[BUG] (NVMEoF) scale-nvmet-start.sh fails to start on TrueNAS SCALE 23

[crutonjohn]

First off, thank you all for your hard work on this project.

I has come to my attention that upon upgrading from TrueNAS SCALE version 22 ("Bluefin") to version 23 ("Cobia") the scale-nvmet-start.sh script fails to run. I believe there were some underlying changes made to the SCALE system Python that is preventing this from working.

Since I depend heavily on NVME-oF working I quickly threw together a script to get this working seen here in a gist that refactors the script a bit to use functions and more importantly sets up a Python virtual environment to get everything working again.

This needs some further investigation but I only have the one TrueNAS system and use it for some important things so I can't really provide any more testing data other than the gist.

If this is found to be problematic for others I can upstream my script to this project. I'm thinking that they should either be broken up into 22-scale-nvmet-start.sh and 23-scale-nvmet-start.sh or perhaps someone could test my version of the script on 22 to verify that it works there as well and we could just keep it as a singular scale-nvmet-start.sh.

Let me know how I can help!

[travisghansen]

I made some slight changes to ensure it works on the upcoming 24 release.

In the config you'll need to set the following moving forward.

nvmetcliPath: /root/.local/bin/nvmetcli

6ada268

24 makes the filesystem ro so it cannot install the nvmetcli script to /usr/sbin/nvmetcli