Skip to content

Helm Values Recipes

Complete, copy-pasteable values files for the situations people actually hit. Pick the one matching your cluster, replace the CHANGE_ME placeholders, install. For what any individual value means, see Helm Chart Deployment; for why a certificate or DNS record behaves the way it does, TLS and DNS.

Your situation Recipe
Cluster already has an ingress controller and cert-manager 1 — Bring your own ingress and issuer
Bare cluster, hostname's DNS record created by hand, port 80 reachable 2 — Chart installs Traefik + cert-manager (HTTP-01)
OVH DNS zone, want the certificate and the A record handled for you 3 — OVH DNS-01 + ExternalDNS
AKS / Azure DNS zone 4 — Azure DNS-01 + ExternalDNS
Admin console only; students use each lab's own portal 5 — Admin-only release
Production: dedicated node pool, real storage class, sizing 6 — Production overlay

Recipes 5 and 6 are overlays: layer them on top of 1–4 by passing both files (-f values-ovh.yaml -f values-prod.yaml), later files winning.

Every recipe needs the secrets from Required secrets; they are spelled out in recipe 1 and abbreviated afterwards.

Recipe 1 — Bring your own ingress and issuer

Nothing is installed but EasyLab. You own the ingress controller, cert-manager and the ClusterIssuer, so you supply the cert-manager.io/cluster-issuer annotation yourself (the chart only adds it when it creates the issuer). NGINX shown; for Traefik drop the ssl-redirect annotation — the chart handles the redirect itself, see Redirecting HTTP to HTTPS.

# values-existing.yaml
secrets:
  adminPassword: "CHANGE_ME"
  # openssl rand -base64 32
  dataEncryptionKey: "CHANGE_ME_BASE64_32_BYTES"

ingress:
  enabled: true
  # kubectl get ingressclass
  className: nginx
  host: easylab.example.com
  tls:
    enabled: true
    secretName: easylab-tls
  annotations:
    cert-manager.io/cluster-issuer: letsencrypt   # your existing issuer
    nginx.ingress.kubernetes.io/ssl-redirect: "true"

persistence:
  jobs:
    size: 5Gi
    storageClass: ""      # "" = cluster default
  data:
    size: 1Gi
    storageClass: ""
helm install easylab oci://registry-1.docker.io/yodamad/easylab-helm \
  --version 1.3.0 \
  -f values-existing.yaml

Recipe 2 — Chart installs Traefik + cert-manager (HTTP-01)

For a cluster with neither. HTTP-01 needs no DNS credentials, but ingress.host must already resolve to the ingress controller's LoadBalancer IP and answer on port 80 for the ACME challenge — so this one still has a manual DNS step. It cannot issue a wildcard certificate; Let's Encrypt only does those over DNS-01 (recipe 3 or 4).

# values-bare-http01.yaml
secrets:
  adminPassword: "CHANGE_ME"
  dataEncryptionKey: "CHANGE_ME_BASE64_32_BYTES"

traefik:
  enabled: true
cert-manager:
  enabled: true
  crds:
    enabled: true

certManager:
  # must match where cert-manager's pods actually run
  namespace: cert-manager
  clusterIssuer:
    create: true
    email: "you@example.com"
    # while testing, avoid production rate limits:
    # server: https://acme-staging-v02.api.letsencrypt.org/directory

ingress:
  enabled: true
  className: traefik      # matches the IngressClass traefik.enabled installs
  host: easylab.example.com
  tls:
    enabled: true
helm install easylab oci://registry-1.docker.io/yodamad/easylab-helm \
  --version 1.3.0 \
  --namespace cert-manager --create-namespace \
  -f values-bare-http01.yaml \
  --wait --timeout 5m

--wait --timeout 5m is not optional here: the ClusterIssuer is created by a post-install hook, and without --wait it can fire before cert-manager's admission webhook is up and fail. --namespace cert-manager --create-namespace is what makes certManager.namespace line up — see Namespaces.

Recipe 3 — OVH DNS-01 + ExternalDNS

The fully automated path on an OVH DNS zone: DNS-01 gets the certificate (wildcards included, no port-80 requirement) and ExternalDNS creates and maintains the A record, so no manual DNS step remains. Both use the same three OVH API credentials, created at api.ovh.com/createToken with GET, POST, PUT and DELETE on /domain/zone/*.

# values-ovh.yaml
secrets:
  adminPassword: "CHANGE_ME"
  dataEncryptionKey: "CHANGE_ME_BASE64_32_BYTES"

traefik:
  enabled: true
cert-manager:
  enabled: true
  crds:
    enabled: true

certManager:
  namespace: cert-manager
  clusterIssuer:
    create: true
    email: "you@example.com"

dns:
  ovh:
    enabled: true
    # the zone itself, NOT the ingress host — ExternalDNS filters on it
    zone: "example.com"
    endpoint: ovh-eu
    credentials:
      applicationKey: "CHANGE_ME"
      applicationSecret: "CHANGE_ME"
      applicationConsumerKey: "CHANGE_ME"
  externalDns:
    enabled: true
    # log the intended changes without calling the OVH API; flip to false once
    # the logs show the record you expect
    dryRun: true

ingress:
  enabled: true
  className: traefik
  host: easylab.example.com    # or *.easylab.example.com for a wildcard
  tls:
    enabled: true

On a cluster that already has cert-manager, one command is enough. On a bare cluster it must be split in two, because the OVH webhook subchart's own PKI is made of cert-manager.io/v1 objects that cannot be validated before the CRDs exist (the Troubleshooting entry on missing cert-manager CRDs has the details):

helm install easylab oci://registry-1.docker.io/yodamad/easylab-helm \
  --version 1.3.0 \
  --namespace cert-manager --create-namespace \
  -f values-ovh.yaml \
  --set dns.ovh.enabled=false --set dns.externalDns.enabled=false \
  --wait --timeout 5m \
&& helm upgrade easylab oci://registry-1.docker.io/yodamad/easylab-helm \
  --version 1.3.0 \
  --namespace cert-manager \
  -f values-ovh.yaml \
  --wait --timeout 5m

Pass -f to both phases — helm upgrade does not inherit the previous release's values, so a bare --set dns.ovh.enabled=true would fall back to chart defaults and fail on the missing credentials. Phase one turns dns.externalDns.enabled off as well: it needs a DNS provider enabled, and the chart refuses to render otherwise.

Then verify, and switch dryRun off:

kubectl logs -n easylab -l app.kubernetes.io/component=externaldns
kubectl describe clusterissuer letsencrypt
kubectl get challenge -A
dig +short easylab.example.com

Only OVH's application authentication method is supported. OAuth2 credentials, external account bindings and per-issuer ACME servers are not exposed; use the upstream chart as its own release if you need them.

Recipe 4 — Azure DNS-01 + ExternalDNS

Same shape as recipe 3 on an Azure DNS zone, using cert-manager's built-in azureDNS solver — no extra webhook, so no two-phase install. One service principal covers both the solver and ExternalDNS; it needs permission to manage records in the zone, which the built-in DNS Zone Contributor role on the zone's resource group grants. This is a separate concern from the subscription-scoped credentials EasyLab uses to provision labs (see Azure cloud provider).

# values-azure.yaml
secrets:
  adminPassword: "CHANGE_ME"
  dataEncryptionKey: "CHANGE_ME_BASE64_32_BYTES"

traefik:
  enabled: true
cert-manager:
  enabled: true
  crds:
    enabled: true

certManager:
  namespace: cert-manager
  clusterIssuer:
    create: true
    email: "you@example.com"

dns:
  azure:
    enabled: true
    zone: "example.com"             # Azure DNS hosted zone name
    tenantId: "CHANGE_ME"
    subscriptionId: "CHANGE_ME"
    resourceGroup: "CHANGE_ME"      # resource group holding the zone
    clientId: "CHANGE_ME"
    clientSecret: "CHANGE_ME"
  externalDns:
    enabled: true
    dryRun: true

ingress:
  enabled: true
  className: traefik
  host: easylab.example.com
  tls:
    enabled: true
helm install easylab oci://registry-1.docker.io/yodamad/easylab-helm \
  --version 1.3.0 \
  --namespace cert-manager --create-namespace \
  -f values-azure.yaml \
  --wait --timeout 5m

Recipe 5 — Admin-only release

An overlay for the common workshop setup: this release serves the admin console only, and each lab gets its own student portal deployed inside the lab's cluster. See In-lab student portals for what each value does.

# values-admin-only.yaml
config:
  mode: admin
  # where portals send students for Microsoft / GitHub / GitLab sign-in;
  # defaults to the ingress host, so this is only needed if they differ
  publicUrl: "https://easylab.example.com"
  # defaults to this release's own image; the lab's nodes must be able to pull it
  portalImage: ""

secrets:
  # not needed with mode: admin — each portal generates its own student password
  studentPassword: ""

Recipe 6 — Production overlay

An overlay with the settings that matter once a release is more than a demo: a dedicated node pool, a named storage class, and sizing. See Pinning to a dedicated node pool for why the node labels also change how labs pre-pull images.

# values-prod.yaml
# keep EasyLab off the pool student workspaces run on
nodeSelector:
  pool: control-plane
tolerations:
  - key: dedicated
    operator: Equal
    value: easylab
    effect: NoSchedule

persistence:
  jobs:
    size: 20Gi          # Pulumi state and job logs grow with lab count
    # name one explicitly rather than relying on the cluster default, and pick a
    # network-attached provisioner — `kubectl get storageclass`
    storageClass: CHANGE_ME
  data:
    size: 2Gi
    storageClass: CHANGE_ME

resources:
  requests:
    memory: "1024Mi"
    cpu: "500m"
  limits:
    memory: "4096Mi"
    cpu: "3000m"

image:
  # pin the image instead of tracking the chart's appVersion. Note the leading
  # "v": app image tags carry it, chart versions do not.
  tag: "v1.3.0"
  pullPolicy: IfNotPresent

# pin the chart's subcharts to the same pool
traefik:
  nodeSelector:
    pool: control-plane
cert-manager:
  nodeSelector:
    pool: control-plane

Next steps

  • Namespaces — which namespace each resource lands in, and the --namespace trap that stops certificates issuing.
  • Values reference — every value the chart accepts.
  • Upgrade and Uninstall — including what helm uninstall leaves behind.
  • Troubleshooting — symptom-first, for when one of these recipes does not behave.