none
Is it better to create a 2nd volume on a Hyper-v cluster or expand Volume 1? RRS feed

  • Question

  • I have a 4tb volume 1 mirrored and am running out of space - no unallocated.  However, there is space in the pool.  I could expand the volume 1 or create a second volume (the way I understand it) and move vm's to that volume.  Is there any advantage in creating a second volume?  Would it be faster for vm's to be in a smaller volume?  I have a software raid so I'm looking for any increase in speed I can get. 
    Monday, December 2, 2019 3:08 PM

Answers

  • Hi,

    I don't believe there exists a "best practice" for the size of the cluster shared volumes (CSVs), as it's mostly up to every organization, remember most environments are different.

    I think the performance difference will barely be noticeable if you run a big CSV or many smaller CSVs, but it can also depend on the RAID configuration. Normally you would go with a RAID 1 or RAID 5 depending on the performance and capacity requirements you have.

    Depending on the storage vendor you have, you can also ask your vendor for recommendations about how to configure the storage for specific workloads.

    I personally prefer to stay within 2-4 TB range for cluster shared volume, while others may like them bigger. 
    In smaller clusters, I normally provision a single CSV for each Hyper-V host in the cluster. If the Hyper-V cluster grows, meaning more nodes added I might then increase the number of CSVs.

    It is also a matter of monitoring the workload of each CSV to see if the system is handling the I\O.

    Some may also go based on their backup, depending on how long it can take to backup/recover, will it take longer to restore one or many CSVs and so on...

    It may also depend what kind of workloads you run on your VMs, if you use for example a storage system that does not only have flash drives, you can run run some workloads on faster disks and others on slower, although nowadays most run on "all flash", but small companies might not.

    You'll also find some general information from Microsoft over here:
    Use Cluster Shared Volumes in a failover cluster
    https://docs.microsoft.com/en-us/windows-server/failover-clustering/failover-cluster-csvs

    Best regards,
    Leon


    Blog: https://thesystemcenterblog.com LinkedIn:

    Monday, December 2, 2019 6:00 PM

All replies

  • You should be able to add an additional disk to the Software RAID set (Assuming it's RAID 0, 10, or 5) to expand the volume, but the amount of performance you'd gain from doing so depends heavily on use case and RAID level. Adding a second volume outside of the RAID array can improve performance of VMs that are moved to it, but that again depends on the RAID level you have and what the VMs are doing.

    If you're using RAID 0 or 10, expanding the volume should give you essentially the same performance boost as adding an additional volume, but you will want to look at your existing Drive utilization to see how much the VMs are utilizing drive IO.


    Adam Brown

    MCSE, CISSP

    Blog: AC Brown's IT World

    Monday, December 2, 2019 5:58 PM
  • Hi,

    I don't believe there exists a "best practice" for the size of the cluster shared volumes (CSVs), as it's mostly up to every organization, remember most environments are different.

    I think the performance difference will barely be noticeable if you run a big CSV or many smaller CSVs, but it can also depend on the RAID configuration. Normally you would go with a RAID 1 or RAID 5 depending on the performance and capacity requirements you have.

    Depending on the storage vendor you have, you can also ask your vendor for recommendations about how to configure the storage for specific workloads.

    I personally prefer to stay within 2-4 TB range for cluster shared volume, while others may like them bigger. 
    In smaller clusters, I normally provision a single CSV for each Hyper-V host in the cluster. If the Hyper-V cluster grows, meaning more nodes added I might then increase the number of CSVs.

    It is also a matter of monitoring the workload of each CSV to see if the system is handling the I\O.

    Some may also go based on their backup, depending on how long it can take to backup/recover, will it take longer to restore one or many CSVs and so on...

    It may also depend what kind of workloads you run on your VMs, if you use for example a storage system that does not only have flash drives, you can run run some workloads on faster disks and others on slower, although nowadays most run on "all flash", but small companies might not.

    You'll also find some general information from Microsoft over here:
    Use Cluster Shared Volumes in a failover cluster
    https://docs.microsoft.com/en-us/windows-server/failover-clustering/failover-cluster-csvs

    Best regards,
    Leon


    Blog: https://thesystemcenterblog.com LinkedIn:

    Monday, December 2, 2019 6:00 PM
  • Thanks for the replies.  This is a raid 5 failover cluster set up by a 3rd party.  All the disks are allocated to this pool.  The pool is 18tb and 8tb are in use (the 4tb volume).  If I add a second volume would that not be also part of the raid 5?  I only have 1 host with 2 nodes and a shared storage volume.  I have 11 virtual machines on this host.  I haven't read your link yet but I'm going to right now.  It's just that I'm running out of time and will have to just pick one solution.  We are a small company and no flash drives. They don't believe in spare equipment here so I don't have anything to practice on.  Nor do they believe in training. I'm learning as I go.  It's all in production.  I do have 1 virtual machine that has a lag in response time.  That's why I was looking for more speed. 
    Tuesday, December 3, 2019 4:17 PM
  • If I add a second volume would that not be also part of the raid 5?

    >> Yes.

     I do have 1 virtual machine that has a lag in response time.  That's why I was looking for more speed

    >> If it's only "one" virtual machine, I would monitor it closely with for example Performance Monitor to find out what could be the bottleneck, it might also be as simple as giving it some more resources, but it could also be something else.


    Blog: https://thesystemcenterblog.com LinkedIn:

    Tuesday, December 3, 2019 4:34 PM