Got a question to a planned power outage in one of our datacenters.
The power is supplied with two independent USPs. Every shelf/controller/switch(FMC) has redundant power supplies connected to both USPs.
Because of a maintenance of the USP, our DC-team has to shut down each USP for about 20 minutes (not at the same time :-))
Of course the filer will detect the power failure, but how will it react? Will it only report via syslog? Or will there be a gracefull shutdown? Is there a timer which forces the HA-systems to takeover the partner which has "failed" psu's?
If only half of PSUs in each filer/shelf are switched off, it will continue to run. The only known exception is FAS2240, which in some configurations requires 3 PSUs to run. So in your scenario there won’t be any shutdown or takeover, unless you have FAS2240.
If you happen to have shelf/controller with PSUs on the same power feed (or - the worst case - remaining PSU suddenly fails) it will most likely result in multiple disk failure and panic; partner will miss disks as well and aggregate won’t come online. It again depends on aggregate layout. It is possible to setup disk groups spanning multiple shelves which can withstand even complete shelf failure. Some customer did it.
If you do determine you will run on degrade power, you might want to turn off autosupport during that time or call support to let them know not to ship psus. With some maintenance like this I have seen a lot of parts ship when not needed.