Backup, Restores, & Snapshots

An important part of running any production db-backed application is making sure your data is backed up and safe, and if the unspeakable happens, it’s straightforward to restore data and get back up and running.

Postgres databases on Fly.io are treated as Fly Apps, which you can read more about in the docs. What that means is backing up data is an exercise in taking snapshots of the Postgres app’s volumes, then restoring the snapshots to a new database server, verifying the restoration, and connecting the application to the restored database.

Finding the database app instance

If you set up your application using fly launch, the name of your database app might be <app-name>-db, or you might have given it a name during the initial configuration. To figure that out for sure, run the following from the root of the project:

fly info

App
  Name     = my-app
  Owner    = personal
  Version  = 40
  Status   = running
  Hostname = my-app.fly.dev

The Name key is the name of your app. In this case, the database app would be my-app-db. Let’s see if that instance exists by appending the -a flag with the name of the database application:

fly info -a my-app-db

App
  Name     = my-app-db
  Owner    = personal
  Version  = 40
  Status   = running
  Hostname = my-app-db.fly.dev

If that doesn’t work, then try to find the database instance by running:

fly postgres list

NAME                       OWNER    STATUS  LATEST DEPLOY
my-app-db       personal running 37m11s ago

and look for the database instance under the NAME column on the list.

Listing snapshots

Fly.io performs daily storage-based snapshots of each of your provisioned volumes. These snapshots can be used to restore your dataset into a new Postgres application.

Snapshots are volume specific, so you need to identify the volume whose snapshot you want to restore from. You can list your volumes by running the volumes list command with your Postgres app name.

fly volumes list -a <postgres-app-name>

ID                   NAME    SIZE REGION ATTACHED VM CREATED AT
vol_x915grn008vn70qy pg_data 10GB atl    b780ce3d    2 weeks ago
vol_ke628r677pvwmnpy pg_data 10GB atl    359d0e24    2 weeks ago

The number of volumes varies depending on how many database replicas were elected when provisioning the database. One primary database and one replica yields 2 volumes.

Once you have identified which volume to target, you can list its snapshots:

fly volumes snapshots list <volume-id>

ID                  SIZE   CREATED AT
vs_2AjJ4lGqQwDbRfxm 29 MiB 2 hours ago
vs_BAARBQxZKl6JKU04 27 MiB 1 day ago
vs_OPQXXna6kA2Qnhz8 26 MiB 2 days ago

The values under the ID columns are what will be used to restore a snapshot.

Identifying your Postgres image version

Depending on when you created your Postgres cluster, it may be running an older image than the default for newly created clusters. Different Postgres major versions may not be fully compatible, so it’s important to use the same version for your restored cluster.

To see your Postgres image and version, run fly image show.

fly image show -a <postgres-app-name>

MACHINE ID      REGISTRY                REPOSITORY      TAG     VERSION DIGEST                        LABELS                                                                          
e286004f696700  registry-1.docker.io    flyio/postgres  14.6    v0.0.41 sha256:3c25db96357a78e827ca7d fly.app_role=postgres_clusterfly.pg-version=14.6-1.pgdg110+1fly.version=v0.0.41

The values under the REPOSITORY and TAG columns reference the image you’ll use to restore the snapshot. Legacy postgres images use the flyio/postgres repository, while new Postgres Flex images use the flyio/postgres-flex repository. In the above example, the machine is running a legacy flyio/postgres:14.6 image.

Restoring from a snapshot

To restore a Postgres application from a snapshot, simply specify the --snapshot-id argument and the --image-ref argument when running the create command as shown below:

fly postgres create --snapshot-id <snapshot-id> --image-ref <image-version>

? App Name: my-app-db-restored
Automatically selected personal organization: Awesome Person
? Select region:  [Use arrows to move, type to filter]
? Select region: sjc (San Jose, California (US))
? Specify the initial cluster size: 2
? Select VM size: shared-cpu-1x - 256
Creating postgres cluster my-app-db-restored in organization personal
Postgres cluster my-app-db-restored created
  Username:    postgres
  Password:    <redacted>
  Hostname:    my-app-db-restored.internal
  Proxy Port:  5432
  PG Port: 5433
Save your credentials in a secure place, you won't be able to see them again!

This provisions and launches a new Fly Postgres database server with the specified snapshot and the same image as your existing database.

Connect the Restored Database

Detach the web application from the current Postgres cluster:

fly postgres detach my-app-db

Then attach it to the new cluster:

fly postgres attach my-app-db-restored

Now your application is pointing at the restored database.