<img src="https://secure.insightfulcloudintuition.com/268016.png" style="display:none;">
Skip to content
akuity.iostaticlogo-4e82a3d473fae8e1041846a7b9e1d3cb-1
Akuity.io

[On-Demand Webinar]

Automated Rollbacks in Kargo: Recovering from Failed Deployments

Jesse Suen, co-creator of Argo CD and Kargo and Akuity's co-founder and CTO, demos automated rollbacks in Kargo Enterprise 1.11. When a stage fails verification, Kargo re-promotes the last verified freight through the standard GitOps path, so a bad release reverts in seconds instead of falling to whoever is on call. After a short Kargo overview, Jesse runs live demos: a promotion caught by a Prometheus error-rate check, and the new auto-promotion holds that stop rollback and auto-promotion from fighting. Community Q&A covers database migrations, audit trails, Argo CD self-healing, ApplicationSets, Terraform, and when to use AKI incident remediation instead.

See automated rollbacks in Kargo Enterprise

Key takeaways

  • Automated rollback restores the last verified version. With rollback on, Kargo tracks a stable freight per stage: the version that last passed verification. When a new promotion fails verification, Kargo re-promotes that freight through the same GitOps path. In the demo, a version with a high HTTP error rate is caught by a Prometheus check and reverted in seconds.
  • Triggers are set per stage. Rollback lives in the project config's promotion policies with a stage selector. You choose whether it fires on failed verification (the default), failed promotion, or both, and whether transient errors count too.
  • Auto promotion holds prevent rollback loops. Also new in 1.11: after a rollback pins a stage to older freight, Kargo holds auto promotion so it can't re-promote a candidate that may still carry the bug. Clearing the hold is manual by design, so a broken version can't slip back in.
  • Rollback is a promotion in reverse. It runs the same steps as rolling forward, so keep promotions idempotent. For steps that aren't safe to repeat, like a database migration, use the new rollback context variable to skip them or the fail step to block the promotion.
  • Kargo rolls back at the stage level, not one workload. Argo Rollouts reverts a single workload's replica set; Kargo reverts the whole stage, including config maps, multiple Argo CD apps, and Terraform, through Git. The two work together; keep Argo Rollouts for traffic shaping.

Helpful Resources:

If you have questions about the Akuity Platform, security and compliance documentation, or would like to request additional information, please contact us at sales@akuity.io or visit https://akuity.io/connect-with-akuity for technical support.