Release recovery

Roll back a known bad release

Lesson 17 of 20 · 12 min · Module 6: Recover a complete service

Use a known-good version and verify it after a candidate fails.

Start this lesson →

Free. No installation: you write and run the code in your browser, and it is checked automatically.

By the end, you’ll be able to…

  • Keep a known-good identity available
  • Use verification to justify the next action

The idea

This practice runbook is an ordered sequence of the teaching actions named in the task. Each action changes or observes the supplied in-memory scenario.

01Keep a known-good identity available

Recovery should refer to a specific earlier artifact. A movable latest reference can make the previous working release ambiguous.

02Use verification to justify the next action

The candidate is unhealthy in this scenario. Record its failed smoke result before choosing the recovery action.

03Verify the recovered behavior

Rollback itself is another change. The final smoke result should demonstrate that the identified known-good release is healthy in the model.

Your turn

Deploy notes:2.0, which the scenario marks unhealthy. Smoke it, roll back to notes:1.0, then smoke again. The first probe must fail and the second must pass.

What the checks look for

  • Observes failure then recovery
  • Returns to the known-good release
Try it in the editor →

Take this with you

Prepare a specific recovery target and verify the result.

Go a little further

Describe the additional checks needed when a release contains a database migration.

Learn Infrastructure & Operations for free

20 short lessons with real code and instant checks.

See the whole course →