Storage recommendation to be used from Proxmox 8 .4 and 9.0

Jun 14, 2022
20
0
6
Hi Team,


We currently have two new Proxmox clusters — one running version 8.4 and another on 9.0 — each with 4 nodes. So far, we’ve been using Directory-based storage, but we’re planning to implement Proxmox Backup Server (PBS) for managing backups.


Our setup includes both containers and virtual machines with high resource utilization, and we need to ensure that snapshots and backups can be taken while the systems are running. Additionally, we have SAN storage connected to these nodes.


We would like your recommendation on:


  1. The best storage type to use for a production environment with our requirements.
  2. The recommended approach for handling storage failover or migration — i.e., if one node goes completely down, how can the storage be seamlessly migrated or accessed from the remaining nodes?

Kindly advise on the optimal configuration for performance, reliability, and snapshot support.
 
Hi @harikrishna075,

The first thing to clarify - there’s no single “best” way to do something in IT. It all depends on your goals, constraints, tolerance for risk, and budget.

Here are the general approaches you can take:

a) Use your existing hardware infrastructure and make the best of it.
With PVE 8 and PVE 9, you can take backups that shouldn’t impact running VMs. However, if your backup infrastructure is undersized or slow, you will likely experience issues.

PVE 9 also introduces the ability to take snapshots on external SAN storage. This feature is currently in preview, with ongoing development. Only you can decide whether it meets your reliability requirements.

Keep in mind that many members of the “big storage cartel” don’t officially support Proxmox VE and may refuse assistance if you need help. You’ll need to decide whether that’s acceptable for your environment.

b) Build a new greenfield environment.
In this case, you can choose one of the fully supported storage types built into PVE (most commonly Ceph), or pick a storage vendor that officially supports PVE.

Finally, your performance criteria are not defined.
Are you optimizing for IOPS, bandwidth, or latency? All three depend on your budget. If you plan to keep your existing compute, network, and SAN, you should already be aware of the performance limits in your environment.

Best of luck,


Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox
 
The best storage type to use for a production environment with our requirements.
Lets begin with the obvious. what are you using now? I presume it some sort of NAS since you're using "directory-based storage."
Additionally, we have SAN storage connected to these nodes.
Is this besides the above? in addition to?

i.e., if one node goes completely down, how can the storage be seamlessly migrated or accessed from the remaining nodes?
Either a NAS or a SAN can handle this, so it looks like you already have it covered.

What pain points are you experiencing with your current solution?
 
Lets begin with the obvious. what are you using now? I presume it some sort of NAS since you're using "directory-based storage."
San storage with directory-based storage
Is this besides the above? in addition to?


Either a NAS or a SAN can handle this, so it looks like you already have it covered.

What pain points are you experiencing with your current solution?
We are planing to take backup using PBS at that time if we are using directory container will be in suspended state which will affect our production.
 
I think we may be experiencing some semantical difference.

"SAN" or Storage Area Networks typically mean shared block storage- eg iSCSI, FC, NVOF or even SAS/SCSI.

"Directory Storage" means (to me) locally mounted filesystem, either on exclusive (local) disk or network storage device (colloquially, NAS, using a network storage protocol such as NFS or SMB/CIFS)

So long as your storage is NOT on local disk, it can be used for cluster purposes.
We are planing to take backup using PBS at that time if we are using directory container will be in suspended state which will affect our production.
Thats a different matter. as long as your storage METHOD allows for Copy of Write which allow snapshots in turn, you can take a backup without disruption of the guest. On "directory" shares that means qcow on-disk format- if you use anything else (eg, raw) snapshots are not supported.
 
On our directory based nfs storage we do snapshots for images raw, qcow2, efi and tpm disks (live while freeze for seconds), not through the webgui but "unsupported" works fine. After generated the snapshots could be backuped to other hosts and running vm's are not further affected by..