Tag Archives: Data De-duplication

Data corruption issue with NTFS sparse files in Windows Server 2016

Microsoft has released a new patch KB4025334 which prevents a critical data corruption issue with NTFS sparse files in Windows Server 2016.  This patch will prevent possible data corruptions that could occur when using Data Deduplication in Windows Server 2016. However this update also a remedy prevent this issue in all applications and Windows components that leverage sparse files on NTFS in Windows Server 2016.

Although this is an optional update, Microsoft recommends to install this KB to avoid any corruptions in Data deduplication although this KB doesn’t provide  a way to recover from existing data corruptions. The reason being is that NTFS incorrectly removes in-use clusters from the file and there is no way to identify what clusters were incorrectly removed afterwards. Furthermore this update will become a mandatory patch in the “Patch Tuesday” release cycle in August 2017.

Since this issue is hard to notice, you won’t be able detected that by monitoring the weekly Dedup integrity scrubbing job. To overcome this challenge this KB also includes an update to chkdsk which will allow you to identify which files are already corrupted.

Identifying corrupted NTFS sparse files with chkdsk in KB4025334

  • First, install KB4025334 on affected servers and restart same. Keep in mind that if your servers are in a failover cluster this patch needs to be applied for all the servers in your cluster.
  • Execute chkdsk in read-only mode which is the default mode for chkdsk.
  • For any possibly corrupted files, chkdsk will provide an output similar to below. Here 20000000000f3 is the file id and make a note of all the file ids of the output.
The total allocated size in attribute record (128, "") of file 20000000000f3 is incorrect.
  • Then you can use fsutil to query the corrupted files by their ids as per below example.
D:\afftectedfolder> fsutil file queryfilenamebyid D:\ 0x20000000000f3
  • Once you run above command, you should get a similar output like below. D:/affectedfolder/TEST.0 is the corrupted file in this case.
A random link name to this file is [file://%3f/D:/affectedfolder/TEST.0]\\?\D:\affectedfolder\TEST.0