Automating Certificate Management in a Kubernetes Environment

KENNETH

2년 ago

Automating Certificate Management in a Kubernetes Environment

Valid SSL/TLS certificates are a core requirement of the modern application landscape. Unfortunately, managing certificate (or cert) renewals is often an afterthought when deploying an application. Certificates have a limited lifetime, ranging from roughly 13 months for certificates from DigiCert to 90 days for Let’s Encrypt certificates. To maintain secure access, these certificates need to be renewed/reissued prior to their expiration. Given the substantial workload of most Ops teams, cert renewal sometimes falls through the cracks, resulting in a scramble as certificates near – or worse, pass – their expiration date.

It doesn’t need to be like this. With some planning and preparation, cert management can be automated and streamlined. Here, we will look at a solution for Kubernetes using three technologies:

In this blog, you’ll learn to simplify cert management by providing unique, automatically renewed and updated certificates to your endpoints.

Certificates in a Kubernetes Environment

Before we get into technical details, we need to define some terminology. The term “TLS certificate” refers to two components required to enable HTTPS connections on our Ingress controller:

The certificate
The private key

Both the certificate and private key are issued by Let’s Encrypt. For a full explanation of how TLS certificates work, please see DigiCert’s post How TLS/SSL Certificates Work.

In Kubernetes, these two components are stored as Secrets. Kubernetes workloads – such as the NGINX Ingress Controller and cert-manager – can write and read these Secrets, which can also be managed by users who have access to the Kubernetes installation.

Introducing cert-manager

The cert-manager project is a certificate controller that works with Kubernetes and OpenShift. When deployed in Kubernetes, cert-manager will automatically issue certificates required by Ingress controllers and will ensure they are valid and up-to-date. Additionally, it will track expiration dates for certificates and attempt renewal at a configured time interval. Although it works with numerous public and private issuers, we will be showing its integration with Let’s Encrypt.

Two Challenge Types

When using Let’s Encrypt, all cert management is handled automatically. While this provides a great deal of convenience, it also presents a problem: How does the service ensure that you own the fully-qualified domain name (FQDN) in question?

This problem is solved using a challenge, which requires you to answer a verification request that only someone with access to the specific domain’s DNS records can provide. Challenges take one of two forms:

HTTP-01: This challenge can be answered by having a DNS record for the FQDN that you are issuing a certificate. For example, if your server is at IP www.xxx.yyy.zzz and your FQDN is cert.example.com, the challenge mechanism will expose a token on the server at www.xxx.yyy.zzz and the Let’s Encrypt servers will attempt to reach it via cert.example.com. If successful, the challenge is passed and the certificate is issued.

HTTP-01 is the simplest way to generate a certificate, as it does not require direct access to the DNS provider. This type of challenge is always conducted over Port 80 (HTTP). Note that when using HTTP-01 challenges, cert-manager will utilize the Ingress controller to serve the challenge token.

DNS-01: This challenge creates a DNS TXT record with a token, which is then verified by the issuer. If the token is recognized, you have proved ownership of that domain and can now issue certificates for its records. Unlike the HTTP-01 challenge, when using the DNS-01 challenge, the FQDN does not need to resolve to your server’s IP address (nor even exist). Additionally, DNS-01 can be used when Port 80 is blocked. The offset to this ease of use is the necessity of providing access to your DNS infrastructure via API token to the cert-manager installation.

Ingress Controllers

An Ingress controller is a specialized service for Kubernetes that brings traffic from outside the cluster, load balances it to internal Pods (a group of one or more containers), and manages egress traffic. Additionally, the Ingress controller is controlled through the Kubernetes API and will monitor and update the load balancing configuration as Pods are added, removed, or fail.

To learn more about Ingress controllers, read the following blogs:

In the examples below, we will use NGINX Ingress Controller that is developed and maintained by F5 NGINX.

Certificate Management Examples

These examples assume that you have a working Kubernetes installation that you can test with, and that the installation can assign an external IP address (Kubernetes LoadBalancer object). Additionally, it assumes that you can receive traffic on both Port 80 and Port 443 (if using the HTTP-01 challenge) or solely Port 443 (if using the DNS-01 challenge). These examples are illustrated using Mac OS X, but can be used on Linux or WSL as well.

You will also need a DNS provider and FQDN that you can adjust the A record for. If you are using the HTTP-01 challenge, you only need the ability to add an A record (or have one added for you). If you are using the DNS-01 challenge, you will need API access to a supported DNS provider or a supported webhook provider.

Deploy NGINX Ingress Controller

The easiest way is to deploy via Helm. This deployment allows you to use both the Kubernetes Ingress and the NGINX Virtual Server CRD.

Add the NGINX repo.

$ helm repo add nginx-stable https://helm.nginx.com/stable
  "nginx-stable" has been added to your repositories

Update the repository.

$ helm repo update
  Hang tight while we grab the latest from your chart repositories...
  ...Successfully got an update from the "nginx-stable" chart repository
  Update Complete. ⎈Happy Helming!⎈

Deploy the Ingress controller .

$ helm install nginx-kic nginx-stable/nginx-ingress 
  --namespace nginx-ingress  --set controller.enableCustomResources=true  
  --create-namespace  --set controller.enableCertManager=true 
  NAME: nginx-kic
  LAST DEPLOYED: Thu Sep  1 15:58:15 2022
  NAMESPACE: nginx-ingress
  STATUS: deployed
  REVISION: 1
  TEST SUITE: None
  NOTES:
  The NGINX Ingress Controller has been installed.

Check the deployment and retrieve the IP address of the egress for the Ingress controller . Note that you cannot continue without a valid IP address.

$ kubectl get deployments --namespace nginx-ingress
  NAME                      READY   UP-TO-DATE   AVAILABLE   AGE
  nginx-kic-nginx-ingress   1/1     1            1           23s
  $ kubectl get services --namespace nginx-ingress
  NAME                      TYPE           CLUSTER-IP      EXTERNAL-IP    PORT(S)                      AGE
  nginx-kic-nginx-ingress   LoadBalancer   10.128.60.190   www.xxx.yyy.zzz   80:31526/TCP,443:32058/TCP   30s

Add Your DNS A Record

The process here will depend on your DNS provider. This DNS name will need to be resolvable from the Let’s Encrypt servers, which may require that you wait for the record to propagate before it will work. For more information on this, please see the SiteGround article What Is DNS Propagation and Why Does It Take So Long?

Once you can resolve your chosen FQDN you are ready to move on to the next step.

$ host cert.example.com
  cert.example.com has address www.xxx.yyy.zzz

Deploy cert-manager

The next step is to deploy the most recent version of cert-manager. Again, we will be using Helm for our deployment.

Add the Helm repository.

$ helm repo add jetstack https://charts.jetstack.io
  "jetstack" has been added to your repositories

Update the repository.

$ helm repo update
  Hang tight while we grab the latest from your chart repositories...
  ...Successfully got an update from the "nginx-stable" chart repository
  ...Successfully got an update from the "jetstack" chart repository
  Update Complete. ⎈Happy Helming!⎈

Deploy cert-manager.

$ helm install cert-manager jetstack/cert-manager 
  --namespace cert-manager --create-namespace 
  --version v1.9.1  --set installCRDs=true 
  NAME: cert-manager
  LAST DEPLOYED: Thu Sep  1 16:01:52 2022 
  NAMESPACE: cert-manager
  STATUS: deployed
  REVISION: 1 
  TEST SUITE: None
  NOTES:
  cert-manager v1.9.1 has been deployed successfully!

  In order to begin issuing certificates, you will need to set up a ClusterIssuer
  or Issuer resource (for example, by creating a 'letsencrypt-staging' issuer).

  More information on the different types of issuers and how to configure them
  can be found in our documentation:

  https://cert-manager.io/docs/configuration/

  For information on how to configure cert-manager to automatically provision
  Certificates for Ingress resources, take a look at the `ingress-shim`
  documentation:

  https://cert-manager.io/docs/usage/ingress/

Validate the deployment.

$ kubectl get deployments --namespace cert-manager
  NAME                      READY   UP-TO-DATE   AVAILABLE   AGE
  cert-manager              1/1     1            1           4m30s
  cert-manager-cainjector   1/1     1            1           4m30s
  cert-manager-webhook      1/1     1            1           4m30s

Deploy the NGINX Cafe Example

We are going to be using the NGINX Cafe example to provide our backend deployment and Services. This is a common example used within the documentation provided by NGINX. We will not be deploying Ingress as part of this.

Clone the NGINX Ingress GitHub project.

$ git clone https://github.com/nginxinc/kubernetes-ingress.git
  Cloning into 'kubernetes-ingress'...
  remote: Enumerating objects: 44979, done.
  remote: Counting objects: 100% (172/172), done.
  remote: Compressing objects: 100% (108/108), done.
  remote: Total 44979 (delta 87), reused 120 (delta 63), pack-reused 44807
  Receiving objects: 100% (44979/44979), 60.27 MiB | 27.33 MiB/s, done.
  Resolving deltas: 100% (26508/26508), done.

Change to the examples directory. This directory contains several examples that demonstrate various configurations of the Ingress controller. We are using the example provided under the complete-example directory.

$ cd ./kubernetes-ingress/examples/ingress-resources/complete-example

Deploy the NGINX Cafe example.

$ kubectl apply -f ./cafe.yaml
  deployment.apps/coffee created
  service/coffee-svc created
  deployment.apps/tea created
  service/tea-svc created

Validate the deployment and Services using the kubectl get command. You are looking to ensure that the Pods are showing as READY, and the Services are showing as running. The example below shows a representative sample of what you are looking for . Note that the kubernetes service is a system service running in the same namespace (default) as the NGINX Cafe example.

$ kubectl get deployments,services  --namespace default
  NAME                     READY   UP-TO-DATE   AVAILABLE   AGE
  deployment.apps/coffee   2/2     2            2           69s
  deployment.apps/tea      3/3     3            3           68s

NAME                 TYPE        CLUSTER-IP       EXTERNAL-IP   PORT(S)   AGE
service/coffee-svc   ClusterIP   10.128.154.225   <none>        80/TCP    68s
service/kubernetes   ClusterIP   10.128.0.1       <none>        443/TCP   29m
service/tea-svc      ClusterIP   10.128.96.145    <none>        80/TCP    68s

Deploy the ClusterIssuer

Within cert-manager, the ClusterIssuer can be used to issue certificates. This is a cluster-scoped object that can be referenced by any namespace and used by any certificate requests with the defined certificate-issuing authority. In this example, any certificate requests for Let’s Encrypt certificates can be handled by this ClusterIssuer.

Deploy the ClusterIssuer for the challenge type you have selected. Although it is out of scope for this post, there are advanced configuration options that allow you to specify multiple resolvers (chosen based on selector fields) in your ClusterIssuer.

ACME Challenge Basics

The Automated Certificate Management Environment (ACME) protocol is used to determine if you own a domain name and can therefore be issued a Let’s Encrypt certificate. For this challenge, these are the parameters that need to be passed:

metadata.name: The ClusterIssuer name, which needs to be unique within the Kubernetes installation. This name will be used later in the example when we are issuing a certificate.
spec.acme.email: This is the email address you are registering with Let’s Encrypt for the purpose of generating certificates. This should be your email.
spec.acme.privateKeySecretRef: This is the name of the Kubernetes secret you will use to store your private key.
spec.acme.solvers: This should be left as-is – it notes the type of challenge (or, as ACME refers to it, solver) you are using (HTTP-01 or DNS-01) as well as what Ingress class it should be applied to, which in this case will be nginx.

Using HTTP-01

This example shows how to set up a ClusterIssuer to use the HTTP-01 challenge to prove domain ownership and receive a certificate.

Create the ClusterIssuer using HTTP-01 for challenges.

$ cat << EOF | kubectl apply -f 
  apiVersion: cert-manager.io/v1
  kind: ClusterIssuer
  metadata:
    name: prod-issuer
  spec:
    acme:
      email: example@example.com
      server: https://acme-v02.api.letsencrypt.org/directory
      privateKeySecretRef:
        name: prod-issuer-account-key
      solvers:
      - http01:
         ingress:
           class: nginx
  EOF
  clusterissuer.cert-manager.io/prod-issuer created

Validate the ClusterIssuer (it should show as ready).

$ kubectl get cluster-issuer
  NAME          READY   AGE
	prod-issuer   True    34s

Using DNS-01

This example shows how to set up a ClusterIssuer to use the DNS-01 challenge to authenticate your domain ownership. Depending on your DNS provider you will likely need to use a Kubernetes Secret to store your token. This example is using Cloudflare. Note the use of namespace. The cert-manager application, which is deployed into the cert-manager namespace, needs to have access to the Secret .

For this example, you will need a Cloudflare API token, which you can create from your account. This will need to be put in the <API Token> line below. If you are not using Cloudflare you will need to follow the documentation for your provider.

Create a Secret for the API token.

$ cat << EOF | kubectl apply -f 
  apiVersion: v1
  kind: Secret
  metadata:
    name: cloudflare-api-token-secret
    namespace: cert-manager
  type: Opaque
  stringData:
    api-token: <API Token> 
  EOF

Create the issuer using DNS-01 for challenges.

$ cat << EOF | kubectl apply -f 
  apiVersion: cert-manager.io/v1
  kind: ClusterIssuer
  metadata:
    name: prod-issuer
  spec:
    acme:
      email: example@example.com
      server: https://acme-v02.api.letsencrypt.org/directory
      privateKeySecretRef:
        name: prod-issuer-account-key
      solvers:
        - dns01:
            cloudflare:
              apiTokenSecretRef:
                name: cloudflare-api-token-secret
                key: api-token
  EOF

Validate the issuer (it should show as ready).

$ kubectl get clusterissuer
  NAME          READY   AGE
	prod-issuer   True    31m

Deploy the Ingress

This is the point we’ve been building towards – the deployment of the Ingress resource for our application. This will route traffic into the NGINX Cafe application we deployed earlier.

Using the Kubernetes Ingress

If you are using the standard Kubernetes Ingress resource, you will use the following deployment YAML to configure the Ingress and request a certificate.

apiVersion: networking.k8s.io/v1 
  kind: Ingress 
  metadata: 
    name: cafe-ingress 
    annotations: 
      cert-manager.io/cluster-issuer: prod-issuer 
      acme.cert-manager.io/http01-edit-in-place: "true" 
  spec: 
    ingressClassName: nginx 
    tls: 
    - hosts: 
      - cert.example.com 
      secretName: cafe-secret 
    rules: 
    - host: cert.example.com 
      http: 
        paths: 
        - path: /tea 
          pathType: Prefix 
          backend: 
            service: 
              name: tea-svc 
              port: 
                number: 80 
        - path: /coffee 
          pathType: Prefix 
          backend: 
            service: 
              name: coffee-svc 
              port: 
number: 80

It’s worth reviewing some key parts of the manifest:

The API being called is the standard Kubernetes Ingress.
A key part of this configuration is under metadata.annotations where we set acme.cert-manager.io/http01-edit-in-place to “true”. This value is required and adjusts the way that the challenge is served. For more information see the Supported Annotations document. This can also be handled by using a master/minion setup.
The spec.ingressClassName refers to the NGINX Ingress controller that we installed and will be using.
The spec.tls.secret Kubernetes Secret resource stores the certificate key that is returned when the certificate is issued by Let’s Encrypt.
Our hostname of cert.example.com is specified for spec.tls.hosts and spec.rules.host. This is the hostname for which our ClusterIssuer issued the certificate.
The spec.rules.http section defines the paths and the backend Services that will service requests on those paths. For example, traffic to /tea will be directed to Port 80 on the tea-svc.

Modify the above manifest for your installation. At minimum, this will involve changing the spec.rules.host and spec.tls.hosts values, but you should review all parameters in the configuration.
Apply the manifest.

$  kubectl apply -f ./cafe-virtual-server.yaml
  virtualserver.k8s.nginx.org/cafe created

Wait for the certificate to be issued. You are looking for a value of “True” for the READY field.

$ kubectl get certificates
  NAME                                      READY   SECRET        AGE
  certificate.cert-manager.io/cafe-secret   True    cafe-secret   37m

Using the NGINX Virtual Server / Virtual Routes

If you are using the NGINX CRDs, you will need to use the following deployment YAML to configure your Ingress.

apiVersion: k8s.nginx.org/v1 
  kind: VirtualServer 
  metadata: 
    name: cafe 
  spec: 
    host: cert.example.com 
    tls: 
      secret: cafe-secret 
      cert-manager: 
        cluster-issuer: prod-issuer 
    upstreams: 
    - name: tea 
      service: tea-svc 
      port: 80 
    - name: coffee 
      service: coffee-svc 
      port: 80 
    routes: 
    - path: /tea 
      action: 
        pass: tea 
    - path: /coffee 
      action: 
pass: coffee

Once again, it’s worth reviewing some key parts of the manifest:

The API being called is the NGINX-specific k8s.nginx.org/v1 for the VirtualServer resource.
The spec.tls.secret Kubernetes Secret resource stores the certificate key that is returned when the certificate is issued by Let’s Encrypt.
Our hostname of cert.example.com is specified for spec.host. This is the hostname for which our ClusterIssuer issued the certificate .
The spec.upstreams values point to our backend Services, including the ports.
The spec.routes defines both the route and the action to be taken when those routes are hit.

Modify the above manifest for your installation. At minimum, this will involve changing the spec.host value, but you should review all parameters in the configuration.
Apply the manifest.

$  kubectl apply -f ./cafe-virtual-server.yaml
  virtualserver.k8s.nginx.org/cafe created

Wait for the certificate to be issued. You should see a status of Valid.

$ kubectl get VirtualServers
  NAME   STATE   HOST                    IP             PORTS      AGE
  cafe   Valid   cert.example.com   www.xxx.yyy.zzz   [80,443]   51m

View the Certificate

You can view the certificate via the Kubernetes API. This will show you details about the certificate, including its size and associated private key.

$ kubectl describe secret cafe-secret
  Name:         cafe-secret
  Namespace:    default
  Labels:       <none>
  Annotations:  cert-manager.io/alt-names: cert.example.com
                cert-manager.io/certificate-name: cafe-secret
                cert-manager.io/common-name: cert.example.com
                cert-manager.io/ip-sans:
                cert-manager.io/issuer-group:
                cert-manager.io/issuer-kind: ClusterIssuer
                cert-manager.io/issuer-name: prod-issuer
                cert-manager.io/uri-sans:Type:  kubernetes.io/tlsData
  ====
  tls.crt:  5607 bytes
  tls.key:  1675 bytes

If you’d like to see the actual certificate and key, you can do so by running the following command. (Note: This does illustrate a weakness of the Kubernetes Secrets. Namely, they can be read by anyone with the necessary access permissions.)

$ kubectl get secret cafe-secret -o yaml

Test the Ingress

Test the certificates . You can use any method that you wish here. The example below uses cURL. Success is indicated by a block similar to what is shown before , which includes the server name, internal address of the server, date, the URI (route) chosen (coffee or tea), and the request ID. Failures will take the form of HTTP error codes, most likely 400 or 301.

$ curl https://cert.example.com/tea
  Server address: 10.2.0.6:8080
  Server name: tea-5c457db9-l4pvq
  Date: 02/Sep/2022:15:21:06 +0000
  URI: /tea
  Request ID: d736db9f696423c6212ffc70cd7ebecf
  $ curl https://cert.example.com/coffee
  Server address: 10.2.2.6:8080
  Server name: coffee-7c86d7d67c-kjddk
  Date: 02/Sep/2022:15:21:10 +0000
  URI: /coffee
Request ID: 4ea3aa1c87d2f1d80a706dde91f31d54

Certificate Renewals

At the start, we promised that this approach would eliminate the need to manage certificate renewals. However, we have yet to explain how to do that. Why? Because this is a core, built-in part of cert-manager. In this automatic process, when cert-manager realizes that a certificate is not present, is expired, is within 15 days of expiry, or if the user requests a new cert via the CLI, then a new certificate is automatically requested. It doesn’t get much easier than that.

Common Questions

What About NGINX Plus?

If you are an NGINX Plus subscriber, the only difference for you will involve installing the NGINX Ingress Controller. Please see the Installation Helm section of the NGINX Docs for instructions on how to modify the Helm command given above to accomplish this.

What Challenge Type Should I Use?

This largely depends on your use case.

The HTTP-01 challenge method requires that Port 80 is open to the Internet and that the DNS A record has been properly configured for the IP address of the Ingress controller. This approach does not require access to the DNS provider other than to create the A record.

The DNS-01 challenge method can be used when you cannot expose Port 80 to the Internet, and only requires that the cert-manager have egress access to the DNS provider. However, this method does require that you have access to your DNS provider’s API, although the level of access required varies by specific provider.

How Do I Troubleshoot Problems?

Since Kubernetes is so complex, it’s difficult to provide targeted troubleshooting information. If you do run into issues, we’d like to invite you to ask us in the NGINX Community Slack (NGINX Plus subscribers can use their normal support options).

Get Started Today

Get started by requesting your free 30-day trial of NGINX Ingress Controller with NGINX App Protect WAF and DoS, and download the always‑free NGINX Service Mesh.

The post Automating Certificate Management in a Kubernetes Environment appeared first on NGINX.

Source: Automating Certificate Management in a Kubernetes Environment