KB2913659 – WS2012 R2 Hyper-V Crashes When Network Traffic Passes Through A Virtual Switch

This KB article was released in January 2014 and is not related to the commonly reported issues with Intel and Emulex 10 GbE NICs. This hotfix is for when Windows Server 2012 R2-based Hyper-V server crashes when network traffic passes through a virtual switch.

Symptoms

Consider the following scenario:

  • You have the Hyper-V server role installed on a computer that is running Windows Server 2012 R2.
  • You create a switch team over a physical network adapter.
    Note The virtual machine queue (VMQ) is enabled on the network adapter.
  • You create a virtual switch over the switch team.
    Note There is no forward extension present.
  • Network traffic passes through the virtual switch.

In this scenario, the computer crashes, and data loss occurs.

A supported hotfix is available from Microsoft.

KB2929078 & KB2929869 – CSV Snapshot Corrupted After File Modifications On WS2012

There are two new KB articles that offer two different, but very similar, hotfixes for this situation.

The first is KB2929078 which deals with a scenario when you delete and then re-create a file on the live volume in Windows Server 2012, the Cluster Shared Volumes (CSV) snapshot is corrupted.

A hotfix is available.

The second article is KB2929869:

Symptoms

Consider the following scenario:

  • You create some files on a Cluster Shared Volume (CSV) in Windows 8 or Windows Server 2012.
  • You take a snapshot of the CSV.
  • You delete the files.
  • You create some files and delete some older snapshots in parallel.

In this situation, the snapshot CSV snapshot file is corrupted.

A second hotfix is also available for this issue.

KB2901896 – WS2012 CSV Cache Causing Poor Performance For Hyper-V VMs

Microsoft has released a hotfix for when CSV block cache causes poor performance of virtual machines on Windows Server 2012 Hyper-V.

Symptoms

Consider the following scenario:

  • You have Hyper-V virtual machines (VM) that are configured on Windows Server 2012 Hyper-V Cluster by using Scale Out File Server as the Storage Solution.
  • The virtual machine .vhdx files are held in Cluster Shared Volume (CSV).
  • CSV Block Cache is enabled on the volume.

In this scenario, virtual machines may experience slow performance.

 

A supported hotfix is available from Microsoft.

KB2935810 – CSV Failover Takes Longer Than Expected In Windows Failover Cluster

Microsoft released a hotfix for WS2012 and WS2012 R2 to deal with a scenario where CSV failover time is longer than expected in Windows failover cluster.

Symptoms

In a Windows failover-cluster that uses Cluster Shared Volumes (CSV), the diff area that is allocated by Volsnap is large and fragmented. In this situation, you encounter the following issues:

  • The failover time on the CSV is longer than expected.
  • The time that Volsnap takes to mount or unmount snapshots is several minutes.

More Information

When a NTFS or ReFS volume is mounted or dismounted, Volsnap iterates through the diff area to mount or unmount the snapshots that belong to that volume. When the diff area allocation becomes large and fragmented, the time that Volsnap takes to mount or unmount operations could be several minutes. Additionally, failover time can be longer than expected.

The resolution is … hmm … long. It is related to two updates:

Two new cluster Physical Disk resource private properties were added, and they can be manipulated to resolve the issue:

  • SnapshotDiffSize: This property controls the maximum diff area size that can be consumed by Volsnap for a Physical Disk resource configured for CSV. Units: In MB (DWORD), Default Value: 0, Maximum Value: 1 TB, The Physical Disk resource must be taken offline/online for changes to take effect.
  • SnapshotAgeLimit: This property is aResource Type private property of the Physical Disk to control the maximum age of a snapshot. Long lived snapshots are a significant contributor to diff area fragmentation. Units: In Days (DWORD), Default Value: 7, Range: 1-60 , This is a global property which affects  all Physical Disk resources. You do not have to take the resource offline or online for it to take effect.

Get-ClusterSharedVolume <Cluster Disk Name> | Set-ClusterParameter snapshotdiffsize <Snapshot Diff Size in MB>

Get-ClusterResourceType "physical disk" | Set-ClusterParameter snapshotagelimit <Snapshot Age in Days>

My advice: leave well alone and only manipulate these settings under the advice of Microsoft support (not some local dude, but actual Premier support).