I had many problems getting the hyper-converged fail-over cluster to work. First i tried to convert a Windows server standard to a Server datacenter cluster node. Something went wrong. Every time i shutdown a node all volumes went offline. They should only be degraded, but still available. The slabs were placed wrong. It was like Storage spaces direct placed slabs anywhere without caring about both must not be on the same node. I started over from scratch. Wiped all drives with Linux wipefs. You cant wipe a storage space drive in Windows. This time i could shutdown a node and the volumes got only degraded. It takes a long time to re-synchronize the volumes. Next problem was that a NVME SSD was not automatically used as cache. I tried to set it as journal with a powershell command. It changed to journal, but no space was used on it. I tried to remove it and add it. Then i remember that i had a Samsung NVME SSD that was used as storage spaces direct cache in another. I removed the Kingston SSD and put in the Samsung. Immediately storage spaces direct turned it into a cache drive. I dont know why it did not like the Kingston. I noticed that writing to the cluster volumes were slow. Turns out that storage spaces direct turns off disk caches if the drives dont have Power Loss Protection. Even a NVME SSD gets slow without write cache.
Categories