HTTP Health Checker

Project roadmap

1. 01

   Define the config and output

   ~1.5h

   Write README.md: the config file (TOML or YAML) listing endpoints with a name, URL, method, expected status codes, optional text the body must contain, a timeout and a latency limit, plus global settings for concurrency and retries; the table and JSON output; and the exit codes (0 all healthy, 1 some unhealthy and 2 config errors). Create the project with tokio, reqwest, serde, toml (or serde_yaml) and clap. Done when: README.md defines the config, output and exit codes.

2. 02

   Load and validate the config

   ~1.5h

   Deserialize the config into structs with Serde, with defaults for optional fields, and validate it: URLs parse and use http or https, names are unique, timeouts are positive, and status lists aren't empty. Report every problem with the endpoint name. Done when: a valid config loads, and a config with several mistakes lists all of them.

3. 03

   Check endpoints concurrently

   ~2h

   Use one shared reqwest::Client and check endpoints concurrently with a limit (for example with futures::stream::iter(...).buffer_unordered(n), or a Tokio Semaphore), applying each endpoint's timeout with tokio::time::timeout. Measure the latency with Instant, read at most a limited amount of the body to check for the expected text, and retry once on connection errors. Done when: checking a list of endpoints runs them in parallel and records status, latency and errors.

4. 04

   Report the results

   ~1.5h

   Classify each result as healthy, slow (over its latency limit), wrong status, missing text, timeout or error. Print an aligned table with colors in a terminal, sorted with problems first, and a summary, or JSON with --json (serialized with Serde). Exit with the right code. Done when: the output clearly shows each endpoint's health, and the exit code matches.

5. 05

   Add watch mode

   ~1.5h

   Add --watch SECONDS, which repeats the checks on a tokio::time::interval, keeps each endpoint's previous state, and prints only changes (such as api-login: healthy to timeout, with the time), plus a short summary each round. Stop cleanly on Ctrl+C. Done when: watch mode reports when an endpoint goes down and when it recovers.

6. 06

   Test it

   ~2h

   Write tests with a local fake server (such as the wiremock crate) returning various statuses, slow responses, bodies with and without the expected text, and dropped connections: check classification, timeouts, the concurrency limit, retries, JSON output and exit codes. Done when: cargo test passes.

Resources

• DocsTokio tutorial
• Docsreqwest documentation
• DocsSerde documentation
• Docsfutures: StreamExt::buffer_unordered
• Docswiremock crate