In a fully automated Storage SDRS cluster, VMDKs are not migrating even though any of the Datastores reaches threshold value (80 % and above)
search cancel

In a fully automated Storage SDRS cluster, VMDKs are not migrating even though any of the Datastores reaches threshold value (80 % and above)

book

Article ID: 443280

calendar_today

Updated On:

Products

VMware vCenter Server

Issue/Introduction

VMDKs are not migrating from the Datastores which has crossed 80% utilization threshold to the Datastores which are less utilized in the Automated SDRS Cluster.

Issue persist even though below settings are in place for the affected SDRS cluster:

- "Enable I/O metric for SDRS recommendations"  is disabled

- "Keep VMDKs together" rule is ON

- Datastore Space utilization threshold is set to 80%

- No VM overrides configured in the Cluster

- Restarting the SDRS scheduler for the cluster.

- Storage DRS is migrating VMs off other datastore except the ones which crossed 80% utilization threshold. 

- Thin provisioned VMs are present in the cluster 

- VMs configured with bigger size disks

Environment

vCenter Server 8.x

ESXi hosts 8.x

Cause

This is an expected SDRS Cluster behavior.
The packing of the disks and the layout of the VMs is such that no moves are possible.

Resolution

MIN_SPACE_UTILIZATION_DIFFERENCE = 5 | This is the minimum required improvement in normalised entitled utilisation.

For the datastore cluster, even physically viable moves (VMs that fit within destination headroom) are rejected because the minimum move needed to fix the violation produces < 5% utilization improvement.

When VMs are configured as thin disk, SDRS adds a space buffer to account for peak space demand (PercentIdleMBInSpaceDemand) and this is set to 25%.

 

Use any of the below Workaround steps:

 

Workaround 1:

Reduce the PercentIdleMBInSpaceDemand from 25% to 10% or lower. That will bring down space utilization to below 80% on all datastores. It will reduce the 'datastoreEntitledSum' and  more space will be available in their environment.

To reduce PERCENT_IDLE_MB_IN_SPACE_DEMAND to 10% :

- Log in to the vSphere Client and navigate to the Datacenter.
- Select your Datastore Cluster from the inventory tree.
- Click the Configure tab.
- Under "Services," select Storage DRS.
- Click the Edit button to open the "Edit Datastore Cluster" dialog.
- Navigate to the Advanced Options section to find and modify PercentIdleMBInSpaceDemand.
- If the setting is not there after creating a datastore with Storage DRS enabled, you will need to add "PercentIdleMBinSpaceDemand" in the option setting and set the value at 10%, in the Advanced Options.

Workaround 2:

Manually migrate large VMs between the LUNs.

 

Additional Information

Storage DRS FAQ

Avoiding VMDK Level Over-Commitment While Using Thin Disks and Storage DRS