# Updates and maintenance

Source: /docs/raptor-mcp/updating-and-maintenance

Raptor Gateway update, repair, PostgreSQL migration, backup and maintenance guidance for customer-managed deployments.

## Choose Updater or Full Setup

Download the current release from the authenticated Raptor Portal. Installed Gateways do not update themselves. An administrator runs the local Setup transaction during an agreed maintenance window.

Use Updater when the existing installation has the required dependencies and PostgreSQL core configuration. If Setup requests missing dependencies, a PostgreSQL configuration transfer or recovery of an interrupted installation, use Full Setup. A package at the installed version offers Repair Raptor; a newer eligible package offers Update Raptor.

Some older installations need a configuration and identity migration before ordinary updates are available. Full Setup checks the existing service, account, HTTPS and database. It preserves the existing Gateway identity and connection settings where that migration is supported. If Setup reports that an assisted migration is required, retain the installation and contact support.

## Prepare the maintenance window

Record the installed version, Gateway address, installation folder, service account and actual storage locations. Notify users and account for active reloads, scheduled notifications, workflow jobs and writeback before stopping the Gateway.

Retain a tested recovery plan for the whole deployment: PostgreSQL, repository files, configuration, certificates and encryption material. The updater creates a Gateway code backup, but deliberately excludes live PostgreSQL data. That backup is not a database backup or a complete disaster-recovery copy.

- Use a database-aware backup or a coordinated stopped snapshot. Do not copy a live PostgreSQL data directory as an ordinary file backup.
- Back up a separately located repository as well as the database that holds its metadata.
- Keep the Gateway encryption key with the protected recovery material. Restoring encrypted state without its original key does not restore access to its secrets.
- Retain the previous release identity, update logs and rollback files until the agreed recovery period ends.

## Run and verify the update

Run the downloaded executable as an administrator on the Gateway machine. Review the detected installation before proceeding. Do not extract a replacement ZIP over a running installation or rebuild a package from an old extracted Setup directory.

1. Open Updater or Full Setup and wait for the installation check. Confirm that the selected installation is the one you intend to maintain.
2. Accept the terms and choose Update Raptor or Repair Raptor. For a supported older-installation migration, supply the requested Windows service credentials and database password when required.
3. Keep Setup open until it reports Raptor is ready. The maintenance transaction checks the Gateway, service, certificate trust and Experience Hub before completion.
4. Open Experience Hub and confirm the expected licence, Qlik connections and users. Check one permitted MCP operation with a normal user, then verify the features your deployment uses.
5. Review schedules and external integrations before returning the Gateway to normal service.

- [Connect AI clients](/docs/raptor-mcp/connecting-ai-clients)
- [InsightOps](/docs/raptor-mcp/insightops)
- [Writeback and workflows](/docs/raptor-mcp/writeback-and-workflows)

## If maintenance fails

Choose Open installation log and retain the reported operation details. Setup can restore an earlier code state after a failed update, but an error or a stopped process does not prove that recovery completed. Confirm the service, version, Hub and user access before reopening the installation.

If Full Setup offers Recover Raptor, use that recovery flow with the original Windows service account. Do not delete journals, quarantine folders or database files to make installation checks pass. If recovery remains required, keep users out and provide the logs to support.

After business writes have resumed, restoring an older database snapshot can discard those changes. Stop writes and agree a recovery or reconciliation plan before returning to an earlier data snapshot.

## Qlik and non-Windows components

Gateway maintenance and Qlik configuration are separate operations. Verify the Qlik-hosted assistant, status extension and sign-in bridge when your deployment uses them. Any replacement of Raptor-owned Qlik assets must use the supported transaction and retain its rollback evidence; customer-created extensions remain outside that operation.

Customer-managed Linux and Docker deployments need an administrator-controlled rollout, an exact package or image identity and retained state backups. Keep the previous configuration and storage mapping with the rollback reference. A PostgreSQL core migration requires its own offline plan; a Windows Updater is not a Linux migration tool.

## Related Links

- [Install and configure](/docs/raptor-mcp/install-and-configure)
- [Troubleshooting](/docs/raptor-mcp/troubleshooting)
- [Qlik connections](/docs/raptor-mcp/qlik-connections)
- [Advanced analytics and SSE](/docs/raptor-mcp/advanced-analytics)