Open lidocaineus opened 12 months ago
Quick update – when tossing in two power supplies, the system wattage probe OID (formally named " amperageProbeReading") remains as 1.3.6.1.4.1.674.10892.5.4.600.30.1.6.1.2 and is labeled "System Board Pwr Consumption". It does enumerate two additional entries in that table as OID 1.3.6.1.4.1.674.10892.5.4.600.30.1.6.1.1 and 1.3.6.1.4.1.674.10892.5.4.600.30.1.6.1.3, however the reading they present are in tenths of an amp, and are labeled "PS1 (or PS2) Current 1 (or 2)".
This makes sense as the two power supplies are redundant and feed a single power input, which would be the "System Board Pwr Consumption" probe; I initially and incorrectly assumed it would present another watt probe. So by default, it looks like you should always use the OID at 6.1.2.
The fix for this is to either set the OID in idrac-input.conf to 1.3.6.1.4.1.674.10892.5.4.600.30.1.6.1.2, or instead of grabbing the specific field, to use inputs.snmp.table instead, and have logic in the grafana dashboard to locate the correct reading. The latter is probably the better way, as another table in the column ("amperageProbeType" / OID .1.3.6.1.4.1.674.10892.1.600.30.1.7) actually specifies what amperageProbeReading presents as a value; it can be a number of things, and although I don't think it changes, I only have a limited sample of servers to test against.
Thoughts on either fix?
On my iDRAC7s with the last available firmware (2.65.65.65) the OID for system-watts appears to now be
1.3.6.1.4.1.674.10892.5.4.600.30.1.6.1.2
The corresponding amperage reading is just before that at
1.3.6.1.4.1.674.10892.5.4.600.30.1.6.1.1
The current OID is listed in the sample idrac-input.conf as .1.3.6.1.4.1.674.10892.5.4.600.30.1.6.1.3. That could possibly be the case on a server with two power supplies; I run them all single-ended and haven't had time to swap one in and check.
These don't appear to be documented very well, possibly due to age of the hardware, but it's in a table at OID 1.3.6.1.4.1.674.10892.5.4.600.30.1.6 identified as "amperageProbeReading" in the iDRAC9 SNMP Reference. I can verify this against an iDRAC8 later this week (as well as with dual power supplies) to see if it's consistent there as well and possibly submit a pull request if one of the OIDs is more reliable with either one or two power supplies.