Requisites for SnapManager for SQL in a SQL Cluster environment? Does SnapInfo should be visible to both nodes?

[SNETAPPUSER]

Hello,

In a MS SQL Server Cluster environment within VMWare, what are the pre-requisites / requirements for deploying NetApp SnapManager SQL? Does it require to have two SnapInfo LUNs? If so, does it require both SnapInfo LUNs to be visible for both SQL Cluster nodes?

Please advice. In the guides things are not very clear with regard to SQL servers clusters.

Thanks

[radek_kubka]

Hi,

To make it 100% clear - you are talking about clustered SQL instance(s), and not just a SQL VM being run on top of ESX cluster, right?

The latest BPG for SMSQL 7 says:

"The SnapInfo LUN must be a cluster disk resource in the same cluster group as the SQL Server instance being backed up by SnapManager"

Have a look at this thread:

https://communities.netapp.com/thread/33849

Thanks,

Radek

[SNETAPPUSER]

I am talking about clustered SQL instances here... and of course they are running on top a ESX cluster.

Referring to this...

"The SnapInfo LUN must be a cluster disk resource in the same cluster group as the SQL Server instance being backed up by SnapManager"

Does it mean that ONLY ONE cluster disk resource (i.e. one shared disk) is enough for SnapInfo? Think of scenario where there is a two node cluster running two SQL instances on each of them individually. In case of a failure to one node, the remaining node take over the resources held by the failed node and operate both SQL instances.

In the above case, how SnapInfo should be allocated? Two separate iSCSI LUNs to individual nodes as cluster resources (i.e. both nodes can see the disks but one node held the ownership) OR a single iSCSI LUN as a cluster resource?

Thank you

[radek_kubka]

As far as I can tell it should be two LUNs, as both SQL nodes can drive backup jobs independently, so each of them needs access to SnapInfo, potentially at the same time.

As per the other thread, I'm actually surprised by the stated requirement that SnapInfo LUNs should be on clustered disks - clearly SMSQL installations on both nodes are independent from each other, so are they SnapInfo catalogues?

[david_ovenden]

This is not my understanding.

As I understand it only the active node with access to the Databases can drive backup jobs, since it requires local access to the Logs and Database LUNs.

Therefore the Snapinfo LUN needs to be a cluster resource in the same cluster role as the SQL disks to ensure it is available on the same node at all times. 

[Andrew_P]

Definitely only one snapinfo lun required... it's part of the resource group which the owning node will keep.

I tend to use mount points on the clusters - single drive letter per resource group that way!

eg:

E:\ (root mount point)

E:\Data - SQL Databases

E:\Log - SQL Logs

E:\SnapInfo

E:\TempDB

This way, full backups take snapshots of both the data and log luns, transaction log backups take a snapshot of the snapinfo lun...

If there is a failover - all the resources must go to the other node.