Hi everyone! I'm in the process of setting up my Proxmox environment and I'm looking for tips on creating a system that's robust, reliable, and scalable for the future. My setup involves using a Dell PowerEdge R430 for compute resources while I'm planning to have storage on a separate node. Here's some of the hardware I'm working with:
**Compute Node (Proxmox Host):**
- Dell PowerEdge R430 with PERC H730 RAID Controller
- 2 Intel Xeon E5-2682 v4 CPUs (32 threads total)
- 64 GB DDR4 ECC RAM
- 2 1.2 TB 10K RPM SAS drives in RAID 1
- Various additional storage options
**Storage Node:**
- Intel i7 11th Gen with 8 GB RAM
- 4 2 TB HDDs running Proxmox Backup Server
**Network Setup:**
- Configuration includes a Firewall leading to the Proxmox host and then to the Storage Node.
I'm primarily seeking advice on the following:
1. How can I optimize this setup for robustness and performance? Should I go with NFS, iSCSI, or ZFS replication?
2. Is my current 1 GbE networking sufficient, or should I consider upgrading to 10 GbE for better performance?
3. What would be the best ZFS setup or caching strategy for my storage node?
4. Any suggestions on redundancy or failover solutions between the compute and storage nodes? Thanks in advance!
3 Answers
It sounds like you're building a solid home lab, but there are a few things to consider. First off, relying on a single compute node does introduce a point of failure; if it goes down, your whole setup is impacted. I recommend thinking about redundancy—maybe plan for a second compute node down the line.
For your storage, using iSCSI over 1 GbE might give you decent performance, but I’d strongly advise moving to 10 GbE or more as your workloads grow; 1 GbE can be quite bottlenecking especially with multiple VMs running. And, absolutely, if you're hitting serious I/O needs, those spinning HDDs may become a limitation for performance. For ZFS, you could start with basic mirroring and consider caching later on if necessary!
ZFS is definitely a great choice, but make sure you have enough RAM—more cache will help with those slower disks. I’d also suggest experimenting with setting up a solid backup solution alongside.
Honestly, with just the one compute node, it’s tough to call it robust. A SAN would do wonders for your reliability, but I get that it might not be feasible right now. Also, splitting compute and storage is a good idea, but with your current hardware, managing network traffic could be challenging—seriously consider upgrading to 10 GbE to prevent bottlenecks.
Your setup is interesting for a company, but I'd be cautious. Maybe since you're limited to the current hardware, focus more on optimizing what you have and less on getting additional nodes for now. Look into iSCSI for storage connection; it's manageable over your existing network for the time being and you can upgrade components like the NICs later on!
Totally agree with the need for better networking. For a server setup that might see growth, 10 GbE should be a bare minimum, and if you're on a budget right now, maybe just start planning for the switch soon.