microlinkhq / queue

The high resilient queue for processing URLs.

High resilient & reliable URLs processing queue.

Motivation

When you are consuming an API (such as Microlink API or using any other API) you need to assume it can be a wide resource and some of them can eventually fail for many and varied reasons: rate limit, timeout, proxy rotation, etc.

Some of these errors can regret simply retrying the original request after waiting a prudent period of time

In any case, you want to have the guaranteed the URL will be successfully processed in an indeterministic moment in the future.

This package presents a small but powerful architecture for processing URLs with guarantees.

Architecture

For getting guarantees that your requests are successfully processed, we are going to use bull, a lightweight FIFO queue backed on redis.

The requests (called jobs) will remain in the queue until a consumer get it and verify is has been successfully processed.

Sender

npm run start:sender

The sender is is who sends the work to process the consumers.

It's exposed using a HTTP server.

For sending a job into the queue, just send it using a GET

$ curl http://localhost:3000\?url\=http://microlink.com\&video

If the job has been added successfully, you will have a 201 Created and the job.id back.

HTTP/1.1 201 Created
{
  "id": "7"
}

You can provide two customizable things as query parameters:

• priority (optional): Priority value. ranges from 1 (highest priority) to MAX_INT (lowest priority).
• delay (optional): An amount of miliseconds to wait until this job can be processed.

Any other query parameter provided will be passed as part of the job.

Receiver

npm run start:receiver

The receivers will process the job pushed by the consumer.

They are a pool of processes, waiting on idle until they have something to consume.

$ npm run start:receiver

receiver=0 status=listening
receiver=1 status=listening
receiver=2 status=listening
receiver=3 status=listening
status=processing id=13
receiver=0 status=processed id=13 status=success

by default, You are going to have one receiver per physical CPU core. Since we delegated on farm, this is easily customizable.

License

queue © microlink.io, released under the MIT License.
Authored and maintained by Kiko Beats with help from contributors.

microlink.io · GitHub microlink.io · Twitter @microlinkhq