NetApp Best Practice and Independent Disks

[joeflint]

Hi, NetApp best practices for VMware recommends, transient and temporary data such as the guest

operating system swap file, temp files and pagefiles, should be moved to a separate virtual disk on a

different datastore as snapshots of this type of data can consume a large amount of storage in a very short

period of time due to the high rate of change (i.e. to create a datastore dedicated to transient and temporary data for all VMs with no other data types or VMDKs residing upon it).

NetApp also recommends configuring the VMDKs residing in these datastores as “Independent Persistent” disks within vCenter. Once configured, the transient and temporary data VMDKs will be excluded from both the VMware vCenter snapshot and the NetApp Snapshot copy initiated by SnapManager for Virtual Infrastructure.

I'd like to understand the impact of following this best practice - can someone advise on the following:

• If the above is implemented:
  
  
  • Will snapshots work via vcenter?
    
    
  • Will snapshots work via the Netapp Snapmanager tool?
    
    
  • Does the snapsot include the entire VM disks? If not what is the consequence for not having the entire VM image?
    
    
  • Can the vcenter snapshot restore ok?
    
    
  • Can the Netapp snapshot restore ok?
    
    
• What impact does the above have on the back process if using a backup product that is reliant to snapshot technology?
  
  

Thanks

[Pieter Du Preez]

Hi Joe

These recommendations is purely to save storage space when replicating or backing up.

For instance you can move your *.vswap file (VM swap file) to a different Datastore. Netbackup can do SMVI Snapshots of the Datastores and with this setup you can exclude that particular datastore

This is also true if you create a dedicated datastore for OS Swap files, and mark them independent so that vCenter do not snap those vmdks.

I have done a project with NetApp on SAP production boxes

We moved all *.vswap files to dedicated datastores and created RDMs for the OS Swap locations

We actually used the NetApp snapdrive technology to quiesce the SQL DB on the RDMs before the RDMs gets snaped, but I won't go into too much detail :smileyhappy:

To answer your questions (see comments in quote)

joeflint wrote:

• If the above is implemented:

• Will snapshots work via vcenter? --- Yes it will - Independent disk gets ignored
  
  
• Will snapshots work via the Netapp Snapmanager tool? --- Yes it will - Snaps the whole LUN/Datastore
  
  
• Does the snapsot include the entire VM disks? If not what is the consequence for not having the entire VM image? -- No - *.vswap file gets created when VM is started (No need to backup)
  
  

                                           - OS Swap location vmdks needs to be recreated in the event of a restore. WIndows will

                                      still boot if the Swap drive is missing, and you can specify the new swap location.

• What impact does the above have on the back process if using a backup product that is reliant to snapshot technology? - These backup products leverage vCenter snapshots and because the vCenter snapshots works  100% this should not be a problem.    
  
  

Hope it helps :smileywink:

Please award points if it does :smileycool:

[joeflint]

Thanks, for the feedback. I think I understand, just to ensure I do, assuming I have a number of VMs with 2 VMDKs; one VMDK for OS and other for data. I should put OS vmdk on separate datastore to data vmdks.

Also, to ensure all swap files are on separate datastore.

[Pieter Du Preez]

Hi

joeflint wrote:

I should put OS vmdk on separate datastore to data vmdks.

Also, to ensure all swap files are on separate datastore.

It is not necessary to put your OS and Data vmdk on separate datastores. OS does not tend to use high IO unless there are OS Memory constraints and the OS swap file gets used. In this case I would create a 3rd vmdk on a separate Datastore and move OS swap file there and make it independent.

Just to make sure we don't miss each other, you should not make your OS or Data vmdk's Independent :smileywink:

OS swapfile - If the OS uses the swap file there might not be enough allocated memory, unless it is a SAP box

VM Swapfile - Yes, on the cluster settings you can specify where to store the VM's swap files

Please award points for helpful answers :smileycool: