Recovery Planning
Protect recovery material and prepare for service or data recovery
Table of Contents
About this guide
Agree on a recovery procedure during deployment and confirm it before an update or network move. Recovery has several distinct parts: restoring service, reverting software, and restoring data. Each needs the appropriate procedure and retained material.
Protect recovery material
The appliance’s secret store uses recovery material that cannot simply be recreated after it is lost. Have the deployment administrator place the required keys and recovery records in an approved, protected location outside the appliance. Verify the authorized recovery staff can access them during an outage.
The recovery record should also cover administrator access, certificate renewal, provider credentials, custom model configuration, network settings, and the installed release. Keep secret values out of ordinary support tickets, screenshots, and shared documentation.
Define the backup scope
| Item | Why it matters |
|---|---|
| Application database | Accounts, conversations, metadata, and other application records. |
| Vector database and document storage | Retrieval indexes and the source data needed for document workflows. |
| Secret-store state and recovery material | Access to the credentials and keys required to start services. |
| Manager configuration and custom overlays | Local accounts, custom models, and application settings as applicable to the deployment. |
| Certificates and network configuration | Trusted access at the correct addresses after recovery. |
| Release bundles and model files | Software and weights needed to rebuild in an isolated environment. |
Have your infrastructure and Ask Sage deployment teams define the exact backup commands, storage targets, consistency requirements, retention, and restore order for your installation. Record the acceptable data loss and outage duration, then validate a restore in an appropriate recovery environment.
The Appliance Manager does not provide a general full-appliance Backup/Restore wizard in the console described by these guides. Release rollback and model-catalog snapshots are not complete backups. An installation SSD does not contain the conversations or datasets created after installation.
Recover from a service outage
- Record the last known working state and the event preceding the outage: power loss, network change, update, or model load.
- Check power and network connectivity to each expected node.
- Open the Manager and inspect Overview, Model Management, and Live events.
- If services are intentionally stopped, use Start Ask Sage and wait for readiness.
- If core services are healthy but chat fails, confirm a model is ready and test it directly through the workspace.
- If the Manager is unavailable or the secret store remains sealed, use the platform-specific console/recovery runbook supplied at handover or contact support.
Do not reinitialize the secret store, reinstall the appliance, or delete persistent data as a general troubleshooting step. Those actions can remove information needed to recover the existing deployment.
Console recovery
Some deployments include the asksage-console terminal menu. An authorized administrator can access it through the supplied console or SSH procedure and use its Lifecycle, Network, Recovery, and Information functions.
Use the console runbook for your hardware and installed release, including the appropriate power-on and recovery sequence. Remote hardware management credentials and recovery keys should remain available outside the failed system.
For an update regression, see Updates and rollback. For deleted or damaged data, involve the recovery owner and use the validated data-restore procedure rather than a software rollback.