Add health endpoints to Dapr Agent for Kubernetes and Dapr probes

[bibryam]

When the agent is run in standalone mode (exposed over HTTP and or PubSub), it should expose standard health endpoints to support Kubernetes and Dapr health checks.

Add the following endpoints:

• /healthz: Overall application health.
• /livez: Kubernetes liveness probe, indicates whether the process should be restarted.
• /readyz: Kubernetes readiness probe, indicates whether the agent can accept traffic.

If all 3 endpoints behave the same way, IMO a single healthz would be sufficient.

Additionally, Dapr always probes /dapr/config for subscriptions. Instead of returning 404, the agent should return an empty list to avoid surfacing errors when no subscriptions are defined.

Reference: https://docs.dapr.io/operations/resiliency/health-checks/app-health/

[JoshVanL]

I don't expect dapr-agents to have an app port for daprd, so /healthzor /dapr/config should not be called by daprd.

[CasperGN]

@bibryam as per our convo, we shouldn't add /healthz as it's a deprecated field.

Let's keep this one to track adding /livez and /readyz endpoints. I know dapr probes /healthz but instead of having /healthz can we suggest users to go with --app-health-check-path pointing to /livez instead?

[Faakhir30]

I'd like to work on this one, from what I understand, we are supposed to:

• expose these two endpoints at AgentRunner level
• documment for daprd to set app-health-check-path flag to /livez
• provide a way (setters) that changes status of liveness and readiness

Hints on how am I supposed to know liveness and readiness of AgentRunner are aprriciated

[sicoyle]

@JeffreyJPZ can you pls comment "/assign" so the issue will assign itself to you?

[JeffreyJPZ]

/assign