A New Setup and Configuration

[89east]

Hello everyone,

Could anyone please contribute to the following? We currently have the following hardware on hand and would like to determine if the proposed configuration below is ideal for our initial deployment.

We have:

- 2 HP DL380 Gen10 servers, each equipped with 64GB of RAM.

Our intended configuration is as follows:

- Proxmox Virtual Environment 8.0.3

We are also considering the following options for our storage pool:

1. A single ZFS pool, which combines four 480GB SSDs configured in RAID Z, resulting in an allocation of approximately 1.2 to 1.4 TB.

2. A second storage pool, consisting of two 4TB SATA drives in a Mirror configuration, providing a 4TB allocation.

If anyone has alternative configurations or suggestions, please feel free to share them.

Thank you in advance!

 

[LnxBil]

If anyone has alternative configurations or suggestions, please feel free to share them.

I really like using special devices, so you could have

• 4x 480 GB SSD in stripped mirror as special device
• 2x 4 TB as mirror

This will speedup your disks (metadata will be on SSD) and you can also use SSD for other stuff. You will not have the same amount of SSD storage available, but it'll be faster.

 

[89east]

Hi and thank you for the feedback

But with your suggested (4x 480 GB SSD in stripped mirror as special device) I will still be able to create a ZFS as I need it to have the Guest machines in High availability in case one of the physical nodes fails.

Thank you

 

[LnxBil]

High availability

There is no high availability with ZFS. You need replication and you will loose all data that is not transfered if one node crashes.

 

[bbgeek17]

@89east you will also want to have a qdevice/vote to allow for quorum (ie 2 votes majority) on node failure. Without it, you will have to go through few extra steps to force your cluster to operate in r/w mode on node failure (in addition to loosing any not-yet-replicated data).

Its very important to understand that unless you are able to do Synchronous replication, you dont have High Availability, you have Data Recovery. With Async replication the RPO is hopefully better than with backup, but this will depend on your link, rate of change and network stability.

Very few people can afford to do Sync replication as it implies a write confirmation on the remote side before the client can continue with IO. This requires a very low latency link and very fast storage.

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox

 

[LnxBil]

Very few people can afford to do Sync replication as it implies a write confirmation on the remote side before the client can continue with IO. This requires a very low latency link and very fast storage.

With local SSD and DRBD will you have a poor-mans-solution that just works. Used it for years many years ago. Untortunately, it's no longer available directly within PVE.