Microsoft Virtualization Discussions

Clarification on SnapManager Hyper-V


Looking to deploy SnapManager Hyper-V on our 3 host cluster and just looking for some clarification on a few things.


We have a 3 host Hyper V cluster that connects to the same Failover Cluster and 6TB CSV to store the VM's. I know I need to install SnapManager Hyper V on all 3 hosts. When I connect to a host in SMHV, I connect to the name of the Failover Cluster and not the 3 individual hosts. This makes sense.


What I am confused on, when I create my Dataset and add the VM's, the VM's will exist on 1 of the 3 hosts and with LiveMigration, they can move around. When a DataSet policy kicks off, what host is taking the backup? The host that owns the VM?


In 2012, the documentation says it supports CSV for SnapInfo LUNs, but then goes into making a cluster group called "smhv_snapinfo" which really isn't a CSV and just an additional Role in Failover Cluster Manager. With a 3 host Failover Cluster, this "smhv_snapinfo" role will only be owned by one of the three hosts, and thus the disk associated to it, will only appear on the single host that is the owner. Does this mean, whatever host is the owner of the "smhv_snapinfo" role, will take the backup of all the VM's...even if the VM's exist on the other two nodes that arent the role owner?


If I move the SnapInfo LUN to a CSV...this will be presented to all 3 hosts. What host will be responsible for taking the backups when the Dataset Policy kicks off?


Just a little confused on this operatiblity, and I can't seem the find the explanation. Just looking for some clarification on this process? Let me know if I am missing a document that goes into this, etc. 



With Windows Server 2012 R2 you will typically select the option to do "distributed application consistent" backups in SMHV. These will leverage the new VSS features that were added in CSV 2.0 so that only one node actually takes a snapshot on the NetApp controller. The CSV 2.0 VSS components will handle quiescing applications on all of the other nodes so they are consistent in the snapshot.


WRT SnapInfo on a CSV, yes, that is supported in the latest version. As is using a NetApp SMB/CIFS share.






Ah...thats what VSS is used for and the "Understanding VSS/How SnapManager for Hyper-V uses VSS" section in the documentation now makes sense. Always knew of VSS conceptually..never used VSS in practice up until learning as I go. 


One final followup...Do the nodes go through any kind of election process on who is making the SnapShot copy or the likes? With a 3 node cluster...when I setup the Dataset/Policy on one node, I am assuming these settings are replicated to SnapManager Hyper V on the other 2 nodes? Also, would it be safe to assume, whatever node is writing the SnapShot is also writing the data to the SnapInfo directory?


Correct, when you edit a dataset on one node, the dataset is replicated to all nodes so that everybody is on the same page in the cluster.


With SMHV and distributed application consistent backups, we pick the cluster name resource owner to be the backing up node. There are lots of other options we could have picked, but since we use cluster name owner for some other things, it seemed to make sense to do it that way here too.


I hope this helps, and best wishes on your new deployment!





Thanks for the information BINGEN...makes sense now!


Ill probably move the SnapInfo to a CIFS volume, but other than that, I believe I finally understanding the workings and operations on how this process flows from start to end.