The warehouse stores max power draw per chassis and max load per UPS. That means we can make a tool to add up the max power draw for all chassis connected to each UPS and report percentage of capacity used.
Do we want to do that? Would it be useful? It's never going to compete with having proper SNMP-enabled UPS:en feeding their load into a time-series database with a graphing web front-end, but for those who don't have that, is it better than nothing? Are the differences between nominal max draw and actual draw too great and fluctuating for it to be of any use? Could those deficiencies be remedied by asking the service processor for the real power draw (at some random moment when the load is assumed to be typical) and insert that value into the warehouse?
Obvious deficiencies in the current code include:
UPS capacity is in VA, but chassis load in W
Chassis load is wildly over-estimated at nominal max
Loose physical hierarchy data model makes for messy code. I think #125 would help, since each chassis, rack, room, and building inherits the UPS in different ways
I don't think we want to require a strict hierarchy here, since not everything is in neat racks in neat server halls
The tool could output a lot more interesting data if it knew more about the hierarchy
Non-chassis loads are not considered
I assume most users will see network equipment dominate this category. Do we want to model switches etc. in the warehouse (automatically generating RANCID config may be nice)? Also, I hear some people have lots of weird non-computer electronics in their racks :)
Names are not well thought out
Version number is not bumped, since I don't know yet if I want to introduce any breaking changes
The warehouse stores max power draw per chassis and max load per UPS. That means we can make a tool to add up the max power draw for all chassis connected to each UPS and report percentage of capacity used.
Do we want to do that? Would it be useful? It's never going to compete with having proper SNMP-enabled UPS:en feeding their load into a time-series database with a graphing web front-end, but for those who don't have that, is it better than nothing? Are the differences between nominal max draw and actual draw too great and fluctuating for it to be of any use? Could those deficiencies be remedied by asking the service processor for the real power draw (at some random moment when the load is assumed to be typical) and insert that value into the warehouse?
Obvious deficiencies in the current code include:
chassis
loads are not considered