Before you begin
- The Administrator role for Backup & Restore
- Access to the Atlas resource group in the Azure portal for database restores and for Recovery Mode
Creating a backup
- In the sidebar, open Administration → Platform → Backup & Restore.
- Under Generate backup, leave Scope on Current tenant only (All tenants is available only to platform administrators).
- To protect the file, turn on Encrypt with password and enter a password of at least 8 characters. The backup is encrypted with AES-256-GCM, and Atlas cannot recover the password for you.
- Under Included modules, keep every module ticked or clear the ones you do not need: Tenants, Settings, Host pools, Deployments, Fortify, Optimizations, Automations, Audit logs, Activity & alerts, and Vault (redacted).
- Click Generate backup. A progress bar and live log show each stage; when it finishes, click Download backup and keep the .atlasbackup file somewhere safe. Copy checksum gives you its SHA-256 fingerprint.
Backup files are deleted from Atlas 24 hours after they are created. Download any backup you want to keep. Vault secrets are never included: the Vault module holds only metadata and Key Vault references.
Automatic backups and history
Platform administrators also see an Automatic backups card: turn on Run backups automatically and set Hours between backups (1 to 24, 24 by default) to take a full backup of all platform data on a schedule. The interval tops out at 24 hours because every backup is deleted 24 hours after it is created, so a longer gap would leave you with no backup to restore. The card shows when the last automatic backup ran. Every backup is listed under Backup history with its date, tenant, size, status, and time until it expires, plus Download, Copy SHA-256 checksum, and Delete actions.
Restoring from a backup
Under Restore from backup, pick a completed backup From history or switch to Upload file and drop in a .atlasbackup file (enter its password if it is encrypted). Leave every Restore scope box clear to restore everything, or tick just the parts you need: Host pool configuration, ADMX / Group Policy baselines, Integration credentials & settings, Alert thresholds & notifications, Automations & schedules, or User & RBAC assignments. Click Preview (dry-run) to see how many rows would be written and which would overwrite existing data, then click Restore and confirm.
A restore overwrites matching rows in the active tenant and cannot be undone; rows that are not in the backup are left alone. The Vault module and the tenant registry are never restored. Atlas checks the file's integrity first and warns you if the backup failed its health check.
Database backups in Azure
Atlas keeps its data in an Azure Database for PostgreSQL flexible server in your subscription, and Azure backs it up automatically. Backups are kept for 7 days unless you chose a longer retention (up to 35 days) when you deployed from the Azure portal, and geo-redundant backup can only be turned on when the server is first created. Point-in-time restore is done from the database server in the Azure portal, within the retention window.
Viewing a VM's backup status
A VM's detail page includes a Backup card that shows whether the VM is protected by Azure Backup. It reads your Recovery Services vaults with your tenant's own app registration and shows the status (Protected, Protection Stopped, Not Protected, No Vault, Permission Required, or Error), the policy, the last backup, and the recovery points, each tagged Snapshot, Vault, or Archive.
This is a status view. Backup policies are configured in Azure Backup, and VM restores are performed there.
Recovery Mode and the startup page
If an update leaves the app broken, Recovery Mode makes Atlas reinstall its current version on the next restart. It is not a switch in the app: set the RECOVERY_MODE application setting to true on the App Service (see the Recovery Mode guide), then set it back to false once the app is healthy. The Recovery Mode Sentinel card on Monitoring → Atlas Health shows when a recovery ran and lets you clear the record afterwards. While the app is starting, or after a crash, the startup page shows the current phase, stage, and progress with a Reload app button and a Health probe link, even before the main app has loaded.