Recover SQL Server from a Stuck Snapshot Backup Suspension

SUSPEND_FOR_SNAPSHOT_BACKUP = ON freezes write I/O until the metadata backup completes or an administrator turns suspension off. If external snapshot orchestration fails between those steps, use a separate connection to detect and thaw the affected scope.

Last updated: October 21, 2026.

SELECT
    SERVERPROPERTY('IsServerSuspendedForSnapshotBackup') AS ServerSuspended,
    SERVERPROPERTY('SuspendedDatabaseCount') AS SuspendedDatabaseCount;

SELECT db_name, suspend_time_ms, is_write_io_frozen
FROM sys.dm_server_suspend_status;

-- Use this only when the whole server was suspended:
ALTER SERVER CONFIGURATION
SET SUSPEND_FOR_SNAPSHOT_BACKUP = OFF;

-- For one database, thaw only that database instead:
-- ALTER DATABASE [Sales] SET SUSPEND_FOR_SNAPSHOT_BACKUP = OFF;

Confirm the intended scope before running either OFF command. Do not thaw a database while storage tooling is still taking a coordinated snapshot; first stop or fail the external snapshot operation.

Detect suspension from an independent session

Microsoft documents sys.dm_server_suspend_status for SQL Server 2022 and later. It shows each suspended database, the session that initiated suspension, elapsed suspension time, and whether write I/O is frozen. Server properties distinguish a server-wide suspension from database-level suspension.

The recovery session must not depend on the snapshot controller that might have lost connectivity. Keep an administrative path available, and alert on a nonzero suspended count or a suspension duration beyond the expected snapshot window. Test the alert and thaw procedure on a disposable database before relying on it in production.

Build a watchdog around the snapshot workflow

The official snapshot-backup workflow pairs suspension with storage snapshot creation and BACKUP ... WITH METADATA_ONLY. Local error handling is useful, but it cannot run if the controlling process disappears or its connection is lost.

Add an independent watchdog with a conservative deadline. It should verify that no authorized snapshot operation is active, record the incident, issue the correctly scoped OFF command, and page an operator. Protect the watchdog credentials and grant only the permissions it needs. A short automatic timeout can corrupt the coordination window by thawing writes while storage is still capturing the snapshot, so base the threshold on measured worst-case duration plus margin.

Continue with restoring a metadata-only snapshot backup, restoring SQL Server to another host, and understanding SQL Server session timeouts.

Related Web Cheat Sheet guides

Sergey Kornilov

Sergey Kornilov