Back up data and rehearse a restore
Documentation for Tutela v3.9.0.
Create a backup, inspect its recorded contents, and rehearse recovery in an isolated deployment. A backup file and an accepted restore request are different from a completed, verified recovery.
Prerequisites
- Administrator access to Settings > Backups.
- An approved backup destination and the permissions required by your installation.
- A separate recovery environment, its operator, and a written plan for returning that environment to its prior state.
- An agreed recovery objective and a list of records to check, such as users, policies and audit evidence.
Create and inspect a backup
- Open Settings > Backups. Review the destination and schedule with the operator. If either source cannot load, retry it before making changes.
- Select Test Destination. A successful check confirms access to that location only. If you edit the destination, run the check again.
- Select Backup Now. Follow its operation status, then locate the result in Backup History. Do not start another backup solely because a response was interrupted; first inspect the operation status.
- Inspect the archive result, contents and excluded stores. Record its identifier, creation time and destination. An unknown record count or storage location remains unknown; do not substitute zero.
- Compare the archive coverage with your recovery plan. Arrange separate recovery for excluded data before treating the archive as a complete recovery point.
The excluded stores include the classifier's model cache. Model artifacts are part of the release, not customer data in the database backup. For the Laya stable release, retain access to the pinned model archive as well as the recorded runtime image digests: the model can be mounted separately from the image. Reprovision those exact artifacts during recovery and verify readiness. Do not assume a database restore recreates the model runtime. The retired System 1 runtime is not a recovery prerequisite.
Read schedule and operation results separately
Scheduler and retention reports the configured schedule and timezone, the next run when supplied, and the last reported run status. A future run time is not a completed backup, and an idle operation does not confirm that the last archive succeeded. Match the last run's backup identifier to its history entry and inspect that archive's result and coverage.
For API integrations, preserve missing schedule or operation timestamps as
missing; do not fill them with the current time. Read last_run.status and its
error alongside the archive record. If operation status is unavailable, use
Retry backup operation status before attempting another backup or restore.
Rehearse recovery
- Confirm that you are signed in to the isolated recovery environment. Restoring overwrites existing data. Check the hostname, archive and target with the operator before confirming.
- Record the target's current state and its own recovery point. Inspect the selected archive's validation result and exclusions.
- In Backup History, select Restore for the intended archive and complete its confirmation once. Restore request accepted means the server accepted an asynchronous operation; remain on the operation status until its outcome is known.
- If the result is Restore outcome unconfirmed, inspect status before retrying. The first restore may already be running. If status is unavailable, stop and involve the operator rather than submitting a duplicate request.
- After recorded completion, sign in to the recovered deployment. Check Service health, selected users and roles, policy configuration, and the audit records included in the archive. Record which stores were excluded and how they were recovered separately.
- Run Verify your installation against the recovered connection. Compare actual recovery time and recovered data with your recovery objectives.
Handle a failed rehearsal
Keep the archive identifier and operation error. Do not overwrite a working deployment with an archive that failed validation or recovery. Have the operator restore the isolated target to its saved state using the agreed recovery procedure, then investigate storage access, archive integrity or the failed store before another attempt.
A failed destination check requires a storage or permission correction. A failed archive creation requires a new successful backup. A failed restore requires a new successful rehearsal; repeating a destination check does not resolve it.
Validate
- Backup History records a successful archive with understood coverage and exclusions.
- The restore operation reports completion, not just acceptance.
- Selected recovered records and a real AI workflow were checked.
- The rehearsal records elapsed time, recovery gaps, and the next rehearsal date.
Next steps
- Include backup results in daily operational checks.
- Use Low disk space when capacity threatens backup completion.
Authorized clean replacement
A clean replacement is different from an upgrade or a restore. Only explicitly authorized disposable installations may have their compute, disks, tenant data and installation-specific backups removed. Confirm resource ownership first and preserve shared infrastructure and unrelated workloads. Provision fresh service secrets and TLS material; do not copy old customer policies, provider connections or demonstration activity. Keep the established first-sign-in bootstrap contract. Customers configure their own policies and providers after deployment.