Recently reminded that of all of the bad ideas in SRE culture, that error budgets really are an outright toxic concept. I've only quadrupled down on that in the last few years. Even just hearing the words is like a signal to help deprogram someone.
In brief:
1./ The word "budget" is way too often interpreted as "a certain amount of customer pain is something you can spend." Absolutely not. Fire people who intentionally invoke customer pain. With haste.
1./ The word "budget" is way too often interpreted as "a certain amount of customer pain is something you can spend." Absolutely not. Fire people who intentionally invoke customer pain. With haste.
2./ The concept fosters a kind of beaurocratic boundary that reinforces the outdated "us and them" model that is at the core of SRE and the shady attempt to claim that separate teams specializing on dev and ops is somehow as effective as devops.
3./ Resilient systems do benefit from a certain low percentage of low-grade errors, to keep anti-bodies and ductility primed. Sometimes a 99.9% success rate is better than 99.999%. But error budgets do not encapsulate this and big modal outages on schedule are not how to get it.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
