Request for Validation of My Proxmox VE 9 Cluster Installation and Configuration

[zahmad]

Dear Proxmox Team and Community,

I have recently completed the deployment of a 3-node Proxmox VE 9 cluster for our production environment at my organization. The cluster was initially deployed on Proxmox VE 8.4 and has since been successfully upgraded to Proxmox VE 9. I would greatly appreciate it if the Proxmox experts could review my architecture and configuration to confirm that I have followed best practices and that the environment is ready for production.

• Proxmox VE: 9
• Cluster Nodes: 3 × HPE ProLiant DL380 Gen10 Plus Servers
• Shared Storage: HPE MSA 2060 SAN
• Storage Connectivity: Dual Fibre Channel (FC) with Multipath configured on all three nodes
• Cluster Communication: Dedicated network
• High Availability (HA): Configured and operational

The HPE MSA 2060 presents two shared LUNs to all three cluster nodes.

• Multipath is configured and operating correctly on all nodes.
• The shared LUNs are configured as LVM storage and are available on every cluster node.
• I am using the SAN-backed shared LVM storage for virtual machine disks.
• I am not using Ceph.
• I am not using lvmlockd or DLM. Based on my understanding of the current Proxmox documentation and my testing, I would like confirmation that this design is appropriate for my deployment.

Everything appears to be functioning correctly.

• Cluster quorum is healthy.
• High Availability is working normally.
• Live migration completes successfully.
• Shared storage is accessible from all nodes.
• Virtual machine creation, migration, and failover testing have been successful.
• Multipath is functioning correctly without any reported issues.

I would appreciate your expert opinion on the following:

1. Is this overall architecture suitable for a production environment?
2. Is using HPE MSA 2060 shared LVM over dual Fibre Channel Multipath a recommended deployment model?
3. Is it correct that lvmlockd and DLM are not required for this storage configuration?
4. Are there any configuration changes or best practices that you would recommend before I migrate all production workloads to this cluster?
5. Is there anything in my configuration that could affect stability, performance, or High Availability in the future?

If further technical information is needed, please let me know which command outputs, configuration files, logs, or validation steps you would recommend. I will provide whatever information is required to help you assess whether the cluster has been configured according to Proxmox best practices.


Our organization is evaluating Proxmox VE as the virtualization platform for our production infrastructure. If the Proxmox team confirms that our installation and configuration follow recommended best practices and are suitable for production, we intend to purchase Proxmox Enterprise Subscriptions for all cluster nodes so that we can benefit from the Enterprise Repository and official SLA-backed support. We want to ensure that our environment is correctly designed before moving our production services and making that investment.

Thank you for taking the time to review our deployment. Any recommendations or suggestions from the Proxmox team and community will be highly appreciated.

Kind regards,

Zahoor Ahmad
Director (IT)
Pakistan Institute for Parliamentary Services (PIPS)
@bbgeek17 @alexskysilk @t.lamprecht @LnxBil @Dunuin @Falk R.

 

[LnxBil]

Is this overall architecture suitable for a production environment?

Yes, we had this kind of setup running for years and started with PVE 3.4

Is using HPE MSA 2060 shared LVM over dual Fibre Channel Multipath a recommended deployment model?

I'm unfamiliar with MSA 2060, but if it has a dual controller setup and you're using either DA FC or two FC switches, it should be fine

Is it correct that lvmlockd and DLM are not required for this storage configuration?

Yes, for a couple of major versions, you don't need locking anymore unless you want to work with the lvm commands yourself. If you don't and PVE is handling everything, this is totally fine.

Are there any configuration changes or best practices that you would recommend before I migrate all production workloads to this cluster?

Have you considered trying lvm chain volumes in order to have snapshots? It's still experimental, though.

Is there anything in my configuration that could affect stability, performance, or High Availability in the future?

No, it's rock solid as long as anything around it that you haven't described is also HA, so at least two network switches and two controllers on the SAN and if applicable two FC switches. Also local RAID1 for OS.

The HPE MSA 2060 presents two shared LUNs to all three cluster nodes.

Why two? Are they both physical volumes in one volumegroup and bound the each SAN controller?