FLASH MODE for Volume Groups and Virtual Machines

  • 5 March 2021
  • 0 replies

Userlevel 2
Badge +4
  • Nutanix Employee
  • 69 replies

The flash mode for VM allows to set the storage tier preference to SSD for a virtual machine or a volume group. 

  • Without flash mode, data for a mission critical application such as a relational database can run out of the room in the SSD tier because other workloads running on the same cluster. 

  • When this happens, the database could potentially migrate to the HDD tier.

  • For extreme latency sensitive workloads, this migration to the HDD tier could negatively affect the read and write performance.


By default, you can use up to 25% of the cluster-wide SSD tier as flash mode space for VMs or VGs.

  • If the data size for flash mode enabled VMs or VGs exceeds 25% of the SSD capacity, the system may down migrate the data. 

  • Before down migration, the flash mode feature tries to preserve the excess data on the SSD tier for some reasonable amount of time so that you can take corrective actions on the cluster and bring it back to stable.

  • To reduce flash mode usage, we can disable the flash mode on some Ms or VGs or add SSDs


If flash mode feature is enabled on the VM, all the virtual disks that are attached to the VM are automatically moved to the SSD tier.


Caution: Enabling flash mode on some VMs might increase their performance, but can lower the performance on other VMs on which this feature is not enabled.


For VMs - This feature is supported for ESXi and AHV

For VGs - This feature is supported for all hypervisors.


Enabling Flash mode:

For Vms -- We can enable the feature on the VM only during the VM update workflow

For VGs -- We can enable the feature on the VGs while creating them.


From cli:

Ssh into the CVM and access acropolis command line (acli)

Execute the following command to enable flash mode feature for the Volume Group

acli> vg.update vg_name flash_mode=true

For more information, please refer to Flash mode for VMs and VGs

This topic has been closed for comments