Skip to main content

retro-template


title: Template retrospective

Short Title of Incident

Date(s) of Incident: YYYY-MM-DD date or date range of incident

Summary

This should be a 1-2 paragraph executive summary of the incident that conveys the systems involved, impact scope, and severity. The goal of this summary is help someone determine if they want or need to read the rest of the report; do not make it too long in an effort to convey everything about the incident in one place.

Description

This is the most important section of your retrospective. This is where you will finally assemble a narrative of the incident. Do not write this section until the retrospective has been conducted.

This narrative does not need to presented in a chronological fashion; you could organize it by the systems or teams involved, by tested hypotheses, or any other way that makes sense. However you organize it, you should make sure that a reader can follow the progression of the incident and understand what happened.

When writing this narrative, be sure to highlight:

  • when the incident started, stopped, or changed its nature and/or severity
  • when and how people made realizations about the nature of the incident
  • when and how people made hypotheses (right or wrong) about the nature of the incident
  • anything that surprised people during the incident
  • any point where people had contradictory perceptions about the incident

Timeline

This should be a minute-by-minute account highlighting critical events over the course of the incident; as much as possible, these should be linked to a primary record of some type, such as a Slack message, system log, or PagerDuty alert. If that is not possible, snipping out a log line or providing some other message for context is important.

The timeline should start with the date of the events and the timezone for the timeline, and then each event should be tagged with a timestamp.

For example:

2020-05-22

All times CDT (GMT -0500).

  • 08:17 - spacecats-fe begins logging messages about being unable to contact spacecats-api
  • 08:21 - cblkwell receives alert that spacecats-fe failed ping test

This timeline should be curated; you don't need every individual message or event cataloged here. However, do not edit out so much that context is lost. Do not remove things that weren't relevant to the incident if they seemed relevant at the time. We want to make sure people who read this timeline see how things looked to the responders at the time. This means including actions that didn't ultimately turn out to be productive, spurious alerts that caused concern but ended up being unrelated, or anything else that may have affected the incident response even if unrelated to the incident (was someone unable to respond because their internet service went down?).