30-Second Rollback.
Every production change deserves a rollback plan that fits on a sticky note. Fill the form. Get the runbook. Print it. Tape it next to the deploy button. If the rollback takes longer than 30 seconds to read, it's too long.
Inspired by Shipping a Killable Feature. No data leaves your browser.
Runbook preview
# 30-SECOND ROLLBACK > Generated by sageafterdark.com/tools/30-second-rollback ## Change [name your change] ## When deployed [fill in time] ## Kill switch [where is the kill switch] ## Rollback command (copy/paste) ``` [exact command to reverse] ``` ## Verify rolled back ``` [command/URL that proves it] ``` ## If it doesn't work, page [who to ping, in what channel, in what order] ## Known blast radius [what this affects, what it doesn't] --- Print this. Tape it next to the deploy button.
Most outages aren't caused by the change itself — they're caused by the time it takes to undo the change. A 12-minute rollback is a 45-minute incident. A 30-second rollback is a tweet.
This generator forces you to write down the four things you'll forget at 2am: what you changed, where the kill switch lives, who to ping if it doesn't work, and how you'll know it actually rolled back.
Print it. Paste it into the PR description. Pin it in the deploy channel. Whatever it takes to make the rollback findable before you need it.