This cluster houses several very large VMs. Each VM has been allocated 30TB of disk space.
From the OS (rhel7) I show a different amount of space used on every host. the discrepancy ranges from 2TB to 8TB.
Here is an example...
$ df -h /nb_pool00
Filesystem Size Used Avail Use% Mounted on
/dev/mapper/datavg-nb_pool00lv 30T 13T 18T 42% /nb_pool00
...yet, from prism i see...
17.17 TiB / 30.1 TiB
...a 4TB discrepancy.
So far i have ruled out the following:
1) Deleted files hanging on. lsof shows only 80MB of files that are held on to right now from things like files being deleted while they were being read.
2) Snapshots. I have no snapshots listed in data protection for these VMs.
As a percentage, 4TB out of 30TB miscalculated is not very large (13% error). However, when considering the raw size of the wasted disk combined with the nuber of VMs, this miscalculation has been putting my cluster at risk.
In VMware there was a way to evoke a process that reclaims blocks that had been written to but then later removed. it was evoked during a storage vmotion or it could be called directly. Does Nutanix have any way to reclaim this unused space, or am i forever doomed to have poor disk utilization? If there is no current solution to this problem, what steps can i take to have this raised with the development team. I find this sort of thing to be an urgent and foundational problem, and a solution here would let Nutanix brag about data density.