Hi,
I want to monitor status of the redundant power supply on a Supermicro
motherboard (H8DGT). It is available via IPMI, but even after patching
src/ipmi.c, I cannot get collectd to monitor that particular sensor. Turns
out this is because the "sensor reading type" is "sensor specific". Even
when I patch ipmi.c to call ipmi_sensor_id_get_reading even for these
sensors, the sensor_read_handler function is simply not called --
presumably because this reading type confuses OpenIPMI which refuses to
work with that. I cannot blame it, but that'[s the HW we've bought a couple
months ago :(.
OK, I can easily get roughly the same result via the command line -- simply
executing `ipmitool sdr get 'PS Status' | grep 'Sensor Reading' | cut -d :
-f 2 | tr -d '[[:space:]]' | tr -d h` is ugly, but works well enough.
However, this command needs root, as /dev/ipmi0 is root:root, 0660.
Now I could set up sudo for this script, I could also have it run from cron
and talk to collectd via socat and the unix sock plugin, I could even write
a C suid wrapper for just this, or perhaps mess with udev rules for making
that device node owned by another group. However, I would have to do all
that just because there are four lines in exec.c which prevent running
plugins as root.
I would like to respectfully ask you to reconsider this design choice.
Please consider my use case, as well as the plugins on the net (and even
collectd.git's own contrib/exec-smartctl). Do we *really* get any benefit
by this forced dropping of root, only to gain it again through a SUID
wrapper? I just fail to see any benefits here.
I've read the git log, apparently this has been the status quo since the
very beginning seven years ago. I'm sure there must have been *some*
reason, but I fail to see it -- please, educate me.
With kind regards,
Jan
--
Trojitá, a fast Qt IMAP e-mail client -- http://trojita.flaska.net/
_______________________________________________
collectd mailing list
[email protected]
http://mailman.verplant.org/listinfo/collectd
