DevOps Retrospectives

How to choose

Think in terms of where you are in the release cycle.

• Before a risky launch or migration. The Pre-Mortem imagines the launch has already failed and asks why. Run it a week or two before cutover, while there's still time to add a feature flag or a rollback plan.
• Right after a release. The Release Retro covers what went smoothly, what broke or slipped, and what you missed before release. Hold it within a week, with the release notes and any incident links open.
• After an outage. The Incident Review builds a timeline, looks at how you detected and responded, and ends with follow-ups. To dig into a single failure, follow it with the 5 Whys. When the cause is spread across tooling, process and environment, the Fishbone keeps all of them in view.
• When delivery feels slow. Kanban Flow looks at where work waited: review queues, flaky test runs, a staging environment everyone needed at once.
• When shortcuts catch up with you. The Tech Debt retro lists the debt the team ran into this sprint, what it costs and what to pay down next. Keep it to debt you actually hit, not the full wish list.

A rhythm that works for many teams: a Pre-Mortem before each major launch, a Release Retro after it, an Incident Review whenever production breaks, and a Tech Debt retro every few months in place of a regular one.

Turning findings into engineering work

DevOps retros produce technical follow-ups: add an alert, write a runbook, fix a flaky test, automate a manual step. They only matter if they get done.

• Write tasks, not themes. "Alert when queue depth passes 10,000" beats "improve monitoring".
• Give each one an owner in the Action Items panel. Retro Rabbit doesn't create tickets in your tracker, so copy the bigger items into your backlog and put the ticket number in the action item.
• Check them first at the next retro. Open the Previous Retro and Old / Unresolved Items tabs. An incident follow-up still open three sprints later is a risk in its own right.
• Bring your own numbers. Retro Rabbit doesn't track deploy frequency, failure rates or recovery times. If you watch them, bring them to the meeting.

On Pro AI (and during the free trial), CarrotIQ Suggest proposes action items from the top-voted notes, a quick first draft for your follow-ups. See retrospective action items that actually get done for how to keep the list short.