Defining environment variables for components

This article explain why and how to define environment variables for components on examples.

Description

Environment variables are an essential part of the elastic.io platform environment. Apart from the ever growing list of system-wide variables, it is possible and rather encouraged, to define custom environment variables.

Why use environment variables?

Let us consider the following use case when a component is developed for a CRM, Shop or ERP. If this particular component is intended to be used only for one specific client then many values can be hard-coded inside the component.

What if the component is developed to work for different clients? All the hard-coded values for the different clients need to be rewritten again and again. This will introduce an unnecessary amount of overhead work and be the potential source of errors.

Here are a couple of examples where the environment variables are implemented.

Example1: Environment Variables as OAuth Client keys

Connecting an external service with elastic.io platform is usually handled through the security credentials which are used to establish a communication between our platform and the external service.

When a system-wide deployed component is used, the login/password pair or API Keys are usually enough. The establishment of a secure connection with the external servers is handled by our system using our own OAuth Client keys and OAuth Apps created by us. For this case no further configuration is necessary.

When the component is deployed separately into your repository on the main tenant or on your own dedicated tenant, then it is necessary to establish custom OAuth keys and create separate OAuth Apps in addition to supplying the usual login/password information.

For example, here is how you can create your own OAuth App for Salesforce. After the creation of OAuth App elastic.io platform needs to be given the specific details such as the OAuth client key and the OAuth client secret so that it can connect to your Salesforce instance. elastic.io Salesforce component is configured to expect these parameters as environment variables.

Exposing envVars in the Node.js code

When you define any environment variables in your code elastic.io platform has no knowledge of them unless they are properly exposed. To expose all the environment variables they must be declared in the component descriptor file component.json, as shown below. This is required to inform the elastic.io platform that these environmental variables are required for the component to operate properly.

In the most cases envVars are defined in the component.json as it is done in the case of Salesforce:

"envVars": {
  "OAUTH_CLIENT_ID": {
    "required": true,
    "description": "Your Salesforce OAuth client key"
    },
"OAUTH_CLIENT_SECRET": {
    "required": true,
    "description": "Your Salesforce OAuth client secret"
   }
}

Following the above definition it is then used further down like this:

"oauth2": {
  "client_id": "",
  "client_secret": "",
  "auth_uri": "https://.salesforce.com/services/oauth2/authorize",
  "token_uri": "https://.salesforce.com/services/oauth2/token"
}

These are just small parts of the whole component.json which can be viewed fully in Salesforce-component GitHub repository.

Please note: the actual clent_id and clent_secret are defined inside theoauth-utils.js and not only in component.json

Example2: Advance case of HTTP(S) PROXY setup

Here is more advanced use case connected with deployment and connection of component behind an HTTP(S) PROXY. This is particularly useful in case of the on-premise installation of elastic.io platform when all the communication needs to go through the PROXY server.

Depending on the used programming language and libraries some of the implementations can vary significantly. For example in the simple case of defining HTTP PROXY. If we follow the simple case of envVars described above then for Node.js code we would need to define it like in component.json:

"envVars": {
  "HTTP_PROXY": {
    "required": true,
    "description": "Your HTTP PROXY and port http://proxy:1234"
   },
  "HTTPS_PROXY": {
    "required": true,
    "description": "Your HTTPS PROXY and port https://proxy:1234"
  }
}

In some cases, this type of definition might not be sufficient. For example curl and wget respect the HTTP_PROXY and HTTPS_PROXY environment variables. Most Node.js libraries do it the same way while Java rather expects http.proxyHost and http.proxyPort system properties defined in the gradle.properties file as it already explained in the Java components behind a proxy guide.

Some Node.js libraries have their own different way to use proxy setups. For example, when Microsoft SDK is used then the necessary setup needs to follow a specific approach used in superagent to send HTTP requests. The problem is that superagent does not respect HTTP_PROXY and HTTPS_PROXY variables as most Node.js libraries do, but wants these configurations to be passed directly to the constructor.