LoadFire is an easy to script load balancer and reverse proxy in NodeJS.
It allows you to write your own pieces of logic as "Patterns" and the core engine takes care of all the proxying logic so you don't have to worry about it.
The real power is that all it's behavior is entirely scriptable in JavaScript (NodeJS). :)
###This allows for many different use cases, such as:
- Dynamic realtime proxying rules (domain mappings stored in Redis for example). Such things are useful for PaaSs
- Add pieces of middleware to your reverse proxy
- Customizable load balancing patterns (Sticky, RoundRobin, ...)
###It supports proxying:
- HTTP/HTTPS
- WebSockets
varhttp=require('http');varloadfire=require('loadfire');varEDITOR_PORTS=[7001,7002,7003,7004];varEDITOR_SERVERS=EDITOR_PORTS.map(function(x){return{host: 'localhost',port: x};});// Start all our different http serversfunctionstartEditorServers(ports){ports.forEach(function(port){// Setup HTTP ServervarhttpServer=http.createServer(function(req,res){// Output the port number the server is running onres.writeHead(200,{'Content-Type': 'text/plain'});res.end(port.toString());});// Listen on assigned porthttpServer.listen(port);});}// Config for our loadfire servervarCONFIG={'resources': [{// resource is some value identify this resource// by default it should be the hostname to matchresource: 'localhost:8000',// List of backends to hitbackends: EDITOR_SERVERS,// Load balancing pattern// As of now a few are builtin// random, roundrobin, stickypattern: 'roundrobin'}],// Server to start loadfire onport: 8000};functionmain(){// Start our http serversstartEditorServers(EDITOR_PORTS);// Setup our load balancer with the above configvarloadServer=loadfire.createServer(CONFIG);// Now start our load balancerloadServer.run();// Check out localhost:8000// Refresh a few times and you'll see different port numbers appear// depending on which http server the requests are proxied to// since we are using the roundrobin pattern it will cycle through them}// Run mainmain();It was initially built to satisfy our needs at FriendCode, and we've been using it in product for months without any issues, so it can be considered as stable.
The API however will be changed soon due to some design decisions aiming to simplify it's API. (Patterns are quite monolithic as of now, they will be split up into different parts: matcher, mapper, balancer, store).