Health Probes

There are two types of probes possible - Liveness and Readiness.

• Liveness - this means we only check if the port is really reachable from outside the pod.
• Readiness - This means we check if the app is full up and can really start accepting requests. Only when this is true, the endpoint of the pod is added to the service or it will remove the pod from service if the readiness check fails. Until then no request will read the pod.

Mental model to remember

• Liveness is like a heartbeat check to check if the application is alive. If not alive, it try to restart the pod. Similar to a health check in a hospital, if the patient isn't alive, we try to resuscitate them.
• Readiness is like a check to see if the application is ready to serve requests. If not ready, it won't send traffic to the pod. Similar to a check before a patient is discharged from the hospital, we want to make sure they're ready to go home and take care of themselves.

Not only at startup

Liveness and readiness probes aren't only checked at startup, but they're continuously checked during the lifecycle of the pod.

Liveness and Readiness are the same in some cases

In frameworks like tomcat, the server is available on the port only when the complete initialization is complete. This means, when the application ready, it's also live. But this isn't true for all cases.

For example, even in tomcat the server maybe fully started but we can implement custom readiness probes, where we can additionally check if certain required microservices are reachable or not and decide the readiness answer. So such case, the app is live but it's not yet ready to really serve requests.