Sign in
 
HomeWindows Server 2012Windows Server 2008 R2Windows Server 2003LibraryForums
Windows Server TechCenter >
> 
> 
Time sync issue event id 51 on Hyper-V VM's

Answered Time sync issue event id 51 on Hyper-V VM's

  • Monday, January 28, 2013 1:18 PM
     
     
    Sign In to Vote
    0

    Hi,

     Hyper-V VM's are having timesync issues, Event 51 Time-Sevice is reported during the whole day/night. This results in the problem that time is always 1~10 seconds off. This is generating problems with some applications.

    We are using Domain structure for time and have disabled the  Hyper-V Time integration service for the VM's

    Our env consists of a root- and child domain. The Root domain PDC is a VM and has no event ID 51 error. This root PDC syncs time with an internetsource.

    The PDC of the child domain (physical hardware) is syncing time with the Root domain PDC and has no event ID 51 error.

    We have  some Hyper-V host, installed with 2008r2 sp1 in the childdomain. These hosts sync time with the physical child domain DC's. The VM's Running on these Hyper-V hosts are having the event id 51 timesync issue.

    These hosts are running 2008R2 SP1 VM's.

    Any Ideas how to get proper time sync for the VM's?

     

All Replies

  • Monday, January 28, 2013 2:40 PM
     
     Answered
    Sign In to Vote
    0

    Windows is not a real-time operating system that guarantees a response in x number of seconds.  I believe that Kerberos, by default, is set up to allow for a 5 minute difference in times between computers in a domain before it will start complaining.

    What is the network between the systems?  Is there a bottleneck in the switches?  Can you allow the applications to have a greater time difference margin?  1 second off causing an application to have problems sounds like the application is expecting something the operating system was not designed for.

    .:|:.:|:. tim

     
  • Tuesday, January 29, 2013 1:34 PM
     
     
    Sign In to Vote
    0

    Windows is not a real-time operating system that guarantees a response in x number of seconds.  I believe that Kerberos, by default, is set up to allow for a 5 minute difference in times between computers in a domain before it will start complaining.

    What is the network between the systems?  Is there a bottleneck in the switches?  Can you allow the applications to have a greater time difference margin?  1 second off causing an application to have problems sounds like the application is expecting something the operating system was not designed for.

    .:|:.:|:. tim

    The systems mentioned are in the same broadcast domain. We checked the switches and there are no problems there.

    I cannot influence the application.

    Physical Win2k8R2 systems, in the same subnet do not experience the issue that the Hyper-V VM's do.

    The Hyper-V host itself on which the problematic VM's run, does not experience the problem either.

    We have TCP offloading and IPv4 offloading enabled on the VM's and Hyper-V systems enabled. Could disabling these options help?


    • Edited by Marzzie Tuesday, January 29, 2013 1:35 PM
    •  
     
  • Tuesday, January 29, 2013 8:31 PM
     
     
    Sign In to Vote
    0

    What does  “w32tm /query /source” return on one of the affected virtual machines?

    If the Hyper-V hosts aren't affected by the drift issue, is there a particular reason not to allow time synchronization via integration services?

    Eric Siron http://www.altaro.com/hyper-v-backup/

     
  • Friday, February 01, 2013 7:29 PM
     
     
    Sign In to Vote
    0

    What does  “w32tm /query /source” return on one of the affected virtual machines?

    If the Hyper-V hosts aren't affected by the drift issue, is there a particular reason not to allow time synchronization via integration services?

    Eric Siron http://www.altaro.com/hyper-v-backup/

    The source shows the Physical DC.
     
  • Friday, February 01, 2013 7:30 PM
     
     
    Sign In to Vote
    0

    We found this article, it suggests setting a flag in SQL and optionally set a regkey. We are going to try that and see if any improvements will be made. The affected VM's have SQL installed.

    http://blogs.msdn.com/b/robertvi/archive/2011/11/11/time-drift-in-vm.aspx

     
  • Friday, February 01, 2013 9:48 PM
     
     
    Sign In to Vote
    0
    Ah, "deep C-states". I've personally only ever seen that be an issue for LiveMigration, but since so many Hyper-V admins have made it a habit to disable this on all their hosts, its quite possible that it affects a lot of other things that don't get seen very often. This is also a BIOS setting in the host. I think it's normally listed as "C1E" or "S3 sleep" or something similar. If you can't find anything related to that, disabling all power management in BIOS would be the heavy-handed approach.

    Eric Siron http://www.altaro.com/hyper-v-backup/


    • Edited by Eric Siron Friday, February 01, 2013 9:50 PM clarity
    •  
     
  • Saturday, February 02, 2013 11:41 PM
     
     Answered
    Sign In to Vote
    0

    VMs get their time from the host by default through integration services, so if the time on the host is off, then you need to start there.

    you fix the host time you fix the problem, and if you can't fix the host, then disable time sync with the host from the VMs, and then they should sync time with the DC. 

    Mohsen Almassud