748563 - [briar-patch] Give concise, human-readable next steps for slaves needing recovery in the kitten emails

Status: RESOLVED WONTFIX

(Infrastructure & Operations Graveyard :: CIDuty, task)

Description

[Chris Cooper [:coop] (he/him)]

I posted about Facebook's auto-remediation system back in the fall:

https://www.facebook.com/notes/facebook-engineering/making-facebook-self-healing/10150275248698920

As much as possible, I want briar-patch to be working towards this goal.

In almost all cases, we know what the next steps are that a human should take for a particular slave. At the very least, we know a first step, e.g. "Is this slave enabled in slavealloc?" Rather than display a list of previous states in the kitten emails, let's map those to a human action, and (where possible) provide a link for someone (buildduty) to get started performing that action.

For example:

I know that the kitten report is in some state of flux right now with the colo move, but two win64 slaves were consistently appearing in the "previously seen" category. By logging into those slaves, I was able to determine that auto-logon was not setup on these slaves, so buildbot was never getting the chance to start. That kind of information should live in a state matrix somewhere so the next time a win64 slave enters that state, we can tell buildduty (Please VNC into this host to make sure auto-logon is setup."). 

In short, I want the report to show me actionable work, with a link to the dashboard that will to allow me to drill-down and get to the info that the report currently displays.

Comment 1

[Chris Cooper [:coop] (he/him)]

I've made a first stab at HTML mail with some bug links here:

https://github.com/ccooper/briar-patch/commit/ec362f277659a24e151dc239937a5bb26a3ea4eb

Comment 2

[Mike Taylor [:bear]]

coop's changes have been merged and tested - doing a test run now on staging

https://github.com/mozilla/briar-patch/commit/ca1a41b981b72d7382283b1da6054be3155ad948