azure-ts-aks-mean

Azure Kubernetes Service (AKS) App Using CosmosDB

Stands up an Azure Kubernetes Service (AKS) cluster and a MongoDB-flavored instance of CosmosDB. On top of the AKS cluster, we also deploy Helm Chart with a simple Node.js TODO app (bitnami/node), swapping out the usual in-cluster MongoDB instance with our managed CosmosDB instance.

Prerequisites

Ensure you have downloaded and installed the Pulumi CLI.

We will be deploying to Azure, so you will need an Azure account. If you don't have an account, sign up for free here. Follow the instructions here to connect Pulumi to your Azure account.

This example deploys a Helm Chart from Bitnami's Helm chart repository, so you will need to install the Helm CLI and configure it:

For Helm v2 users:

$ helm init --client-only
$ helm repo add bitnami https://charts.bitnami.com/bitnami

For Helm v3 users:

$ helm repo add bitnami https://charts.bitnami.com/bitnami

Now, install dependencies:

npm install

Running the App

1. Create a new stack:

   $ pulumi stack init
   Enter a stack name: azure-mean

2. Set the required configuration variables for this program:

   $ pulumi config set azure:environment public
   $ pulumi config set password --secret [your-cluster-password-here]
   $ ssh-keygen -t rsa -f key.rsa
   $ pulumi config set sshPublicKey < key.rsa.pub
   $ az login

3. Perform the deployment:

   Note: Due to an issue in Azure Terraform Provider, the creation of an Azure Service Principal, which is needed to create the Kubernetes cluster (see cluster.ts), is delayed and may not be available when the cluster is created. If you get a "Service Principal not found" error, as a work around, you should be able to run pulumi up again, at which time the Service Principal replication should have been completed. See this issue and this doc for further details.

   $ pulumi up
   Updating stack 'azure-mean'
   Performing changes:
   
        Type                                         Name                   Status      Info
    +   pulumi:pulumi:Stack                          azure-mean-azure-mean  created     1 warning
    +   ├─ azure:core:ResourceGroup                  aks                    created
    +   ├─ azure:ad:Application                      aks                    created
    +   ├─ azure:ad:ServicePrincipal                 aksSp                  created
    +   ├─ azure:ad:ServicePrincipalPassword         aksSpPassword          created
    +   ├─ azure:cosmosdb:Account                    cosmosDb               created
    +   ├─ azure:containerservice:KubernetesCluster  aksCluster             created
    +   ├─ pulumi:providers:kubernetes               aksK8s                 created
    +   ├─ kubernetes:core:Secret                    mongo-secrets          created
    +   └─ kubernetes:helm.sh:Chart                  node                   created
    +      ├─ kubernetes:core:Service                node-node              created
    +      └─ kubernetes:extensions:Deployment       node-node              created
   
   ---outputs:---
   cluster        : "aksclusterbfb9388b"
   frontendAddress: "40.76.25.71"
   
   info: 12 changes performed:
       + 12 resources created
   Update duration: 14m33.922322098s
   
   Permalink: https://app.pulumi.com/hausdorff/azure-mean/updates/1

   We can see here in the ---outputs:--- section that our Node.js appwas allocated a public IP, in this case 40.76.25.71. It is exported with a stack output variable, frontendAddress. We can use curl and grep to retrieve the <title> of the site the proxy points at.

   $ curl -sL $(pulumi stack output frontendAddress) | grep "<title>"
       <title>Node/Angular Todo App</title>>

Next steps

One of the interesting aspects of this example is the way it demonstrates how easy it is to use Azure resources to configure Kubernetes resources, without the need for intermediate APIs such as OSBA. In particular, this example uses the connection strings exposed by the CosmosDB instance to configure the bitnami/node Helm Chart to connect to CosmosDB, instead of creating and connecting to an in-cluster MongoDB instance.

In index.ts, we see the MongoDB-flavored CosmosDB resource definition:

// Create a MongoDB-flavored instance of CosmosDB.
const cosmosdb = new azure.cosmosdb.Account("cosmosDb", {
    kind: "MongoDB",
    resourceGroupName: config.resourceGroup.name,
    location: config.location,
    consistencyPolicy: { ... },
    offerType: "Standard",
    enableAutomaticFailover: true,
    geoLocations: [ ... ]
});

And then subsequently, in the same file, we see that we use this CosmosDB object to create a Kubernetes Secret containing the connection credentials, which is then consumed by the bitnami/node Helm chart to connect.

// Create secret from MongoDB connection string.
const mongoConnStrings = new k8s.core.v1.Secret(
    "mongo-secrets",
    {
        metadata: { name: "mongo-secrets" },
        data: mongoHelpers.parseConnString(cosmosdb.connectionStrings)
    },
    { provider: k8sProvider }
);

// Boot up nodejs Helm chart example using CosmosDB in place of in-cluster MongoDB.
const node = new k8s.helm.v2.Chart(
    "node",
    {
        repo: "bitnami",
        chart: "node",
        version: "4.0.1",
        values: {
            serviceType: "LoadBalancer",
            mongodb: { install: false },
            externaldb: { ssl: true, secretName: "mongo-secrets" }
        }
    },
    { providers: { kubernetes: k8sProvider }, dependsOn: mongoConnStrings }
);