Open-source PostgreSQL backup tools compared, 2026

Eight self-hosted tools that put PostgreSQL backups somewhere other than the database server, compared on what matters after the first week: whether anything checks the file is usable, where it can go, what deletes old copies, and who tells you when it stops.

Everything below comes from each project's own README or documentation and from the GitHub and Docker Hub APIs, all read on 2026-10-19; the sources are under each tool. "Not documented" means the project's documentation does not say, not that the feature is missing. We make Dumpling and dumpling-backup, which is in the list; it is marked, and it is the newest and least used tool here.

Which one fits

• You want a dashboard and real restore tests: Databasus. It is the one backup tool here that restores into a throwaway database and reports row counts.
• You back up many kinds of database on several machines: Portabase, for one dashboard with agents next to each database, or db-backup if a configuration file suits you better than a UI.
• You only need rotating dumps on one host: postgres-backup-local, with something else copying the volume off the machine. A backup on the same disk as the database is not off-site.
• You want dumps in your own bucket from a stateless container, cron job or CI: dumpling-backup (ours). It checks every archive as it streams, deletes only files it wrote, and needs no disk. It is new; weigh that.
• You already have backups and need evidence they restore: Restoredrill, described below, alongside whatever makes the backups.
• You run one of the postgres-backup-s3 images: Plan a move. The eeshugerman repository is archived with no PostgreSQL 17 or 18 image, and the schickling image on Docker Hub has not been rebuilt since 2022.

What "verified" means, three levels

• Checksum. Proves the file was not changed after it was written. A dump that was truncated before the checksum was taken passes.
• Archive check. The restore tool must accept the file: pg_restore --list parses a PostgreSQL archive and lists what is in it, and a MySQL dump must end with its completion line. Cheap enough to run on every backup. Does not prove the data loads into a server with your extensions and roles.
• Test restore. The file is restored into a throwaway database and checked. The strongest evidence, at the cost of a database's worth of disk and time per check.

Whichever a tool does, a restore drill by a person, a few times a year, is what proves the whole procedure works.

Not a backup tool, but worth knowing: Restoredrill

Restoredrill: A Go CLI and GitHub Action (MIT, v0.4.0 of 2026-09-03, 113 stars) that takes an existing PostgreSQL or MySQL 8 dump, restores it into a throwaway container, runs checks you define (row counts, SQL assertions, how old the newest data is) and writes a JSON evidence report. It makes no backups itself. Pair it with any of the tools above if an auditor wants proof of restores.

How this list was made

The tools are the self-hosted PostgreSQL backup projects with the most stars or Docker pulls that we found, plus our own. Managed services (SimpleBackups, SnapShooter and others) and physical-backup tools for running your own server (pgBackRest, Barman, WAL-G) are left out: they answer different questions. Star and pull counts are a snapshot; pulls include automated redeploys and say little about how many people use a tool. If something here is wrong or out of date, write to support@getdumpling.dev and it will be corrected with the date.