The `wmftkbot` IRC bot is now logging detected experiment changes, as described at the [[ https://wikitech.wikimedia.org/wiki/Test_Kitchen/Troubleshooting | Test Kitchen/Troubleshooting page ]].
There are some quality-of-life improvements to add to this bot, plus it should be added to the `main` branch of the `mpic` repo. This should be pretty easy - maybe 2-3 points in total.
- Add a `README.md` that points the user to the instructions in the bootstrap shell script
- Get it to pass CI - YAML and `.gitignore` and other
- Have it mention multiple handles (including `dr0ptp4kt`) if it is encountering flaps accessing MPIC
- Add some additional `on_*` handlers and log a little bit more to the application log
- Align verbiage for two different types of alerts (they're close, but could use small tweaks)
- Rebase, ensuring `gitmodules` retained
- Add `X-` header just like Varnish poller.
- Send an "I'm still alive" message every weekday at a reasonable time (it should bug probably multiple team members, including `dr0ptp4kt`).
- Ensure that it can't spill errors rapidly - for exception handling, even have it take breaks if possible, or rate limit the frequency of logged exceptions. There's an edge case where if the service is restarted and a human is logged into IRC with the ID at the same time, it can generate a number of log entries during the ~15 seconds of startup time; the code has tired to prevent against this, but there's some condition in the underlying Python IRC package that can cause this, so a basic guard will be helpful.
- Test Kitchen renaming somehow, and update URL so it uses fewer redirects now that TK URLs are in place