docs.vmst.io
About

Backups

We backup the persistent data storage of vmst.io multiple times per day/week and to different locations.

Databases

Posts made to vmst.io are stored in backend PostgreSQL databases with ValkeyDB used as a replacement for Valkey as a key-value store and timeline cache.

  • The primary backup method for our PostgreSQL database cluster is via Digital Ocean's managed backups, which provide transaction-level rollback functionality in addition to nightly backups.
  • In addition we use pg_dump with some custom scripts that performs additional full database backups every weekend, which is then and replicated twice across geographies.

Data in the Valkey databases is mostly ephemeral but to avoid problems a snapshot taken periodically to restore in the event of a service failure.

Media

  • Media data is stored on S3-compatible storage which is then synced directly to another object store via the rclone utility.
  • This is done using some custom scripts that process each task and then fire off notifications to our backend channels.
  • Media backups currently run every day.
  • Only the latest copy of media data is retained.

Configuration

  • All configuration files for core applications, documentation, and web clients are stored on GitHub with changes committed there before being applied to servers.

Docker Image

We have a customized container image available from both Docker and GitHub container registries used for backup purposes. This image is designed for backup, replication, and maintenance of container-based Mastodon implementations. It includes rclone, s3cmd, PostgreSQL, and Valkey utilities.

  • The version tags for the image represent the version of rclone used.
  • The image is based on Debian Trixie.
  • PostgreSQL utilities are based on version 17.
  • Valkey utilities are based on version 7.

There are /root/backups and /root/scripts directories suitable for mounting and processing relevant files.

In order to use rclone, you'll need to mount an existing rclone.conf file to /root/.config/rclone/rclone.conf.

AI/ML

Large language models, like ChatGPT, is a type of "artificial intelligence" program designed to understand and generate human-like text by processing vast amounts of written language. While the creation and use of large language models are not inherently unethical or illegal, there are many important questions about how companies obtain their data to train their models and how that data is used.

Bots

If you're interested in starting a Bot account, we encourage you to reach out to a member of our staff ahead of time to make sure that it's an appropriate fit for our site. nce created, it must be clearly designated as a Bot in the account bio and have an identification of who maintains it and how they can be contacted.