Experimenting with 2 nodes (+ QDevice) in my homelab. Old PCs, no enterprise-grade hardware. I have an LXC with a bindmount to a local disk (500GB, no shared storage available) on node1. The disk is not setup as a proxmox data store. Data rarely changes on this disk. Similar local disk available in node 2. Hmmm, need to sync those somehow. PVE-zsync can't use recursion, so I can't use that one. Also, I need to reverse the direction of the sync, when I move the container to node2. Perhaps not automatically though, or without deletions...
My idea is to keep the disks in sync using rsync node1->node2, as long as my LXC is running on node1. Of course, the container is unaware of the underlying proxmox host. I'd like to get your feedback on this idea:
Have some kind of file on node1/disk/direction.txt that is excluded from the sync.
Have the same kind of file on node2/disk/direction.txt, also excluded from the sync.
In my LCX, read the file and sync based on the direction read.
I admit it's wacky idea, but I strongly feel that gluster, CEPH and so on would be far too over the top.
What do you think? How did / would you solve this?
I also thought about syncthing installed directly on the proxmox nodes, but I'd prefer to not mess around with them.
Other ideas are welcome.
My idea is to keep the disks in sync using rsync node1->node2, as long as my LXC is running on node1. Of course, the container is unaware of the underlying proxmox host. I'd like to get your feedback on this idea:
Have some kind of file on node1/disk/direction.txt that is excluded from the sync.
Have the same kind of file on node2/disk/direction.txt, also excluded from the sync.
In my LCX, read the file and sync based on the direction read.
I admit it's wacky idea, but I strongly feel that gluster, CEPH and so on would be far too over the top.
What do you think? How did / would you solve this?
I also thought about syncthing installed directly on the proxmox nodes, but I'd prefer to not mess around with them.
Other ideas are welcome.