Filters
How to search
  • Plain words find names, texts and examples.
  • "two words" finds the exact phrase.
  • --name or -x finds an option.
  • #word looks only inside commands.
  • @grep finds lessons and labs that use grep.
  • Filters: cat:network · in:labs · in:notes · level:rhcsa · ch:8 · root:yes. Put ! before a word to leave it out. OR between words means one of them.
  • ↑ ↓ to move · Enter to open · Esc to close · / to search

Observability & SREDevOps series

5 postmortem mistakes

Most postmortems are written by tired people who want to close the ticket. That is exactly why the same five mistakes appear in review after review.

Quick reference

1. ❌ Name the engineer — It ends the conversation early. ✅ Name the condition — Fix the guardrail that allowed it.

2. ❌ Action with no owner — A wish, not a task. ✅ One named owner — They accept it in the document.

3. ❌ No deadline — It slips for a quarter. ✅ A date in the tracker — Even if the date later moves.

4. ❌ Fix lives in a wiki — Nobody reads it at 3 a.m. ✅ Change code or config — A test that fails next time.

5. ❌ Review stays in the team — Four people learn, nobody else. ✅ Publish and discuss — The organisation learns from one night.

Explained

The five

  1. Blame. Naming the engineer who ran the command ends the learning, because everyone else stops talking. Name the condition that made the command look safe: the missing guardrail, the alert nobody saw. \2. No owner. An action with no name is a wish. Every action gets one person, not a team, and that person accepts it in the document. \3. No deadline. An action with no date is a wish with a longer horizon. Put a date in the tracker, even if it slips later. \4. No code change. If the fix lives in a wiki page, it will not hold. The strongest action changes a pipeline, a default, a check or a test. \5. Not shared. A review read by four people fixes four people's habits. Post it where the organisation reads, and repeat the finding somewhere public.

What they share

Each one is a way to finish the document without changing the system. The document looks complete, the incident returns, and the next review repeats the same paragraphs.

The honest caveat: a blameless review is not a consequence-free one. It removes blame from people so the facts can surface; performance problems are handled elsewhere, privately.

The lesson: a postmortem is finished when the system is different, not when the document is signed.

In one line

Change the system, not the name. A review is finished when the system is different.

Check the actions: owner, date, and a merged change

More notes: Observability & SRE

Open in the notes app: search, filters, copy This note on GitHub