Free scpace discrepancy between Nimble and VMWare

[Fred Blum]

We are using Nimble HF20C with OS 6.0.0.200-927971-opt and VMWare 7.0.3. 

We are seeing a difference in free space representation between the Nimble Volume and the VMWare Volume. On the Nimble Datstore1 there was still 1TB of free space while VMWare gives and error during backup as only about 150MB is free. according tpo VMWare 

Overall Nimble has 18% free. Nimble Node 1 17,5% 12,3 used of 14,8TB and Node 2 18,5%.

Node1 says it still has 2,6TB free and the datastore1 on Node1 says it has 1TB free.  

The WMware datastore1 says it only has about 150MB free giving an error when creating a snapshot during backup. I removed some lifecycle snpashots that were not automatically removed. That gives me say 150GB of free space but it still doesn't add up to the 1TB as reported by the Nimble.

Any idea's?

Regards,

Fred

[Fred Blum]

Used 16,6TB of 14,8TB?

When I look at the storage pool than currently used would be over 16,6TB but that is I believe before savings. So effectively used 12,3TB. The usable size of a node is 14,8TB.

I found a post on block reclamation. Do we need to periodically reclaim blocks on the VMware volume?

Regards,

Fred

[Fred Blum]

Block reclamation would be the other way round. So the SAN showing more used space compared to the OS. 

I did not see any VMWare alerts regarding Consolidation Needed and there are no snapshots showing anymore in the vcenter GUI.  The datastore1 VM csv export shows a kolom header first line Space used and second line Consolidation neeeded. There are no backups currently running. 

When I add them up I see the used space difference again as compared to the SAN. 

[Sheldon Smith]

Hello,

It's hard to say without doing a deep dive on the system. Offhand, I'd chalk it up to deduplication and compression by the Nimble. Have you asked Nimble Support?

And on an side note, 6.0.0.300 is now GA.

[mamatadesaiNim]

Please give a call to Nimble Support.  They can help look into this and unravel the mystery.

[Fred Blum]

The backup succeeded without any errors or warnings. The Datastore1 now shows even more free space on the Nimble Gui. VM Ware vcenter has not changed. 

I am not sure if it is a WM Ware or Nimble issue. I have contacted Nimble Storage support.

Regards,

Fred

[PaulBraun1]

Hi Fred,

please open a Nimble CASE (support@nimblestorage.com) to determine this issue.

Warm regard,

Paul

[Fred Blum]

I have contacted Nimble Storage Support. They refered me to the Block Reclamation support doc. But in our case it is the other way round,, VM Ware shows more space consumption than the Nimble GUI. and in VMFS6 auto space reclamation should correct that automatically, 

They are now refering to VM Ware.

[Fred Blum · Accepted solution]

The Datastore that has the biggest difference also has more disks that are thick provisioned but have lazy zeroed instead of eager zeroed., so not claimed at storage level. The free space of thick provisioned disks are reserved in VMWare while cakulated as available on the Nimble Volume. 

So to solve it we can change it to eager zeroed so the space is actually claimed at storage level or keep it in mind that this accounts for the difference at VMWare level and at Volume level.

Thick provisioning is normally used for performance reasons with high IO disks as expanding the disk takes time. So lazy zeroed seems to be a internal contradiction in this case.  So we might also go for thin provisioned when the performance improvement is unnoticable to users and processes.  

Regards,

Fred

[plasticwax]

@Fred Blum 

any chance you can post link to this space reclaim document? 

why are HPE engineers so useless that they recommend the WORST version of software Nimble has ever produced? for the love of god dont go past 5.2.1.900

[Fred Blum]

We are at version 6.0.0.400-991061-opt running stable for years. This was not a matter of space reclamation but of  lack of knowledge of VMware on my side. See the solution give.n. 

VMWare reserves space in it's tables but not really on the SAN when thick provisioned is used with lazy zeroed.  So if you really need thick provisioned and use eager zeroed the space will also be claimed on the SAN. 

If you want to reclaim space on VMWare and have thick provisioned datastores with lazy zeroed than changing it to thin provisioned will bringe the values of VMWare and the SAN in line.