Product Announcements

Using Datastore maintenance mode for migration virtual machines to Datastore clusters

By Frank Denneman, Sr. Technical Marketing Architect

Storage DRS introduced the feature Datastore Maintenance mode and this new feature can be compared to Host Maintenance Mode. When a datastore is placed in Maintenance Mode all registered virtual machines on that datastore, are migrated to the other datastores in the datastore cluster.  Datastore maintenance mode can be used to safely migrate virtual machines out of the datastore for storage array related maintenance operations such as migrating a LUN to another RAID group, however it is also an excellent feature that can help you during a storage migration project migrating virtual machines from standalone VMFS3 datastores to VMFS5 datastores aggregated in a Storage DRS datastore cluster.

Moving virtual machines manually out of a VMFS3 datastore to a datastore in a datastore cluster is a time-consuming operation. Fortunately by selecting the datastore cluster, Initial placement will help you select the appropriate datastore, however you still need to manually start a Storage vMotion process for each virtual machine and manage the migrations until the last virtual machine is migrated.

Datastore maintenance mode on a VMFS3 datastore
By adding a VMFS3 datastore to a datastore cluster and placing the datastore in maintenance mode Storage DRS will take care of the tedious process described above. When the VMFS3 datastore inside a datastore cluster is placed into maintenance mode, migration recommendations are generated to empty the datastore. Storage DRS reviews the virtual machines on the datastores and distributed them across the datastores in the datastore cluster. Storage DRS finds an optimal placement based on the space utilization of the virtual machine and move them to selected datastores while considering the space utilization threshold of the datastore cluster.

Depending on the Storage DRS automation mode, it will generate a list of recommendations (Manual mode) or will automatically execute migrations of the virtual machines (Fully Automated).

Next steps
After all the virtual machines are migrated, Storage DRS will indicate that the datastore is in maintenance mode. At this point you can remove the datastore from the datastore cluster.  If the datastore is going to be reused, I recommend to reformat the empty datastore to VMFS5 filesystem. By reformating the datastore instead of upgrading, the datastore is using the unified blocksize. One of the advantages of having datastores configured with a unified blocksize is the possible improvement of storage vMotion lead times. If the array is VAAI-enabled, and the datastores use the same blocksize, Storage vMotion can leverage the FS3DM –hardware offload datamover.  Using hardware offload will offer maximum performance while creating the least host CPU and memory overhead. Hardware offload will decrease the lead time of the Storage vMotion process allowing Datastore clusters to reach a balanced steady state faster, impacting I/O performance of virtual machines in the datastore cluster in a positive manner.

If you want to know more about the the FS3DM datamover, I recommend reading the article “Blocksize impact” and “Storage vMotion Performance Difference” on Yellow-Bricks.

After giving the datastore a fresh coat of VMFS5, add it to the datastore cluster and add another VMFS3 datastore and rinse and repeat the datastore maintenance mode operation on another loaded VMFS3 datastore.

Be aware that Storage DRS will only generate migration recommendations for virtual machines that are registered in vCenter; orphaned virtual machines, ISOs or other non-related files will not be migrated. When executing migration recommendations, vCenter only considers the remaining datastores of the datastore cluster as destinations. Datastores outside the cluster are not considered as suitable and compatible destinations.