Re: [Ipmitool-devel] cacti graphs

[Duncan L. <du...@ic...>]

Eric A. Hall wrote:
> That's what I'm using now, with regex pulling out the name and value. This
> works pretty well actually since it gives me something stiff to measure
> against (I can catch errors through the absence of vertical bar
> characters, for example).
>
> What I am requesting is to provide an additional parameter to that
> command, so that I can request a single sensor.
>
> The first question is whether or not that would actually make anything
> take less time. Is there support in the IPMI protocol for fetching a
> single sdr or do you ahve to pull them all? If the latter then there won't
> be any extra efficiency introduced.
>
>   

Yes and no... The actual Get Sensor Reading IPMI command does just
take a sensor number (from the second column in the sdr output) and
return a reading for that sensor.  However given the flexibility of the
IPMI spec this isn't guaranteed to be unique for different record types. 
The only truly "unique" key for a sensor is its identifier string, and the
only way to do a lookup based on that is to read the SDR until found.

If you're feeling adventurous you can try the CVS source and use the
new "sensor reading" command that I just added.  When used together
with the SDR cache and even CSV output format (with -c) it should
make for relatively fast and easy readings to parse:

# ipmitool -S cache.sdr -c sensor reading "CPU"
CPU,39

# ipmitool -S cache.sdr -c sensor reading "FAN 1"
FAN 1,8000

-duncan