Microsoft VSS 10 second timeout

[Hasslehogg]

Hi

I wonder if anyone can help please? We are comparing SRM + Mirrorview A + replication manager to VEEAM replication for our DR. We have approx 100 VM's and our DR site is under 10 mile away connected via a 1GB Lan Extension.

We already have Veeam in place for backups and faster local recoverability on our production sites, but are now looking at DR and replication side of Veeam

We have approx 5 SQL 2005 Std VM's that have high IO requirements running on vSphere 5 U1, we have tested mirrorview with replication manager in the past to get application consistency but sometimes (not always) the VSS timed out depending on the IO of the box at the time, but we could never get it to succeed every time.

I have been told this is a hardcoded Microsoft VSS timeout setting which is set to 10 seconds, so no matter which tool i use, if it uses the Microsoft VSS writer, then it will have the same problem.

So, I assume if the above statement is correct, Veeam will have the same issue not being able to complete the 10 seconds VSS timeout if IO is high?
Has anyone came across this problem or have a workaround please?
Which interface Veeam uses, is it the microsoft VSS writer or the SQL VDI interface as SQL VDI is the better performing of the 2?

I must stress i havent yet tested Veeam replication with our SQL servers, because i was told it wouldnt work, not because of a Veeam problem, but the Microsoft VSS Timeout of 10 seconds on high IO boxes, so if this isnt true or Veeam can get around it then excellent, and we would then consider for DR as well as backups

any advice or info would be greatly appreciated

thanks

[Vitaliy S.]

Hi Michael,

I have been told this is a hardcoded Microsoft VSS timeout setting which is set to 10 seconds, so no matter which tool i use, if it uses the Microsoft VSS writer, then it will have the same problem.

The only timeout I am aware of is a timeout of 60 seconds that Microsoft has hard-coded to keep OS in a frozen state. This state is mandatory if you want to take a snapshot of you VM in a consistent state.

You cannot extend or change this timeout, but what you do control is an application preparation freeze timeout in Veeam B&R. As far as I remember our default value is 900 seconds, so this should be more than enough for any highly transactional application to prepare itself for a freeze operation.

Btw, I have just found the corresponding KB article that you might find useful:
http://www.veeam.com/kb_articles.html/KB1377

Thanks!

[Hasslehogg]

Vitaliy

as always thank you very much for the info

thanks

[Cokovic]

The only timeout I am aware of is a timeout of 60 seconds that Microsoft has hard-coded to keep OS in a frozen state. This state is mandatory if you want to take a snapshot of you VM in a consistent state.

Exchange has a hard coded timeout of 20 seconds as far as i know

[Vitaliy S.]

Maybe, but my point was that it does no matter what timeout Microsoft has for keeping its OS or application in the frozen state, the preparation timeout is a different story which really depends on the backup application you're using. Furthermore, even 20 seconds is a lot for a snapshot creation operation, as it literally takes a couple of seconds to create a snapshot in a healthy environment.

[tsightler]

Furthermore, even 20 seconds is a lot for a snapshot creation operation, as it literally takes a couple of seconds to create a snapshot in a healthy environment.

Well, even healthy environments will see a timeout if there is consistent transactional load. Microsoft VSS looks for a "quiet" time prior to even starting the backup operation, if the server is consistently busy it may not find such a "quiet" time to even start the VSS operation. The VSS operation will end in a timeout error. Some users have reported success with the following registry key:

HKLM\Software\\Microsoft\\Windows NT\\CurrentVersion\\SPP\\CreateTimeout

The "CreateTimeout" key should be added as a DWORD 32-bit, and it's value is in milliseconds. The default value for older Windows editions was 10 minutes, but I think it has been reduced significantly with 2008R2, to something like 1 minute. You might try increasing this value if you are using VSS on a system that is consistently busy. Many users and KB articles for backup products suggest using 1200000 which is 20 minutes.

[Cokovic]

Well, even healthy environments will see a timeout if there is consistent transactional load. Microsoft VSS looks for a "quiet" time prior to even starting the backup operation, if the server is consistently busy it may not find such a "quiet" time to even start the VSS operation. The VSS operation will end in a timeout error. Some users have reported success with the following registry key:

HKLM\Software\\Microsoft\\Windows NT\\CurrentVersion\\SPP\\CreateTimeout

The "CreateTimeout" key should be added as a DWORD 32-bit, and it's value is in milliseconds. The default value for older Windows editions was 10 minutes, but I think it has been reduced significantly with 2008R2, to something like 1 minute. You might try increasing this value if you are using VSS on a system that is consistently busy. Many users and KB articles for backup products suggest using 1200000 which is 20 minutes.

This sounds really interesting as i do have actually a support case with Veeam and Microsoft running regarding VSS issues on our new Exchange 2010 DAG cluster. I'm going to take a closer look at this reg key. Thanks for this info Tom.

[Hasslehogg]

Tom

thanks for that info, i will try that if i run into problems

thanks again