“BACKUP failed to complete the command” is a summary, not the root cause. Preserve the messages immediately before it, then inspect the SQL Server error log and the backup application log for the operating-system or database error that ended the operation.
Last updated: November 9, 2026.
EXEC sys.sp_readerrorlog 0, 1, N'Backup';
EXEC sys.sp_readerrorlog 0, 1, N'Operating system error';
SELECT TOP (20)
database_name, backup_start_date, backup_finish_date,
type, is_copy_only
FROM msdb.dbo.backupset
ORDER BY backup_start_date DESC;Run the log queries on the Database Engine instance that performed the backup. A missing row in msdb.dbo.backupset confirms that the backup did not finish, but it does not identify why it failed.
Separate database errors from destination errors
Microsoft documents sys.sp_readerrorlog as the supported way to filter current and archived SQL Server or Agent logs. Record the full error number, path, time, and any preceding I/O message. Then check free space, path length, antivirus interference, storage availability, and whether another process holds the file.
For disk or network destinations, the backup destination guidance says the SQL Server service account needs read and write access. Your interactive Windows account and the account running SSMS are not the identities that write the backup.
Reproduce with a small controlled backup
After preserving the failed job evidence, test a small database to a new local path owned by the SQL Server service account. If local succeeds and the share fails, focus on share and NTFS permissions, name resolution, and the service identity. If every destination fails, review database state, storage health, and the first SQL error. Compare the job owner and proxy configuration when SQL Server Agent launches a third-party backup command.
Do not repeatedly restart a mirroring or availability workflow until a standalone full backup succeeds. A retry can overwrite useful logs or leave multiple large partial files. Once the cause is fixed, create a fresh backup with checksum, verify it, and perform a test restore before using it to seed another replica.
