Observability - Readiness Probes

Question 1

What POD Status exist and what do they mean?

Answer 1

• Pending: when first created, scheduler tries to find a node for the pod, if none found stays in pending state
• ContainerCreating: once scheduled, images required are pulled and containers are started
• Running: Container running, till process fulfilled successfully etc,

Question 2

What are POD Conditions and what do they do?

Answer 2

• complement pod status
• array of true or false status that tell us the state of a pod
  Instances:
• PodScheduled: pod scheduled on a node
• Initialized: when pod is initialized
• ContainersReady: when all containers in the pod are ready
• Ready: pod itself is read with everything, sometimes applications might still be ‘warming up’

Described in ‘kubectl describe pod …’

Question 3

How does Kubernetes know that the application inside a container is actually running and available for input?

Question 4

What kind of Readiness Probes exist?

Answer 4

• HTTP Test against URL/Endpoint
• TCP Test to see if port is open
• Exec command, to run a custom script that checks if application is ready

Question 5

How do we define a Readiness probe?

Answer 5

pod-yaml:
~~~
…
spec:
containers:
- name:
readinessProbe: (http)
httpGet:
path: /api/ready
port: 8080
(optional) initialDelaySecons: 10
(optional) periodSeconds: 5
(optional) failureTreshold: 8
///

readinessProbe: (tcp)
tcpSocket:
port: 3306
///

readinessProbe: (exec)
exec:
command:
- cat
- /app/is_ready
~~~

Question 6

What is the consequence of a ReadinessProbe?

Answer 6

Kubernetes does not set Status Ready as soon as pod is ready
- tests instead, whether the api respons positively
- until then, no traffic is forwarded to the pod

Question 6

How are Readiness Probes useful in a multi-pod environment

Answer 6

• without, container would be shown ready, while app is still warming up
• service would immediately start routing requests to the still starting server