Package: munin-plugins-core
Version: 2.0.28-1
Severity: normal

Here is the old web page of my sensors_temp plugin, last month (backup)

Field   Internal name   Type    Warn    Crit    Info
Core 0  temp2   gauge   105.0   105.0    
Core 1  temp3   gauge   105.0   105.0    
Core 2  temp4   gauge   105.0   105.0    
Core 3  temp5   gauge   105.0   105.0    
SYSTIN  temp6   gauge   65      75       
CPUTIN  temp7   gauge   75.0    80.0     
AUXTIN  temp8   gauge   75.0    80.0     

Last week, I made a global system update.

Here is the same page today:

Field   Internal name   Type    Warn    Crit    Info
SYSTIN          temp1   gauge   0.0     0.0
CPUTIN          temp2   gauge   75.0    80.0     
AUXTIN          temp3   gauge   75.0    80.0     
PCH_CHIP_TEMP   temp4   gauge           
PCH_CPU_TEMP    temp5   gauge           
PCH_MCH_TEMP    temp6   gauge   20:65    10:75
temp1           temp7   gauge           90.0     
Core 0          temp8   gauge   105.0   105.0    

Forget about missing lines like temp1; I have disable some fields in
server conf. The core problem is that some probes have changed name:
SYSTIN used to be temp6 and is now temp1; implies two problems:
new SYSTIN inherits the past RRD of temp1 (which was probably a stupid
probe since I had it disabled).

Other way to look at it: temp8 was AUXTIN and is now core0; before the
update the curve represents external probe; after the update, internal
CPU. So, a given rrd(curve) will represent two different probes,
depending on the date we are looking at.

This is a severe problem: I had the machine setup with warnings and
criticals for security reasons. After update, all my settings are
messed-up; what means, the configuration I made to wanr me about failing
parts is lost; if I had not dig in Munin, some critical part of the
server could have failed without a single warning. In my machine, when
any fan is broken, the machine has to stop (and stop before a component
overheats; I had a PSU fan diyng; when PSU overheated, it came into
short circuit, and injected high voltage into the HDD: bad PSU fan
killed my disk !!!).

Munin is not responsible for sensors package updates, and potential
changes; but the plugin needs to provide a way to keep track of which
field is pointing to which probe. And have that stable over years.

If plugin can not record things this way, it needs to provide a manual
way to associate sensors-probes to munin-field-name.

-- System Information:
Debian Release: stretch/sid
  APT prefers testing
  APT policy: (500, 'testing')
Architecture: amd64 (x86_64)

Kernel: Linux 4.8.0-2-amd64 (SMP w/4 CPU cores)
Locale: LANG=en_GB.UTF-8, LC_CTYPE=en_GB.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Init: systemd (via /run/systemd/system)

Versions of packages munin-plugins-core depends on:
ii  munin-common  2.0.28-1
ii  perl          5.24.1~rc4-1

Versions of packages munin-plugins-core recommends:
ii  libnet-snmp-perl  6.0.1-2

Versions of packages munin-plugins-core suggests:
pn  conntrack                <none>
pn  libcache-cache-perl      <none>
ii  libdbd-mysql-perl        4.041-1
pn  libnet-dns-perl          <none>
pn  libnet-netmask-perl      <none>
pn  libnet-telnet-perl       <none>
ii  libxml-parser-perl       2.44-2+b1
ii  python                   2.7.11-2
pn  ruby | ruby-interpreter  <none>

-- no debconf information

Reply via email to