Appendix B: Practice Exam
Goal: You solve 17 exam-style tasks in 120 minutes and grade yourself. · Duration: 120 min · Prerequisite: Days 0–17
Rules of Play
- Time: 120 minutes, set a timer. You pass with 66 out of 100 points.
- Allowed: the Kubernetes docs (
kubernetes.io/docs),helm --help,kubectl explain. Not allowed: this book during the exam time. - Result files go into
~/pb/. - The solutions are at the end. Only read them after the time is up.
- Cluster: the HA cluster from Day 16 with Traefik, MetalLB and the
StorageClass
nfs.
Preparation
On cp1 (the sabotage for tasks 14 and 15):
mkdir -p ~/pb
k create ns pb-broken
k create deployment api -n pb-broken \
--image=registry.k8s.io/e2e-test-images/agnhost:2.53 \
-- /agnhost netexec --http-port=8080
k set resources deployment api -n pb-broken --requests=memory=64GiOn worker2:
sudo systemctl stop kubeletStart the clock.
Exercise 1: Deployment and Service (4 points)
Create the namespace pb-app. In it, a Deployment
web with nginx:1.29 and three replicas, plus a
ClusterIP Service web on port 80.
Save your progress? Sign up for free
Exercise 2: Update and Rollback (5 points)
Update web to nginx:1.29-alpine with the
change cause alpine. Then roll back to the previous
version.
Save your progress? Sign up for free
Exercise 3: Configuration (4 points)
ConfigMap app-cfg with COLOR=blue, Secret
app-sec with TOKEN=lab-token. A Pod
cfg (busybox:1.37, sleep 3600)
gets both as environment variables COLOR and
TOKEN.
Save your progress? Sign up for free
Exercise 4: RBAC for CI (6 points)
ServiceAccount ci in pb-app. There, it may
read Deployments (get, list,
watch) and create and delete Pods – nothing else.
Save your progress? Sign up for free
Exercise 5: A New User (6 points)
For the user anna (group ops), generate a
key and CSR, have Kubernetes sign it and save the certificate as
~/pb/anna.crt. anna may read everything cluster-wide.
Save your progress? Sign up for free
Exercise 6: NetworkPolicy (5 points)
Only Pods with the label role=client from
pb-app may access the Pods of web, and only on
port 80. Test with one Pod with and one without the label.
Save your progress? Sign up for free
Exercise 7: Ingress (5 points)
An Ingress web (class traefik) routes
web.lab to the Service web. It must work
despite Exercise 6:
curl -H 'Host: web.lab' 172.16.0.242/ returns
200.
Save your progress? Sign up for free
Exercise 8: Gateway API (6 points)
Gateway pb-gw (class traefik, HTTP listener
port 8000) and an HTTPRoute api that routes
api.lab to the Service web.
Save your progress? Sign up for free
Exercise 9: Shared Storage (5 points)
PVC shared (500Mi, ReadWriteMany, default
class) and a Pod writer that mounts it at
/data and writes ok to
/data/probe.txt.
Save your progress? Sign up for free
Exercise 10: StorageClass (4 points)
StorageClass nfs-retain: the same provisioner and the
same parameters as nfs, but reclaim policy
Retain and binding only with the first Pod.
Save your progress? Sign up for free
Exercise 11: Scheduling (5 points)
Label worker1 with tier=fast. A Pod
fast (nginx:1.29, request 100m CPU and 64Mi,
memory limit 128Mi) may only run on nodes with tier=fast –
via node affinity.
Save your progress? Sign up for free
Exercise 12: Autoscaling (4 points)
web should scale between 2 and 6 replicas, target 60 %
CPU.
Save your progress? Sign up for free
Exercise 13: etcd Backup (9 points)
Back up etcd to ~/pb/etcd-snap.db and verify the
snapshot.
Save your progress? Sign up for free
Exercise 14: A Deployment Won’t Come Up (8 points)
The Deployment api in pb-broken is not
fully available. Find the cause and fix it without deleting the
Deployment.
Save your progress? Sign up for free
Exercise 15: A Node Doesn’t Report In (10 points)
A node is NotReady. Bring it back.
Save your progress? Sign up for free
Exercise 16: Helm (5 points)
Install the chart podinfo/podinfo in version 6.17.0 as
release pi into the new namespace pb-helm,
with two replicas.
Save your progress? Sign up for free
Exercise 17: Kustomize (9 points)
Under ~/pb/kust/base there is a Deployment
api (nginx:1.28); generate it with
--dry-run. An overlay ~/pb/kust/prod sets
namespace pb-prod, prefix prod-, image
nginx:1.29 and two replicas. Apply the overlay.
Stop the clock.
Save your progress? Sign up for free
Solutions
Solution 1
k create ns pb-app
k create deployment web -n pb-app --image=nginx:1.29 --replicas=3
k expose deployment web -n pb-app --port=80Solution 2
k set image deployment/web -n pb-app nginx=nginx:1.29-alpine
k annotate deployment web -n pb-app kubernetes.io/change-cause="alpine"
k rollout status deployment web -n pb-app
k rollout undo deployment web -n pb-app
k get deploy web -n pb-app -o jsonpath='{.spec.template.spec.containers[0].image}{"\n"}'Expected (last line): nginx:1.29
Solution 3
k create configmap app-cfg -n pb-app --from-literal=COLOR=blue
k create secret generic app-sec -n pb-app --from-literal=TOKEN=lab-token
cat <<'EOF' | k apply -f -
apiVersion: v1
kind: Pod
metadata:
name: cfg
namespace: pb-app
spec:
containers:
- name: cfg
image: busybox:1.37
command: ["sleep", "3600"]
env:
- name: COLOR
valueFrom:
configMapKeyRef:
name: app-cfg
key: COLOR
- name: TOKEN
valueFrom:
secretKeyRef:
name: app-sec
key: TOKEN
EOF
k exec -n pb-app cfg -- env | grep -E 'COLOR|TOKEN' | sortExpected:
COLOR=blue
TOKEN=lab-token
Solution 4
Warning, trap:
k create role ci-role --verb=get,list,watch --resource=deployments.apps --verb=create,delete --resource=pods
grants all verbs on all resources –
Deployments would then be deletable. Two rules require YAML:
k create sa ci -n pb-app
cat <<'EOF' | k apply -f -
apiVersion: rbac.authorization.k8s.io/v1
kind: Role
metadata:
name: ci-role
namespace: pb-app
rules:
- apiGroups: ["apps"]
resources: ["deployments"]
verbs: ["get", "list", "watch"]
- apiGroups: [""]
resources: ["pods"]
verbs: ["create", "delete"]
EOF
k create rolebinding ci-rb -n pb-app --role=ci-role --serviceaccount=pb-app:ci
k auth can-i list deployments -n pb-app --as=system:serviceaccount:pb-app:ci
k auth can-i delete deployments -n pb-app --as=system:serviceaccount:pb-app:ciExpected: yes, then
no.
Solution 5
cd ~/pb
openssl genrsa -out anna.key 2048
openssl req -new -key anna.key -subj "/CN=anna/O=ops" -out anna.csr
cat <<EOF | k apply -f -
apiVersion: certificates.k8s.io/v1
kind: CertificateSigningRequest
metadata:
name: anna
spec:
request: $(base64 -w0 anna.csr)
signerName: kubernetes.io/kube-apiserver-client
usages: ["client auth"]
EOF
k certificate approve anna
k get csr anna -o jsonpath='{.status.certificate}' | base64 -d > anna.crt
openssl x509 -in anna.crt -noout -subject
k create clusterrolebinding anna-view --clusterrole=view --user=anna
k auth can-i list pods -A --as=annaExpected:
subject=O = ops, CN = anna
yes
Solution 6
cat <<'EOF' | k apply -f -
apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
name: web-from-client
namespace: pb-app
spec:
podSelector:
matchLabels:
app: web
policyTypes:
- Ingress
ingress:
- from:
- podSelector:
matchLabels:
role: client
ports:
- port: 80
EOF
k run c1 -n pb-app --image=busybox:1.37 -l role=client -- sleep 3600
k run c2 -n pb-app --image=busybox:1.37 -- sleep 3600
k exec -n pb-app c1 -- wget -qO- -T3 web | grep -o '<title>.*</title>'
k exec -n pb-app c2 -- wget -qO- -T3 webExpected:
<title>Welcome to nginx!</title>
wget: download timed out
Solution 7
k create ingress web -n pb-app --class=traefik --rule="web.lab/*=web:80"
curl -s -o /dev/null -w '%{http_code}\n' -H 'Host: web.lab' 172.16.0.242/Expected: 504 – Traefik is not a
role=client Pod, so the policy from Exercise 6 locks it
out. Additionally allow the controller’s namespace:
cat <<'EOF' | k apply -f -
apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
name: web-from-traefik
namespace: pb-app
spec:
podSelector:
matchLabels:
app: web
policyTypes:
- Ingress
ingress:
- from:
- namespaceSelector:
matchLabels:
kubernetes.io/metadata.name: traefik
ports:
- port: 80
EOF
curl -s -o /dev/null -w '%{http_code}\n' -H 'Host: web.lab' 172.16.0.242/Expected: 200
Solution 8
cat <<'EOF' | k apply -f -
apiVersion: gateway.networking.k8s.io/v1
kind: Gateway
metadata:
name: pb-gw
namespace: pb-app
spec:
gatewayClassName: traefik
listeners:
- name: http
protocol: HTTP
port: 8000
---
apiVersion: gateway.networking.k8s.io/v1
kind: HTTPRoute
metadata:
name: api
namespace: pb-app
spec:
parentRefs:
- name: pb-gw
hostnames: ["api.lab"]
rules:
- matches:
- path:
type: PathPrefix
value: /
backendRefs:
- name: web
port: 80
EOF
k get gateway pb-gw -n pb-app
curl -s -o /dev/null -w '%{http_code}\n' -H 'Host: api.lab' 172.16.0.242/Expected: PROGRAMMED True and
200 (the policy from Solution 7 applies here too).
Solution 9
cat <<'EOF' | k apply -f -
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: shared
namespace: pb-app
spec:
accessModes: ["ReadWriteMany"]
resources:
requests:
storage: 500Mi
---
apiVersion: v1
kind: Pod
metadata:
name: writer
namespace: pb-app
spec:
containers:
- name: w
image: busybox:1.37
command: ["sh", "-c", "echo ok > /data/probe.txt; sleep 3600"]
volumeMounts:
- name: d
mountPath: /data
volumes:
- name: d
persistentVolumeClaim:
claimName: shared
EOF
k exec -n pb-app writer -- cat /data/probe.txtExpected: ok
Solution 10
cat <<'EOF' | k apply -f -
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
name: nfs-retain
provisioner: nfs.csi.k8s.io
parameters:
server: 172.16.0.126
share: /srv/nfs/k8s
reclaimPolicy: Retain
volumeBindingMode: WaitForFirstConsumer
EOF
k get sc nfs-retainExpected (abridged):
NAME PROVISIONER RECLAIMPOLICY VOLUMEBINDINGMODE
nfs-retain nfs.csi.k8s.io Retain WaitForFirstConsumer
Solution 11
k label node worker1 tier=fast
cat <<'EOF' | k apply -f -
apiVersion: v1
kind: Pod
metadata:
name: fast
namespace: pb-app
spec:
affinity:
nodeAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
nodeSelectorTerms:
- matchExpressions:
- key: tier
operator: In
values: ["fast"]
containers:
- name: fast
image: nginx:1.29
resources:
requests:
cpu: 100m
memory: 64Mi
limits:
memory: 128Mi
EOF
k get pod fast -n pb-app -o jsonpath='{.spec.nodeName}{"\n"}'Expected: worker1
Solution 12
k set resources deployment web -n pb-app --requests=cpu=100m
k autoscale deployment web -n pb-app --cpu=60% --min=2 --max=6Without a CPU request, the HPA would stay at
<unknown>.
Solution 13
sudo etcdctl --endpoints=https://127.0.0.1:2379 \
--cacert=/etc/kubernetes/pki/etcd/ca.crt \
--cert=/etc/kubernetes/pki/etcd/server.crt \
--key=/etc/kubernetes/pki/etcd/server.key \
snapshot save /home/ubuntu/pb/etcd-snap.db
sudo etcdutl snapshot status /home/ubuntu/pb/etcd-snap.db -w jsonExpected (abridged):
Snapshot saved at /home/ubuntu/pb/etcd-snap.db
{"hash":3699137888,"revision":6933,"totalKey":1386,"totalSize":26976256,"version":"3.7.0
"}
~ does not work reliably with sudo – use
absolute paths.
Solution 14
k get pods -n pb-broken
k describe pod -n pb-broken $(k get pods -n pb-broken \
--field-selector=status.phase=Pending -o name | head -1) | grep FailedScheduling
k set resources deployment api -n pb-broken --requests=memory=64Mi
k rollout status deployment api -n pb-brokenExpected (abridged):
api-898b99767-pzxjd 0/1 Pending 0 2m32s
api-c4994b678-gfxrm 1/1 Running 0 2m32s
Warning FailedScheduling ... 0/3 nodes are available: 3 Insufficient memory. ...
deployment "api" successfully rolled out
A 64Gi memory request – no node has that much. The old Pod kept running because the rollout never finished.
Solution 15
k get nodesOn the affected node (worker2):
systemctl is-active kubelet
systemctl is-enabled kubelet
sudo systemctl start kubeletExpected: inactive,
enabled – so only start is needed. On
cp1:
k wait --for=condition=Ready node/worker2 --timeout=120s.
Solution 16
helm install pi podinfo/podinfo -n pb-helm --create-namespace --version 6.17.0 \
--set replicaCount=2
k get deploy -n pb-helmExpected: pi-podinfo with
2/2 after a short while.
Solution 17
mkdir -p ~/pb/kust/base ~/pb/kust/prod
cd ~/pb/kust
k create deployment api --image=nginx:1.28 $do > base/deploy.yaml
printf 'resources:\n- deploy.yaml\n' > base/kustomization.yaml
cat > prod/kustomization.yaml <<'EOF'
resources:
- ../base
namespace: pb-prod
namePrefix: prod-
images:
- name: nginx
newTag: "1.29"
replicas:
- name: api
count: 2
EOF
k create ns pb-prod
k apply -k prod
k get deploy -n pb-prod -o custom-columns=NAME:.metadata.name,REPLICAS:.spec.replicasExpected:
deployment.apps/prod-api created
NAME REPLICAS
prod-api 2
Scoring
Count the points of the tasks that are completely and verifiably solved – there are no half points.
| Points | Assessment |
|---|---|
| 85–100 | exam-ready |
| 66–84 | passed, but barely: review your weak spots |
| under 66 | review the days covering the missed tasks and start over |
Clean up:
k delete ns pb-app pb-broken pb-helm pb-prod
k delete clusterrolebinding anna-view
k delete csr anna
k delete sc nfs-retain
k label node worker1 tier-Checkpoint
k get ns | grep -c '^pb-'Expected: 0
Good luck in the exam.