nmdctl behavior questions

[phixxy]

Not sure if this is the correct place to post this or if I should have opened an issue.

Currently when you install nonraid but before you have created a superblock if you run nmdctl status -o json you get an output of
Error: nonraid module is not loaded Superblock file not found: /nonraid.dat To create an array with this new superblock, run: nmdctl --super /nonraid.dat create

Is this intended behavior? I was expecting a json response, but if this is intended behavior I can work with this as well.

Also when you do run nmdctl --super /nonraid.dat create and then nmdctl status -o json the response has a "disks" object with a "none" parity disk. Is this also intended behavior or should the disks object be empty? Thanks.

[qvr]

I guess the error is intended in that sense, that it's a very early stage check that all nmdctl commands have, done before even the arguments for status are parsed:
https://github.com/qvr/nonraid/blob/main/tools/nmdctl#L1824-L1828

Similarly wrong arguments will also cause "plain-text" output. If we were to just wrap the error in a json format (and then I guess we'd also need to handle prometheus output too), the response would still be very different from actual status json response, so special handling would anyway be needed in scripts using the output. So not sure if that would make sense, if scripts using the output can handle this anyway.

About the "none" parity disk, that's definitely intended for the human status output, as running the array permanently without an parity does not really make sense, so I want to highlight that the parity is missing and that's why it's always shown there.

While I think it could make sense to not have it in json output, the "none" parity gets added in to the data structures in the general data collection functions, so the actual format functions all use the same data structures and the "none" parity would need to be separately skipped here:
https://github.com/qvr/nonraid/blob/main/tools/nmdctl#L1357-L1391

Though without testing I'm not sure if that will have some side effects when for example disks fail or are unassigned for replacement etc.

[phixxy]

That would be incredibly helpful. I was looking at the source of nmdctl, checking if something like that already exists, and was considering modifying the create command myself, but I am not very comfortable in bash. I am also not great at web development but I am sure I could get something basic working. I would probably build it with flask because I am more comfortable with python, though I don't really have experience with flask either. This project is something I am very interested in working on though so it is a good excuse to learn whatever I need to learn to make this more useful.

[phixxy]

I have something very basic working for now, exciting. The only thing I have come across that isn't straightforward is:

1. Running nmdctl stop with mounted disks gives an interactive prompt asking if you want to unmount all filesystems first, maybe this should throw an error, or unmount disks and stop array?

Still keeping this in the discussions area since it isn't really an issue with nmdctl, just the behavior of it. I can obviously handle this by doing the unmounting myself before attempting to stop the array (which is what I am going to do either way). Also thanks for the create options! I haven't used them yet, but that is going to help a ton!

[qvr]

There is actually already the global unattended mode parameter, -u, which will prevent these commands from asking interactive things and defaulting to safe operation, for stop it will error out:

# nmdctl stop
Warning: Some filesystems are mounted. Unmount them before stopping the array.
Do you want to unmount all filesystems? (y/N):
Array stop cancelled by user
# nmdctl -u stop
Error: Cannot stop array with mounted filesystems in unattended mode
# echo $?
1

[qvr]

Also, if you are stopping the array from your webui, you should also remove the array state file /var/lib/nonraid/array.running after successful stop, otherwise the nonraid systemd service will force a parity check on the next reboot.

[phixxy]

I appreciate the tips! I worked on the webui a bit today and I got quite a bit working. I think I can get something basic with most nmdctl features done very soon.

So far it can:
Display array disks
Display unassigned disks
Create an array
Start/Stop the array (except on a NEW_ARRAY state)
Format disks in the array
Run parity checks

I am happy with how this is coming along and it has been a fun project to work on. So thanks again for all the work you have done on nonraid to make this possible.