After System Center 2016 came out i went straight on to setting up DPM 2016 with Modern Backup Storage. This had been running fine for a while. A bit slow for the DPM console to respond after doing a synchronization .
On December 27th i decided to install KB3206632 on the Storage server hosting the disks and the VHDX with dedupe enabled on the storagepool volume. The next day i had lot’s of emails saying the backups had failed. I noticed that the vm had rebooted. So i tried to resume backups and do synchronizations where it needed. After about 10 minutes i lost communication with the DPM VM server, due to it rebooting.
Now what i noticed in the logs for the storagepool was that it was loosing connection with the disks. These where the attached VHDX files. Now i started troubleshooting and trying diffrent things. Nothing helped. All of a sudden after too many times the storagepool lost disks, the volume now was in raw format. This is a error that happens if the ReFS fails to many times. So i removed all backup and the volume from dpm and reformatted to NTFS, then added back in. Now i noticed as soon as the backup reached the size of my SSD tier on the storage server the VHDX disks got dismounted from the VM server.
With checking event viewer logs on the DPM server, Hyper-V host and the storage server i could see that there was issues writing to the storage pool when this happen. I still had access to it, and i could copy files to it locally. If i did a large copy i would see the performance drop quite a lot after a while. Indicating some issue with flushing of data from the SSD to the HDD and you could also see it in perfmon.
I contacted Microsoft on the 28th to get some help about this issue. Sometimes these things take time. So i did a lot of troubleshooting. Yesterday the 5th of January i checked in windows update for a new patch on the storage server and there was KB3213522. Installed this, shut down the VM, rebooted the Storage Server and tried setting up the backup again. And it’s been running perfect for almost a day now. No issues.
So this must be a bug in the KB3206632 with storage spaces and tiering. I can’t find any related issues for this, but will update this post as soon as i get an answer from Microsoft.