I develop an application with nodejs and react. I use dotenv for configuration in my different environment.
I use TFS 2017 for build and release my application.
What is the best practise for add my .env file of production environment?
Production configs can be difficult to maintain. You can use any module or even write your own custom module to load environment variable in to your application.
However, maintaining production .env files locally for each product (i.e. committing them or putting them inside your docker image is a bad idea. If you ever happen to change some of those configs, you will have to manually change in the code for each application, build docker image (if required) and redeploy. If there are just a couple of applications, it might be easy. But if the number of related applications grow in size (as usually happens in case of microservice based architecture) - all of them sharing credentials, ips for database connections etc. it becomes a hectic task to manually change code and redeploy all the applications.
Usually, developers tend to keep a central repository of all credentials and environment variables. For examples, AWS offers parameter store and MS Azure offers Azure Key vault. In such a case, all the parameters are fetched during start time, and all we have to do is restart the application.
So usually people set only one global varibale NODE_ENV (as prod or dev), dynamically fetch all environment variables based on NODE_ENV running something like env $(node read-env-variables.js) node app.js.
Related
I've been using K8S for a year or so and continue to revisit a problem.
My app is running in K8S and I now need to debug it. I have a NodeJS App that I'm asking about. But similar questions could be asked about Java SpringBoot apps (but this question is just for NodeJS).
I want to use my favorite IDE (IntelliJ or VSCode) to run the app but the app is currently getting it's configuration (inside K8S) using ConfigMaps and Secrets.
(Q) Is there a "best practice" or "pattern" that can be followed that supports the DRY principle and has configuration in one place that can be used for both K8S and when running locally.
Background
I have a NodeJS app that I decided to use ENVIRONMENT variables to hold configuration information because that worked well in IntelliJ IDE, in Docker and in K8S.
I used npm dotenv and created .env.local, .env.stage, .env.prod files to support running in different environments. This worked well enough until it was running in K8S and someone wanted to tweak the configuration and didn't believe that rebuilding the image was the best way to support this. Instead the K8S experts told me I should use ConfigMaps and Secrets, so I converted from the dotenv approach to use the K8S ConfigMaps and Secrets.
I kept the old .env files around just in case and I can use them but the source code no longer call uses dotenv package.
require('dotenv').config()
process.env.myConfigVariable
So I need to either add that code back to support debugging, or manually set the environment variables. I'm wondering if there is a better approach.
I have yaml files templates to make it easy to recreate the deployment from scratch if/when needed.
.env.local
deploy/
helm/
create-configmap.yaml
create-secret.yaml
src/
common/*
appMain.js
Some of the approaches I've considered:
(a) Accept it and have two configs (one for local and one for K8S). Leave the code for dotenv but don't deploy a .env file when deploying to K8S.
(b) Run local k8s (like minikube or k3s) and use my ConfigMap and Secrets as I would with K8S. I then need to figure out how to connect from my IDE to the local K3S environment and open ports in the k3s environment to support this. Some solutions include: Bridge to Kubernetes, YouTube Video Remote Debugging in Kubernetes with Cloud Code,
Debug Java Microservices in Kubernetes with IntelliJ, and I'm sure several others.
(c) Use a JSON config file instead of dotenv. For example, use a JSON config file for everything and map that to /app/config.json and that same config file can be used in both environments. I could have config-local.json, config-stage.json, and config-prod.json to support the different environments.
(d) You tell me. What's another way?
Thanks!
I have an Angular4 web app, deployed on Azure. Now I want to deploy this app to other environments on Azure: one for testing, one for acceptance and one for production. Every environment has different API endpoints and may have other variables, like Application Insights. All those environments run Angular in production mode.
The way Angular advises you to do this, is by the Enviroment files (environment.test.ts, enviroment.acc.ts, environment.prod.ts). I could configure all the different API endpoints in those files, and run my build with --prod for production for example.
But that is not the way I want to do this. I want to use the exact same application package deployed to test for my acceptance environment, without rebuilding the project. In Visual Studio Online, this is also really simple to configure.
The point is: how can I make my API endpoints differ per environment in that way?
The way I want to do this, is by the App Settings in Azure. But Angular can't get to those environment variables because it's running on the client side. Node.js is running on serverside and could get those App Settings - but if that's the way I need to do it, how do I make Node.js (used in Angular4 CLI) to send those server variables to the client side? And what about performance impact for this solution?
How did you fix this problem for your Angular4 apps on Azure? Is it just impossible to fix this problem with the Azure App Settings?
For everyone with the same question: I didn't fix this problem the way I described above.
At the end, I did it the way Angular wants you to do it: so rebuild for dev, rebuild for acc and rebuild for prod.
In Visual Studio Online, at build time, it builds and tests our code and it saves the uncompiled/unminified code. At release time, it builds en tests it again and releases it to the right environment with the right environment variables (--prod for example).
I don't think there is another way to fix this.
The solution is pretty old school but it works! Although you can use branching or tag for this purpose instead of cloning the code to the package.
The best solution as you said is Azure app settings will be saved as environment variable so you should implement an API with node.js and share the variables you want.
Of course there is an impact because of additional http call, but it's just one time at application start which is about max 5ms and depends on each program policy whether is impact or not.
Another option could be move the variables to the JSON file in the asset folder, and change it at deploy runtime with release pipeline. that's easier implementation but the disadvantage is you will have to use release variables instead of app settings and if you have config changes you will have to update the variable value first and redeploy it, although that works most of the times but sometimes you want to change just like a connection string and you will have to redeploy.
I've come across people using both methods to do config management.
What are the pros and cons of each approach?
If I have a lot of variables that I store in my config object, will I have to set them all one by one in an upstart script before executing the node app?
You generally use envvar to keep an application stateless. The same codebase should work in dev, staging, test and production environment.
You will put var like MySQL config, API keys, if log is enabled or not, if debug is on or not ...
Config file are used for variables which are not dependent of the environment. For instance, name of the application, number of items per page ...
I guess you can use config.json file for storing big configs. ENV I've usually use for passing application port or something very important for normal application start. For example if you use some external lib, it's better to be able to pass custom path to lib executor in ENV.
P.S. You should never save config.json in SVN.
I'm learning nodejs, when I work with tutorial like expressjs and krakenjs, I saw these terms many times:
Production and Development Environment
My understand is, those are config for production and development time, but why are they in same code? If the app/website is published, does it still have the development config inside?
I still don't get the idea...
You are correct that they are configuration files that are used either when doing development on your own box or when the code is running in production. You're development config file may contain a connection string to a database running on your development machine whereas the production connection string will probably point to a big, fast server with production data on it.
And yes, it's fine to have both files in your code at the same time, even when it's running in production as long as you make sure the system is using the correct one. You tell the system which one to use by setting the NODE_ENV environment variable to either dev or prod. Kraken will look at the value inside NODE_ENV and use the appropriate config file. This is a pretty common thing to do in node.
I am building a Node.js application and need to store database credentials (and other runtime properties) in such a way that they can be read when deployed on Heroku. My source is available in a public GitHub repository.
I am currently using environment variables, configured using heroku config:add, but am looking to understand if there are any alternatives. I would potentially like to use Cloud9 IDE, but it does not currently support environment variables.
Another option is to store the parameters in a config. file, but I believe the file would need to be checked in to Git (and as such, be publicly available) in order to be pushed to Heroku.
Thanks for your help.
ENV vars are generally considered the way to go, and the way Heroku do it themselves for database_urls and the like.
As you and your app are the only people with access to the env vars, you're generally OK security wise.
Putting credentials in Git or similar is a bad idea as it's another place that needs to be secured.
The one way I know of to solve the problem for development using command-line arguments. These can be specified in your run/debug configuration. You can then access the parameters in process.argv. Of course this means that they will be stored in your Cloud9IDE dev environment. You could then use the ENV variables in a retail production. This will at least prevent the credentials from being visible in source or config files.