I have a cluster with a number of nodes (EC2 instances) which, under considerable load, can die and be cycled back into the cluster by RKE. I want to run Longhorn on both those nodes and a few other, much more reliable ones. I have three nodes dedicated to Longhorn, one in each availability zone, and am storing my RWX volume replicas there. I can attach my volume to one of these dedicated nodes, but the moment I mount the volume to a deployment by using a PVC, the volume attaches to any random node Longhorn is installed on, not necessarily the one I selected through the UI.
Is there any way I can reliably schedule the Share Manager pod to a given dedicated node? I realize the drawbacks with high availability and NFS, but would like to avoid running the pod on any of my more unreliable nodes.
Just want to confirm that this feature still does not exist before opening an enhancement issue as I have another use case that would benefit from this feature.
How can I maintain a balance between high availability requirements and the need to ensure critical components like the Share Manager pod are consistently hosted on reliable nodes within my cluster?