Question

A CloudOps engineer maintains a runbook using AWS Systems Manager Automation to update a configuration file across 500 EC2 instances. During a recent execution, a bad change was pushed to all instances simultaneously, causing a widespread outage before anyone noticed. The engineer must modify the Automation execution so that changes are applied gradually and automatically stop if too many instances fail. Which approach BEST meets this requirement?

Accepted Answer

Run the Automation document with rate control by specifying a concurrency (MaxConcurrency) and an error threshold (MaxErrors) targeting the fleet.

Answer

Split the fleet into separate SSM Run Command invocations and manually run them one instance at a time.

Answer

Attach a CloudFormation rolling update policy to the Automation document to batch the instance updates.

Answer

Enable State Manager to reapply the configuration every 30 minutes so failures are eventually corrected.