For proxmox, I’d say to use a single 256gb or lower ssd, go with either ext4 or zfs (won’t make any difference, I prefer ext4 for single drive, also more tested and stable in the linux world, but zfs should be fine as the root partition too). You don’t need zfs caching, neither l2 arc nor ZIL. Those are features more for enterprise users, or users with heavy things running on their servers, or with many, many users. Literally more cost for no benefit.
Since you already have the 1tb SSD, use it both for proxmox boot and root partitions and for VM’s root volumes. Use the zpool for bulk data storage (2nd vdisk for VMs, or SMB / NFS share from proxmox).
For 3 drives, no. Just go with 2 disk mirror. For 5 drives in raid-z, that’s the sweet-spot between capacity and redundancy IMO (and 6 for raid-z2, with less capacity, but a tiny bit more redundancy, which is a bit much for most labs anyway).
I’d say:
- mirror for photos & personal videos
- single disk for movies
- backup server with 2 disk mirror for important data and single disk for unimportant data (like movies), the latter which you can of course skip
If you’re feeling adventurous, you can buy an odroid hc4 and use the 2 HDD slots for important data backups and a (powered) usb hdd for your video collection (or, you know… just slap the USB inside your main system and back up the movies there, without having to go through your network). If not, an odroid h3 with the type-1 case should do (you could even run proxmox backup server on the h3). That’s if you want to save some power by powering off your backup server and its spinning rust that’ll just idle. Also, technically more secure to use a different box, so if something catastrophic happens to the main server, the backup server was intact (and powered off).
Do be mindful that if you want to expand in the future, you’ll want mirrors. You can have 2 disks in mirror now and add 2 or 4 more disks, each in 2-way mirror, then stripe all 4 / 6 of them (note that your pool won’t be balanced, meaning that data will still reside on the single original mirror, but if you delete data and restore it, it’ll be accessible again - for picture contents, it wouldn’t be a problem and you don’t need to worry about it being unbalanced).
Back to the backup server, if you want to save a buck, you can just run backups on a different pool, that will not be affected by the main pool. And you can take snapshots and zfs-send them to the other pool, all locally, avoiding the network, which is nice. So you can basically ignore the hc4 / h3 if you don’t care about expanding with more drives (and instead later expanding with higher capacity drives).