Q: We just survived an outage. What monitoring changes should the post-mortem include?

Q: We just survived an outage. What monitoring changes should the post-mortem include?

A: Every incident is free tuning data. Run this section in your post-mortem, no blame, just gaps.

— Time-to-detect: how long from real failure to first alert? If it was slow, tighten the interval or fix the check.
— Did the right alert fire? If the homepage check passed while checkout was broken, you need a deeper transaction check.
— Did the right person get paged, fast? If not, the escalation chain had a gap.
— Was the status page updated, and how late? Bake the first update into the runbook.
— Add one new monitor for the specific failure mode you just hit so it never surprises you twice.

Follow-up: schedule the change. A post-mortem with action items nobody dates is a post-mortem you'll repeat.

Got a question? Drop it in the comments.