Project Requestko

What is this app

Simple async service that tries to forward one GET request to another server and return response.

Little bit about architecture

This app makes use of asynchronous nature of Python to handle as much load possible only using main thread. Server used to serve this app isuvicorn Framweork used to create the app is Starlette. Why not FastAPI, Django, Flask etc. well I wanted to use something lightweight and native to async, Starlette shines there. Why not FastAPI? while it's great, it has features that bring overhead that I mostly don't need for this project.

Why not gunicorn? Well this is something that Iwould later add to scale the app, I would use combination of gunicorn with uvicorn workers to make use of processes and async. But I think for now this is enough as showcase, and here we don't really have anything that is CPU bound, so we are doing only I/O where async should be enough.

ALso there are different approaches to scaling this app but I won't discuss everything here it would be pretty long README.md

How to run this service locally

• Position yourself in the directory requestko-project and simply run docker compose up this comand will bring up:
  • Promehteus server
  • Grafana
  • Requestko service

I won't show pictures in text below as they arn't nice, take look for yourself

Out of the box Prometheus should be able to scrape Requestko service for metrics and you should see something like in the picture prometheus.png.

After that you should go to Grafana (port 3000), log in with default username and password (which are admin and admin), add new DataSource like in the picture datasource.png. The URL for the Prometheus you can get from simple docker inspect where you filter for network address. Next step is to import Requestko-grafana.json dashboard. To that simply use import feature, but importantly you should change:

      "datasource": {
        "type": "prometheus",
        "uid": "l1Fhpl_7z"
      },

each snippet like this, to:

      "datasource": {
        "type": "prometheus",
        "uid": "HERE_ENTER_WHATEVER_IS_YOUR_UID_OF_YOUR_SOURCE"
      },

Now everything should work.

Requestko service

Requestko service will be brought up inside docker container and binded to your localhost:80 address so you don't need to check for containers IP address.

Enpoints

Requestko has 3 defined endpoints you can reach intentionally:

EDNPOINT: /api/smart
METHOD: GET
REQUESTED QUERY PARAMETERS: 
         PARAMETER_NAME: timeout
         PARAMETER_TYPE: integer [0, infinity] - represents milliseconds


EXAMPLE CALL:
curl -X GET http://localhost:80/api/smart?timeout=100

EDNPOINT: /metrics
METHOD: GET
DESCRIPTION: Prometheus metrics are exposed here

EDNPOINT: /schema
METHOD: GET
DESCRIPTION: OpenAPI schema with more details about possible responses for the route /api/smart can be checked here

What should I add, and what needs to be done better

• Write tests
• Install aiohttp with speedups
• Handle exceptions better
• Clean the code
• Handle edge cases
  • what if someone in timeout sends negative number
  • better serialization/deserialization of requests and responses
  • Check if payload returned from Exponea matches specific structure
  • ...
• better error logging (mainly Exceptions)
• Add automatic test stage, linting, static code check etc.

Why haven't I added this you may ask, well answer is not enough time and this service will never run in production at least it shouldn't :D