Comment

Comment by ipnon

The simplest solution is to compensate the on-call engineer, either by paying them 2 times their hourly rate per hour on-call, or by accruing them an hour of vacation time per hour on-call. This works because it incentivizes all parties to minimize the amount of time spent in on-call alert.

Management is incentivized to minimize time spent in alert because it is now cheaper to fix the root-cause issues instead of having engineers play firefighter on weekends. Long-term, which is the always the only relevant timeline, this saves money by reducing engineer burnout and churn.

Engineers are also incentivized to self-organize. Those who have more free time or are seeking more compensation can volunteer for more on-call. Those who have more strict obligations outside of work thus can spend less time on alert, or ideally none at all. In this scenario, even if the root cause is never addressed, usually the local "hero" quickly becomes so inundated with money and vacation time that everyone is happy anyway.

It doesn't completely eliminate the need for on-call or the headaches that alerts inevitably induce but it helps align seemingly opposing parties in a constructive manner. Thanks to Will Larson for suggesting this solution in his book "An Elegant Puzzle."

parent ipnon 2 years ago

Discussion

Live from the official Hacker News API

A modern reader for Hacker News · SvelteKit + Skeleton