MS says VSS can corrupt data under certain circumstances

[wentum]

Hello,
today I came across a posting at the german http://www.heise.de website. ( http://www.heise.de/newsticker/meldung/ ... 68641.html )

It basically says that MS has found a bug in VSS which corrupts data under certain circumstances.
Here is a link to the MS posting ( http://support.microsoft.com/kb/2748349#appliesto)

And now I'm asking myself if we as Veeam User are affected, too?

Ciao
Joerg

[Gostev]

Hi Joerg, no - Veeam users are not affected by this bug. While our product does use VSS to quiesce file system and applications (in preparation for VMware snapshot), we do NOT use the actual VSS snapshot (where this corruption sits) to pull the data from. Instead, the data is read from the actual disk image (from VMware snapshot).

Specifically, referring to the KB article, this is what Veeam does NOT do:

the data that is read ... from a shadow copy

Thus, Veeam users are safe. On the other hand, the article is correct stating every legacy agent-based backup solution that uses VSS might be affected, because those do pull data direct from VSS snapshots.

Thanks!

[joncz]

Gostev,

Can you clarify the susceptibility to Veeam on Hyper-V?

Thx.

[Gostev]

Hi Jon, Windows Server and Hyper-V versions that we support simply do not have this bug, as per the affected OS list in the above-linked KB article. Thanks!

[Spex]

Hi Joerg, no - Veeam users are not affected by this bug. While our product does use VSS to quiesce file system and applications (in preparation for VMware snapshot), we do NOT use the actual VSS snapshot (where this corruption sits) to pull the data from. Instead, the data is read from the actual disk image (from VMware snapshot).

Are you absolute sure, that this is correct?

MS VSS works with copy on write - so the consistent state during a valid snapshot consists from the unaltered blocks on the original disk plus the changed blocks that have been written to the vss snapshot location.
Since there is a small time gap between vss snap and vmware snap there are blocks written to the vss storagearea that are important for the consistent state.

[Gostev]

Since there is a small time gap between vss snap and vmware snap there are blocks written to the vss storagearea that are important for the consistent state.

I am not sure what implementation you are describing here, but this is not how Veeam works. In our case, VMware snap happens before VSS snap (but once the whole system is fully quiesced and ready for that VSS snap).