bullmq

a httpproxythrottlingmessagejobqueuedomainrouter.

What is it?

Its a enormous pillow to put in front of your webserver. Prevent server/application-panic by limiting and rerouting webrequests to different areas/servers.

Isnt there already software for this?

There are many enterprise-ish packages out there, but I wanted something deadsimple to configure. Next to that AMQP-support (not finished) will be able to scale out to big servicebusses like RabbitMQ e.g.

Concept

Here's a sequencediagram (because sequencediagrams are cool :) hope it makes sense though)

A simple configuration

Lets have a look at the conf.d/myproject.js file:

exports.proxy = function(config){  
  config["localhost:8080"]        = ["http://127.0.0.1:80"];
  config["mydomain.org:8080"]     = ["http://127.0.0.1:80"];
  config["www.mydomain.org:8080"] = ["http://127.0.0.1:80"];
  config["ws.mydomain.org:8080"]  = ["http://127.0.0.1:8111"];
  return config;
}

Above are domainredirects: localhost on 8080 will redirect to port 80. Other domains can be entered (for example ws. for websocket traffic). Multiple arrayvalues will result in random round-robin behaviour (quickndirty scalable).

exports.limiters = function(config){
  config["localhost:8080"]     = { unit: "minute", rate:90 };  // 90 requests per minute 
  config["localhost:8080/foo"] = { unit: "minute", rate:20 };  // overrides 90 with 20 
  return config;
}

Above you can limit per domain and/or url. The units are: day,hour,minute,second. This is handy to ensure your projects will not overload your server. For example, one could setup 3 subdomains for api-calls:

freemium.api.com
basic.api.com
premium.api.com

Just set different rates for these subdomains, and bullmq will reply to clients like so when rate are exceeded:

{succes:false, code: 429, msg:"Too Many Requests - your IP is being rate limited",data:{}}

And now for workers:

exports.queue = function(config){  
  config["localhost:8080"]    = {
    "/foo": { 
      queue: "/request_web",       // put '/foo' in queue 'request_web'
      timeout: 500                 // generate timeout if no redis subscribers respond within 500 milliseconds
    }      
  };
  return config;
}

Here you can queue a request for later processing using nodejs bull (b2b apis/messaging/long running tasks). If a redis- or amqp-subscriber does snot respond with a 'completed'-event to the task, bullmq will reply:

{succes:false, code: 104, msg:"server could not handle your request in time",data:{}}

If a subscriber does respond within the timeout-timeframe, it can respond with a "queued", "processing" or "ok" reply. Example: user wants to encode 4 video's, clients are firing hundreds of push requests to the databases etc.

Syslog

logging can be directed to stdout and/or syslog. Here's some syslog output :

May  9 10:01:41 bullmq [26639]: ACCEPT => www.yourdomain/tpl/front/gfx/search_black.png from 94.214.179.213 (bucket:232.06
May  9 10:01:41 bullmq [26639]: ACCEPT => www.yourdomain/lib/flop/site/tpl/gfx/terminal.empty.gif from 94.214.179.213 (buc
May  9 10:01:45 bullmq [26639]: ACCEPT => www.yourdomain/js/?hash=MjAxMS0wNi90dHlyZWNvcmQtMTMwODMxODg4Mw== from 65.49.68.1
May  9 10:01:45 bullmq [26639]: ACCEPT => www.yourdomain/js/?hash=MjAxMS0wNi90dHlyZWNvcmQtMTMwNjk2NjA2NQ== from 65.49.2.17
May  9 10:01:48 bullmq [26639]: ACCEPT => www.yourdomain/lib/flop/js/gfx/flop.button.white.png from 65.49.68.165 (bucket:2
May  9 10:01:48 bullmq [26639]: ACCEPT => www.yourdomain/lib/flop/js/gfx/flop.skin.gif from 65.49.2.175 (bucket:227.572666
May  9 10:01:50 bullmq [26639]: ACCEPT => www.yourdomain/r/MjAxMy0xMC90dHlyZWNvcmR0dHktMTM4MTU3NjE5Nnw4MHgyNA== from 157.5
May  9 10:02:10 bullmq [26639]: ACCEPT => www.yourdomain/ from 94.214.179.213 (bucket:227.02533333333335)
May  9 10:02:11 bullmq [26639]: ACCEPT => www.yourdomain/lib/core/functions.js from 94.214.179.213 (bucket:226.07126666666