Re: Outages
Many orks ago I worked for a borged company that was migrating their (fairly small) computer room into a larger purpose-built one, along with all the servers from the borgee company.
I'd done the thermal and power load calculations and mentioned that the 3 aircon units [1] would be running full-time just to keep up. And the UPS was no-where near the capacity required (could do either companies stuff, just not both).
I was ignored. Oh well, I'd got my CYA material so just got on with it.
We moved our stuff in, life was good. Borgee-company started moving their stuff in. About 1/3 of the way through, the UPS tripped due to too much load. So I was told to run it on bypass for the moment (to get the migration done as we were giving up the lease on our old building) and they'd look at upgrading it afterwards.
Ho hum. Again, stored carefully in my CYA material store. We finished migrating and had a few days of uptime.
Then one of the PFYs noticed that one of the aircon units was dribbling water onto one of the racks. We couldn't shut it down (thermal capacity didn't allow that) so did a Heath Robinson duct with plastic sheeting and duct tape to carry the water away from the rack - forgetting that duct tape isn't particularly waterproof and, if soaked, ceases to be sticky.. And, after a week or so, promptly came unstuck.
So we went round, shut down every server that wasn't business-critical and got the aircon serviced at eye-wateringly 'emergency callout' rates (because, yes, they hadn't bothered to get a maintenance contract in place..)
Then, one day, the building had a complete power outage. And, the UPS, being in bypass mode, did nothing other than beep. And the room got very, very quiet (other than the UPS beeping..)
Needless to say, every server was set to 'resume previous power state' so, when the power came back on, every server tried to come back up at the same time.
UPS gives up in disgust as the oversurge protection kicks in.
Eventually we get everything back (manually power off each rack suppy board, reboot UPS, turn each rack back on one by one) and the IT Director starts looking at upgrading the UPS. Which can't be upgraded because they decided to buy a smaller unit and exapnded it to the max. So the whole UPS has to be replaced.
I left there a couple of months later.
[1] Original idea was that we could run on two units with the third on standby to be used when one of the others needed maintenance. Not only were all the coolant pipes running directly along the top of the racks rather than across them but the units themselves were not powerful enough.