Database Snapshot Backups

In addition to our existing native backups (backup dumps), we are excited to introduce snapshots (also called disk backups) for all our database offerings (MySQL, PostgreSQL, MongoDB, MinIO and Redis).

This was made possible and is powered by the Kubernetes CSI driver and works across currently supported providers: Northflank PaaS, GCP, AWS, bare-metal (K3s Longhorn) and Azure.

Native vs Snapshot Backups

Native (Dump) backups make use of dumping tools for the datastore (e.g. MongoDB uses mongodump).

• Backup creates a full copy of the database
• Well-suited if the backups need to be inspected or exported
• Can be imported via URL or file upload
• Restores do not lead to a database container restart but the database will be unreachable whilst restoring

Snapshot (Disk) backups take a snapshot of the whole volume at the point of the backup.

• If multiple backups are created, they rely on the previous backup as only the difference between snapshots is stored
• Great for recurring/scheduled backups as they only store the difference of the file system since the last snapshot
• Cannot be inspected or exported
• Restoring a disk backup creates a new disk from the latest snapshot, once the database is restarted with the new disk your database is resumed exactly how it was when the snapshot was taken

We have already seen users implementing Northflank cron jobs to trigger regular snapshots and hourly native backups via the Northflank API, excited to see how this evolves!

Other features & fixes

• Added handling for rendering containers with unknown status and improved sorting of the container list to always list running containers first
• Changing service template during creation no longer wipes previously entered data
• Added support for comments on the environment editor
• Fixed an issue where login page required a refresh if Google SSO was used with an invalid account
• Fixed an issue where service environment editor would be stuck on load if a user was missing secret groups permissions
• Fixed service header responsiveness
• Fixed an issue where sign up with GitLab would fail
• Fixed an issue where team members missing the billing read permission could not create new resources
• Fixed an issue where some builds configured with Buildkit would fail
• Fixed sorting of branches on the build selector
• Fixed the environment editor lagging on some lower specification machines, or those under CPU throttling conditions