00 - Assignment

The Task

Write an HTTP server. You can choose to implement it either in Python or Go. The server should serve one endpoint:

/api/smart

Requirements for the endpoint:

The endpoint performs up to 3 HTTP requests to the Exponea Testing HTTP Server and returns the first successful response.
Instead of firing all 3 requests at once, the endpoint fires only a single request at the beginning.
- If there is a successful response within 300 milliseconds, the endpoint returns this response and does not fire other requests.
- If there is no successful response within 300 milliseconds, it fires another 2 requests. Then it returns the first successful response from any of the 3 requests (including the first one).
The endpoint accepts a timeout parameter - an integer value specifying time in milliseconds. The endpoint always returns a response within the given timeout. If the timeout is not specified, the endpoint should respond with the first successful response from our testing server.
If there is no successful response within this timeout, the endpoint returns an error.

The endpoint returns a JSON response.
A successful response is defined as a response that returns HTTP status code 200 and has a valid payload.
Your endpoints should return a JSON response that will contain successful responses from Bloomreach testing server.

The Bloomreach Testing HTTP Server serves a single endpoint:

GET https://exponea-engineering-assignment.appspot.com/api/work

It performs some work (taking roughly 100-600 ms) and then returns the time it took to respond as a JSON response. Example response: {“time”: 160}

Note: The server is not very reliable , which means that it does not always return a successful response.

Performance - we expect that the requests to ExponeaTesting HTTP Server will be done concurrently. We will test how your server implementation behaves under load - both in terms of requests per second, and the number of concurrent requests.
Robust implementation - your server should behave correctly in presence of errors. Exponea Testing HTTP Server is infamously known for being unreliable. Also, the implementation should not leak resources (stuck threads / goroutines / connections for unknown time).
Readable, maintainable implementation - for example, prefer well-known open source libraries, with licenses that allow them to be used commercially.

Is the task too easy for you? Don’t worry, we have prepared a few extras for you:

Write tests for your implementation.
Instrument your code with tools that will make debugging production easier - logging, monitoring, tracing. We need a way to check if Bloomreach Testing HTTP Server is behaving well in production.
Include Dockerfile that can be used to build and run the server.
Attach a discussion about code behavior - knowledge cases, how it behaves in certain conditions, performance characteristics, resource requirements, etc. How many concurrent requests can the server handle? How would you protect the server against being overloaded?