Unfortunately the unmap and trim commands that esx issues doesn’t work on NFS based storage.  you have to run a tool like sdelete or fill the drive and purge it to reclaim the space.

I’m having the same throuble, but with linux nfs clients. I would like to offer a high available and low latency nfs storage.

I tried nutanix files, but I didnt like it because the high availability/load balancing features are poor.

So I created a storage container and shared it among linux nfs clients. I’ve tried sdelete, but it isnt practical, because sdelete is too slow. Is there another way? I did a nfs_ls in the storage container… no files listed, but prism reports storage container is full. Is there a acli or ncli command to free space?

Free space was restored about 3 hours later… so if you have frequent file creation/exclusion you need to set additional container space, proportionally to “3 hours of growth” to avoid run out space.

@DASR what you describe is basically working as designed. The mechanism is discussed in the article Why does my Nutanix cluster not show free or reclaimed space after deleting many VMs or files?

With many enterprise storage systems deletions are not high priority. This is only mostly true with Nutanix since additional work to clear free space is automatically triggered once a threshold is reached.

Generally speaking, we need enough free space in the cluster to be able to rebuild RF2 resiliency (or RF3 if configured) when a node (or two) fails. This is referred to as “rebuild capacity.”

For more detail on space usage recommendations on Nutanix, see the article “Recommended guidelines for maximum storage utilization on a cluster”

As to the initial question around unmap, ESXi, reclaiming free space, and sdelete there is still some work being done on this but at present running sdelete as needed is the best bet. The technical hurdle is specific to vdisks in an NFS datastore so the same problem wouldn’t be seen with volume groups, or for VMs on AHV.