Hello,
We have a running Exchange 2013 CU6 deployment on a 2-Server 2012 R2 Hyper-V Cluster with FibreChannel Shared Storage. (6 Discs, Raid6)
The average Read using Raw data is at about 450 MB/s. 5 Days ago Exchange started behaving slow and finally almost stopped responding. After searching several hours it turned out that the Disc Queue length on the Hyper-V Hosts (both) was jumping between 3 - 60! while the Disk I/O was at about 5mb/s at maximum.
Doing a new raw transfer test was showing a transfer of 2-3 mb/s at a packet size of 1MB and Perfmon showing 100% Highest active time. Response time went up from 20-30ms to 500-800 ms .
We had made no changes either to the Host or the Exchange for about 2 Month. The Database growth during the past 2 Years running it on The FC Raid was at about 25 GB.
Finally we moved Exchange away to a new Host with local Raid and the same happened there while the FC Raid wen´t back to normal Operation. This was tried several Times with the same result. Wherever the Exchange VM was running the I/O Performance dropped like 99%
Exchange VM is configured with
16 GB RAM fixed (Usually using 12 Gigs)
6 Virtual Processors (Host is a 6 Core Xeon, Overall Usage between 5-25 %)
Fixed size 100 GB Drive for the System
Fixed size 300 GB Drive for the Mail Database (actually ~100 GB Size)
Network is a Broadcom 1GB NIC and VMQ is disabled due to the VMQ Feature not Supported on this card (didn´t even work with the latest drivers)
We are searching now for more than 5 Days for the "magic switch" that resumes operation back to normal.
Are there any known but not documented limmits for Database Sizes running on Virtual disc or does anyone has other Hints/Tips how to fix this issue ?
Best regards
Andreas