Data corruption due to unreliable SMB (CIFS) shares

Challenge

Storages that use the SMB (CIFS) protocol without the continuous availability and write-through features can cause data corruption in case of technical outages. This can lead to restore or backup problems.

Cause

If shares operate without the continuous availability feature in the write-though mode, data can end up in the “undefined” state in one of the many caches (RAM, hard disk cache, etc.) between the source and the disk. Many SMB (CIFS) NAS filers handle data on the “best effort” to gain better performance instead of safe data handling. 

Solution

SMB 3.0 continuous availability or transparent failover ensures high availability of the SMB file share.
Write-through capabilities with SMB guarantee that data written to the share will be written to the persistent storage (hard disks, SSDs, etc.). That means that the storage system only acknowledges a successful write after the data is stored successfully to the physical storage.


Requirements for the SMB NAS filer:

• Option A: Windows failover cluster with the file server role installed and the share configured in the “Continuous Availability” mode.
• Option B: NAS storage with at least two controllers supporting SMB 3.0, “Continuous Availability” and “write-through” support.

These requirements are similar to other applications that require reliable storage (e.g. SQL or Hyper-V).

More Information

The Veeam storage level corruption guard (Health Check) can detect corrupted backup files.
See this Microsoft Tech-Community article for more information on SMB share high availability.