Question

A retail company runs its checkout service on a regional managed instance group (MIG) behind an external HTTP(S) load balancer. The team uses a rolling update to deploy new instance template versions. During a recent release, error rates spiked several minutes after the update completed, and operators had to manually recreate instances from the previous template, extending the outage. The architect wants a release process that automatically limits exposure to a bad version and enables fast, low-risk rollback with minimal manual steps. What should the architect recommend?

Accepted Answer

Use a canary rolling update by adding the new instance template as a second version in the MIG with a small target size, monitor health, and increase the new version's size only after validation—rolling back by removing the canary version.

Answer

Increase the MIG's maxUnavailable to 100% so the rolling update replaces all instances at once, reducing the time the two versions coexist and simplifying rollback.

Answer

Disable autohealing during deployments so that faulty new instances are not automatically recreated, allowing operators more time to investigate before rolling back.

Answer

Deploy the new template to a separate unmanaged instance group and manually swap the load balancer backend once testing is complete.