Skip to content

waldner/cert-manager-webhook-he

Repository files navigation

cert-Manager ACME DNS01 Webhook Solver for HE DNS

Go Report Card Releases LICENSE

A webhook to use HE DNS as a DNS01 ACME Issuer for cert-manager.

Installation

The webhook must be installed in the same namespace where cert-manager is running (usually cert-manager). See the following paragraph for an explanation of what "use secrets" or "use environment variables" mean.

To install with helm from the registry, run:

# to use secrets for credentials:
$ helm upgrade --install --namespace cert-manager --set auth.useSecrets=true cert-manager-webhook-he oci://ghcr.io/waldner/charts/cert-manager-webhook-he

# to use environment variables
$ helm upgrade --install --namespace cert-manager \
   --set auth.heUsername=myusername \
   --set auth.hePassword=mypassword \
   --set auth.heApiKey=myapikey \
   cert-manager-webhook-he oci://ghcr.io/waldner/charts/cert-manager-webhook-he

If you want to install from the repo checkout:

$ git clone https://github.com/waldner/cert-manager-webhook-he.git
$ cd cert-manager-webhook-he

# to use secrets for credentials:
$ helm upgrade --install --namespace cert-manager \
     --set auth.useSecret=true cert-manager-webhook-he deploy/cert-manager-webhook-he

# to use environment variables:
$ helm upgrade --install --namespace cert-manager \
   --set auth.heUsername=myusername \
   --set auth.hePassword=mypassword \
   --set auth.heApiKey=myapikey \
   cert-manager-webhook-he deploy/cert-manager-webhook-he

Check the logs with

$ kubectl get pods -n cert-manager --watch
$ kubectl logs -n cert-manager cert-manager-webhook-he-xxxxx

Concepts and configuration

The webhook can work in two modes: login and dynamic-dns (explained later). Whatever method you choose, in the Issuer YAML the configuration options must be under the dns01.webhook path (see examples below). Also regardless of the mode, the webhook can read its credentials either from environment variables (the default) or from kubernetes Secrets.

The main difference is that credentials passed via environment variables are static and can only be changed by redeploying the container, while credentials stored in secrets can be updated by just updating the secrets (or creating new ones), then referencing them from the Issuer.

If you want to use multiple accounts, or be able to set per-issuer credentials, you should use secrets. If, on the other hand, you only have a single set of credentials that you want to use everywhere, using environment variables is appropriate.

Whether the webhook reads the credentials from environment variables (the default) or from secrets is determined by the auth.useSecrets variable of the Helm chart, which you can override when you deploy the chart.

Choosing to use secrets or environment variables has implications for the deployment, since when using secrets additional permissions will be given to the webhook service account to be able to read secrets (see below for details).

login mode

In login mode, the TXT record(s) are created and deleted by logging into the HE DNS control panel using the normal user credentials. The credentials needed for this mode are the HE DNS control panel username and password. If you store them in a secret, they must be associated respectively to the username and password keys in the secret data. Example secret:

apiVersion: v1
kind: Secret
metadata:
  name: he-credentials
type: Opaque
stringData:
  username: "myHEusername"
  password: "myHEpassword"

If you use environment variables, you must pass them as auth.heUsername and auth.hePassword when deploying the Helm chart.

Here's a sample Issuer configuration for the login mode:

apiVersion: cert-manager.io/v1
kind: Issuer
...
  solvers:
    - dns01:
        webhook:
          solverName: he
          groupName: acme.xdb.me
          config:
            heUrl: "https://dns.he.net"   # URL for operations. Default (and probably the only valid value): "https://dns.he.net"
            method: "login"               # method to use. "login" is also the default
            # only if you use secrets
            credentialsSecretRef:
              name: "my-secret"           # name of secret. Default: "he-credentials"
              namespace: "myns"           # optional namespace for the secret. If not given, the secret is
                                          # looked for in the issuer namespace.
                                          # For a ClusterIssuer, specify this or the release namespace (eg,
                                          # `cert-manager`) will be used.

dynamic-dns mode

In dynamic-dns mode, the TXT record(s) are never created or deleted, but instead you need to pre-create an aptly named TXT entry (eg, _acme-challenge.mydomain.com) in the domain control panel and the webhook will update/overwrite it with the actual key needed to solve the ACME challenge. To do this, you also need to generate or set an API key (done via the control panel for the record) that will be used for the dynamic update requests.

NOTE: The dynamic-dns mode cannot do concurrent validations (it's always the same TXT record that gets updated), so it should only be used in environments where you expect only a single challenge at a time for each domain, and you know in advance the name of the TXT record to update.

For more information, see the section "Dynamic TXT records" here.

For this mode, the only credential you need is the API key. If you want to use a secret, you store it in the apiKey field. Here's an example:

apiVersion: v1
kind: Secret
metadata:
  name: he-credentials
type: Opaque
stringData:
  apiKey: "skdhjfkdhkjs"

If you pass it via the environment, you must use the auth.heApiKey variable when deploying the Helm chart.

Here's a sample Issuer configuration for the dynamic-dns mode:

apiVersion: cert-manager.io/v1
kind: Issuer
...
  solvers:
    - dns01:
        webhook:
          solverName: he
          groupName: acme.xdb.me
          config:
            heUrl: "https://dyn.dns.he.net" # URL for operations. Default: "https://dyn.dns.he.net"
            method: "dynamic-dns"           # method to use.

            # Only if you use secrets
            apiKeySecretRef:
              name: "my-secret"             # name of secret. Default: "he-credentials"
              namespace: "myns"             # optional namespace for the secret. If not given, the secret is
                                            # looked for in the issuer namespace.
                                            # For a ClusterIssuer, specify this or the release namespace (eg,
                                            # `cert-manager`) will be used.

Access control for secrets

If using secrets, there is the option to limit the namespaces the webhook will be able to access, and also the name of the secrets it will have permission to read. This is done by setting the helm variables rbac.secretNamespaces (a list of namespaces, by default [default]) and rbac.secretNames (a list of names, by default [he-credentials]). If you want to be able to read secrets in any namespace, pass an empty list for rbac.secretNamespaces, and a ClusterRole will be created instead of a Role (use with caution).

Development

IMPORTANT NOTE: only the login mode is conformant with the cert-manager requirements, as it allows for multiple simultaneous DNS01 challenges (and thus TXT records) in a single domain. The dynamic-dns mode cannot do that (it's always the same TXT record that gets updated), so it should only be used in environments where you expect only a single challenge at a time for each domain.

For the same reason, a dynamic-dns mode test is not included in the test suite, as it expects the TXT record to be removed for the test to be declared successful (dynamic-dns mode merely overwrites the key with new values every time; the TXT record is never removed).

Fortunately, in the actual runtime cert-manager doesn't check that a given record is deleted or not after the challenge, so that's why you can use the dynamic-dns method if you want (but not run tests with it), modulo the above notice.

Running the test suite

Conformance testing is achieved through Kubernetes emulation via the kubebuilder-tools suite, in conjunction with real calls to HE on a test domain, using valid credentials or API token stored in secrets.

The test configures a _acme-challenge-test TXT entry, attempts to verify its presence, and removes the entry, thereby verifying the Prepare and CleanUp functions.

To run the test suite, you must create two files under testdata/he. One (let's call it config.json) represents the configuration fragment that will be used by the webhook; the other one (secret.yaml) must contain the HE credentials (username and password). There are examples of both files under testdata/he.

Once the files are in place, run the test suite with:

TEST_ZONE_NAME=yourdomain.com. make test

You can also set VERBOSE=1 (or to any other nonempty value) to see debug messages (note that this increases verbosity for all components):

VERBOSE=1 TEST_ZONE_NAME=yourdomain.com. make test

Have a look at main_test.go in case you want to customize the test suite.