vmotion and compute-only nodes with Simplivity

[ar1000]

Hi all,

I'm dealing with an issue where a customer has added compute-only nodes to Simplivity cluster, and is trying to use those compute nodes by vmotioning an existing VM. Simplivity maintains host/vm rules in vCenter to pin the VM to specific hosts for performance reasons. So what happens is that the customer moves a VM to the compute node and then DRS moves the VM back to the host specified by the host/vm rule! 

I feel that simply removing the VM from the DRS rules that are maintained by Simplivity is not the best way to solve it but I'm not able to find documentation on how to deal with this issue. Any insights would be very much appreciated!

[support_s]

System recommended content:

1. HPE SimpliVity OmniStack Interoperability Guide 3.7.10U1

2. HPE SimpliVity OmniStack Interoperability Guide 4.1.1U1

Please click on "Thumbs Up/Kudo" icon to give a "Kudo".

Thank you for being a HPE valuable community member.

[ICPMuenchen]

I think this is not a simplivity issue. It's more a VMware / DRS issue.

Of course you need to change DRS settings as well. Otherwise this will happen and the vm always moves back to the host regarding your DRS rules.

Have the compute nodes also been added to DRS settings? Or how is the DRS setup?

[ar1000]

Compute nodes are added to the Simplivity cluster. DRS is enabled on the cluster level.

The issue is that the compute nodes are not utilized because Simplivity created a large number of host and VM groups in attempt to pin VMs to specific hosts (two hosts per VM), and only Simplivity storage hosts are members of the DRS groups.

So I guess the obvious solution would be to add compute nodes to each host group, but is this a recommended way to fix this?

[ar1000]

Apparently this is just how Simplivity 's IWO (Intelligent Workload Optimizer) works. It "automatically creates cluster virtual machine groups, cluster host groups, and affinity rules in DRS that bring the virtual machines closer to the host with the source virtual machine." Obviously this interferes with VMware DRS' ability to load balance based on cluster CPU and memory utilization.

The documentation further states that "in some high load environments DRS will ignore IWO" but unfortunately provides no further information on conditions when IWO is ignored by DRS.

It appears that the recommended way to ensure that VMs are evenly distributed across all nodes in the cluster is to turn off IWO and get rid of all custom DRS groups: https://community.hpe.com/t5/Around-the-Storage-Block/How-VM-Data-Is-Managed-within-an-HPE-SimpliVity-Cluster-Part-3/ba-p/7033153, Scenario 3. Otherwise, it's a manual process of managing DRS rule assignments.