Hey everyone! I'm setting up a Proxmox environment for a small business using Nextcloud and would love to get your insights before finalizing everything. Here's a quick breakdown:
I've got my main server at a family member's place in Guadalajara, Mexico, which has stable power and good internet. It's currently a Ryzen 3 Pro 2200G with 32GB RAM, but I'm soon upgrading to a Ryzen 9 3950X for more power. Right now, I've got three VMs running:
- Nginx Proxy Manager (2 cores, 4GB)
- A GPU VM with Jellyfin and around 30 containers (currently 4 cores, will increase to 8 after the upgrade with 16GB RAM, and an RX 580 GPU passthrough)
- The crucial Nextcloud VM (2 cores now, upgrading to 4 with 8GB RAM)
I store Nextcloud data on a ZFS mirror (2x 2TB WD Blue SSDs) for redundancy, while my homelab data lives on an 18TB HDD (single disk, media is redoable, so I'm not too worried about that).
For disaster recovery, I have two backup PCs located at different sites (one in the office and one at home) running Proxmox VE and Proxmox Backup Server. They connect to the main server via Tailscale VPN.
Here's the plan:
- Local backups every 2 hours (using vzdump to the 18TB HDD).
- PBS sync to both backup PCs after each backup through Tailscale.
- If the main server goes down, I'll manually restore the Nextcloud VM on the backup PC with the latest sync.
- Update Cloudflare's CNAME to point to the backup location.
- I'm aiming for a downtime of 30-60 minutes.
Monitoring runs on an InterServer VPS using n8n and Uptime Kuma, checking everything through Tailscale IPs, so I don't have to worry about dynamic public IPs. If something goes down, n8n sends a Discord message and email.
I prefer a manual failover approach; given we have only 10-15 users, I worry auto failover might cause issues like data corruption.
My backup PCs aren't super powerful – one is an i7-7700 with 8GB RAM and a 4TB HDD, and the other is a Ryzen 3 2200G with 8GB RAM, a 512GB SSD, and a 4TB HDD. During a failover, the Nextcloud VM would have about 6GB RAM – will that suffice?
I've put together a detailed PDF of the architecture, storage layout, backup strategy, and failover steps for anyone interested: [Link to PDF]
I'm primarily looking for feedback on:
1. Is my backup strategy solid enough?
2. Should I stick to manual failover for this scale?
3. Are there any issues with running PBS alongside PVE on the same machine?
4. Will 8GB RAM on the backup PCs during failover be sufficient?
5. Is there anything blatantly wrong or missing in my setup?
6. Would you trust this setup for a small business?
Any and all advice would be greatly appreciated, even if it's just to say I should do something differently! Looking to get this right before we bring on users. Thanks in advance!
2 Answers
Sounds like you're on the right track! Your backup system with vzdump and PBS over Tailscale is well thought out. I’d suggest that manual failover makes sense for just 10-15 users. Automated failover could lead to issues you’d rather avoid. Regarding PBS running with PVE, it should work fine, just ensure PBS has enough resources—at least 2 cores and 4GB of RAM would be wise.
8GB RAM during failover should be acceptable for Nextcloud, just confirm you're not running additional VMs while you're doing a restore. Testing your restore process before needing it is crucial—run a full restore to make sure your RTO is as expected.
Your setup looks solid for the number of users you have. For your backup interval, it might be worth considering tightening it to every hour or even 30 minutes just for Nextcloud, since it’s business-critical. Incremental backups using PBS after the initial setup are usually minimal in terms of bandwidth, so that shouldn’t be a major concern.
Make sure to actually test your disaster recovery plan. The 30-60 minutes you’ve planned for manual restores is realistic, but it can stretch longer without proper documentation. Writing a runbook with exact steps would be helpful for when the moment arises. Don’t forget about power outages—having a UPS for each backup site can prevent corruption during a sync. Also, B2 or another offsite cloud backup could be a good idea for an additional safety net!
Thanks for the tips! Definitely will test the process out.
Will do! I appreciate your help with that!