not the warning value; only solution to this

9 Apr 2024 · *not*

      Hi,
the temperature check in xymon's version of bookworm is broken in a
rather strange way. The check is located in
/usr/lib/xymon/client/ext/temp and fails for all NVMe disks that contain
several temperature sensors:
For example the first NVMe in Lynx which holds its temperature values in
/sys/block/nvme0n1/device/hwmon0/:
files   name        value   min         max         crit
temp1_* Composite   27.85   -273.15     86.85       87.85
temp2_* Sensor 1    27.85   -273.15     65261.85    n/a
temp3_* Sensor 2    31.85   -273.15     65261.85    n/a
The inner logic of the temperature check works as follows to calculate
the values for red and yellow:

if there is a crit and a max value use them
if there is a max and a mid value use them
if there is a max and a min value, use them
if there is only a max file, use it for both
if there is only a crit file, use it for both.

The sensor 'Composite' uses max and crit as they're both available.
'Sensor 1' and 'Sensor 2' however do only provide max and min. Therefor
these values are being used but lead to 'yellow' warnings as the min
value actually isn't an upper boundary as assumed but a lower boundary.
The linux kernel documentation
(https://docs.kernel.org/hwmon/sysfs-interface.html) also outlines that
every file using 'min' in their name is a low threshold:
 The common scheme for files naming is: &lt;type>&lt;number>_&lt;item>. Usual 
types for sensor chips are "in" (voltage), "temp" (temperature) and
"fan" (fan). Usual items are "input" (measured value), "max" (high
threshold, "min" (low threshold).
The proposed fix would be to either use max value for yellow and red or
to at least sanity check whether min is below zero and in that specific
case only use the max value for both:
In /usr/lib/xymon/client/ext/temp beginning on line 182:
     my ($red, $yellow);
     if (-r $crit_file and -r $max_file) {
         $red     = read_one_chomped_line_from_file($crit_file);
         $yellow  = read_one_chomped_line_from_file($max_file);
     } elsif (-r $max_file and -r $mid_file) {
         $red     = read_one_chomped_line_from_file($max_file);
         $yellow  = read_one_chomped_line_from_file($mid_file);
     } elsif (-r $max_file and -r $min_file) {
         $red     = read_one_chomped_line_from_file($max_file);
         #TODO: min_file contains the lower temperature boundary and
         #      *not* the warning value; only solution to this would
         #      be to set either yellow to red or to at least do that
         #      when yellow is below 0 for example.
         $yellow  = read_one_chomped_line_from_file($min_file);
         $yellow = $yellow > 0 ? $yellow : $red;
         # Alternative solution: do not use min at all v1:
         #$red = $yellow = read_one_chomped_line_from_file($max_file);
         # Alternative solution: do not use min at all v2: remove 
this 'elsif'
} elsif (-r $max_file) {
$red = $yellow = read_one_chomped_line_from_file($max_file);
} elsif (-r $crit_file) {
$red = $yellow = read_one_chomped_line_from_file($crit_file);
}
There are three ways to solve this:

sanitize min by checking whether the value is below 0 and in that case
use the max value
use the max value in any way
completely remove the 'elsif' that reads min and max as the next
'elsif' just reads and uses max

Thanks in advance!
Best regards
Christoph

nvme temperature check broken in Debian bookworm

zechner＠vrvis.at

jeremy＠laidman.org

damien＠makelofine.org

not the warning value; only solution to this

be to set either yellow to red or to at least do

when yellow is below 0 for example.

Alternative solution: do not use min at all v1:

Alternative solution: do not use min at all v2:

tags

participants (3)