Handling reboots when running commands

If you use Run Command, a capability of Amazon Systems Manager, to run scripts that reboot managed nodes, we recommend that you specify an exit code in your script. If you attempt to reboot a node from a script by using some other mechanism, the script execution status might not be updated correctly, even if the reboot is the last step in your script. For Windows managed nodes, you specify exit 3010 in your script. For Linux and macOS managed nodes, you specify exit 194. The exit code instructs Amazon Systems Manager Agent (SSM Agent) to reboot the managed node, and then restart the script after the reboot completed. Before starting the reboot, SSM Agent informs the Systems Manager service in the cloud that communication will be disrupted during the server reboot.

Create idempotent scripts

When developing scripts that reboot managed nodes, make the scripts idempotent so the script execution continues where it left off after the reboot. Idempotent scripts manage state and validate if the action was performed or not. This prevents a step from running multiple times when it's only intended to run once.

Here is an outline example of an idempotent script that reboots a managed node multiple times.

$name = Get current computer name
If ($name –ne $desiredName) 
    {
        Rename computer
        exit 3010
    }
            
$domain = Get current domain name
If ($domain –ne $desiredDomain) 
    {
        Join domain
        exit 3010
    }
            
If (desired package not installed) 
    {
        Install package
        exit 3010
    }

Examples

The following script samples use exit codes to restart managed nodes. The Linux example installs package updates on Amazon Linux, and then restarts the node. The Windows Server example installs the Telnet-Client on the node, and then restarts it.