For the default longhorn storage class, should my

replica's

be the same count as the number of work nodes, or would it be one less than the number of worker nodes (n-1)? Thinking if I deploy a Pod/PVC and it gets assigned to a node, wouldn't the replica's only need to exist on the worker nodes that are not running the Pod workload?

Replicas indicates the total number of places the data is stored. So if you want the data on 3 nodes (1 on the node with the pod, and 2 more on other nodes), you would set replicas to 3.

TY

sometimes people get confused, thinking that replicas or copies refers to only additional copies of the data, as in a read-only database replication scheme, where you have a primary and then replicas. that is not the case here.

Thank you very much for the explanation. I'm new to using longhorn and have been trying to set it up on a sandbox cluster and have run into all sorts of issues. I keep going back to the docs and I'm not finding clear answers or explanations on how this works. A lot of my problems I believe to be related to the initial setup of my disk on my nodes and linking those for Longhorn to use. All the videos and everything I've come across go straight into installing it and using it but I feel like my initial setup is lacking and can't find resources to help me understand how the initial setup should look

Converged storage can be really resource intensive. Longhorn really wants to run on hosts with SSD and 10GBE. If you’re running it with less than that on anything except for a very lightweight evaluation workload I would expect to see it being pretty unhappy.