This message was deleted.

Harvester defaults to 3 replicas in the default Longhorn storage class. If you have 3 RKE2 VMs, each with 200 GB disks, then Longhorn would set aside 3*200 GB = 600 GB. That said, they are sparse, so if your RKE2 nodes only write 20GB each to their disks, then Longhorn will only have written 3*20GB = 60GB to disk across your Harvester cluster.

Assuming a fully populated image, each RKE2 VM replica is 200GB, correct? So, if there are 3 replicas each of 3 200GB disks, that's 200*2*3? Math wasn't always my strong suit so maybe I'm overlooking something obvious.

heh, I was wondering where the 2 came from

typing, apparently, isn't a strength, either

Don't forget to consider any hardware RAID redundancy you might have underneath 🙂

Pour some salt on that wound, why don't you! 😛

If the footprint is a little too high you can sacrifice some availability and put the RKE2 VMs on a Longhorn storage class with 2 replicas instead, and just nail down a backup/restore workflow

I am running harvester+RKE2 at home and I just turned down my VM replicas to 1. I'm managing the cluster with rancher and using the Auto Replace functionality to recreate nodes if they go unhealthy, and things have been okay. I wouldn't recommend for an enterprise solution though

Yeah, I'll look at options to reduce the footprint. The realization of how much storage is being consumed just dawned on me and I wanted to be sure that I was considering the situation correctly.

If they are only RKE2 nodes and ultra high availability is not a concern, I would suggest turning down the replicas and backing up the cluster via state snapshot or some other method instead of relying on the vm disk redundancy and prioritize multiple volume replicas for data volumes

That seems like a wise course of action.

In my homelab I've done away with multiple replicas and I back up VM disks to my NAS periodically if I have been too lazy to write a file-level backup solution for w/e the workload is (which is often lol)