Simple script to create backups of labeled VMs in Harvester, and provide smart aging of backups.

harvester-auto-backup scans your harvester cluster looking for for a specific label, and when it finds it on a VM creates a backup of that VM. It then checks the age of any existing backups and deletes them according to the retention policy.

• Harvester cluster with kubectl access
• Configured backukp target in Harvester
• Ability to edit the yaml of VMs in Harvester

A valid backup target must be configured in Harvester. This can be done by following the Harvester documentation.

Install the CronJob by applying the manifest from this repository:

kubectl apply -f https://raw.githubusercontent.com/jaevans/harvester-auto-backup/main/deploy.yaml

Or download and modify deploy.yaml to suit your needs.

auto-backup takes the following arguments, specified in the CronJob resource, to configure its behavior:

The CronJob is configured to run every night at 02:00 UTC by default. You can modify this by editing the deploy.yaml file and changing the CronJob's spec.schedule field.

The VM label is configured in the CronJob's spec.jobTemplate.spec.template.spec.containers[0].args[0] field. You can modify this by editing the deploy.yaml file and changing the field to any valid label you want to use to mark VMs for backup.

A "smart" retention policy is applied to existing backups of the VMs. The retention policy is as follows:

• Keep all backups for the past week
• Keep one backup per week for the past 3 months
• Keep one backup per month for the past year
• Delete all backups more than a year old

These thresholds are currently hardcoded in the auto-backup script, and can be adjusted by modifying the retention variables keep_weekly_after, keep_monthly_after, and delete_after. All backups yunger than keep_weekly_after are kept, weekly backups younger than keep_monthly_after are kept, and then monthly backups are kept until delete_after.

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.