Problem

Coder workspaces on AWS currently lack persistent state across workspace stop/start cycles beyond basic EC2 stop functionality. Users need the ability to snapshot their workspace state and resume from selected snapshots.

Proposed Solution

Create a Terraform module that enables AMI-based snapshots of EC2 instances when workspaces are stopped, with user-specified labels and selection options for workspace start.

Implementation Steps

1. Create AMI snapshots on workspace stop

   • Use aws_ami_from_instance resource triggered by coder_workspace.me.transition == "stop"
   • Tag AMIs with workspace owner, name, label, and creation timestamp

2. Add user parameters for snapshot control

   • Implement coder_parameter resources for:
     • Enable/disable snapshot functionality
     • Custom snapshot labels
     • Option to start from previous snapshots

3. Retrieve available snapshots

   • Use aws_ami data source with filters on Coder-specific tags
   • Format snapshot metadata for selection dropdown

4. Modify instance creation

   • Create a local.ami_id variable that selects user snapshot or default AMI
   • Update aws_instance resource to use this dynamic AMI ID
   • Add lifecycle { ignore_changes = [ami] } to prevent Terraform conflicts

5. Implement optional cleanup

   • Consider aws_dlm_lifecycle_policy for snapshot retention management
   • Add configuration options for retention periods and counts

Key Considerations

• Users need IAM permissions for AMI creation/management
• AMI creation takes time; handle workspace stop gracefully
• Implement cost control through snapshot cleanup policies
• Ensure proper tagging for snapshot organization