[Freeipmi-devel] Re: ipmi sensor thresholds

[Anand Babu]

Latest update on "SDR#12 Proc Bd sensor ALERT" issue:
=====================================================
BUG:
----
  Most SR870BN4 (Tiger4) systems's Proc Bd sensors exceeds normal
  min/max limits.
  
  Example
  12: Proc Bd +1.2V (Voltage): 1.22 V (low=1.18/nom=1.20/high=1.21)

Solution:
---------
  After carefully going through the IPMI specification, I now
  understand that interpreting normal min/max fields to calculate
  ALERT status is a BUG. Also current health-check routine interprets
  to only {OK, ALERT, FAILED} values. This is wrong.

FIX:
----
  Correct implementation should interpret {non-critical, critical and
  non-recoverable} threshold limits to calculate {OK, Non-critical,
  Critical, Non-recoverable, Unspecified fault, Unknown} states.

  Available limits:
   - Nominal reading
   - Normal maximum 
   - Normal minimum
   - Sensor maximum reading
   - Sensor minimum reading
   - Upper non-recoverable threshold
   - Upper critical threshold
   - Upper non-critical threshold
   - Lower non-recoverable threshold
   - Lower critical threshold
   - Lower non-critical threshold

  Possible states:
   [OK]                - monitored parameters within normal operating
                         ranges 
   [Non-critical]      - `warning': hardware outside normal operating
                         range 
   [Critical]          - `fatal' :hardware exceeding specified ratings
   [Non-recoverable]   - `potential damage': system hardware in
                         jeopardy or damaged 
   [Unspecified fault] - fault detected, but severity unspecified
   [Unknown]           - status not available/unknown

  Interpretation rules:
   [OK]              : < non-critical
   [Non-critical]    : => non-critical < critical
   [Critical]        : => critical < non-recoverable 
   [non-recoverable] : => non-recoverable 

Because sensors readings in SR870BN4 exceeding normal min/max limits
falls under [OK] state according to specs, there is no need to fix SDR
firmware. How ever we will continue to develop "sdr-ed" tool as a part
of the GNU FreeIPMI project for completeness.

-- 
Anand Babu
Free as in Freedom <www.gnu.org>