Cristian Edwards

A piece of virtualization with Hyper-V and SCVMM

Live Migration fail and source node reboot on AMD processor servers

Live Migration fail and source node reboot on AMD processor servers

  • Comments 7
  • Likes

Recently i found several customers that are running Hyper-V R2 under AMD Family 10h processors and they experienced server reboots on the source host that start the migration. There isn’t any information on the event viewer or any dump cause the root cause is a hardware problem. This crash was reported by AMD and our product team help to fix the issue with a software workaround. This issue mostly happens with virtual machines with 2 or more processors. Most of the customers who reported the issue were able to live migrate VMs with 1 virtual processors.

http://support.microsoft.com/kb/981618

Comments
  • We started experiencing this after moving to our cluster the other day. It is most annoying.

    Any idea on when the hotfix will go out normally? We only have 1 AMD host in our test setup which means we would need to put the patch into production for testing.

  • Hi Stephen,

    You can download the hotfix from the posted link kb ( 981618 ) . One question. Do you have only one AMD host? Are you sure you are facing the same issue?

  • We have 3 AMD hosts in a cluster, 1 I can use for testing. Live Migration will regularly cause the a crash. System Bios reports a CPU CHK fault.

    I have just tried the hotfix and it won't install saying it is not applicable for my computer.

    CPUZ picks up the cpus are AMD Opteron 2376, Shanghai Core.

  • Can you check these file versions?

    File name File version File size Date Time Platform

    Hvax64.exe 6.1.7600.16561 643,584 27-Mar-2010 02:26 x64

    Hvix64.exe 6.1.7600.16561 707,072 27-Mar-2010 02:26 x64

    Hvax64.exe 6.1.7600.20678 643,584 27-Mar-2010 02:26 x64

    Hvix64.exe 6.1.7600.20678 707,072 27-Mar-2010 02:26 x64

  • I had this problem on Windows Server 2008 R2 and installed the hotfix. This fixed my problem BUT.... I have now made a complete reinstall of the hyper-cluster with "slipstreamed" Windows Server 2008 R2 Servicepack 1 and I am havning this problem again. Has anyone else had this problem? We are running the Cluster nodes on HP DL185 G5 with AMD 2376 CPU. We have updated ALL bios and firmware on the servers.

    Regards

    Anders Månsson

    Senior Execute Consultant

    AddLevel AB

  • Hi Anders,

    The fix is already inlcluded on SP1. are you 100% sure that is the same issue? Can you describe in what scenarios are you reproducing the issue?

  • Hi Cristian,

    Accually the problem has changed a bit. Before SP1 and before the hotfix the source host rebooted when doing live migrations. Now after SP1 the target host reboots.  I have located the virtual guest thats the bad guy and it has 2 virtual CPU:s. The guests OS with one virtual CPU migrates fine.

    Regards

    Anders Månsson

    Senior Execute Consultant

    AddLevel AB

Your comment has been posted.   Close
Thank you, your comment requires moderation so it may take a while to appear.   Close
Leave a Comment