RAM monitoring apparently has a hole

[pboothe]

The hole should be discovered and plugged.

From https://github.com/m-lab/ops-tracker/issues/1085

[stephen-soltesz]

Two possible strategies:

• can we poll the DRAC for this information?
• can we change kernel configuration to look in the “right” place so node-exporter reports the metrics we need?

[nkinkade]

Maybe we need something like an idrac_exporter? Or something that feels like banging a round peg into a square hole would be to try to piggy back on the reboot-service, which already logs into every DRAC periodically. Seems like not a huge stretch to think it could login, like it already does, and then also check for reported errors.