Installation

In the top-level project directory, run

go mod tidy

Usage

To start monitoring endpoints, run:

go run main.go <config_file>

where <config_file> is the path to your YAML configuration file.

Identifying Issues

General Approach

To identify issues, I started by reading over the requirements and then reading over the code to understand how it currently works, making note of anything that seemed like it would violate the requirements.

Then, I wrote a combination of automated unit and integrations tests, trying to make sure the test suite covered all of the project requirements. Using these tests I was able to confirm that the issues I identified by reading the code were actually issues, and was also able to discover some issues I hadn't initially seen. I would then update the code as necessary to resolve the issues, and then confirm that the associated test cases passed after the fixes were applied.

Discovered Issues and Solutions:

extractDomain

• extractDomain makes mistakes in some cases
• identified issues by reading the code
• confirmed by writing test cases (TestExtractDomain) to cover suspected failure cases (and other basic test cases to prevent regression)
• issues with original code:
  • included port numbers in domain
  • if path contained //, would not parse the domain correctly (it would use the part of the path after the last // as the domain)
• updated extractDomain to specifically look for and remove http:// or https:// prefix, instead of splitting on //
• updated extractDomain to remove port numbers if present, by splitting the authority section of the URL on : and only keeping the string before the first :

checkHealth

• checkHealth would not report a failure if request took more than 500 ms:

  • identified issue by reading the code
  • confirmed issue by adding test cases for checkHealth (TestCheckHealth), one of which included the server taking more than 500 ms to respond
  • added Timeout value of 500 ms to http.Client used by checkHealth, now checkHealth cancels the request after 500 ms and reports that the request failed

• checkHealth was sending entire Endpoint struct as request body, should only be sending actual body

  • discovered while writing TestPostOK test, when adding check to confirm that body matches what is provided in the YAML file
  • fixed by updating checkHealth to only use the actual body as the request body

monitorEndpoints

• when checking all endpoints the checkHealth calls are serialized, which could easily take more than 15 seconds if the total latency of all of the endpoints is high enough

  • moved logic for checking all endpoints to checkEndpoints function, which checks health of each endpoint exactly once
  • added TestCheckEndpoints test to confirm that, with starter code, time to check all endpoints exceeds 15s with 100 endpoints with 250ms latency each
  • updated checkEndpoints function to check all endpoints in parallel, with each checkHealth call in a separate goroutine, ensuring we can check a large number of endpoints while staying well with the 15s interval
    • also updated DomainStats to use atomic integers to prevent race conditions when multiple checkHealth calls are run in parallel for endpoints belonging to the same domain

• we sleep for 15 seconds once all endpoints have been checked, which means our actual health check period will exceed 15s (i.e. the actual time period will be time_to_check_all_endpoints + 15s)

  • identified by reading the code, confirmed by adding test case that has slow-to-respond server and fails when time interval between consecutive iterations is not between 14600ms and 15400ms (TestSlow)
  • updated code to wait to log stats until 15 seconds have passed since the stats were last logged, rather than 15 seconds since the stats were last collected