Skip to content
Hands-on IT IT Drills

Day 4: Workloads at Exam Speed

Goal: You roll Deployments out and back, configure applications and pick the right workload type – all imperatively and fast. · Duration: approx. 60 min · Prerequisite: Day 2

Briefing

CKA: Workloads and Scheduling (15 %) – rolling updates and rollbacks, ConfigMaps and Secrets, self-healing workloads. You know the basics; today speed is what counts.

Type Purpose Create with
Deployment stateless apps k create deployment
DaemonSet one Pod per node rework Deployment YAML
StatefulSet stable identity, own storage YAML
Job / CronJob one-off / scheduled k create job / cronjob
mkdir -p ~/cka/tag04 && cd ~/cka/tag04
k create ns tag04
k config set-context --current --namespace=tag04

Drill 1: Rolling Update with a Change Cause

k create deployment shop --image=nginx:1.28 --replicas=3
k rollout status deployment shop
k set image deployment/shop nginx=nginx:1.29
k annotate deployment shop kubernetes.io/change-cause="nginx 1.29"
k rollout status deployment shop
k rollout history deployment shop

Expected (abridged):

deployment.apps/shop image updated
Waiting for deployment "shop" rollout to finish: 1 out of 3 new replicas ...
deployment "shop" successfully rolled out
REVISION  CHANGE-CAUSE
1         <none>
2         nginx 1.29

The container is named nginx because k create deployment names it after the image. You have to specify that name with k set image.

Save your progress? Sign up for free

Drill 2: A Broken Rollout and the Way Back

k set image deployment/shop nginx=nginx:9.99
k annotate deployment shop kubernetes.io/change-cause="kaputtes Image" --overwrite
k get pods -l app=shop

Expected (after about 20 s):

NAME                    READY   STATUS             RESTARTS   AGE
shop-59456556cd-ktjpx   1/1     Running            0          25s
shop-59456556cd-l2bdq   1/1     Running            0          26s
shop-59456556cd-mzxp2   1/1     Running            0          23s
shop-6fbc884fbc-jzvlq   0/1     ImagePullBackOff   0          20s

The old Pods keep running – the rolling update stalls, but the app is not down.

k rollout undo deployment shop
k rollout status deployment shop
k rollout history deployment shop
k get deployment shop -o jsonpath='{.spec.template.spec.containers[0].image}{"\n"}'

Expected:

deployment.apps/shop rolled back
deployment "shop" successfully rolled out
REVISION  CHANGE-CAUSE
1         <none>
3         kaputtes Image
4         nginx 1.29
nginx:1.29

Revision 2 has disappeared: a rollback copies the old revision as the new revision 4. To go back to a specific revision, use k rollout undo deployment shop --to-revision=1.

Save your progress? Sign up for free

Drill 3: Mounting a ConfigMap and a Secret

k create configmap app-config --from-literal=MODE=prod --from-literal=LOG_LEVEL=info
k create secret generic db-cred --from-literal=DB_PASSWORD=lab-password
k run app --image=busybox:1.37 $do --command -- sh -c 'sleep 3600' > app.yaml

Extend the container part in app.yaml so that it looks like this:

apiVersion: v1
kind: Pod
metadata:
  name: app
spec:
  containers:
  - name: app
    image: busybox:1.37
    command: ["sh", "-c", "sleep 3600"]
    envFrom:
    - configMapRef:
        name: app-config
    env:
    - name: DB_PASSWORD
      valueFrom:
        secretKeyRef:
          name: db-cred
          key: DB_PASSWORD
    volumeMounts:
    - name: cfg
      mountPath: /etc/app
  volumes:
  - name: cfg
    configMap:
      name: app-config
k apply -f app.yaml
k wait --for=condition=Ready pod/app
k exec app -- env | grep -E 'MODE|LOG_LEVEL|DB_PASSWORD' | sort
k exec app -- ls /etc/app

Expected:

DB_PASSWORD=lab-password
LOG_LEVEL=info
MODE=prod
LOG_LEVEL
MODE

envFrom imports all keys as variables, and the volume stores each key as a file. Secrets are only base64-encoded, not encrypted:

k get secret db-cred -o jsonpath='{.data.DB_PASSWORD}' | base64 -d; echo

Expected: lab-password

Save your progress? Sign up for free

Drill 4: DaemonSet from a Deployment YAML

There is no k create daemonset. The fast way: generate a Deployment YAML and rework it.

k create deployment logger --image=busybox:1.37 $do -- sh -c 'sleep 3600' > ds.yaml
sed -i -e 's/^kind: Deployment/kind: DaemonSet/' -e '/replicas:/d' \
  -e '/strategy:/d' -e '/^status:/d' ds.yaml
k apply -f ds.yaml
k get ds logger
k get pods -l app=logger -o custom-columns=NAME:.metadata.name,NODE:.spec.nodeName

Expected:

daemonset.apps/logger created
NAME     DESIRED   CURRENT   READY   UP-TO-DATE   AVAILABLE   NODE SELECTOR   AGE
logger   2         2         2       2            2           <none>          15s
NAME           NODE
logger-6gdgk   worker1
logger-k8fnp   worker2

Only two Pods: cp1 carries a taint (Day 5 and the Challenge).

Save your progress? Sign up for free

Drill 5: StatefulSet with a Headless Service

cat <<'EOF' | k apply -f -
apiVersion: v1
kind: Service
metadata:
  name: db
spec:
  clusterIP: None
  selector:
    app: db
  ports:
  - port: 80
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: db
spec:
  serviceName: db
  replicas: 3
  selector:
    matchLabels:
      app: db
  template:
    metadata:
      labels:
        app: db
    spec:
      containers:
      - name: nginx
        image: nginx:1.29
EOF
k rollout status statefulset db
k get pods -l app=db

Expected:

partitioned roll out complete: 3 new pods have been updated...
NAME   READY   STATUS    RESTARTS   AGE
db-0   1/1     Running   0          6s
db-1   1/1     Running   0          4s
db-2   1/1     Running   0          2s

Each Pod has a stable DNS name through the headless Service:

k run dnstest --image=busybox:1.37 --rm -it --restart=Never -- \
  nslookup db-1.db.tag04.svc.cluster.local

Expected (abridged):

Name:   db-1.db.tag04.svc.cluster.local
Address: 192.168.235.143

Delete db-1 – it comes back with the same name:

k delete pod db-1
k wait --for=condition=Ready pod/db-1
k get pods -l app=db

Expected:

NAME   READY   STATUS    RESTARTS   AGE
db-0   1/1     Running   0          15s
db-1   1/1     Running   0          1s
db-2   1/1     Running   0          11s

Save your progress? Sign up for free

Drill 6: Job and CronJob

k create job pi --image=busybox:1.37 -- sh -c 'echo 3.14159'
k wait --for=condition=Complete job/pi
k logs job/pi
k create cronjob tick --image=busybox:1.37 --schedule='*/1 * * * *' -- date

Expected:

job.batch/pi created
job.batch/pi condition met
3.14159
cronjob.batch/tick created

After a little over a minute:

k get jobs

Expected:

NAME            STATUS     COMPLETIONS   DURATION   AGE
pi              Complete   1/1           4s         56s
tick-29915854   Complete   1/1           5s         33s

Save your progress? Sign up for free

Drill 7: Self-Healing with Probes

cat <<'EOF' | k apply -f -
apiVersion: v1
kind: Pod
metadata:
  name: healthy
spec:
  containers:
  - name: app
    image: busybox:1.37
    command: ["sh", "-c", "touch /tmp/ok; sleep 3600"]
    livenessProbe:
      exec:
        command: ["test", "-f", "/tmp/ok"]
      periodSeconds: 5
      failureThreshold: 2
    readinessProbe:
      exec:
        command: ["test", "-f", "/tmp/ok"]
      periodSeconds: 5
EOF
k wait --for=condition=Ready pod/healthy
k exec healthy -- rm /tmp/ok
k get events --field-selector involvedObject.name=healthy --sort-by=.lastTimestamp

Expected (after about 20 s, abridged):

26s   Normal    Started     pod/healthy   Container started
17s   Warning   Unhealthy   pod/healthy   Liveness probe failed:
17s   Normal    Killing     pod/healthy   Container app failed liveness probe, ...

The restart creates /tmp/ok again, and the Pod becomes healthy:

k wait --for=condition=Ready pod/healthy --timeout=90s
k get pod healthy

Expected:

NAME      READY   STATUS    RESTARTS     AGE
healthy   1/1     Running   1 (1s ago)   42s

Liveness restarts the container; readiness only takes it out of the Service.

Save your progress? Sign up for free

Exercise: Speed Drills

  1. Scale shop to 5 replicas. Target time: 10 s
  2. Trigger the CronJob tick once by hand right now as the Job manual. Target time: 20 s
  3. Set maxSurge: 1 and maxUnavailable: 0 for shop. Target time: 60 s
  4. Generate the YAML of a Secret api-key with KEY=sk-lab-1234. Target time: 20 s

Solution

k scale deployment shop --replicas=5
k create job manual --from=cronjob/tick
k patch deployment shop \
  -p '{"spec":{"strategy":{"rollingUpdate":{"maxSurge":1,"maxUnavailable":0}}}}'
k create secret generic api-key --from-literal=KEY=sk-lab-1234 $do

Expected (excerpt):

deployment.apps/shop scaled
job.batch/manual created
deployment.apps/shop patched
apiVersion: v1
data:
  KEY: c2stbGFiLTEyMzQ=
kind: Secret

Save your progress? Sign up for free

Checkpoint

k get deployment shop -o jsonpath='{.spec.strategy.rollingUpdate}{"\n"}'

Expected: {"maxSurge":1,"maxUnavailable":0}

k get statefulset db

Expected: READY shows 3/3.

k exec app -- cat /etc/app/MODE; echo

Expected: prod

Challenge

The DaemonSet logger should also run on cp1.

Hint: The Pod has to tolerate the control plane’s taint.

Solution

cat > tol.yaml <<'EOF'
spec:
  template:
    spec:
      tolerations:
      - key: node-role.kubernetes.io/control-plane
        operator: Exists
        effect: NoSchedule
EOF
k patch ds logger --patch-file tol.yaml
k rollout status ds logger
k get pods -l app=logger -o custom-columns=NAME:.metadata.name,NODE:.spec.nodeName

Expected:

daemonset.apps/logger patched
daemon set "logger" successfully rolled out
NAME           NODE
logger-2z6l7   cp1
logger-mgntl   worker1
logger-vnmnj   worker2

Clean up:

k config set-context --current --namespace=default
k delete ns tag04

Save your progress? Sign up for free

Exam Tip

  • k set image needs the container name, not the Deployment name. Look it up with k get deploy shop -o jsonpath='{.spec.template.spec.containers[*].name}'.
  • DaemonSet: generate Deployment YAML, change kind, delete replicas and strategy.
  • Remember these docs pages: “Configure a Pod to Use a ConfigMap”, “Configure Liveness, Readiness and Startup Probes”.

Enterprise Note

  • In production, rollouts run through GitOps (Argo CD, Flux); rollout undo is the emergency lever, and afterwards the state belongs in Git.
  • No plaintext Secrets in Git – use External Secrets, Sealed Secrets or OpenBao.
  • For databases, use an operator (CloudNativePG) instead of your own StatefulSet.
  • Use probes with care: a liveness probe that is too strict restarts Pods under load and makes outages worse.