Day 12: CRDs, Operators and the Interfaces
Goal: You extend the Kubernetes API with your own types, install an operator and find your way around CRI, CNI and CSI on the node. · Duration: approx. 60 min · Prerequisite: Day 11
Briefing
CKA: “Understand CRDs, install and configure operators” and “Understand extension interfaces (CNI, CSI, CRI, etc.)” (part of the 25 %).
kubelet --CRI--> containerd --> runc (start containers)
kubelet --CNI--> Calico (Pod network)
kubelet --CSI--> nfs.csi.k8s.io (volumes)
API server + CRD --> new type; operator = controller that implements it
mkdir -p ~/cka/tag12 && cd ~/cka/tag12
k create ns tag12
k config set-context --current --namespace=tag12Drill 1: A resource type of your own
cat <<'EOF' | k apply -f -
apiVersion: apiextensions.k8s.io/v1
kind: CustomResourceDefinition
metadata:
name: backups.lab.example.com
spec:
group: lab.example.com
scope: Namespaced
names:
plural: backups
singular: backup
kind: Backup
shortNames: ["bk"]
versions:
- name: v1
served: true
storage: true
additionalPrinterColumns:
- name: Schedule
type: string
jsonPath: .spec.schedule
- name: Retention
type: integer
jsonPath: .spec.retentionDays
schema:
openAPIV3Schema:
type: object
properties:
spec:
type: object
required: ["schedule"]
properties:
schedule:
type: string
retentionDays:
type: integer
minimum: 1
maximum: 365
EOF
k api-resources --api-group=lab.example.comExpected:
customresourcedefinition.apiextensions.k8s.io/backups.lab.example.com created
NAME SHORTNAMES APIVERSION NAMESPACED KIND
backups bk lab.example.com/v1 true Backup
The name of a CRD is always
<plural>.<group>.
Save your progress? Sign up for free
Drill 2: Objects of the new type
cat <<'EOF' | k apply -f -
apiVersion: lab.example.com/v1
kind: Backup
metadata:
name: nightly
spec:
schedule: "0 2 * * *"
retentionDays: 30
EOF
k get bk
k explain backup.specExpected (shortened):
backup.lab.example.com/nightly created
NAME SCHEDULE RETENTION
nightly 0 2 * * * 30
GROUP: lab.example.com
KIND: Backup
VERSION: v1
...
FIELDS:
retentionDays <integer>
schedule <string> -required-
The columns come from additionalPrinterColumns,
explain from the schema.
Save your progress? Sign up for free
Drill 3: The schema protects you
cat <<'EOF' | k apply -f -
apiVersion: lab.example.com/v1
kind: Backup
metadata:
name: broken
spec:
retentionDays: 999
EOFExpected:
The Backup "broken" is invalid:
* spec.retentionDays: Invalid value: 999: spec.retentionDays in body should be less than
or equal to 365
* spec.schedule: Required value
A CRD only stores data. For a Backup to actually back
something up, you would need a controller that watches such objects – an
operator.
Save your progress? Sign up for free
Drill 4: Install an operator
cert-manager is a widely used operator: it reads
Certificate objects and creates Secrets with certificates
from them.
helm install cert-manager oci://quay.io/jetstack/charts/cert-manager --version v1.21.3 \
-n cert-manager --create-namespace --set crds.enabled=true
k wait -n cert-manager --for=condition=Available deploy --all --timeout=180s
k get crd | grep cert-manager.ioExpected (shortened):
Pulled: quay.io/jetstack/charts/cert-manager:v1.21.3
STATUS: deployed
deployment.apps/cert-manager condition met
deployment.apps/cert-manager-cainjector condition met
deployment.apps/cert-manager-webhook condition met
certificaterequests.cert-manager.io 2026-11-17T22:22:11Z
certificates.cert-manager.io 2026-11-17T22:22:11Z
...
issuers.cert-manager.io 2026-11-17T22:22:11Z
The chart comes from an OCI registry, so no
helm repo add is needed.
Save your progress? Sign up for free
Drill 5: Use the operator
cat <<'EOF' | k apply -f -
apiVersion: cert-manager.io/v1
kind: Issuer
metadata:
name: selfsigned
spec:
selfSigned: {}
---
apiVersion: cert-manager.io/v1
kind: Certificate
metadata:
name: shop-cert
spec:
secretName: shop-cert-tls
dnsNames:
- shop.lab
issuerRef:
name: selfsigned
kind: Issuer
EOF
k get issuer,certificate
k get secret shop-cert-tls -o jsonpath='{.data.tls\.crt}' | base64 -d \
| openssl x509 -noout -ext subjectAltName -enddateExpected:
NAME READY AGE
issuer.cert-manager.io/selfsigned True 9s
NAME READY SECRET AGE
certificate.cert-manager.io/shop-cert True shop-cert-tls 9s
X509v3 Subject Alternative Name: critical
DNS:shop.lab
notAfter=Feb 15 22:22:40 2027 GMT
And the core of every operator – it restores the desired state:
k delete secret shop-cert-tls
k get secret shop-cert-tlsExpected (after a few seconds):
secret "shop-cert-tls" deleted from tag12 namespace
NAME TYPE DATA AGE
shop-cert-tls kubernetes.io/tls 3 8s
Save your progress? Sign up for free
Drill 6: CRI – containers without kubectl
On worker1. The kubelet talks to containerd over this socket:
sudo grep containerRuntimeEndpoint /var/lib/kubelet/config.yaml
sudo crictl pods --namespace calico-system | head -3
sudo crictl ps --name calico-node
sudo crictl images | grep -E 'pause|nginx'Expected (shortened):
containerRuntimeEndpoint: unix:///var/run/containerd/containerd.sock
POD ID CREATED STATE NAME NAMESPACE
5e0527d980ee1 40 minutes ago Ready calico-typha-7d4f7659c4-xbf9k calico-sy...
2beaedfdaae8f About an hour ago Ready whisker-86cb858779-qmvbx calico-sy...
CONTAINER IMAGE CREATED STATE NAME ...
663954e153291 b424ff80b1e98 About an hour ago Running calico-node ...
docker.io/library/nginx 1.28 fda7399e81045 62.9MB
docker.io/library/nginx 1.29 5dfe511714e1f 63MB
registry.k8s.io/pause 3.10.2 4a83b15d3ecfe 321kB
A container’s logs straight from the runtime – your tool when the API server is dead (Day 14):
sudo crictl logs --tail 2 $(sudo crictl ps --name calico-node -q)Expected: two log lines from Calico. Kubernetes
containers live in containerd in the namespace k8s.io –
sudo ctr -n k8s.io containers list shows them as well.
Save your progress? Sign up for free
Drill 7: CNI and CSI on the node
Still on worker1:
sudo ls /etc/cni/net.d
ls /opt/cni/bin | head -8
sudo grep -o '"type":"[a-z-]*"' /etc/cni/net.d/10-calico.conflist | sort -uExpected:
10-calico.conflist
calico-kubeconfig
LICENSE
README.md
bandwidth
bridge
calico
calico-ipam
dhcp
dummy
"type":"calico"
"type":"calico-ipam"
"type":"portmap"
/etc/cni/net.d says which plugin;
/opt/cni/bin holds the programs. If the file is missing,
the node stays NotReady.
Back on cp1 – which CSI drivers exist, and which are
registered on worker1?
k get csidrivers
k get csinode worker1 -o jsonpath='{.spec.drivers[*].name}{"\n"}'Expected (shortened):
NAME ATTACHREQUIRED PODINFOONMOUNT ... MODES AGE
csi.tigera.io true true ... Ephemeral 69m
nfs.csi.k8s.io false false ... Persistent 8m41s
csi.tigera.io nfs.csi.k8s.io
Save your progress? Sign up for free
Exercise: Speed Drills
- List all CRDs of the group
cert-manager.io. Target time: 15 s - Show which container runtime, including version, each node uses. Target time: 30 s
- On
worker1, find the ID of the runningkube-proxycontainer. Target time: 30 s
Solution
k get crd -o name | grep cert-manager.io
k get nodes \
-o custom-columns=NAME:.metadata.name,RUNTIME:.status.nodeInfo.containerRuntimeVersionOn worker1:
sudo crictl ps --name kube-proxy -qExpected (excerpt):
customresourcedefinition.apiextensions.k8s.io/certificaterequests.cert-manager.io
...
NAME RUNTIME
cp1 containerd://2.4.2
worker1 containerd://2.4.2
worker2 containerd://2.4.2
Save your progress? Sign up for free
Checkpoint
k get bk nightly -o jsonpath='{.spec.retentionDays}{"\n"}'Expected: 30
k get certificate shop-cert -o jsonpath='{.status.conditions[0].status}{"\n"}'Expected: True
Challenge
Use cert-manager to create your own small CA: a self-signed CA
certificate lab-ca, on top of it an Issuer of
type CA, and with that a certificate for api.lab.
Hint: The CA certificate needs
isCA: true; the CA issuer points to its Secret with
ca.secretName.
Solution
cat <<'EOF' | k apply -f -
apiVersion: cert-manager.io/v1
kind: Certificate
metadata:
name: lab-ca
spec:
isCA: true
commonName: lab-ca
secretName: lab-ca-tls
issuerRef:
name: selfsigned
kind: Issuer
---
apiVersion: cert-manager.io/v1
kind: Issuer
metadata:
name: lab-ca
spec:
ca:
secretName: lab-ca-tls
---
apiVersion: cert-manager.io/v1
kind: Certificate
metadata:
name: api-cert
spec:
secretName: api-cert-tls
dnsNames:
- api.lab
issuerRef:
name: lab-ca
kind: Issuer
EOF
k get certificate
k get secret api-cert-tls -o jsonpath='{.data.tls\.crt}' | base64 -d \
| openssl x509 -noout -issuerExpected:
NAME READY SECRET AGE
api-cert True api-cert-tls 5s
lab-ca True lab-ca-tls 5s
shop-cert True shop-cert-tls 3m
issuer=CN = lab-ca
Clean up (cert-manager stays):
k config set-context --current --namespace=default
k delete ns tag12
k delete crd backups.lab.example.comSave your progress? Sign up for free
Exam Tip
- “Install the operator X” usually means: a Helm chart with a given
version, then check the CRDs with
k get crdand create a sample object. k api-resources --api-group=…andk explain <kind>.specshow how to write an unfamiliar object – faster than any docs.crictlis available on every node:ps,pods,logs,inspect.
Enterprise Note
- Operators run with broad permissions – read their ClusterRoles before installing.
- CRDs with multiple versions need conversion webhooks; a CRD upgrade belongs in the maintenance plan.
- Container runtime, CNI and CSI rarely change, but check the compatibility matrix with every Kubernetes upgrade.