Backing up your server

For any real-world use, you should make sure to regularly backup your Mastodon server.

Overview

Things that need to be backed up in order of importance:

1. PostgreSQL database
2. Application secrets from the .env.production file or equivalent
3. User-uploaded files
4. Redis database

Failure modes

There are two failure types that people in general may guard for: The failure of the hardware, such as data corruption on the disk; and human and software error, such as wrongful deletion of particular piece of data. In this documentation, only the former type is considered.

A lost PostgreSQL database is complete game over. Mastodon stores all the most important data in the PostgreSQL database. If the database disappears, all the accounts, posts and followers on your server will disappear with it.

If you lose application secrets, some functions of Mastodon will stop working for your users, they will be logged out, two-factor authentication will become unavailable, Web Push API subscriptions will stop working.

If you lose user-uploaded files, you will lose avatars, headers, and media attachments, but Mastodon will work moving forward.

Losing the Redis database is almost harmless: The only irrecoverable data will be the contents of the Sidekiq queues and scheduled retries of previously failed jobs. The home and list feeds are stored in Redis, but can be regenerated with tootctl.

The best backups are so-called off-site backups, i.e. ones that are not stored on the same machine as Mastodon itself. If the server you are hosted on goes on fire and the hard disk drive explodes, backups stored on that same hard drive won’t be of much use.

Backing up application secrets

Application secrets are the easiest to backup, since they never change. You only need to store .env.production somewhere safe.

Backing up PostgreSQL

PostgreSQL is at risk of data corruption from power cuts, hard disk drive failure, and botched schema migrations. For that reason, occasionally making a backup with pg_dump or pg_dumpall is recommended.

For high-availability setups, it is possible to use hot streaming replication to have a second PostgreSQL server with always up-to-date data, ready to be switched over to if the other server goes down.

Backing up user-uploaded files

If you are using an external object storage provider such as Amazon S3, Google Cloud or Wasabi, then you don’t need to worry about backing those up. The respective companies are responsible for handling hardware failures.

If you are using local file storage, then it’s up to you to make copies of the sizeable public/system directory, where uploaded files are stored by default.

Backing up Redis

Backing up Redis is easy. Redis regularly writes to /var/lib/redis/dump.rdb which is the only file you need to make a copy of.