Scale down AWS resources during non-work hours to save costs. Inspired by the brilliant kube-downscaler.

• Currently supports the following AWS resource types:
  • EC2 Auto Scaling Groups
  • ECS Services
• Flexible scheduling with timezone support
• Resource exclusion via tags
• Grace periods for newly created resources
• Dry-run mode for testing

1. Install:

pip install .

1. Configure AWS credentials:

export AWS_ACCESS_KEY_ID=your_access_key
export AWS_SECRET_ACCESS_KEY=your_secret_key
export AWS_REGION=your_region

1. Run the downscaler:

aws-downscaler --default-uptime="Mon-Fri 08:00-18:00 America/New_York"

Time definitions accept a comma-separated list of specifications in two formats:

1. Recurring specifications:

<WEEKDAY-FROM>-<WEEKDAY-TO-INCLUSIVE> <HH>:<MM>-<HH>:<MM> <TIMEZONE>

Example: Mon-Fri 07:30-20:30 America/New_York

2. Absolute specifications:

<YYYY>-<MM>-<DD>T<HH>:<MM>:<SS>[+-]<TZHH>:<TZMM>

Example: 2024-04-01T08:00:00-04:00

Resources can be configured using AWS tags:

• downscaler:uptime: Override uptime schedule
• downscaler:downtime: Override downtime schedule
• downscaler:exclude: Exclude resource from scaling
• downscaler:exclude-until: Temporarily exclude until timestamp
• downscaler:downtime-scale: Override downtime scale factor

• --dry-run: Print actions without making changes
• --debug: Enable debug logging
• --once: Run once and exit
• --interval: Loop interval in seconds (default: 60)
• --default-uptime: Default uptime schedule
• --default-downtime: Default downtime schedule
• --grace-period: Grace period for new resources in seconds
• --include-resources: Resource types to manage (comma-separated)
• --exclude-resources: Resource patterns to exclude (comma-separated)
• --downtime-scale: Scale factor during downtime (0-100, default: 0)

Scale down all resources outside work hours:

aws-downscaler --default-uptime="Mon-Fri 08:00-18:00 America/New_York"

Scale down specific resource types:

aws-downscaler --include-resources="asg,ecs"

Contributions and bug reports are welcome! Feel free to open issues, submit pull requests or contact me if you need any support.

This project is licensed under the MIT License.