A Pod Has A Status And A Lifecycle

• Pending
  • the scheduler is trying to figure out where to place a pod
  • the reason for a pending pod can be found in kubectl describe bod the-pod
• ContainerCreating
  • pulling an image
  • on a node
• Running
  • until the container is terminated or completed

Conditions Help Understand a Pod State

These are boolean conditions

• PodScheduled
• Initialized
• ContainersReady
• Ready

These can be discovered in the kubectl describe pod the-pod under the Conditions section.

The Ready Condition Can Be Misleading

Applications in containers may not be "ready" but the container, outside of the application, is ready.
This can feel misleading

Create Readiness Probes for Containers

Readiness Probe is a familiar term for "real-world" k8s containerized apps to be ready. K8s will use a custom-defined readinessProbe to wait until the readinessProbe "tells k8s" that the container within the pod is "ready" via some custom logic we, the container/pod devs, create:

• try building a "ready" api on a web app (/healthz or something)
• try testing a db connecting with tcp on the default port (3306 perhaps)
• try building a "readiness" script to run on a more back-end-y service

Configure a ReadinessProbe in a pod def file

With this, K8s does not set the container status to "ready" until k8s gets a successful response from an api

apiVersion: v1
kind: Pod
metadata:
  name: a-dummy-api
  labels:
    name: horse
spec:
  containers:
  - name: api-box
    image: api-box-img
    ports: 
      - containerPort: 8080
    # THIS!
    readinessProbe:
      httpGet:
        path: /api/ready
        port: 8080

2 more options:

# tcp
readinessProbe:
  tcpSocket:
    port: 3306

# exec a command
readinessProbe:
  exec:
    command:
      - cat
      - some/command/checking/ready_Status

More config can be made in these custom readinessProbes:

# an httpGet example
readinessProbe:
  httpGet:
    path: /api/ready
    port: 8080
  
  # wait to start probing
  initialDelaySeconds: 10
  
  # frequency of readiness probe
  periodSeconds: 6

  # how many failures to allow
  failureThreshold: 8