You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
With Kubernetes 1.16 becoming more common I think it's wise to utilize the newly introduced startupProbe. To date Kowl exposes just one endpoint for probes, the /health endpoint. In order to make sure that Kowl doesn't receive traffic before it has connected to all brokers in the Kafka cluster we use the /health endpoint for the readiness probe.
Kowl is capable of handling transient connection issues (to Kafka) and therefore shouldn't be killed by any container orchestrator unless it's unrespsovie. Thus I propose to introduce another endpoint for probes:
/health - Returns static ok from the webserver. This is used to check whether the HTTP server is still alive
/startup - Returns ok if connectivity to (all) Kafka Brokers is given. This probe can be used for readiness probes but if possible a startup probe should be used in favour of using a readiness probe at all.
The text was updated successfully, but these errors were encountered:
With Kubernetes 1.16 becoming more common I think it's wise to utilize the newly introduced startupProbe. To date Kowl exposes just one endpoint for probes, the
/health
endpoint. In order to make sure that Kowl doesn't receive traffic before it has connected to all brokers in the Kafka cluster we use the/health
endpoint for the readiness probe.Kowl is capable of handling transient connection issues (to Kafka) and therefore shouldn't be killed by any container orchestrator unless it's unrespsovie. Thus I propose to introduce another endpoint for probes:
The text was updated successfully, but these errors were encountered: