We have a 2012 R2 Hyper-V Cluster, all patched to date. This cluster has 6 nodes, each one connected via FC to a 3PAR 7400c All-Flash array.
Each cluster node hosts about 20 VMs and 2 or 3 CSV (so each VM role has direct access to their storage via FC as it is its own Coordinator Node). With that in mind, we installed BEX 15 FP5 in each cluster node so that we could backup the VMs directly through FC and not via Ethernet (it's way faster and we do have a lot of data). Since implementation, eveything is going smooth (with some issues sometimes) with all cluster nodes except one. Every cluster node has either a FC Tape Library or a SAS Drive (all LTO6/LTO5, all firmwares up-to-date).
This cluster node, VHT06, shows erratic behavior. It starts the granular job, but hangs in a random VM in "Snapshot Processing". We can't cancel the job unless we stop the "vmms.exe" service or reboot the cluster node. If we manage to cancel the job, the VM gets stuck in "Backing up..." within Hyper-V manager and can't reboot. We thought it could be the tape library, so we changed it. Still the same issue. We even evicted and reprovisioned the cluster node two times.
Backup through RAWS works fine.