To use NVMe/TCP, I have to manually log into every worker node to install nvme-cli and load the right kernel modules. This is a lot of extra work, especially for large clusters. The process is described in the Trident documentation.

The Solution:
The trident-node DaemonSet should handle this setup automatically.
The trident-node-prep utility, which already prepares nodes for iSCSI, should be updated to also prepare them for NVMe. When a user enables the NVMe feature, trident-node should take care of:

1. Installing nvme-cli and related packages.
2. Loading the nvme-tcp kernel module.

This would make enabling NVMe much simpler and remove the need for manual configuration on each node.