HashiCorp Vault: Loading Environment Variables using VaultSidecar for Node.js API Servers

DevOps May 28, 2021

Hey Everyone! Today’s post is a continuation to last week’s post on installing Vault, configuring Kubernetes Auth Method, and the KV secrets engine. If you haven’t read it yet, check it out here. This guide is going to use my Node.js API Server template express-ts-boilerplate available here. It is a work in progress template written in TypeScript with more features and documentation coming soon, feel free to use it as a base and build upon it, Pull Requests are welcome!

What do we currently have?

  • Single-Node MicroK8s Cluster — Refer to my Home Lab Infrastructure post here for complete details.
  • Vault running in standalone mode on Kubernetes
  • KV secrets engine enabled
  • Kubernetes Auth Method configured

What do we need?

  • Helm Chart annotations for Vault Sidecar
  • Service Account for the Node Server
  • Loading the env vars into the application

Creating a Namespace

Let’s start by creating a namespace for this demo.

kubectl create namespace vault-demo

Creating a Service Account

Next, let’s create a service account for our application to use.

kubectl create sa vault-demo-sa -n vault-demo

Adding Environment Variables to Vault

Log into Vault if you haven’t done so already. If you are following along from the previous post, use your initial root token to login.

export VAULT_ADDR=https://vault.example.com

vault login

Now add your environment variables to vault on a path used by the KV store, in my previous post, we configured it on the path kv/ so I will be using the path kv/env/boilerplate here. If you are using my boilerplate code, the required variables are listed in the example.env file.

vault write kv/env/boilerplate SQL_USER=devuser SQL_PASS=devpassword SQL_NAME=demo_db SQL_HOST=mysql SQL_PORT=3306 NOSQL_USER=devuser NOSQL_PASS=devpassword NOSQL_NAME=demo_db NOSQL_HOST=mongodb NOSQL_PORT=27017

Creating a Policy and a Demo Role

Now we need to create the policy and role for our application to use. I will be calling them demo-policy and demo-role respectively. The policy grants full read permissions to all secrets stored in the KV secrets engine. Change the path for tighter access control as needed.

vault policy write demo-policy -<<EOF 
path "kv/*" 
{  capabilities = ["read"]

vault write auth/kubernetes/role/demo-role \
    bound_service_account_names=vault-demo-sa \
    bound_service_account_namespaces=vault-demo \
    policies=demo-policy \

Here, we are permitting the vault-demo-sa ServiceAccount in the vault-demo namespace to authenticate with this role.

Testing Authentication

You can test if the role is working as expected by trying to log in as the service account as follows.

demo_secret_name="$(kubectl get serviceaccount vault-demo-sa -n vault-demo -o go-template='{{ (index .secrets 0).name }}')"

demo_account_token="$(kubectl get secret ${demo_secret_name} -n vault-demo -o go-template='{{ .data.token }}' | base64 --decode)"

vault write auth/kubernetes/login role=demo-role jwt=$demo_account_token

If it returns a token, everything is fine.

Adding Annotations and ServiceAccount to the helm chart

Now that the Vault preparation is done, we need to add our annotations and specify our service account in the helm chart. Edit your values.yaml file and set the following.

  # Specifies whether a service account should be created
  create: false
  # Annotations to add to the service account
  annotations: {}
  # The name of the service account to use.
  # If not set and create is true, a name is generated using the fullname template
  name: "vault-demo-sa"

  vault.hashicorp.com/agent-inject: "true"
  vault.hashicorp.com/role: "demo-role"
  vault.hashicorp.com/agent-inject-secret-envvars: "kv/env/boilerplate"
  vault.hashicorp.com/agent-inject-template-envvars: |
    {{ with secret "kv/env/boilerplate" -}}
    {{ range $k, $v := .Data }}
    {{ $k }}={{ $v }}
    {{ end }}
    {{- end }}

In the ServiceAccount section, we are instructing the chart to not create an account, and to instead, use the service account name provided.

In the annotations, the first line instructs the Vault Agent Injector to inject the vault sidecar into this pod. The second line defines the role to authenticate with, and the third and fourth lines indicate the path of the secret, and the template with which to expose it in the mounted file.

Here, the part after agent-inject-secret- is taken as the name for this secret. It is mounted as a file at the path /vault/secrets/<name> inside the containers. The template allows us to transform the secret before writing it into the file. The template above takes the key and value from the secret and writes it in the form key=value which is the standard .env format.

Setting up the Node Application to use the mounted file

I use dotenv to load .env files into the environment, so I will be checking if the vault mounted secret exists at /vault/secrets/envvars. If the file exists, then it is used by dotenv, If it doesn’t exist, the default location is used.

// use different env file when running with vault sidecar
if (fs.existsSync('/vault/secrets/envvars')) {
  const envLoad = dotenv.config({path: '/vault/secrets/envvars'})
  if (envLoad.error) {
    throw envLoad.error
} else {
  const envLoad = dotenv.config()
  if (envLoad.error) {
    throw envLoad.error

This code segment is part of my config file config/config.ts in the boilerplate template source code.

Here is how the code looks as part of the entire config file.

import * as dotenv from "dotenv";
import * as fs from "fs";

interface ServerConfig {
  port: number;
  sql_name: string;
  sql_user: string;
  sql_pass: string;
  sql_host: string;
  sql_port: number;
  nosql_name: string;
  nosql_user: string;
  nosql_pass: string;
  nosql_host: string;
  nosql_port: number;

// use different env file when running with vault sidecar
if (fs.existsSync('/vault/secrets/envvars')) {
  const envLoad = dotenv.config({path: '/vault/secrets/envvars'})
  if (envLoad.error) {
    throw envLoad.error
} else {
  const envLoad = dotenv.config()
  if (envLoad.error) {
    throw envLoad.error

const config: ServerConfig = {
  port: normalizePort(process.env.port || 3000),
  sql_user: process.env.SQL_USER || "devuser",
  sql_pass: process.env.SQL_PASS || "devpassword",
  sql_name: process.env.SQL_NAME || "pmplerp_items",
  sql_host: process.env.SQL_HOST || "mysql",
  sql_port: normalizePort(process.env.SQL_PORT || 3306),
  nosql_user: process.env.NOSQL_USER || "devuser",
  nosql_pass: process.env.NOSQL_PASS || "devpassword",
  nosql_name: process.env.NOSQL_NAME || "pmplerp_items",
  nosql_host: process.env.NOSQL_HOST || "mysql",
  nosql_port: normalizePort(process.env.NOSQL_PORT || 27017)

function normalizePort(val: string | number) {
  const port = parseInt(<string>val, 10);

  if (port >= 0) {
    // port number
    return port;

  throw new Error("Port number Invalid")
export default config;

Deploy Helm Chart to Cluster

Once all this is in place, you can go ahead and deploy the chart to your cluster. If you are following along with my template repo, you can use the following command.

helm upgrade --install --wait boilerplate ./chart -n vault-demo

If you want to know more about the vault agent injector, refer to the injector docs here:

Agent Sidecar Injector Overview | Vault by HashiCorp
The Vault Agent Sidecar Injector is a Kubernetes admission webhook that adds Vault Agent containers to pods for consuming Vault secrets.


Tanuj Ravi Rao

99% of the time my brain is thinking blah, meh, why, huh, WTF, food, and computers. The other 1% I’m usually asleep.

Great! You've successfully subscribed.
Great! Next, complete checkout for full access.
Welcome back! You've successfully signed in.
Success! Your account is fully activated, you now have access to all content.