[lm-sensors] 3.1.2 sensord duplicate RRD name problem

Jean Delvare khali at linux-fr.org
Sat May 15 09:55:09 CEST 2010

Hi Sergey,

On Sat, 8 May 2010 21:19:10 +0300, Sergey Kvachonok wrote:
> On 5/6/10, Jean Delvare <khali at linux-fr.org> wrote:
> > On Thu, 6 May 2010 14:20:56 +0200, Jean Delvare wrote:
> >> I have a patch ready, I'll give it some testing to make sure I got it
> >> right, and then I'll post it here.
> >
> > Andre, can you please comment on the patch below? Sergey, any chance
> > you could test it? It works for me on two different machines, but I may
> > not have covered all possible cases.
> smartd starts ok, RRD contains temp1 and temp1_0. Still no idea how to
> tell which is which, but at least it starts and logs everything.

You can ask sensord to generate a sample CGI script (option -g).
Looking at the generated code will show which label corresponds to
which entry name in the rrd file.

I know it is not exactly user-friendly. I would vastly prefer if the
entry names in the rrd file were self-explanatory, for example
coretemp_isa_0000_temp1, coretemp_isa_0001_temp1, etc. The current
approach to resolution of name collisions is poor. The problem is that,
if we change it, then users upgrading to the next version of lm-sensors
will have to regenerate their rrd base from scratch. Even worse, they
will have to figure out by themselves what went on and how to fix it.
So this change isn't to be done lightly. We have to think of an upgrade
path and document it. And we have to make sure we get the new naming
scheme right, so that we don't have to change it again in the next 10

Another thing I don't like with sensord is that, the way it stores its
data, any ignore statement added or removed from the configuration file
forces one to recreate a rrd base. Same with chips added or removed
(which happens to me frequently because I have external thermal
sensors.) Unfortunately I am not familiar enough with rrd to propose
improvements on that front. But if anyone has ideas how this could be
improved, please speak up.

Jean Delvare

More information about the lm-sensors mailing list