asked on Storage optimization for SCVMM 2012R2
I can see there are options for Dynamic Optimization of hosts in SCVMM 2012 R2, and that these are based on memory consumption and processor consumption. I can see there are options for Power Optimization as well. But how do I set up the equivalent for Storage DRS (that exists in VMWare) for SCVMM2012R2? In other words; I want to set up Dynamic optimization of storage and load balancing of storage based on storage consumption, but I can't find descriptions of this or articles about this on the internet. If it is of any concern; the SCVMM 2012 R2 environment that I want to set this up in uses clustered hosts that are clustered with Microsoft Failover Clustering. This cluster in turn uses CSV disks (Clustered Shared Volumes).
Hyper-VVirtualizationMicrosoft Virtual Server
Last Comment
itnifl
ASKER
We are using MPIO, but it is not IO load balancing I am wanting to control. I want a storage load balancing feature that auto migrates vhdx files to other CSV disks when such a disk fills up, so to even out the storage data usage among the CSV disks in the cluster.
Ah, not possible.
How we do things:
Smaller clusters with 4-12 VMs: We set up a dedicated LUN for each VHDX file with ownership of the LUNs distributed either between nodes (Asymmetric Cluster via Storage Spaces) or controllers (DAS like MD3220 SAS, P2000 MSA SAS). Each VM has an OS VHDX and a Data VHDX.
For clusters with more than that our storage is on a Scale-Out File Server cluster with Storage Spaces. We tend to set up one large Space (LUN) or two Spaces (LUNs) for our CSVs since SoFS takes care of load via SMB Multi-Channel.
One can Live Migrate VHDX files between CSVs without bringing the VMs down.
Oh, and we _don't_ thin provision our setups. All VHDX files are FIXED. Having an overprovisioned CSV is a recipe for disaster IMNSHO.
EDIT: I almost forgot. We don't snapshot either. That creates a differencing disk that continues to grow until the snapshot is deleted which merges the differencing disk back into the parent.
How we do things:
Smaller clusters with 4-12 VMs: We set up a dedicated LUN for each VHDX file with ownership of the LUNs distributed either between nodes (Asymmetric Cluster via Storage Spaces) or controllers (DAS like MD3220 SAS, P2000 MSA SAS). Each VM has an OS VHDX and a Data VHDX.
For clusters with more than that our storage is on a Scale-Out File Server cluster with Storage Spaces. We tend to set up one large Space (LUN) or two Spaces (LUNs) for our CSVs since SoFS takes care of load via SMB Multi-Channel.
One can Live Migrate VHDX files between CSVs without bringing the VMs down.
Oh, and we _don't_ thin provision our setups. All VHDX files are FIXED. Having an overprovisioned CSV is a recipe for disaster IMNSHO.
EDIT: I almost forgot. We don't snapshot either. That creates a differencing disk that continues to grow until the snapshot is deleted which merges the differencing disk back into the parent.
ASKER
Philip Elder: Thank you for your reply. Yes, thin provisioned disks are not recommended by Microsoft in production environments. We will have to have strict control of the over provisioning that is being done, constant surveilance. If there is no way to auto live migrate VMs from LUNs that have exceeded a threshold, then i will just have to create a script or something that does this for me at frequent intervalls.
This is a bit of a disapointment. I am used to VMWare having Storage DRS that can automatically load balance and live migrate virtual machine files to other datastores if thresholds are exceeded, without any fuzz or complications.
This is a bit of a disapointment. I am used to VMWare having Storage DRS that can automatically load balance and live migrate virtual machine files to other datastores if thresholds are exceeded, without any fuzz or complications.
In current versions the onus is on us to make sure we have our storage ducks all lined up and not overcommitted. ;)
In a team environment where different folks have control over various components this can be a bit of a challenge.
If granular permissions are required then look into AzMan and the abilities it has to delegate authority for the various virtualization roles. If one does not need authority to delete/remove VMs then AzMan is a great way to help keep things under control.
In a team environment where different folks have control over various components this can be a bit of a challenge.
If granular permissions are required then look into AzMan and the abilities it has to delegate authority for the various virtualization roles. If one does not need authority to delete/remove VMs then AzMan is a great way to help keep things under control.
ASKER
More administrators = more people to keep in the loop of information = more administrative difficulties = larger chance for administrative error. In my opinion there should only be a couple of administrators with full administrative privileges in a virtualization environment. If people are not able to watch out for what they are doing in the environment, they should not have privileges to handle the system or systems should be set up to control this for them.
In this thread I am only concerned with Hyper-V's equivalent to VMWare's SDRS, which seems to be more then lacking.
In this thread I am only concerned with Hyper-V's equivalent to VMWare's SDRS, which seems to be more then lacking.
ASKER CERTIFIED SOLUTION
THIS SOLUTION ONLY AVAILABLE TO MEMBERS.
View this solution by signing up for a free trial.
Members can start a 7-Day free trial and enjoy unlimited access to the platform.
ASKER
I ended up writing a script that I run every 15 minutes. It checks to see if a datastore is below a certain threshold, then migrate the largest vm to the largest available space. If that space can not hold the vm, then migrate the smallest vm to the largest available space. If that space cant hold the vm, then we are out of space, no migration is done.
Is MPIO being used?
If yes, then that's probably the answer to what you are asking.
We use MPClaim from an elevated command prompt to manage storage.
If SMB Multi-Channel is being used then that's a bit different. Scale-Out File Server and Hyper-V would take care of the I/O moving across the channels.