container to determine if it's capable of performing flag evaluations. Liveness and readiness probes are commonly used to periodically check the status of the container and take corrective measures when an unhealthy state is detected.
Proposal
Add a liveness probe compatible http or gRPC endpoint to flagd. This endpoint should return a success status when the process has started and all sync providers have loaded. Once flagd has this endpoint, the OpenFeature Operator can then be configured to support liveness and readiness probes when flagd is injected into a workload.
Questions
Resources