How to use *Thick* disks efficiently in production without excessive creation time

ccesario Lv1Posted Oct-14-2025 22:56

Hello everyone,

I'm facing a challenge with disk provisioning on Sangfor HCI.

In my environment, I usually use Thick Provisioned disks (as I do on VMware, Hyper-V, or Proxmox) because I prefer to allocate all the required storage upfront — avoiding thin growth behavior.

However, I noticed that on Sangfor HCI, both Dynamic and Thin Provision disk types actually behave as thin provisioned, while only the Pre-Allocated type behaves like thick provisioned (reserving all the disk space at once).

The issue is that when I create a pre-allocated disk (for example, 5 TB), the process takes an extremely long time — more than 5 hours, even tested on two different storage back-ends.

My questions are:

Is there any recommended method to use thick disks efficiently in production without this long creation time?

Does Sangfor provide any alternative or optimization for pre-allocated disks (for example, background initialization or lazy allocation)?

Is there a way to have the real space consumption correctly reflected for thin/dynamic disks, similar to other hypervisors?

Any suggestions or best practices would be greatly appreciated.

Thanks in advance!

Zonger has solved this question and earned 10 coins.

Posting a reply earns you 2 coins. An accepted reply earns you 20 coins and another 10 coins for replying within 10 minutes. (Expired) What is Coin?

Enter your mobile phone number and company name for better service. Go

In Sangfor HCI Dynamic/Thin disks allow storage over provisioning by design, means total virtual disk size can exceed physical capacity. There is currently no hard limit to prevent this unlike with Pre-Allocated (thick) disks, which reserve space immediately.

To maintain control admins should use storage usage alerts, set vStore quotas to cap per tenant capacity and monitor datastore utilization closely. While strict prevention of over-provisioning isn't yet supported.

Sangfor has indicated that future versions will include quota enforcement to balance flexibility and capacity safety.
Is this answer helpful?
Zonger Lv5Posted Oct-15-2025 05:47
  
Thick Disk Creation Time

In Sangfor HCI the Pre-Allocated disk type is the true thick-provisioned option as it reserves the full space immediately on creation.

The long creation time (e.g. 5 TB = ~5 hours) is expected because the system initializes and zeroes all allocated blocks to ensure data consistency and performance predictability.

This is a safety and reliability mechanism not a fault.

Best practice: For large disks, perform pre-allocation during off-peak hours or pre-stage disks before service deployment.

Optimization or Alternative

Currently Sangfor does not support lazy or background initialization for thick disks.

If deployment speed is more critical than full pre-zeroing Sangfor recommends using Dynamic disks.

Dynamic disks in Sangfor HCI automatically allocate blocks on demand and are optimized to minimize fragmentation making them suitable for most production workloads.

Space Reporting for Thin/Dynamic Disks

Sangfor HCI storage currently shows logical allocation not actual space usage at the VM level.

Actual space consumption can be monitored at the storage pool level in the HCI console.

There’s no per-VM real-time “used space” reporting yet this is by design as storage deduplication and compression dynamically adjust usage behind the scenes.


Recommended Practice

Use Pre-Allocated disks only for workloads requiring deterministic I/O performance like databases.

Use Dynamic disks for general production VMs as they are safe, efficient and perform comparably in most cases.

Always deploy thick disks ahead of time if mandatory as pre-allocation time scales linearly with capacity.
ccesario Lv1Posted Oct-15-2025 18:46
  
Last edited by ccesario Oct-21-2025 23:22.

Hi @Zonger,

Thank you for the detailed explanation, it makes perfect sense.

However, in my case, I can’t always choose the best time to create pre-allocated disks — sometimes new VMs or storage expansions are required during peak hours, so the long creation time becomes a real bottleneck in production.

Regarding the Dynamic disk type, I understand it’s optimized and performs well, but one concern I have is about capacity control.

When using Dynamic disks, Sangfor HCI allows creating large VMs even if the total physical storage capacity is insufficient - meaning it’s possible to over-provision the datastore far beyond the real available space.

With Pre-Allocated disks, this issue doesn’t happen because all the space is reserved immediately, ensuring that a 30 TB datastore will never be over-committed.

So my question is:
Is there any way to improve capacity control when using Thin or Dynamic disks — for example, to prevent allocating more virtual space than physically available in the datastore?

This would be extremely helpful for environments where Dynamic disks are preferred for performance, but where strict storage control is still required.

Thank you again for the support and clarification!
Zonger Lv5Posted Oct-15-2025 18:55
  
In Sangfor HCI Dynamic/Thin disks allow storage over provisioning by design, means total virtual disk size can exceed physical capacity. There is currently no hard limit to prevent this unlike with Pre-Allocated (thick) disks, which reserve space immediately.

To maintain control admins should use storage usage alerts, set vStore quotas to cap per tenant capacity and monitor datastore utilization closely. While strict prevention of over-provisioning isn't yet supported.

Sangfor has indicated that future versions will include quota enforcement to balance flexibility and capacity safety.
ccesario Lv1Posted Oct-22-2025 00:37
  
Hi @Zonger,

Thank you for the clarification — that makes complete sense.

As I mentioned earlier, I tested the Pre-Allocated mode specifically to maintain tighter control over capacity usage, but the creation time is extremely long, which makes it impractical during production hours.

Regarding the vStore quotas you mentioned, I assume these only apply when using the SCP (Sangfor Cloud Platform) or virtual storage pools.
In my case, I’m using external storage connected via iSCSI, and from what I can see, these quota settings aren’t available or enforced in that configuration.

Thanks again.
Newbie183591 Lv1Posted Jan-14-2026 15:13
  
Thin provisioned is an elastic of storage it is better for DB and long running servers.

I Can Help:

Change

Board Leaders

NyxZale...

Weekly Sharers

Newbie5...

Weekly Questioners