Skip to content
Hands-on IT IT Drills

Day 9: Ingress, Gateway and TLS

Goal: You find Traefik in RKE2, read its access log and error log, and tell 404, 503 and the wrong certificate apart by their cause. · Duration: approx. 75 min · Prerequisite: Day 7

Briefing

  • RKE2 1.37 ships Traefik as a DaemonSet with hostPort 80/443 on every node. You test with curl --resolve directly against cp1.
  • With Traefik, 404 means: no router matches. Whether the Ingress was ignored entirely or is broken is in the error log (ERR).
  • 503 means: the router matches, but there is no ready Pod behind the Service.
  • Wrong certificate: Traefik serves its TRAEFIK DEFAULT CERT.
Fault Symptom Where the answer is
Wrong ingressClassName 404, log is silent get ingress, get ingressclass
Wrong Service port 404 Traefik log Cannot create service
No ready Pods 503 no available server get pods, events
Wrong TLS Secret TRAEFIK DEFAULT CERT Traefik log, Secret keys
Certificate expired certificate has expired openssl s_client
HTTPRoute stuck 404 .status.parents[].conditions

Drill 1: Find Traefik and make it talkative

# on cp1
mkdir -p ~/debug/tag09 && cd ~/debug/tag09
kubectl -n kube-system get ds rke2-traefik | cut -c1-72
kubectl get ingressclass
kubectl -n kube-system get ds rke2-traefik \
  -o jsonpath='{.spec.template.spec.containers[0].ports}{"\n"}' \
  | tr '{' '\n' | grep host

Expected:

NAME           DESIRED   CURRENT   READY   UP-TO-DATE   AVAILABLE   NODE
rke2-traefik   3         3         3       3            3           kube
NAME                CONTROLLER                      PARAMETERS   AGE
traefik (default)   traefik.io/ingress-controller   <none>       104m
"containerPort":8000,"hostPort":80,"name":"web","protocol":"TCP"},
"containerPort":8443,"hostPort":443,"name":"websecure","protocol":"TCP"}]

Traefik belongs to the RKE2 chart rke2-traefik. You pass your own values to it via a HelmChartConfig with the same name: access log on, colors in the log off, dashboard and Gateway API provider enabled.

# on cp1
cat > traefik-config.yaml <<'EOF'
apiVersion: helm.cattle.io/v1
kind: HelmChartConfig
metadata: {name: rke2-traefik, namespace: kube-system}
spec:
  valuesContent: |-
    log:
      noColor: true
    accessLog:
      enabled: true
    ingressRoute:
      dashboard:
        enabled: true
    providers:
      kubernetesGateway:
        enabled: true
EOF
kubectl apply -f traefik-config.yaml

The helm-controller starts a Job helm-install-rke2-traefik, then the DaemonSet rolls out again – about a minute in total. Check the result in the arguments, not in the HelmChartConfig:

# on cp1
kubectl -n kube-system rollout status ds/rke2-traefik
kubectl -n kube-system get ds rke2-traefik \
  -o jsonpath='{.spec.template.spec.containers[0].args}' | tr ',' '\n' \
  | grep -E 'accesslog=|noColor|kubernetesgateway"'
kubectl get gatewayclass
kubectl get gateway -A

Expected:

daemon set "rke2-traefik" successfully rolled out
"--providers.kubernetesgateway"
"--log.noColor=true"
"--accesslog=true"
NAME      CONTROLLER                      ACCEPTED   AGE
traefik   traefik.io/gateway-controller   True       26s
NAMESPACE     NAME              CLASS     ADDRESS   PROGRAMMED   AGE
kube-system   traefik-gateway   traefik             True         27s

Warning: The chart in RKE2 1.37 (rke2-traefik 42.1.301, Traefik 3.7) knows log and accessLog. It silently ignores the keys logs.general and logs.access found in other docs – which is why you always check the arguments.

Save your progress? Sign up for free

Drill 2: 404 – Traefik does not feel responsible

Break it:

# on cp1
kubectl create namespace tag09
cat > shop.yaml <<'EOF'
apiVersion: apps/v1
kind: Deployment
metadata: {name: shop, namespace: tag09}
spec:
  replicas: 2
  selector: {matchLabels: {app: shop}}
  template:
    metadata: {labels: {app: shop}}
    spec:
      containers:
      - name: whoami
        image: traefik/whoami:v1.11
        ports: [{name: http, containerPort: 80}]
---
apiVersion: v1
kind: Service
metadata: {name: shop, namespace: tag09}
spec:
  selector: {app: shop}
  ports: [{name: http, port: 80, targetPort: http}]
EOF
cat > ingress.yaml <<'EOF'
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata: {name: shop, namespace: tag09}
spec:
  ingressClassName: nginx
  rules:
  - host: shop.labor.test
    http:
      paths:
      - path: /
        pathType: Prefix
        backend:
          service: {name: shop, port: {number: 80}}
EOF
kubectl apply -f shop.yaml -f ingress.yaml
kubectl -n tag09 rollout status deploy/shop

Symptom:

# on cp1
curl -s -i --resolve shop.labor.test:80:172.16.0.126 http://shop.labor.test/ | head -1

Expected:

HTTP/1.1 404 Not Found

Investigate: The access log of the Traefik Pod on cp1, then the Ingress itself:

# on cp1
T=$(kubectl -n kube-system get pods -l app.kubernetes.io/name=rke2-traefik \
  --field-selector spec.nodeName=cp1 -o name)
kubectl -n kube-system logs $T --since=1m | tail -1
kubectl -n tag09 get ingress shop
kubectl -n tag09 describe ingress shop | grep Events

Expected (line wrapped):

172.16.0.126 - - [20/Nov/2026:08:21:04 +0000] "GET / HTTP/1.1" 404 19 "-" "-" 2 "-"
  "-" 0ms
NAME   CLASS   HOSTS             ADDRESS   PORTS   AGE
shop   nginx   shop.labor.test             80      3s
Events:            <none>

Cause: The two "-" before the duration are router name and server URL – no router matched. The Ingress asks for the class nginx, which does not exist. Traefik ignores it without comment: no event, no ERR line – just the access log entry without a router.

Fix:

# on cp1
kubectl -n tag09 patch ingress shop --type=merge \
  -p '{"spec":{"ingressClassName":"traefik"}}'

Verify:

# on cp1
curl -s --resolve shop.labor.test:80:172.16.0.126 http://shop.labor.test/ | head -1
kubectl -n kube-system logs $T --since=10s | tail -1

Expected (line wrapped):

Hostname: shop-555c9f66b5-dmqg6
172.16.0.126 - - [20/Nov/2026:08:21:16 +0000] "GET / HTTP/1.1" 200 407 "-" "-" 3
  "tag09-shop-shop-labor-test@kubernetes" "http://10.42.0.49:80" 9ms

Why? Without ingressClassName the default class traefik would have applied. A wrong class is worse than none. The empty ADDRESS column, on the other hand, is not a fault: rke2-traefik is a ClusterIP Service without an external address.

Save your progress? Sign up for free

Drill 3: 404 despite the right class

Break it:

# on cp1
kubectl -n tag09 patch ingress shop --type=json -p '[{"op": "replace",
  "path": "/spec/rules/0/http/paths/0/backend/service/port/number", "value": 8080}]'

Symptom:

# on cp1
curl -s -i --resolve shop.labor.test:80:172.16.0.126 http://shop.labor.test/ | head -1

Expected:

HTTP/1.1 404 Not Found

Investigate:

# on cp1
kubectl -n kube-system logs $T --since=1m | grep ERR | tail -1
kubectl -n tag09 get svc shop -o jsonpath='{.spec.ports}{"\n"}'

Expected (line wrapped):

2026-11-20T08:25:53Z ERR Cannot create service error="service port not found"
  ingress=shop namespace=tag09 providerName=kubernetes serviceName=shop
  servicePort=&ServiceBackendPort{Name:,Number:8080,}
[{"name":"http","port":80,"protocol":"TCP","targetPort":"http"}]

Cause: The Ingress names port 8080 of the Service, but the Service only has 80. Traefik then does not build the router at all – to the client this looks like Drill 2, but the log says it plainly.

Fix: Refer to the port name, as on Day 7:

# on cp1
kubectl -n tag09 patch ingress shop --type=json -p '[{"op": "replace",
  "path": "/spec/rules/0/http/paths/0/backend/service/port",
  "value": {"name": "http"}}]'

Verify:

# on cp1
curl -s --resolve shop.labor.test:80:172.16.0.126 http://shop.labor.test/ | head -1

Expected:

Hostname: shop-555c9f66b5-9cf7g

Save your progress? Sign up for free

Drill 4: 503 – the route is there, but nobody is ready

Break it: Recreate the Deployment with a readiness probe on a port nobody listens on:

# on cp1
kubectl -n tag09 delete deploy shop
cat > shop-bereit.yaml <<'EOF'
apiVersion: apps/v1
kind: Deployment
metadata: {name: shop, namespace: tag09}
spec:
  replicas: 2
  selector: {matchLabels: {app: shop}}
  template:
    metadata: {labels: {app: shop}}
    spec:
      containers:
      - name: whoami
        image: traefik/whoami:v1.11
        ports: [{name: http, containerPort: 80}]
        readinessProbe:
          httpGet: {path: /, port: 8080}
          periodSeconds: 5
EOF
kubectl apply -f shop-bereit.yaml

Symptom:

# on cp1
curl -s -i --resolve shop.labor.test:80:172.16.0.126 http://shop.labor.test/ \
  | sed -n '1p;$p'

Expected:

HTTP/1.1 503 Service Unavailable
no available server

Investigate:

# on cp1
kubectl -n kube-system logs $T --since=15s | tail -1
kubectl -n tag09 get pods -l app=shop
kubectl -n tag09 events --types=Warning | tail -1

Expected (lines wrapped):

172.16.0.126 - - [20/Nov/2026:08:27:09 +0000] "GET / HTTP/1.1" 503 20 "-" "-" 6
  "tag09-shop-shop-labor-test@kubernetes" "-" 0ms
NAME                    READY   STATUS    RESTARTS   AGE
shop-5d77bb949d-blrv6   0/1     Running   0          20s
shop-5d77bb949d-tzh92   0/1     Running   0          20s
3s (x6 over 19s)   Warning   Unhealthy   Pod/shop-5d77bb949d-tzh92   Readiness probe
  failed: Get "http://10.42.2.45:8080/": dial tcp 10.42.2.45:8080: connect:
  connection refused

Cause: This time the router matches (its name is in the access log), but the server URL is "-": no Pod is ready, and the EndpointSlice lists both with ready: false (Day 5).

Fix:

# on cp1
kubectl -n tag09 patch deploy shop --type=json -p '[{"op": "replace",
  "path": "/spec/template/spec/containers/0/readinessProbe/httpGet/port",
  "value": "http"}]'
kubectl -n tag09 rollout status deploy/shop

Verify:

# on cp1
curl -s -i --resolve shop.labor.test:80:172.16.0.126 http://shop.labor.test/ | head -1

Expected:

HTTP/1.1 200 OK

Why port 8080 and not a wrong path? whoami answers every path with 200 – a probe on /gesund (‘healthy’) would succeed. A probe is only as good as what the application really checks there.

Save your progress? Sign up for free

Drill 5: The wrong certificate

Break it: A self-signed certificate valid for one day, stored as a Secret with the key names that openssl suggests:

# on cp1
openssl req -x509 -newkey rsa:2048 -nodes -days 1 -subj "/CN=shop.labor.test" \
  -addext "subjectAltName=DNS:shop.labor.test" -keyout shop.key -out shop.crt
kubectl -n tag09 create secret generic shop-tls \
  --from-file=cert.pem=shop.crt --from-file=key.pem=shop.key
kubectl -n tag09 patch ingress shop --type=merge -p \
  '{"spec":{"tls":[{"hosts":["shop.labor.test"],"secretName":"shop-tls"}]}}'

Symptom:

# on cp1
curl -sS -o /dev/null --resolve shop.labor.test:443:172.16.0.126 \
  https://shop.labor.test/ 2>&1 | head -1
echo | openssl s_client -connect 172.16.0.126:443 -servername shop.labor.test \
  2>/dev/null | openssl x509 -noout -subject -enddate

Expected:

curl: (60) SSL certificate problem: self-signed certificate
subject=CN = TRAEFIK DEFAULT CERT
notAfter=Nov 20 08:27:27 2027 GMT

Your own certificate would have reported self-signed too. Only openssl s_client shows which certificate arrives: Traefik’s built-in one.

Investigate:

# on cp1
kubectl -n kube-system logs $T --since=1m | grep 'Error configuring TLS' | tail -1
kubectl -n tag09 get secret shop-tls -o jsonpath='{.type} {.data}' | cut -c1-40

Expected (line wrapped):

2026-11-20T08:27:45Z ERR Error configuring TLS error="secret tag09/shop-tls is missing
  the following TLS data entries: tls.crt, tls.key" ingress=shop namespace=tag09
  providerName=kubernetes
Opaque {"cert.pem":"LS0tLS1CRUdJTiBDRVJ

Cause: Ingress controllers expect the keys tls.crt and tls.key. If they are missing, Traefik takes the default certificate, and HTTPS still “works” with curl -k.

Fix:

# on cp1
kubectl -n tag09 delete secret shop-tls
kubectl -n tag09 create secret tls shop-tls --cert=shop.crt --key=shop.key

Verify:

# on cp1
echo | openssl s_client -connect 172.16.0.126:443 -servername shop.labor.test \
  2>/dev/null | openssl x509 -noout -subject -enddate
curl -sS --cacert shop.crt --resolve shop.labor.test:443:172.16.0.126 \
  https://shop.labor.test/ | head -1

Expected:

subject=CN = shop.labor.test
notAfter=Nov 21 08:27:36 2026 GMT
Hostname: shop-75888d6697-kwx9n

Save your progress? Sign up for free

Drill 6: The expired certificate

Break it: OpenSSL 3.0 (Ubuntu 24.04) does not know -not_after yet, but -days -1 produces a certificate that expired yesterday:

# on cp1
openssl req -new -key shop.key -subj "/CN=shop.labor.test" \
  -addext "subjectAltName=DNS:shop.labor.test" -out alt.csr
openssl x509 -req -in alt.csr -key shop.key -copy_extensions copy -days -1 \
  -out alt.crt
kubectl -n tag09 create secret tls shop-tls --cert=alt.crt --key=shop.key \
  --dry-run=client -o yaml | kubectl replace -f -

Symptom:

# on cp1
curl -sS -o /dev/null --cacert alt.crt --resolve shop.labor.test:443:172.16.0.126 \
  https://shop.labor.test/ 2>&1 | head -1

Expected:

curl: (60) SSL certificate problem: certificate has expired

Investigate: What does the server deliver, and what is in the Secret?

# on cp1
echo | openssl s_client -connect 172.16.0.126:443 -servername shop.labor.test \
  2>/dev/null | openssl x509 -noout -dates -checkend 0
kubectl -n tag09 get secret shop-tls -o jsonpath='{.data.tls\.crt}' | base64 -d \
  | openssl x509 -noout -enddate -ext subjectAltName

Expected:

notBefore=Nov 20 08:28:09 2026 GMT
notAfter=Nov 19 08:28:09 2026 GMT
Certificate will expire
notAfter=Nov 19 08:28:09 2026 GMT
X509v3 Subject Alternative Name:
    DNS:shop.labor.test

Cause: notAfter lies in the past. -checkend 0 answers the question “is it still valid right now?” – with a number of seconds instead of 0 it becomes a monitoring check (“does it expire within 7 days?”).

Fix:

# on cp1
kubectl -n tag09 create secret tls shop-tls --cert=shop.crt --key=shop.key \
  --dry-run=client -o yaml | kubectl replace -f -

Verify:

# on cp1
echo | openssl s_client -connect 172.16.0.126:443 -servername shop.labor.test \
  2>/dev/null | openssl x509 -noout -checkend 0

Expected:

Certificate will not expire

Why replace and not apply? The Secret was created with create; apply would work but warn that the last-applied-configuration annotation is missing. Traefik loads the new certificate without a restart.

Save your progress? Sign up for free

Drill 7: Gateway API – the route tells you what it lacks

The Gateway API sets conditions where Ingress stays silent.

Break it: An HTTPRoute attached to the chart’s Gateway in kube-system, with a wrong backend:

# on cp1
cat > route.yaml <<'EOF'
apiVersion: gateway.networking.k8s.io/v1
kind: HTTPRoute
metadata: {name: shop, namespace: tag09}
spec:
  parentRefs:
  - {name: traefik-gateway, namespace: kube-system}
  hostnames: [gw.labor.test]
  rules:
  - backendRefs:
    - {name: shop-v2, port: 80}
EOF
kubectl apply -f route.yaml

Symptom:

# on cp1
curl -s -i --resolve gw.labor.test:80:172.16.0.126 http://gw.labor.test/ | head -1

Expected:

HTTP/1.1 404 Not Found

Investigate:

# on cp1
J='{range .status.parents[*].conditions[*]}{.type}={.status} {.reason}: {.message}'
kubectl -n tag09 get httproute shop -o jsonpath="$J{\"\n\"}{end}"
kubectl -n kube-system get gateway traefik-gateway \
  -o jsonpath='{.spec.listeners[0].allowedRoutes}{"\n"}'

Expected (line wrapped):

Accepted=False NotAllowedByListeners:
ResolvedRefs=False BackendNotFound: Cannot load HTTPBackendRef tag09/shop-v2:
  getting service: service "shop-v2" not found
{"namespaces":{"from":"Same"}}

Cause: Two faults, two conditions. The listener of the chart’s Gateway only accepts routes from kube-system (from: Same), and the Service shop-v2 does not exist.

Fix: Your own Gateway in the application’s namespace instead of bending the chart’s Gateway, plus the right backend. Port 8000 is Traefik’s web entrypoint:

# on cp1
cat > gateway.yaml <<'EOF'
apiVersion: gateway.networking.k8s.io/v1
kind: Gateway
metadata: {name: shop-gw, namespace: tag09}
spec:
  gatewayClassName: traefik
  listeners:
  - {name: web, port: 8000, protocol: HTTP, hostname: gw.labor.test}
EOF
kubectl apply -f gateway.yaml
kubectl -n tag09 patch httproute shop --type=json -p '[
  {"op": "replace", "path": "/spec/parentRefs/0", "value": {"name": "shop-gw"}},
  {"op": "replace", "path": "/spec/rules/0/backendRefs/0/name", "value": "shop"}]'

Verify:

# on cp1
kubectl -n tag09 get gateway shop-gw
J='{range .status.parents[*].conditions[*]}{.type}={.status} {.reason}'
kubectl -n tag09 get httproute shop -o jsonpath="$J{\"\n\"}{end}"
curl -s --resolve gw.labor.test:80:172.16.0.126 http://gw.labor.test/ | head -1

Expected:

NAME      CLASS     ADDRESS   PROGRAMMED   AGE
shop-gw   traefik             True         4s
Accepted=True Accepted
ResolvedRefs=True ResolvedRefs
Hostname: shop-75888d6697-kwx9n

Save your progress? Sign up for free

Drill 8: The dashboard

The dashboard runs on the internal entrypoint traefik (port 8080) and is reachable only via port forward. From your machine, use an SSH tunnel to cp1:

# on cp1
kubectl -n kube-system port-forward ds/rke2-traefik 9000:8080 &
sleep 2
curl -s -o /dev/null -w '%{http_code}\n' http://127.0.0.1:9000/dashboard/
# on your machine
ssh -N -L 9000:127.0.0.1:9000 [email protected]

Expected: 200 on cp1, then http://localhost:9000/dashboard/ in the browser:

Traefik dashboard: entrypoints, routers and services without errors

Under HTTP Routers you see every rule Traefik has built from Ingress, IngressRoute and HTTPRoute – with entrypoints and target Service:

HTTP routers: Ingress shop (web, metrics and websecure), HTTPRoute gw.labor.test, dashboard

Why? Whatever is missing in the dashboard, Traefik did not build (Drills 2 and 3). The router tag09-shop-shop-labor-test is attached to web and metrics: an Ingress without an entrypoint annotation is attached to all entrypoints except the internal traefik; for websecure, on which the RKE2 chart enables TLS, Traefik builds a separate router websecure-… – even for an Ingress without spec.tls (verified live with admin from the challenge). The router list is also available as JSON at http://127.0.0.1:9000/api/http/routers. Leave the port forward running – the challenge needs the router API; the cleanup ends it. Until then it prints Handling connection for 9000 between your outputs on every request.

Save your progress? Sign up for free

Exercise: Speed Drills

  1. Show which Ingress class is the default. Target time: 10 s
  2. Show the last two errors of the Traefik Pod on cp1. Target time: 30 s
  3. Show when the certificate served for shop.labor.test expires. Target time: 30 s
  4. Show Type, Status and Reason of the conditions of the HTTPRoute shop. Target time: 30 s

Solution

# on cp1
kubectl get ingressclass
kubectl -n kube-system logs $T | grep ERR | tail -2 | cut -c1-88
echo | openssl s_client -connect 172.16.0.126:443 -servername shop.labor.test \
  2>/dev/null | openssl x509 -noout -enddate
kubectl -n tag09 describe httproute shop | sed -n '/Conditions/,/Controller Name/p' \
  | grep -E 'Reason|Status|Type'

Expected:

NAME                CONTROLLER                      PARAMETERS   AGE
traefik (default)   traefik.io/ingress-controller   <none>       117m
2026-11-20T08:28:48Z ERR Unable to load HTTPRoute backend: Cannot load HTTPBackendRef ta
2026-11-20T08:28:48Z ERR Unable to load HTTPRoute backend: Cannot load HTTPBackendRef ta
notAfter=Nov 21 08:27:36 2026 GMT
      Reason:                Accepted
      Status:                True
      Type:                  Accepted
      Reason:                ResolvedRefs
      Status:                True
      Type:                  ResolvedRefs

Save your progress? Sign up for free

Checkpoint

# on cp1
curl -s --resolve shop.labor.test:80:172.16.0.126 http://shop.labor.test/ | head -1
curl -sS --cacert shop.crt --resolve shop.labor.test:443:172.16.0.126 \
  https://shop.labor.test/ | head -1
curl -s --resolve gw.labor.test:80:172.16.0.126 http://gw.labor.test/ | head -1

Expected: three whoami answers, for example:

Hostname: shop-75888d6697-g8zbd
Hostname: shop-75888d6697-kwx9n
Hostname: shop-75888d6697-kwx9n

Challenge

A colleague creates an Ingress admin and reports: “http://admin.labor.test returns 404, but the Ingress is there and the class is right.” Find the cause without the Traefik log and repair it.

# on cp1
cat > admin.yaml <<'EOF'
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
  name: admin
  namespace: tag09
  annotations:
    traefik.ingress.kubernetes.io/router.entrypoints: websecure
spec:
  rules:
  - host: admin.labor.test
    http:
      paths:
      - path: /
        pathType: Prefix
        backend:
          service: {name: shop, port: {name: http}}
EOF
kubectl apply -f admin.yaml

Hint: Ask the dashboard’s router API which entrypoint the router is attached to.

Solution

# on cp1
curl -s -o /dev/null -w '%{http_code}\n' --resolve admin.labor.test:80:172.16.0.126 \
  http://admin.labor.test/
curl -s http://127.0.0.1:9000/api/http/routers | jq -c \
  '.[] | select(.name | test("admin")) | {name, entryPoints, tls}'
curl -sk -o /dev/null -w '%{http_code}\n' \
  --resolve admin.labor.test:443:172.16.0.126 https://admin.labor.test/

Expected (line wrapped):

404
{"name":"tag09-admin-admin-labor-test@kubernetes","entryPoints":["websecure"],
  "tls":{"options":"default"}}
200

The annotation binds the router to websecure (HTTPS) only. Over HTTP no router matches. If HTTP should work as well, remove the annotation:

# on cp1
kubectl -n tag09 annotate ingress admin \
  traefik.ingress.kubernetes.io/router.entrypoints-
sleep 3
curl -s -o /dev/null -w '%{http_code}\n' --resolve admin.labor.test:80:172.16.0.126 \
  http://admin.labor.test/

Expected:

ingress.networking.k8s.io/admin annotated
200

Save your progress? Sign up for free

Cleanup

The HelmChartConfig stays: you will need the access log, dashboard and Gateway provider again in the competition.

# on cp1
kill %1
kubectl delete namespace tag09
kubectl get gateway -A

Expected:

namespace "tag09" deleted
NAMESPACE     NAME              CLASS     ADDRESS   PROGRAMMED   AGE
kube-system   traefik-gateway   traefik             True         12m

Interview Note

  • 404 or 503 from the Ingress – what’s the difference? 404: no router matches (class, host, path, entrypoint, broken backend). 503: the router matches, but there is no ready endpoint.
  • HTTPS serves the wrong certificate? openssl s_client -servername shows what arrives; TRAEFIK DEFAULT CERT means: the Secret is missing or has the wrong keys.
  • Why Gateway API? Separation of roles (Gateway vs. Route) and conditions with reasons instead of a silent 404.

Enterprise Note

  • In production, certificates come from cert-manager; expiry warnings come from a metric (certmanager_certificate_expiration_timestamp_seconds) rather than from openssl by hand.
  • In RKE2, change values only via HelmChartConfig. If you edit the DaemonSet directly, you lose the change on the next Helm reconciliation.
  • Ingress controller behind a load balancer: you only get the real client IP with Proxy Protocol or externalTrafficPolicy: Local.
  • Never expose the dashboard to the outside without authentication.
  • Gateway API in depth (TLS listeners, ReferenceGrant, separation of roles) is covered in “Gateway API hands-on”.