Probe failure and container restart behavior in Kubernetes - Commands & Configuration

Introduction

Sometimes containers in Kubernetes stop working properly. Probes help check if a container is healthy. If a probe fails, Kubernetes can restart the container to fix problems automatically.

When you want to automatically restart a web server if it stops responding.

When you need to check if a database container is ready before sending traffic.

When you want to detect if a container is stuck and recover it without manual intervention.

When you want to avoid sending requests to a container that is not ready to serve.

When you want to monitor container health and improve application reliability.

Config File - pod.yaml

pod.yaml

apiVersion: v1
kind: Pod
metadata:
  name: example-pod
spec:
  containers:
  - name: example-container
    image: nginx:1.23
    livenessProbe:
      httpGet:
        path: /healthz
        port: 80
      initialDelaySeconds: 5
      periodSeconds: 5
      failureThreshold: 3
    readinessProbe:
      httpGet:
        path: /ready
        port: 80
      initialDelaySeconds: 3
      periodSeconds: 5
      failureThreshold: 3

This pod runs an nginx container with two probes:

• livenessProbe: Checks if the container is alive by requesting /healthz on port 80. If it fails 3 times, Kubernetes restarts the container.
• readinessProbe: Checks if the container is ready to accept traffic by requesting /ready on port 80. If it fails, the container is removed from service endpoints.
• initialDelaySeconds: Wait time before starting probes after container starts.
• periodSeconds: How often to run the probe.
• failureThreshold: Number of failures before action.

Commands

This command creates the pod with the configured liveness and readiness probes. Kubernetes starts the container and begins health checks.

Terminal

kubectl apply -f pod.yaml

Expected Output

pod/example-pod created

Check the pod status to see if it is running and ready after probes start working.

Terminal

kubectl get pods example-pod

Expected Output

NAME          READY   STATUS    RESTARTS   AGE

example-pod   1/1     Running   0          10s

View detailed pod information including probe results and container restarts if any probe failed.

Terminal

kubectl describe pod example-pod

Expected Output

Name:         example-pod

Namespace:    default

Status:       Running

Containers:

  example-container:

    Container ID:   docker://abcdef123456

    Image:          nginx:1.23

    State:          Running

    Ready:          True

    Restart Count:  0

    Liveness:       http-get http://:80/healthz delay=5s timeout=1s period=5s #success=1 #failure=3

    Readiness:      http-get http://:80/ready delay=3s timeout=1s period=5s #success=1 #failure=3

Events:

  Type    Reason     Age   From               Message

  ----    ------     ----  ----               -------

  Normal  Scheduled  1m    default-scheduler  Successfully assigned default/example-pod to node-1

Key Concept

If a liveness probe fails repeatedly, Kubernetes restarts the container to fix it automatically.

Common Mistakes

Not setting initialDelaySeconds causing probes to run before the container is ready.

Probes may fail immediately and cause unnecessary restarts.

Set initialDelaySeconds to allow the container time to start before probes begin.

Using the same probe for both readiness and liveness checks.

Readiness and liveness have different purposes; mixing them can cause traffic routing issues or restarts.

Configure separate readiness and liveness probes tailored to their specific checks.

Setting failureThreshold too low causing frequent restarts on temporary glitches.

Containers may restart unnecessarily, causing instability.

Choose a reasonable failureThreshold to tolerate brief failures before restarting.

Summary

Create a pod with liveness and readiness probes to monitor container health.

Use kubectl apply to deploy the pod and kubectl get pods to check status.

Describe the pod to see probe results and container restart counts.

Practice

1. What happens when a Kubernetes liveness probe fails for a container?

easy

A. Traffic to the container is stopped but it is not restarted.

B. The container is restarted by Kubernetes.

C. The container continues running without changes.

D. The container is removed permanently.

Solution

1. Step 1: Understand liveness probe purpose

   Liveness probes check if a container is alive and healthy.

2. Step 2: Effect of liveness probe failure

   If the liveness probe fails, Kubernetes restarts the container to fix the issue.

3. Final Answer:

   The container is restarted by Kubernetes. -> Option B

4. Quick Check:

   Liveness probe failure = container restart [OK]

Hint: Liveness probe failure always triggers container restart [OK]

Common Mistakes:

• Confusing liveness with readiness probe effects
• Thinking container is deleted permanently
• Assuming no action on failure

2. Which of the following is the correct YAML snippet to define a liveness probe using HTTP GET on path /health at port 8080?

easy

A. livenessProbe:
httpGet:
path: health
port: 8080
initialDelaySeconds: 5
periodSeconds: 10

B. livenessProbe:
exec:
command: ["curl", "/health"]
initialDelaySeconds: 5
periodSeconds: 10

C. livenessProbe:
tcpSocket:
port: 8080
initialDelaySeconds: 5
periodSeconds: 10

D. livenessProbe:
httpGet:
path: /health
port: 8080
initialDelaySeconds: 5
periodSeconds: 10

Solution

1. Step 1: Check HTTP GET probe syntax

   The correct syntax uses httpGet with path starting with a slash and port as a number.

2. Step 2: Validate each option

   livenessProbe: httpGet: path: /health port: 8080 initialDelaySeconds: 5 periodSeconds: 10 correctly uses /health and port 8080. livenessProbe: exec: command: ["curl", "/health"] initialDelaySeconds: 5 periodSeconds: 10 uses exec with curl incorrectly. livenessProbe: tcpSocket: port: 8080 initialDelaySeconds: 5 periodSeconds: 10 uses tcpSocket which is different. livenessProbe: httpGet: path: health port: 8080 initialDelaySeconds: 5 periodSeconds: 10 misses the leading slash in path.

3. Final Answer:

   livenessProbe: httpGet: path: /health port: 8080 initialDelaySeconds: 5 periodSeconds: 10 -> Option D

4. Quick Check:

   HTTP GET probe with correct path and port = livenessProbe: httpGet: path: /health port: 8080 initialDelaySeconds: 5 periodSeconds: 10 [OK]

Hint: HTTP path must start with '/' and port is a number [OK]

Common Mistakes:

• Missing leading slash in path
• Using exec command incorrectly
• Confusing tcpSocket with httpGet

3. Given this pod spec snippet, what happens if the readiness probe fails but the liveness probe passes?

readinessProbe:
  httpGet:
    path: /ready
    port: 80
  initialDelaySeconds: 5
  periodSeconds: 10
livenessProbe:
  httpGet:
    path: /live
    port: 80
  initialDelaySeconds: 5
  periodSeconds: 10

medium

A. The container is restarted immediately.

B. The container is deleted and recreated.

C. The container is marked not ready and removed from service endpoints.

D. Nothing happens; container continues serving traffic.

Solution

1. Step 1: Understand readiness probe role

   Readiness probe controls if the container receives traffic by marking it ready or not.

2. Step 2: Effect of readiness probe failure

   If readiness probe fails but liveness probe passes, container stays running but is removed from service endpoints, so no traffic is sent to it.

3. Final Answer:

   The container is marked not ready and removed from service endpoints. -> Option C

4. Quick Check:

   Readiness fail = no traffic, container runs [OK]

Hint: Readiness failure stops traffic, liveness failure restarts container [OK]

Common Mistakes:

• Confusing readiness failure with container restart
• Thinking container is deleted
• Assuming no effect on traffic

4. You notice your container keeps restarting frequently. The liveness probe is configured with initialDelaySeconds: 5 and periodSeconds: 10. What is a likely cause and fix?

medium

A. The initial delay is too short; increase initialDelaySeconds to allow startup time.

B. The period is too long; decrease periodSeconds to check faster.

C. The probe path is incorrect; change it to /healthz.

D. The container image is corrupted; rebuild the image.

Solution

1. Step 1: Analyze probe timing settings

   If initialDelaySeconds is too short, the probe runs before the app is ready, causing false failures.

2. Step 2: Adjust initial delay

   Increasing initialDelaySeconds gives the container time to start before probing, preventing unnecessary restarts.

3. Final Answer:

   The initial delay is too short; increase initialDelaySeconds to allow startup time. -> Option A

4. Quick Check:

   Short initialDelay causes early probe failure [OK]

Hint: Increase initialDelaySeconds if container restarts too soon [OK]

Common Mistakes:

• Changing periodSeconds instead of initialDelaySeconds
• Assuming probe path is always wrong
• Blaming container image without checking probes

5. You have a container with both readiness and liveness probes. The readiness probe fails intermittently, but the liveness probe always passes. How should Kubernetes behave and what is the best way to improve user experience?

hard

A. Kubernetes removes the pod from service during readiness failure but does not restart it; improve by tuning readiness probe thresholds.

B. Kubernetes restarts the container on readiness failure; improve by disabling readiness probe.

C. Kubernetes ignores readiness probe failures if liveness passes; no action needed.

D. Kubernetes deletes the pod on readiness failure; improve by increasing liveness probe timeout.

Solution

1. Step 1: Understand probe roles on intermittent readiness failure

   Readiness failure removes pod from service endpoints but does not restart container if liveness passes.

2. Step 2: Improve user experience by tuning readiness probe

   Tuning thresholds like failureThreshold or periodSeconds reduces false negatives and prevents unnecessary traffic disruption.

3. Final Answer:

   Kubernetes removes the pod from service during readiness failure but does not restart it; improve by tuning readiness probe thresholds. -> Option A

4. Quick Check:

   Readiness fail = no traffic; tune probe to reduce false failures [OK]

Hint: Tune readiness probe to avoid false failures and traffic loss [OK]

Common Mistakes:

• Thinking readiness failure restarts container
• Disabling readiness probe instead of tuning
• Ignoring intermittent failures