Good morning all,

I’m a recent adopter of Observium so still finding my feet and slowly learning how best to set it up.  However, I have noticed that since installing it, it was reporting that my main UPS management card was rebooting every 6 hours.  The UPS is an Eaton 9SX 5000i + battery pack, with Network-MS management card, firmware HF which is the latest available for card tech level 16.

Dabbling more, it seems that it’s the discovery process that is causing it, specifically when it walks the xupsOutput table (.1.3.6.1.4.1.534.1.4) in the Powerware MIB, the UPS management card immediately stops responding and reboots.  It actually appears to be crashing on doing a GetNext on .1.3.6.1.4.1.534.1.4.7.0, which is beyond the defined values in that table.  The UPS continues to operate, but the management card stops.  This also means that Observium has only managed to discover the ‘input’ voltage – output, bypass and current/power readings all just read as zero as the card has stopped responding when it tries to read those.  The poll runs fine and gets data from the input voltage sensor.

Clearly this is not an Observium issue: I can reproduce the reboot with a simple snmpwalk on the commandline, which should be a valid operation on a published MIB.  I do have a ticket open with the manufacturer, although the first response was ‘does it work with our management software?’.  It may even be a fault on my unit, but I don’t have more than one of these to test.

I’m assuming there’s a reason for the discovery module to need to walk the tables rather than just read the specific values – presumably bigger UPS units might have multiple metered AC inputs and outputs rather than just the one?

I could disable sensor discovery, which will stop the reboots but that won’t let it learn about the additional sensors that it still hasn’t found. Unless I can manually insert data into the database for the poller to pick up?

Alternatively, is there a way for me to override the detection of a Powerware UPS and use the generic UPS MIB rather than the XUPS MIB?

Any other thoughts?

Steve.