vague drift
(1) what happens to the data on storage returned to NFS datastore, for example when a VM is deleted? Is the storage simply marked available for provisioning to new requests for storage or does it first go through some kind of erasure or zero'ing-out process?
(2) Is the process for NFS volumes the same as for VMDKs?
Appreciate your inputs
neat shoal
with NFS - ONTAP handles clean up.
Can you claify on #2 ? you mean if there is a delete inside the vmdk?
vague drift
#2, I meant, in the case of just regular NFS volumes mounted on Linux clients.
I know ONTAP handles clean up. But, Just wanted to know details, if ONTAP just simply mark blocks as available to use for next write requests?
Also, how often or when the "space reclaimation" (space deletion) will be kicked off?
neat shoal
it's a background process.
IIRC it operates on % free for when it kicks off.
obsidian steppe
If you delete files on your NFS-client and your volume is thin-provisioned, it should be pretty simple as mentioned.
But if you mean inside files (like with NFS datastores) you would need NFSv4.2 deallocate support which I'm currently not 100% sure if ONTAP supports it.
glacial olive
The issue is that inside the vmdk is a file system like ntfs. When you delete a file, if I recall it’s not really deleted, it can be recovered. Hence ONTAP can’t clean up. There are tools out there to clean up free space and really delete files. Then if the vmdk is thin provisioned, ONTAP will do its magic if efficiencies are enabled
vague drift
@glacial olive So, if I deleted files from a VM, that part of blocks wouldn't be released for reuse and couldn't be reflect by volume show or df commands?
glacial olive
Think about the file system INSIDE the vm. It’s all dependent on that. If you place a 10g file with random bits (not sparse and not empty) on a windows file system inside a vm it will consume 10g
If you now do a normal file delete, what happens? The blocks do not change, but the file system will manipulate pointers to say those blocks are available to reuse if needed.
ONTAP has no idea what is going on there!
Once the vm actually erases the blocks, maybe using some secure wipe, that equates to a file system change that ONTAP sees. If it zeros out those blocks, ONTAP will use dedupe/compression to reduce footprint (if enabled)
vague drift
So, Unlike deleting files in a VM, in the case of deleting entire VM for instance, ESXi hosts will issue an order to request ONTAP and then disk array to release the part of blocks for reuse, right?