Hi All,
We are running a SCVMM 2012 Hyper-V Cluster with a full SSD SAN, and seeing some massive disk spikes. Disk I/O Counts would go from a nice 5000-10000 up to 300000+ in a gradual increase. We can isolate which node the IO counts are coming from using perfmon on physical disk and selecting the disk number for the CSV. However, this doesn't match up with 'Hyper-V Virtual Storage Device' IO readings when you add all instances. So we cannot seem to determine which VM (if any) is causing the issue. This doesn't happen often enough to try moving around all VMs to try to isolate which one it is. We have a number of linux guest OSs, and they tend to fail kernel panic when the disk IO is too heavy so it's imperative that we get this fixed.
The nodes themselves are just installs of Windows 2008 R2 SP1 with hyper-v and failover clustering installed. Apart from a Hyperic agent for monitoring installed directly on the nodes there is nothing else running. We are really struggling with this one, so any suggestions would be helpful!
Cheers,
- Will Kruss