Software Development Kit and Command Line Interface for spawning streaming stateless HTTP microservices for any programming language or arbitrary binary.
At it's core, microcule maps HTTP request response streams to the STDIN STDOUT streams of a function in any arbitrary programming language or any compiled binary. It's reminiscent of CGI, but utilizes additional STDIO streams, does not attempt to parse STDOUT for HTTP response methods, and ships with streaming plugins for extending your microservices. microcule is an old concept rethought and improved with the latest industry standard toolings.
If you are using Amazon Lambda or other cloud function hosting services like Google Functions or hook.io, you might find microcule a very interesting option to remove your dependency on third-party cloud providers. microcule allows for local deployment of enterprise ready microservices. microcule has few dependencies and will run anywhere Node.js can run.
This project is the component which several production services, including hook.io, use to spawn real-time arbitrary streaming microservices in response to streaming HTTP requests. It's been battle-hardened with over two years of development and it's largest installation is now managing 8000+ microservices.
You are encouraged to use this module as-is, or modify it to suite your needs. If you are interested in contributing please let us know by opening a Pull Request.
Features
Modular
Only require the functionality you need.
microculeitself is actually just a collection of HTTP middleware Plugins presented as a CLI tool. This is essential, as all it's features and functionality are de-coupled with an industry standard API. You are encouraged to use the microcule binary shipped with this project, or use microcule plugins programmatically in your existing application.
Universal
Supports Serverless functions in 20 programming languages! ( and counting )
Supports Serverless with standard Unix or Linux tools ( like ls, echo, or tail -f )
Full support for mapping HTTP -> STDIO streams
Unix first. No custom APIs or buffered context
Intelligent HTTP Request Parsing
microcule can parse any kind of request data based on the incoming request's content-type header and will always stream requests by default.
Query
JSON
Form
Multipart
Streaming
Binary
Even binary data works great! Here is an example of resizing in image in JavaScript or Bash.
Consistent
Spawns functions or arbitrary binaries in response to HTTP requests
Uses a fresh system process per request per execution
State of services clears on every request
Service source code is immutable ( unless configured otherwise )
Uses build step for compiled languages
Versatile
Ships with microcule binary for starting HTTP microservice servers
Scripting support for HTTP request / response API ( differs per language )
Can serve any arbitrary binary like echo, ls, and tail -f as streaming HTTP microservices
Can optionally accept STDIN and process scripts with Unix style pipes ( useful for using functions in DevOps! )
Extendable
Should theoretically be able to work with all programming languages
Should work with any existing Unix or Linux tool
Provides simple customizable interfaces for programming language environments
Uses Plugin System based on standard node.js HTTP middlewares
Simplistic design makes it very easy to add features / add languages / create new plugins
No Containers
By design, microcule ships with no container or OS virtualization
Since it makes no assumptions about worker environment, microcule will work with any Container or Virtual Machine solutions
Isolates state of microservice per system process and request ( stateless service requests )
microcule can be optionally extended through a simple app.use() based plugin architecture. Plugins are standard Node.js Express.js middlewares. This means you can use any existing Node.js middleware as a microcule plugin, or re-use any microcule plugin as a middleware in any existing Node application.
Available Plugins
bodyParser - Intelligent streaming body parser ( JSON / form / multipart / binary )
cronScheduler - Schedule function calls on a timer using a simple Cron syntax
compile - Compile microservice source code into binary ( optional dependency to spawn )
logger - Basic extendable request / response logger function
mschema - Adds mschema validation to incoming request parameters
sourceGithubGist - Pulls microservice source code from a Github Gist
sourceGithubRepo - Pulls microservice source code from a Github Repository
spawn - Spawns instance of microservice code in a new isolated child process. Maps HTTP req/res to process STDIN/STDOUT
rateLimiter - Extendable request rate limiter. Holds rate limits in-memory or in Redis.
For Express based plugins example, see: ./examples/express-plugins.js
Since plugins are standard Node.js middlewares, writing custom plugins is very easy.
Languages
microcule supports 20+ programming languages. The best way to get started is to visit 100+ Microcule Examples.
Want to see a new language added? Simply open a pull request or open an issue. We are glad to add it!
Installation
npm install -g microcule
This will install the microcule binary globally on your system.
Usage
Usage: microcule [command] [options] Commands: help Display help Options: -t, --timeout <number> Sets max timeout of service in milliseconds -h, --host <value> Host to listen on -p, --port <number> Port to listen on -l, --language <value> Target programming languages -w, --watch <bool> Reloads source files on every request ( dev only ) -v, --version Output the version number
By default, microcule will attempt to start a listening HTTP server based on a microservice file path.
Each call to microcule will automatically start a listening HTTP server on port 3000, additional instances of microcule will auto-increment the port to 3001, 3002, etc.
Service target language is automatically detected based on the file extension of the service. This can be overridden using the --language option.
Note: For certain languages ( such as Babel ), the first microservice request to microcule may take additional time as it will perform an initial compile and cache step.
Note: Please see Babel Support for additional Babel configuration
Node Middleware Usage
microcule is designed to work as set of standard Node.js HTTP middlewares. This is very useful and universal in that you can re-use any existing Node.js tooling ( like Express.js middlewares ) with any microcule plugins ( like spawn or compile ).
Examples
see: ./examples for more examples
node express-simple.js
var microcule = require('microcule');var express = require('express');var app = express();var nodeService = function testService (opts) { var res = opts.res; console.log('logging to console'); res.json(opts.params);};var handler = microcule.plugins.spawn({ code: nodeService, language: "javascript"});app.use(handler);app.listen
Spawning arbitrary compiled binaries
Sometimes you may need to spawn a precompiled arbitrary binary instead of a script based microservice. microcule supports two ways of spawning arbitrary binaries.
Executing arbitrary binaries from microcule Bash services
You can create a new bash script and execute your binary in a bash sub-shell using microcule's built-in support for Bash scripts.
Simply create a bash script and use microcule to spawn it. This will work for any binary. Here is a simple curl example:
Spawning arbitrary binaries directly from microcule
In some cases, creating a bash sub-shell is not ideal and you'll want to directly spawn your precompiled binary.
This is very easy, simple call microcule.spawn using the bin and argv options. Here is an example of spawning echo
var handler = microcule.plugins.spawn({ bin: 'echo', argv: ['hello', 'world']});
Chaining Services as Middlewares
Since v5.1.0 and above, microcule is able to compose multiple functions together as middlewares to create composite functionality. Since every service is dealing with streaming http, it's easy to chain them in a row.
In order to chain multiple services, simply call them as standard Node.js middlewares in the order you want them to execute.
app.use([logger(), basicAuthHandler, bashServiceHandler, nodeServiceHandlerA, nodeServiceHandlerB], function (req, res) { console.log("No services ended response, made it to end"); // It's good to have a catch-all handler at the end in case none the services ended the request res.end('caught end')});
For full example see: ./examples/express-chain-services.js
Multiple Microservices Per Server Instance
In some configurations you may want to safely run multiple kinds of microservices on one server instance ( a small monolith ). microcule is designed exactly for this use case.
Since every incoming service request will spawn a separate process, microcule can safely and easily handle spawning multiple types of microservices at once without affecting the state of other services.
If you look at the ./examples/http-server-simple.js file, you will see that spawn() can be used as a standard Node.js or Express HTTP middleware. For multiple services per server, simply map the spawn() method to any custom routes you may want to define.
You can also stack multiple express apps together for multiple microservices with separate routes. see: ./examples/express-multi-language.js
Additional language support is both planned and welcomed. Please open a Pull Request if you wish to see a specific language added
Creating a custom plugin
Creating a custom plugin is very simple. Just code the microcule plugin the same you would any other Node.js middleware. Feel free to use any of the other existing HTTP middlewares in the Node.js ecosystem.
custom-logger.js
module.exports = function loggerMiddleware (config) { // here the function handler can be configured based on a `config` object return function loggerHandler (req, res, next) { console.log('running service ' + req.url); next(); }}
Once you've created a new plugin, simply require() it, and call app.use(customLogger({})). That's it! There are no magic or surprises with how plugins work in microcule.
See: ./examples/express-plugins.js for more details.
SSL
If you wish to start HTTPS SSL enabled services, simply adjust the following configuration properties in ./config/index.js
{ http: { https: true, // enables https / ssl, requires key, cert, ca key: "ssl key as string", cert: "ssl cert as string", ca: "ssl ca as string", sslRequired: false, // redirects all http traffic to https, optional onlySSL: false // will only start https server with no unprotected http interface, optional }}
Security
Running untrusted microservice code in a safe way is a complex problem. microcule is only intended to isolate a small part of the entire untrusted source code execution chain.
If you intend to use this module to run untrusted source code please consider the following:
What this module does isolate
Microservice state
Microservice errors
Stream / Socket errors
HTTP Request state
Every incoming HTTP request will spawn a new system process to run the microservice instance, which then sends data to the HTTP response. Every microservice execution will happen in a fresh systems process.
All errors that can possibly happen during the execution of a microservice should be trapped and isolated to not affect any other microservices.
What this mode does NOT isolate
Server Memory
Server CPU
Server file-system
Process CPU
Process Memory
microcule cannot make any guarantees about the isolation of the server or spawned processes. All microservices will have default access to the server's file-system and child processes.
To ensure isolation of the server file-system, you would want to use the microcule binary in a chroot jail, or another similar container solution.
To ensure isolation of the server memory and cpu, you will want to use the microcule binary in a virtualized environment capable of monitoring and managing resource usage per process.
Coffee-Script
In order to Coffee-Script based microservices, you must install the following packages:
npm install coffee-script
Babel
In order to run Babel / ES6 / ES7 microservices, you must install the following packages: