Identify the service's actual name, executable, account, dependencies and vendor recovery guidance. Record the current settings. If a cluster, application supervisor or another management system owns its lifecycle, use that system's supported recovery process instead of adding competing restart behaviour.

Inspect the recent failure in Event Viewer. Fix an invalid password, inaccessible database or missing file directly. A rapid restart loop does not repair a persistent configuration error.

1. Open services.msc with the required administrative permissions. Select the exact service and open Properties.
2. Open the Recovery tab. Review the actions for the first, second and subsequent failures.
3. Choose an appropriate action for each stage. For a service that safely restarts, allow a bounded initial restart sequence and define the later escalation or no-action state.
4. Set a restart delay that gives dependencies time to recover and avoids a tight loop. Choose the failure-count reset period deliberately rather than copying values from another service.
5. Review whether recovery should also apply to stops reported with an error. This is different from a normal successful stop and depends on how the service reports failure.
6. Apply the settings and arrange a controlled test according to the application vendor's procedure. Do not crash a live data-writing process merely to test the checkbox.
7. Verify the resulting service state, application function and recorded recovery event. Make sure repeated failures produce an alert or operational review.

Record the chosen delay, failure sequence and owner. Microsoft documents a problem involving trigger-start services with continuous recovery and a zero reset period, which reinforces the need to review those settings in context. Use the specific vendor remedy when that scenario matches your system.

• Investigating a Windows service failure with Event Viewer
• Running a scheduled PowerShell task with least privilege
• Finding a Windows Server CPU or memory bottleneck
• Record and review changes to customer hosting services

• Microsoft: performance, event and service data
• Microsoft: service crash and recovery behaviour
• Microsoft: service failure action conditions