• This workshop is designed to be run on Google Cloud Shell. It may work in other environments without modification, but the materials are only tested and guaranteed against Cloud Shell.

• You must have a Google Cloud Platform account and be authenticated as a project owner. Again, if you are using Cloud Shell, this happens automatically. If you are running locally, you will need to download and install the Google Cloud SDK, and then authenticate to Google Cloud appropriately.

• There are places where this workshop sacrifices "best practices" for "things feasible to complete in the duration of a workshop". In particular, this workshop generates self-signed SSL certificates and does not encrypt the resulting keys. For more details on a production-hardened setup, please see the Vault production hardening docs.

• You must clone this repo:

  git clone https://github.com/sethvargo/vault-kubernetes-workshop
  cd vault-kubernetes-workshop
  

The first step is to install Vault. To install the Vault binary in the current working directory, run the install script.

./scripts/00-install-vault.sh

This will install Vault using the sethvargo/hashicorp-installer Docker container which verifies GPG signatures and checksums from the download.

This represents a "best practices" installation for installing secure software like Vault. By verifying the signature of the SHASUMs and then verifying the SHASUMs themselves, we guarantee both the integrity of the download and ensure the binary has not been tampered with beyond it's original publishing. For added security, you can download and compile Vault yourself from source, but that is out of scope for this workshop.

Cloud Shell does not persist things in /usr/local/bin or /usr/bin between sessions. As such, Vault will be installed in ~/bin. We recommend installing other binaries or software you need between sessions in this folder.

By default, a new Google Cloud project does not have many services enabled. Enable the required services (this only needs to be done once per project):

./scripts/01-enable-services.sh

This will make the necessary calls to enable the enable the right APIs on your project. This process can take some time, but it is idempotent (you can run it multiple times to achieve the same result).

Vault requires a storage backend to persist its data. This workshop leverages Google Cloud Storage. Vault does not automatically create the storage bucket, so we need to create:

./scripts/02-setup-storage.sh

Cloud Storage bucket names must be globally unique across all of Google Cloud. The bucket will be named "${PROJECT}-vault-storage".

For security purposes, it is not recommended that other applications or services have access to this bucket. Even though the data is encrypted at rest, it is best to limit the scope of access as much as possible.

Vault will leverage Google Cloud KMS to encrypt its unseal keys for auto-unsealing and auto-scaling purposes. We must create the KMS key in advance:

./scripts/03-setup-kms.sh

It is a best practice to create a limited, dedicated service account that has only the required permissions. Create a dedicated service account in the project and grant it the most minimal set of permissions, in particular:

• The ability to read/write to the Cloud Storage bucket created above
• The ability to encrypt/decrypt data with the KMS key created above
• The ability to generate new service accounts (not required to use Vault, but helpful if you plan to use the Vault GCP secrets engine)

./scripts/04-create-iam-service-account.sh

Next we need to create the Google Kubernetes Engine (GKE) cluster which will run Vault. It is recommended that you run Vault in a dedicated namespace or (even better) a dedicated cluster and a dedicated project. Vault will then act as a service with an IP/DNS entry that other projects and services query.

./scripts/05-create-k8s-cluster.sh

This will create the cluster and attach the service account created in the previous step to the cluster. It also ensures the cluster has the correct oauth scopes.

This step creates and reserves a regional public IP address. In a future step, we will attach this reserved IP address to a Kubernetes load balancer. For now, we will just reserve the dedicated IP address.

./scripts/06-create-public-ip.sh

We use a regional IP address instead of a global IP address because global IPs perform load balancing at L7 whereas regional IP addresses perform load balancing at L4. Ideally we do not want the load balancer to perform TLS termination, and let Vault manage TLS, etc. While recent versions of Vault do support advanced routing like X-Forwarded-For headers, it is still a better practice to let Vault fully manage the TLS and thus use an L4 load balancer.

This step is not required if you are comfortable assigning your Vault Kubernetes service an ephemeral IP or will manage it via an external DNS service.

This is arguably the most complex and nuanced piece of the workshop - generating Vault's certificate authority and server certificates for TLS. Vault can run without TLS, but this is highly discouraged. This step could be replaced with a trusted CA like Let's Encrypt, but that is out of scope for this workshop.

./scripts/07-create-certs.sh

This will create the Certificate Authority (ca.key, ca.crt) and Vault certificate (vault.key, vault.crt). In a future step, we will put these values in a Kubernetes secret so our pods can access them.

Next we create the config map and secrets to store our data for our pods. The insecure data such as the storage bucket name and IP address are placed in a configmap. The secure data like the TLS certificates are put in a Kubernetes secret.

./scripts/08-setup-config.sh

The next step is to actually deploy Vault as a StatefulSet on Kubernetes. The reason we use a StatefulSet is two-fold:

1. It guarantees exactly one service starts at a time. This is required by the vault-init sidecar service.

2. It gives us consistent naming for referencing the Vault servers (which is nice for a workshop).

./scripts/09-deploy-vault.sh

Vault will automatically be initialized and unsealed via the vault-init service.

Even though Vault is deployed, it is not publicly accessible. We need to create a Kubernetes Service Load Balancer to forward from the IP address reserved in the previous steps to the pods we just created.

./scripts/10-deploy-lb.sh

The load balancer listens on port 443 and forwards to port 8200 on the containers.

For production-hardened scenarios, you may want to include firewall rules to limit access to Vault at the network layer.

Lastly, we need to configure our local Vault CLI to communicate with these newly-created Vault servers through the Load Balancer. Since we used custom TLS certificates, we'll need to trust the appropriate CA, etc. This will:

• Set VAULT_ADDR to the Load Balancer address

• Set VAULT_CAPATH to the path of the CA cert created in previous steps for properly verifying the TLS connection

• Set VAULT_TOKEN to the decrypted root token by decrypting it from KMS

./scripts/11-setup-comms.sh

At this point, the local Vault CLI is configured to communicate with our Vault cluster. Verify by running vault status:

vault status

Next we will explore techniques for retrieving static (i.e. non-expiring) credentials from Vault. The kv secrets engine is commonly used with legacy applications which cannot handle graceful restarts or when secrets cannot be dynamically generated by Vault.

./scripts/12-setup-static-kv.sh

This will:

1. Enable the KV secrets engine
2. Create a policy to read data from a subpath
3. Store some static username/password data in the secrets engine

Try reading back the secret by running:

vault kv get kv/myapp/config

You can also read the data via a request tool like curl.

curl -k -H "x-vault-token:${VAULT_TOKEN}" "${VAULT_ADDR}/v1/kv/myapp/config"

Next we are going to create another Kubernetes cluster. There is no requirement that our Vault servers run under Kubernetes (they could be running on dedicated VMs or as a managed service). It is a best practice to treat the Vault server cluster as a "service" through which other applications and services request credentials. As such, moving forward, the Vault cluster will be treated simply as an IP address. We will not leverage K8S for "discovering" the Vault cluster, etc.

To put it another way, completely forget that Vault is running in Kubernetes. If it helps, think that Vault is running in a PaaS like Heroku instead.

Next create the Kubernetes cluster where our services will actually run. This is completely separate from the Vault K8S cluster. In fact, on GCP, is it recommended that you run these in completely separate projects. For the purpose of this workshop, we will run them in the same project.

./scripts/13-create-another-cluster.sh

This will provision a new Kubernetes cluster named "my-apps". We will deploy all future apps and services in this cluster.

Unlike the previous cluster, this cluster does not attach a service account.

In our cluster, services will authenticate to Vault using the Kubernetes auth method. In this model, services present their JWT token to Vault as part of an authentication request. Vault takes that signed JWT token and, using the token reviewer API, verifies the token is authenticated. If the authentication is successful, Vault generates a token and maps a series of configured policies onto the token which is returned to the caller.

./scripts/14-create-service-account.sh

This will create a dedicated service account named "vault-auth" and grant that service account the ability to communicate with the token reviewer API.

Next we need to configure the Vault cluster to talk to our new Kubernetes cluster ("my-apps"). We will need to give Vault the IP address of the cluster, the CA information, and the service account to use for accessing the token reviewer API.

./scripts/15-setup-vault-comms-k8s.sh

This process will:

1. Look up the service account JWT token (this is how Vault will talk to the Kubernetes API)

2. Extract the Kubernetes host from the local kube configuration (this where Vault will make API requests)

3. Extract the Kubernetes CA from the local kube configuration (this is how Vault will authenticate requests)

4. Enable the Kubernetes auth method in Vault

5. Give Vault the service account JWT, host, and CA so that Vault can communicate with Kubernetes

There are other techniques for retrieving some of these values, but leveraging kubectl makes it easy to script.

Typically this process is done by a security team or operations team. We need to configure Vault to map an application or service in Kubernetes to a series of policies in Vault. That way, when an application successfully authenticates to Vault via its JWT token, Vault knows which policies to assign to the response.

./scripts/16-create-kv-role.sh

This will create a role named "myapp-role" that permits pods in the "default" namespace with the "default" service account to receive a Vault token that has the "myapp-kv-rw" policy attached.

This is one of the most common techniques for injecting Vault secrets into an application.

1. An init container pulls the service account JWT token and performs the auth mechanism for that service account. If successful, it stores the resulting Vault token in somewhere on a shared volume mount.

2. A tool like Consul Template runs as the first container. This tool uses the Vault token acquired by the init container and makes the appropriate API calls to Vault based off of a template file. The template file can reference one or more Vault credentials. Consul Template writes the rendered file with the secrets from Vault to a shared volume mount which the app reads.

3. The app reads credentials. In this example, our application is a dummy application that just reads the contents of /etc/secrets/config repeatedly.

./scripts/17-run-kv-sidecar.sh

Verify the app is authenticating and retrieving secrets from Vault:

./scripts/kubectl-logs.sh kv-sidecar

Next we configure Vault to generate dynamic credentials. Vault can generate many types of dynamic credentials like database credentials, certificates, etc. For this example, we will leverage the GCP secrets engine to dynamically generate Google Cloud Platform CloudSQL MySQL users.

./scripts/18-setup-dynamic-creds.sh

This will:

1. Create a CloudSQL database

2. Enable the database secrets engine

3. Configure the database secrets engine

4. Create a "role" which configures the permissions the SQL user has

5. Create a new policy which allows generating these dynamic credentials

6. Update the Vault Kubernetes auth mapping to include this new policy when authenticating

This process can take up to 10 minutes to complete.

In this example, we follow the same pattern as the static KV secrets, but our sidecar application will pull dynamic credentials from Vault. In this case, we will be creating a database password.

./scripts/19-run-db-sidecar.sh

This also configures a command to run which will signal the application when the underlying service account changes. This is important as we need to notify the application (which is not aware of Vault's existence) that it should reload its configuration.

Verify the app is authenticating and retrieving secrets from Vault:

./scripts/kubectl-logs.sh db-sidecar