bunkerity / bunkerweb

🛡️ Open-source and next-generation Web Application Firewall (WAF)
https://www.bunkerweb.io
GNU Affero General Public License v3.0
6.46k stars 363 forks source link

[FEATURE] Add cert-manager Support on Kubernetes Integration #585

Open polarstack opened 1 year ago

polarstack commented 1 year ago

What's needed and why ? Hi All

I see the certbot-dns-* examples which could cover at least the second use case (e.g. ../examples/certbot-dns-cloudflare/docker-compose.yml), but as far as I understands it needs you to mount the "certs" Volume to bunkerweb, scheduler and a custom certbot container with the corresponding config. But not sure how I would implement that on Kubernetes. Using Kubernetes Secrets and Ingress Annotations would make it more natively on that Integration.

Implementations ideas (optional) The Documentation for cert-manager is here: https://cert-manager.io/docs/ But the Installation and Configuration of cert-manager can be out of scope

cert-manager stores the key and crt in a Kubernetes Secret:

# k3s kubectl describe secrets/dummy-secret  -n dummy-namespace
Name:         dummy-secret-tls-0
Namespace:    dummy-namespace
Labels:       controller.cert-manager.io/fao=true
Annotations:  cert-manager.io/alt-names: myapp.example.com
              cert-manager.io/certificate-name: dummy-secret-tls-0
              cert-manager.io/common-name: myapp.example.com
              cert-manager.io/issuer-group: cert-manager.io
              cert-manager.io/issuer-kind: ClusterIssuer
              cert-manager.io/issuer-name: my-issuer-example

Type:  kubernetes.io/tls

Data
====
tls.key:  xxxx bytes
tls.crt:  yyyy bytes

On the Ingress the mapping happens in the annotations section:

Annotations:                   cert-manager.io/cluster-issuer: my-issuer-example
                               cert-manager.io/private-key-rotation-policy: Always

Not a specialist as I'm still learning, but I guess the Ingress Annotation triggers cert-manager which then stores the crt/key as a secret. Finally the ingress controller (e.g. Traefik) picks it up and deploys/configures the TLS termination. Maybe it's also triggered by the Ingress Controller itself. See for example the official Kubernetes NGinx Ingress Chart: https://github.com/kubernetes/ingress-nginx/blob/afd1311f8529c21fdf6621bf683bec814e698f1d/charts/ingress-nginx/templates/admission-webhooks/cert-manager.yaml

As one can have multiple issuer, I would suggest to leave that as a matter of cert-manager and define the secret only:

annotations:
   bunkerweb.io/myapp.example.com_CLUSTER_ISSUER_CERTIFICATE: dummy-secret-tls-0

Finally the BunkerWeb Scheduler(?) would pick up the secret and store it in /certs/ like it does for example on http, server-http, modsec etc. with the ConfigMap Feature.

Hopefully I was able to explain the need simply, otherwise please let me know if I should elaborate. If you think this is an edge case and doesn't map your Roadmap, don't worry about it and close the issue :-)

fl0ppy-d1sk commented 1 year ago

Hello @polarstack,

That's an interesting idea. IMO it's even a must have.

Looks like we should parse the tls part of the ingress resource but we need to dig deeper.

fl0ppy-d1sk commented 10 months ago

Hello @polarstack,

Quick update : we now support tls section of Ingress in 1.5.5. Don't hesitate to test it by yourself.

Keeping this issue open because we still need to document (and test) the cert-manager integration.

schmittse commented 1 week ago

Hi,

The secret loading is a start but not enough for the webapp challenge.

When you deploy a new Ingress using let's encrypt (annotation cert-manager.io/cluster-issuer: letsencrypt) and the secret does not exists, the certmanager spin a new nginx pod and a new ingress listening only on port 80 for the let's encrypt challenge :

apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
  annotations:
    kubernetes.io/ingress.class: nginx
    nginx.ingress.kubernetes.io/whitelist-source-range: 0.0.0.0/0,::/0
  generateName: cm-acme-http-solver-
  labels:
    acme.cert-manager.io/http-domain: "123456789"
    acme.cert-manager.io/http-token: "987654321"
    acme.cert-manager.io/http01-solver: "true"
  name: cm-acme-http-solver-b2n7j
  namespace: hello-world
  ownerReferences:
  - apiVersion: acme.cert-manager.io/v1
    blockOwnerDeletion: true
    controller: true
    kind: Challenge
    name: lets-hello-world-bw-1-1580467882-4097216534
    uid: 389d03bc-9634-4641-8f6c-20c2bb07c647
spec:
  ingressClassName: nginx
  rules:
  - host: hello-world.example.com
    http:
      paths:
      - backend:
          service:
            name: cm-acme-http-solver-lglzx
            port:
              number: 8089
        path: /.well-known/acme-challenge/qRh5NAZCNVFnq0qkNQx3vGDHhPbBUN0n5Cjm8Hlcc_A
        pathType: ImplementationSpecific

So we have 2 Ingresses for the same domain ; one with path: / and one with the .well-known path seen above.

In the Nginx Ingress Controller, both Ingress are merged into a server directive :

[...]
server {                               
  server_name hello-world.example.com ;

  listen 80 proxy_protocol ;
  [...]

  location /.well-known/acme-challenge/qRhcNAZCNVFnq9qFNQx3vGDHhCbBUN0S5CjZ8Hlcc_A/ {
    [...]
    set $proxy_upstream_name "hello-world-cm-acme-http-solver-lglzx-8089";
    [...]
  }
  [...]
  location / {
    [...]
    set $proxy_upstream_name "hello-world-hello-world-service-8888";
    [...]
  }
}

For what I could found in your generated config files, I only get the / location.

location / {
  etag off;
  set $backend1069 "http://hello-world-service.hello-world.svc.cluster.local:8888";
  [...]
}