cancel
Showing results for 
Search instead for 
Did you mean: 

Help setting up NBU & STU's for Two-Node MSCS Media Server Cluster (NBU not clustered) with Two Tape Libraries

Michael_Haslam
Level 3
Hello,

We currently have an NBU 6.5.1 environment across two physical rooms, with a tape library in each room and a windows MSCS cluster that contains a node in each of the rooms also which are acting as "SAN Media Servers" (enterprise clients). SSO is licensed.

NBU is not under cluster control.

A single tape STU is defined for each of the media server nodes for the library in the same room as the media server node.

The cluster Virtual name is also configured as a storage unit but it is only possible to select one of the robots in that particular STU therefore we face a problem when the cluster fails over to the node which is in a different room from the defined robot in the virtual STU the backup will either run over the network and not at all due to the storage unit not being available.

We have defined the virtual cluster name as an app cluster in EMM and added each physical node as a media server entry into the app cluster in EMM. We havent added each physical node as an app cluster entry in EMM.

Is anyone using a similar setup or could someone offer advice on how we could perhaps configure this to get it working - have we configured EMM fully, does this need to run under cluster control with NBU installed with cluster resource groups in MSCS or are we missing anything?

Any help would be brilliant.

Thanks

Mike
10 REPLIES 10

Deeps
Level 6
You menttioned that there is a tape library in both the rooms. Does that mean both nodes have a seprate tape library attachet to them ? Further you mentioend SSO is licensed , that means both nodes of cluster  have access to same tape library .

For failover to work in AppCluster , we need to have a strorage unit configured with virtual name of the cluster and that means that tape library need to be access and shared between both the nodes of cluster.

Michael_Haslam
Level 3
Robot 0 is in one physical room 1

Robot 1 is in the other physical room 2

SAN media server in room 1 uses a tape drive within robot 0

SAN media server in room 2 uses a tape drive within robot 1

Both nodes do have access to the tape libraries however we must ensure that backups occur over fibre and not network so each server should only use the tape library in their respective room and when the cluster fails over to a specific node the cluster STU needs to be able to use the correct robot for that node.

There is a storage unit defined for each SAN media server using the robot & drive in their own room, and as mentioned before there is also an STU defined for the virtual cluster name however when this is created you can only select ONE robot number so we need to guarantee that the correct one is used when running a backup policy for the cluster which is using the virtual cluster STU.

Anyone ?

Cheers

Srikanth_Gubbal
Level 6
Certified
Deeps is right ....

first let me be very clear on licensing, SSO is based on drive, not on server, single drive you can zone to n number of servers.

so taking into this account, zone you tape libraries so that master server( two nodes), media server 1 and media server 2 see the drives.

now make the master server as robotic control host for both the tape libraries, so that u will have 4 storage units created, 2 for tape libraries one and 2 for tape library two.

now when in normal backup window use local tape libraries connected to each media server, when in case of single media server failover, just chaange the storage unit in the policy or go for storage unit group so that it automatically takes the other tape library for backup.

this was tested and put in place, let me know incase of more doubts.

Will_Restore
Level 6
"so taking into this account, zone you tape libraries so that master server( two nodes), media server 1 and media server 2 see the drives."

Shouldn't 'master' be 'media' in this sentence?

Srikanth_Gubbal
Level 6
Certified
i am not sure what he is trying is acheive form media server cluster, when NBU is not in cluster mode, i see master server on top of media server's and normal failover from master server in terms of storage units..

i better recommened him to go for master server cluster and media server load balancing design, where master server will be in cluster and media servers under the master, tape drives would be zoned to all the servers(master and media) and robotic control host would be the master server, so when your one of your media server is down, just change the storage unit for that policy, you need not even recycle the services, in normal scenreio use media server load balancing for proper media server utlization.

Will_Restore
Level 6
the given scenario is unusual.  Sounds like he wants each media server to only talk to it's local library thus my confusion  why you want to share drives on the master.
Perhaps the original poster can elaborate.

Srikanth_Gubbal
Level 6
Certified

inorder to make the master server as robotic control host u need to share the drives to master server.

Michael_Haslam
Level 3
Sorry i gave slightly duff info...

MSCS Server node (also an NBU media server) in Room 1 uses TLD 1 which is in a different room from the media server
MSCS Server node (also an NBU media server) in Room 2 uses TLD 0 which is in a different room from the media server

This is the design and such that if one room is destroyed the other room will have the backup of the other servers to restore. This design is not changeable.

The two servers above are a Microsoft Application Cluster as mentioned before and they are not running as an NBU Cluster.

Each server HAS to backup to the tape library in the other room.  The issue is for a backup of the Cluster drives it must always use the correct STU so that a) it backs up to the correct tape library in the correct room and b) it does not do a network backup. We have therefore defined an STU for each node with the specific tape libary in each STU as per the design requirements.

The only option we can think of (we do not really want to have to keep changing the storage unit in the cluster backup policies) is that we create a Storage Unit group that contains two STU's (each STU will use the required library I have referred to above).

i.e. MSCS cluster with Node A and Node B
Within NetBackup:
Node A with configured storage unit STU-A
Node B with configured storage unit STU-B

create storage unit group with STU-A and STU-B.  If node A is the active node in the cluster then I am hoping the storage unit group will use STU-A due to the fact that STU-B would not be available, and vice versa if node B was the active node in the cluster it would use STU-B ? Not knowing the default behaviour of the tape STU's and having no window to test this currently I cannot confirm if this would work?

Srikanth_Gubbal
Level 6
Certified
where is the master server here.

Will_Restore
Level 6
never heard of that