How to backup and restore a charmed deployment¶

This guide covers the complete process of backing up and restoring a Landscape deployment to a new Juju model using the Landscape Server charm and Charmed PostgreSQL. This guide uses Charmed PostgreSQL 14 because it is compatible with the 24.04 LTS version of the Landscape Server charm.

You might use this process when migrating your Landscape deployment to new infrastructure, performing a major upgrade, or moving to a different cloud environment.

For general backup and restore operations within an existing deployment, see Backup and restore.

Backup¶

Prepare your deployment¶

Export the current Juju model configuration:
```
juju export-bundle | tee bundle.yaml
```
This writes the bundle to bundle.yaml.

Back up the service.conf file on each landscape-server unit, replacing <unit-id> with the actual unit ID (e.g., 0, 1, 2):

mkdir -p service-conf
juju ssh landscape-server/<unit-id> -- sudo cat /etc/landscape/service.conf > service-conf/<unit-id>.conf

Record the PostgreSQL operator password:
```
juju run postgresql/leader get-password username=operator
```
You need this later to dump and restore the databases.
Get the PostgreSQL leader unit’s IP address:
```
juju status
```
Pause all of the Landscape Server units and keep them paused for the entire backup and restore procedure:
```
juju run landscape-server/0 pause
```
Repeat for each landscape-server unit (landscape-server/1, landscape-server/2, etc.).

If the service resumes during the process, connected clients may lose data.

Backup the databases¶

Create a backup directory on the PostgreSQL leader unit:

juju ssh postgresql/leader -- "sudo mkdir -p /var/snap/charmed-postgresql/current/backup && sudo chmod 755 /var/snap/charmed-postgresql/current/backup"

Set environment variables for the database credentials:
```
export PG_PASSWORD="<operator-password>"
export PG_HOST="<postgres-ip>"
```
Replace <operator-password> with the password you recorded and <postgres-ip> with the PostgreSQL leader IP.

Dump each database. Repeat this step for each database and each PostgreSQL unit in your deployment:

for DB_NAME in \
    landscape-standalone-main \
    landscape-standalone-package \
    landscape-standalone-account-1 \
    landscape-standalone-resource-1 \
    landscape-standalone-knowledge \
    landscape-standalone-session; do
    juju ssh postgresql/leader -- "sudo PGPASSWORD=$PG_PASSWORD charmed-postgresql.pg-dump -d $DB_NAME -U operator -h $PG_HOST -f /var/snap/charmed-postgresql/current/backup/$DB_NAME.dump -F directory"
done

Confirm that each dump directory contains data:

juju ssh postgresql/leader -- "sudo du -sh /var/snap/charmed-postgresql/current/backup/*/"

Change ownership of the backup directory so it can be copied:

juju ssh postgresql/leader -- "sudo chown -R ubuntu:ubuntu /var/snap/charmed-postgresql/current/backup"

Export the backup files from the PostgreSQL unit to your local backup location:

juju scp -- -r postgresql/leader:/var/snap/charmed-postgresql/current/backup .

Restore¶

Create a new deployment¶

Create a new Landscape model and deploy Landscape the same way you did originally. Keep the Juju configuration consistent with your backup. Start with a single PostgreSQL unit and scale later if needed.

Note

The service.conf file is generated automatically by the Landscape Server charm based on its configuration. The backup you created earlier is for reference only and the new deployment will have different passwords, endpoints, and database connection details. If your original deployment had custom settings in service.conf, you can add them to the new deployment using the additional_service_config charm configuration option.
Wait for the model to finish deploying. All units should reach active status:
```
juju wait-for model <new-model-name> --timeout 3600s --query='forEach(units, unit => unit.workload-status == "active")'
```
This waits up to 1 hour for all units to become active. Adjust the timeout as needed for your deployment.
Switch to the new model:
```
juju switch <new-model-name>
```
Pause Landscape Server on all units:
```
juju run landscape-server/0 pause
```
Repeat for each unit if your deployment has multiple landscape-server units.

Restore the databases¶

Import the database backup files to the PostgreSQL leader unit:
```
juju scp -- -r backup postgresql/leader:/tmp/
```

Copy the backup files into the snap-accessible path:

juju ssh postgresql/leader -- "sudo cp -r /tmp/backup /var/snap/charmed-postgresql/current/"

Get the new PostgreSQL leader unit’s operator password:

juju run postgresql/leader get-password username=operator

Get the PostgreSQL leader unit’s IP address:
```
juju status
```
Set environment variables for the new database credentials:
```
export PG_PASSWORD="<operator-password>"
export PG_HOST="<postgres-ip>"
```
Replace <operator-password> with the new password and <postgres-ip> with the new PostgreSQL leader IP.

Restore each database. Repeat this step for each database and each PostgreSQL unit in your deployment:

for DB_NAME in \
    landscape-standalone-main \
    landscape-standalone-package \
    landscape-standalone-account-1 \
    landscape-standalone-resource-1 \
    landscape-standalone-knowledge \
    landscape-standalone-session; do
    juju ssh postgresql/leader -- "sudo PGPASSWORD=$PG_PASSWORD charmed-postgresql.pg-restore -U operator -h $PG_HOST -d $DB_NAME /var/snap/charmed-postgresql/current/backup/$DB_NAME.dump -c"
done

Scale PostgreSQL to the desired number of units (if needed):
```
juju add-unit postgresql -n <num-units>
```

Wait for all units to become active again:

juju wait-for model <new-model-name> --timeout 3600s --query='forEach(units, unit => unit.workload-status == "active")'

Clear the hash ID databases and regenerate them:

juju ssh landscape-server/leader -- "sudo rm -rf /var/lib/landscape/hash-id-databases/*"
juju run landscape-server/leader hash-id-databases --wait=30m

Resume Landscape on all units:
```
juju run landscape-server/0 resume
```
Repeat for each unit if your model has multiple landscape-server units.

Verify the restore was successful¶

Verify that the deployment is healthy:
```
juju status
```
The landscape-server unit(s) should be active.
Confirm the restore was successful:

Check that all services are running:
```
juju ssh landscape-server/leader -- "sudo lsctl status"
```
Compare the new service.conf with your backed up version to verify configuration:
```
juju ssh landscape-server/leader -- "sudo cat /etc/landscape/service.conf"
```
Remember the service.conf file is automatically generated by the charm and won’t match your backup exactly (it will have different passwords, hosts, etc.).

Check that data was restored by querying the database:
```
export PG_PASSWORD="<operator-password>"
export PG_HOST="<postgres-ip>"
```
```
juju ssh postgresql/leader -- "echo 'SELECT COUNT(*) FROM account;' | sudo PGPASSWORD=$PG_PASSWORD charmed-postgresql.psql -U operator -h $PG_HOST -d landscape-standalone-main"
```
Access the Landscape UI to verify your data is present.

Tear down the old model (optional)¶

After confirming your restored deployment is working correctly and all data has been transferred, you can safely destroy the original model:

juju destroy-model <original-model-name>

Caution

This removes all data from the original model and cannot be undone. Ensure the migration was successful before destroying the model.