cancel
Showing results for 
Search instead for 
Did you mean: 

NetBackup & Hyper-V CSV's - question re: backing up machines hosted on different nodes

Lynne_Seamans
Level 6

We have a Hyper-V cluster with 3 machine nodes, using a CSV, and being a relatively new NetBackup installation, we're testing backing up the virtual machines on this cluster. 

Policy type "Flash Backup Windows", Hyper-V cluster name is for "Hyper-V server"..   it does see all the virtual machines and does seem to know which ones are High Availabliliy.

Problem is, when we start a backup of one of these virtual machines, it does a snapshot, and we notice that the ownership of the CSV is switched to whichever node in the cluster the virtual machine being backed up happens to be on at the moment.

So, we're wondering what will happen when we specify multiple virtual machines...  I mean, you can't make a policy per host, because you really don't know which VM lives on which host at any particular moment.   And it's VERY impractical to back up the virtual machines one-by-one, since we have 150 of them (or there abouts) and it would take way too long doing a snapshot for each one, and then backing it up.

Doing research, I found this thread:

https://www-secure.symantec.com/connect/forums/simultaneous-backup-vms-single-csv-multiple-nodes

which seems to be addressing similar issues, but really without a conclusion at least as far as NetBackup is concerned.  In NetBackup you disable the "Advanced Open File Option" on the client atributes section of the Master Server properties.  Even if that were the answer, what client would you type in?  All the nodes? or the cluster name?  Would that even help?

Is this worth opening a ticket for?

4 REPLIES 4

Lynne_Seamans
Level 6

Read my Symatec NEtBackup for Hyper-V Guide (7.1) a little more closely, and found on pg 36 :

Notes on CSV backup and restore

Windows Server 2008 R2 includes a new feature for Hyper-V called Cluster Shared

Volumes (CSV). CSV allows multiple virtual machines to share volumes (disk

LUNs). CSV also allows live migration of a running virtual machine from one

Hyper-V server to another without user interruption.

Multiple virtual machines can be created on a single cluster shared volume (CSV)

and owned by different nodes. (Hyper-V servers are configured as nodes in the

cluster.) All nodes can access the CSV simultaneously.

NetBackup can back up the virtual machines that are configured in cluster-shared

volumes.

Note the following:

■During the backup, the cluster-shared volume (CSV) enters the online state

(“Backup in progress, Redirected access"). The cluster node that performs the

backup becomes the owner of the CSV.

■Multiple nodes cannot back up a CSV simultaneously. When a node backs up

a virtual machine on a CSV, the attempt by another node to back up the same

virtual machine fails.

You can use the CSV Timeout in Minutes option in the policy to adjust how

long NetBackup waits for another backup of the same CSV to complete.

See “CSV Timeout in Minutes” on page 29.

■A single node containing two virtual machines can back up both virtual

machines simultaneously, even if they use the same CSV. As long as both

virtual machines reside on the same node, simultaneous backups are allowed 

 So that explains what I'm seeing, but still, how do folks get around this?  It would be lovely to just say "one policy per node", but with Hyper-V and CSV's, virtual machines could move around..  What to do??!?

 

 

KristjanS
Level 2

Im not sure I understand your concern 100% but I think what you should be doing is creating a policy for the cluster not per node.

In the policy attributes you should tic the "Hyper-V server" box and enter the cluster name rather then a specific node. If you have it configured like that and select "New client" on the policy then you should see the cluster and all the nodes when selecting "Browse for Virtual Machines"

In this configuration it doesnt matter which node the VM is on.

However the limitation that only one node can backup VMs inside a CSV is still there. So when the backup window starts one node will put the CSV in a "Backup in progress, redirected mode" and try to backup all its VMs. The other nodes will wait their turn or fail after 300 minutes. I think its the default timeout and you can change it.

Im no expert so take this information for what it is. This is just stuff that I have learned in my own brief experience with implementing Netbackup in our enviroment.

Lynne_Seamans
Level 6

Kristjan, actually, I think you DO understand my question.  You read it better than I wrote it.

Have indeed specified the cluster name in the policy rather than the specific node, and was encountering JUST the problem.  For want of a better term, the individual clients backups kept trying to "steal ownership" of the CSV by putting it in that "Backup in progress, redirected mode".  Thne subsequent backups are failing.

So, what I think you're saying is really all I have to do is increase the "CSV Timeout in minutes" attribute to 300? 

Definitely worth a try !!

 

THANK YOU !!!

GregSchmalhofer
Not applicable

Kristjan,

Thanks for your reply in helping us resolve this issue (I'm working with Lynne above on this issue). There are times when the backup ends either with a failure or perhaps normally we are still left with the CSV state of "Backup in progress, redirected mode". Do you have any recommendations as to how to get out of this "Backup in progress, redirected mode" state? The last time when we had the redirected mode we moved all the virtual servers off the box that was the owner of the CSV and tried to reboot that single host node but the CSV services would not end and the host node had to be shutdown hard. This caused the CSV to not be available to any of the other host nodes and all the virtual servers crashed. Any guidance would be much appreciated.

 

Thanks,

Greg