Control: severity -1 wishlist

> I often use a custom livecd verision of Debian with various utilities
> pre-installed, so they can be easily used offline, or without needing to
> reinstall them after each reboot.

Such a livecd is a non-standard (but not non-important!) usage, and it
already customizes other facets of packages it includes.  What about
configuring that livecd to not start such daemons?

> But often, if I boot this same livecd on a machine that does not have cxl
> or nvm dimm devices, deamon will complain, and fail at boot (with big red
> warning), and sytemd will continue restarting it afaik (or maybe not, it
> looks like it tries to run it only once on my system, then gives up).

That's news to me.  I don't run systemd myself (heck, it's so buggy it even
fails to boot my main workstation, and I have a severe dislike of it for
other reasons), and I've included the .service from upstream as-is, after
very superficial testing.  The daemon is supposed to gracefully exit if
there's nothing to monitor:

2023-06-19T06:01:41.874004+02:00 valinor ndctl: ndctl monitor daemon started
2023-06-19T06:01:41.874062+02:00 valinor ndctl: no dimms to monitor, exiting

I had even suggested/implemented some improvements here to upstream, but not
surprisingly 99% of my testing was done with sysvinit and openrc.

Still, it's surprising the upstream would fail to notice what you report:

> root@debian:~# systemctl status ndctl-monitor.service 
> × ndctl-monitor.service - Ndctl Monitor Daemon
>      Loaded: loaded (/lib/systemd/system/ndctl-monitor.service; enabled; 
> preset: enabled)
>      Active: failed (Result: exit-code) since Wed 2023-06-28 22:11:35 UTC; 
> 25min ago
>    Duration: 89ms
>     Process: 2412 ExecStart=/usr/bin/ndctl monitor (code=exited, status=250)
>    Main PID: 2412 (code=exited, status=250)
>         CPU: 9ms
> 
> Jun 28 22:11:35 debian systemd[1]: Started ndctl-monitor.service - Ndctl 
> Monitor Daemon.
> Jun 28 22:11:35 debian ndctl[2412]: no dimms to monitor, exiting
> Jun 28 22:11:35 debian systemd[1]: ndctl-monitor.service: Main process 
> exited, code=exited, status=250/n/a
> Jun 28 22:11:35 debian systemd[1]: ndctl-monitor.service: Failed with result 
> 'exit-code'.

Maybe this redness is a regression?

> It is not a big deal, but something that is not perfect in my opinion.

Seems like something to investigate once tuits are more abundant.
 
> For my livecd, I can clearly modify some systemd unit files manually
> (using build scripts for live-build), but I wonder if defaulting to
> running a daemon by default is a good idea in general. Most of the time
> probably yes, but sometimes not.

The rule in Debian is to start a daemon if it's installed; eg. Red Hat has
the opposite custom.  There's usually no point in installing daemons that
are not needed; and it those rare cases you do, the admin can disable them
from starting.  This sounds like a good modification for the livecd.

> As of smartmontools you mentioned. I have exactly same issue. I need
> smartctl tools for troubleshooting various computers on my livecd, but I
> DO NOT WANT smartd to start. For very similar reason I opened a bug
> against smartmontools few days ago: https://bugs.debian.org/1039454

As their user count is many orders of magnitude larger than that of ndctl,
and so is diversity of supported hardware (NVDIMMs are found only in fat
servers), they also have far more collective brain cells.  I'd thus wait
for their decision then copy it.

> Feel free to downgrade to wishlist and think if this makes sense at all.

Done this for now.


Meow!
-- 
⢀⣴⠾⠻⢶⣦⠀ You should never, ever, degrade a human being by saying they're
⣾⠁⢠⠒⠀⣿⡁ a worthless waste of food and air.
⢿⡄⠘⠷⠚⠋⠀
⠈⠳⣄⠀⠀⠀⠀ You should also never anthropomorphize spammers and telemarketers.

Reply via email to