CloudConsole

UPDATE_ROLLBACK_FAILED

Updated 17 November 2026 · How we write and check guides

Short answer

An update failed, CloudFormation tried to roll back to the previous state, and the rollback itself failed for at least one resource, usually because something was changed or deleted outside CloudFormation or the stack's role lacked a permission. Fix that resource and run continue-update-rollback; if it cannot be fixed, continue the rollback while skipping that resource.

Fix and continue the rollback, in 3 steps: 1. Correct the underlying problem; 2. Ask CloudFormation to retry the rollback…
The diagnosis and fix steps from this page, in the order to try them; commands and details are below.

What happened

The sequence is always the same: UPDATE_IN_PROGRESS → some resource fails → UPDATE_ROLLBACK_IN_PROGRESS → a resource cannot be returned to its old configuration → UPDATE_ROLLBACK_FAILED. In this state the stack cannot be updated until the rollback is completed. Your resources are in a mixed state: some at the new configuration, some at the old one, one or more in between.

Find the resource that blocked the rollback, not the one that started it:

aws cloudformation describe-stack-events --stack-name orders-prod \
  --query "StackEvents[?ResourceStatus=='UPDATE_FAILED'].[Timestamp,LogicalResourceId,ResourceStatusReason]" \
  --output table

The most recent UPDATE_FAILED events (during the rollback) are the ones to fix.

Typical reasons a rollback fails

Reason in the eventWhat is going on
Resource not found / does not existSomeone deleted it manually (a security group, a Lambda version, a parameter); CloudFormation cannot restore what it cannot see
Access denied for the stack's service roleThe role was changed during the update, or never had permission to restore the old configuration
ECS service did not stabiliseThe previous task definition no longer starts (image deleted, secret removed)
Limit exceededRolling back needs to create a resource and a quota is full
Custom resource did not respondA Lambda-backed custom resource crashed or never sent its response; CloudFormation waits until the timeout

Fix and continue the rollback

  1. Correct the underlying problem: recreate the deleted resource with the same name, restore the role's permissions, push back the missing image, or fix the custom resource function.
  2. Ask CloudFormation to retry the rollback:
    aws cloudformation continue-update-rollback --stack-name orders-prod
  3. Wait for it to finish and confirm the status is UPDATE_ROLLBACK_COMPLETE:
    aws cloudformation wait stack-rollback-complete --stack-name orders-prod
    aws cloudformation describe-stacks --stack-name orders-prod \
      --query 'Stacks[].StackStatus' --output text

If the stack's service role is itself the problem, pass a working one with --role-arn.

When you cannot fix the resource: skip it

You can tell CloudFormation to stop trying to roll back specific resources. It marks them as rolled back and finishes:

aws cloudformation continue-update-rollback --stack-name orders-prod \
  --resources-to-skip OrdersService OrdersTaskSecurityGroup

For resources in nested stacks, use NestedStackName.LogicalId. Be careful: skipped resources are now out of step with what CloudFormation believes. The real resource keeps whatever state it was left in, and a later update may fail or behave unexpectedly. After recovering, run drift detection and fix any differences by hand or with a follow-up update. Only skip resources you have checked.

Avoid it next time

  • Do not change stack-managed resources outside CloudFormation; detect it with drift detection (see detect-stack-drift).
  • Never remove permissions from the stack's service role in the same update that relies on them.
  • Keep old container images and Lambda artefacts long enough to roll back to.
  • Make custom resources always send a response, including in error paths.
  • Preview risky updates with change sets before executing them.

Questions

How do I fix UPDATE_ROLLBACK_FAILED in CloudFormation?

Find the resource whose rollback failed in the stack events, fix the cause, then run aws cloudformation continue-update-rollback. If it cannot be fixed, continue the rollback with --resources-to-skip.

Is it safe to skip resources when continuing a rollback?

It finishes the rollback, but the skipped resources may not match the template afterwards. Check them and run drift detection before the next update.

Can I update a stack in UPDATE_ROLLBACK_FAILED?

No. You must first get it to UPDATE_ROLLBACK_COMPLETE with continue-update-rollback.

CloudConsole CodePipeline screen on a phone, for following up on UPDATE_ROLLBACK_FAILED away from your desk
When this happens away from your desk

A stack stuck in rollback failure blocks every following deploy, so you want to know the moment it happens. CloudConsole, a phone app for AWS, is built to show deployment failures and the alarms that follow on your phone, so you can judge the impact before you are back at a desk.

Did it start after a deploy? What changed? on the app’s alarm screen lists the deploys and configuration changes from the hour before the alarm fired.

About the app → · Download on the App Store

Sources