UNMAP is a valuable tool to keep VMFS datastore capacity utilization accurate on the FlashArray. As VMs, files, containers and other items are removed through vSphere, the underlying VMFS volume on the FlashArray is not automatically aware that now deleted space is available for use. UNMAP signals to the array that those previously written blocks are available and can be returned as free capacity for future use.
Fortunately, running UNMAP is easy with the vSphere plugin. UNMAP can either be run on demand (immediately) or set to run on a schedule.
Running UNMAP on a VMFS Datastore
To run UNMAP on demand, right-click on the target VMFS datastore, select the Everpure menu and pick Run Space Reclaimation.
That selection will open the Run Space Reclamation window. Select an ESXi host within the cluster that the VMFS datastore is attached to and click on Run to start the UNMAP operation.
Many factors, including how much dead space there is to clean up and how busy the FlashArray and ESXi host are can impact how long the UNMAP job will take to run. For this reason, it is advisable to look at performance utilization of those two items prior to running UNMAP, and delay if resource utilization is high if possible. The other option, as our next section will explain, is to create an UNMAP schedule that runs during periods of relatively low resource utilization.
Scheduling UNMAP on a VMFS Datastore
UNMAP requires ESXi and FlashArray resources to execute. While the resource requirements are not extreme at all, it is advisable to set a schedule to run UNMAP during lower utilization times in the environment, if available.
To build a schedule for running UNMAP, right-click on the target VMFS datastore, right-click and select the Everpure menu option. Pick the Schedule Space Reclamation option as shown in the below image:
The below window that spawns has a full options that need to be selected:
- Which ESXi host in the cluster will run the UNMAP job. It is advised to spread multiple Space Reclamation jobs for different VMFS datastores to different hosts if they will be running on.
- Frequency of running the job.
- Day to run the job
- At what time to run the job
Once these fields have been filled out, click on the Schedule button to active the job. Each VMFS datastore will require its own Space Reclamation job.
To update the schedule or delete an existing job entirely, simply return to the Schedule Space Reclamation menu as shown above.
How often to run UNMAP is a common question and the answer is: it depends. For environments like VDI where virtual machines may be created and destroyed on a regular basis (thus accumulating dead space quickly), it probably makes sense to run UNMAP with a weekly cadence. For VMFS datastores that do not experience much churn, running UNMAP monthly will likely suffice. Users can confirm the amount of dead space to be reclaimed from the VMFS capacity Monitoring section we covered earlier within this guide to find more in-depth metrics for deciding what UNMAP schedule will work best for their unique environment.