Has anyone come up with a good solution to use large scale fibrechannel attached SAN storage on Harvester, in the same way vmware VMFS would? The context is converting/migrating a 15node/1000VM vmware cluster to Harvester, and the majority of workloads are traditional monolithic VMs. So far I've explored; • Presenting FC luns to Longhorn, but this means I must present unique LUNs to each node, and either; a. Replicate all data - which costs VM IO performance, doubles the IO to the SAN, and relies on SAN deduplication. b. Have a single replica - which solves the performance, but drastically limits VM portability across nodes to optimize CPU/Memory consumption, and means I have a significant outage if I lose a node (or even need to perform maintenance on it). • SAN CSI driver - Which solves both of the above limitations by provisioning and exporting a dedicated LUN per volume, but does not scale to 1000 VMs (and 1500+ volumes). Each node's FC card limits would quickly be exceeded, and I'm warned that a large concurrent activity like a node failure would likely smash the SAN's API to remount all of the LUNs to new nodes. Keen to hear any alternatives people have come up with!

I would switch to Iscsi and use the vendors csi. If the array is good enough it will scale without issues.. longhorn will not

we integrated HPE alletra SAN with harvestor and it auto-creates volumes directly on the SAN whenever we deploy VMs. We havent tried the migration process yet but I am guessing it shouldnt be any different from new VM creation.