Backup policies
A policy should identify the protected database, backup destination, schedule, retention behavior and any verification expectations. Use separate policies when workloads have materially different recovery or storage requirements.
- Choose the registered device and approved database source.
- Select the databases covered by the policy.
- Define the backup frequency and storage location.
- Review retention and verification settings.
- Save the policy and confirm it is delivered to the intended agent.
Scheduling
Schedule backups around application activity, database maintenance and available storage or network capacity. A schedule should be frequent enough for the business recovery objective but realistic for database size and infrastructure performance.
- Confirm host and SQL Server time settings.
- Avoid overlapping heavy maintenance tasks where possible.
- Check that the agent remains online during the planned window.
- Investigate missed scheduled jobs rather than relying only on later successful backups.
Retention
Retention controls how long completed backup sets remain available. It does not replace capacity planning, offline copies or legal retention decisions. Choose retention according to recovery needs, storage limits and the organization’s approved policy.
Before shortening retention, confirm that older recovery points are no longer required and that deletion behavior is understood.
Backup monitoring
Review job state, start and finish times, error details, file availability and verification results. A green status should still be correlated with expected database coverage and recovery-point timing.
- Investigate failed, missed, unusually short or unusually long jobs.
- Confirm expected databases continue to appear in policy scope.
- Review agent heartbeats and connectivity.
- Escalate repeated failures with exact timestamps and error messages.
Operational review
Perform a regular review of protected devices, active policies, recent jobs, available storage and verification status. Document exceptions and ownership rather than allowing warnings to become permanent background noise.