Welcome to our community

Greeting to all.

We have a rack with 2 HSV200 controllers, 8 disk shelves, 2 SAN switches,a management server, and a Blade (bl25p) enclosure with 8 blades and 2 GBE2 switches.
Because of dodgy power supplies etc we had to completely redo the power to the server room and it was decided to install Switched Rack PDUs from APC into all racks. These each have 24 outlets divided into 2 regions of 12 by two circuit breakers. So our Blade/SAN enclosure now has one on each side and each component is plugged into each APC strip. Each strip can handle 32 amps, the trips go at 16 amps.

A couple of weeks back a PDU for one of the blades blew and took out the 12 outlets on one side of the rack. This meant that 3 disk shelves, a management server and another blade PDU lost power. Due to the redundancy no data was lost but it led me to thinking that were another PDU to short we could have three disk shelves fail simultaneously.

So my questions are: -

1) How does this setup differ from the recommended hp power management? Would I expect a blade PDU to take out the other components in the way described? Am I wrong to feel a little uncertain about this.

2) If 3 disk shelves lose power, clearly the RAID sets are compromised. What would happen when the power is restored? Have I lost all data or does the HSV flag these as shelf failures and thus assume that when power is restored it can continue to use the RAIDsets. I think I know what happens but need confirmation.

3) We are running CA. What does the other site think of this?

TIA

Hi Tom,
Hmm - I think I'm reading this correctly - but you have 12 outlets on one power strip?? I have no experiance of APC equipment and my first thoughts would be that I think that the power config you have would probably not be recommended by HP!

If you are concerned with Blades blowing outlets then maybe you need to reconfig your power again so that all the EVA components are NOT on the same power as the blades ??

It is not unusual for one "item" blowing to take out a whole strip - I just think light bulb blowing - sometimes the breaker trips sometimes it dosn't.

As for No2 - I think you mean you have lost TOTAL power to 3 shelves? If so then I think you would have a problem! You stated you have a 2C8D HSV200, well assuming you have an full RAC config?? [NO Vraid 0 either] then you should be able to survive the total loss of ONE shelf not three. The EVA would just see all the shelves/disks disappear. This should not put any of the disks in a failed state and they should be OK when power is restored. If you stopped all I/O straight away then the EVA may be able to recover from this situation ??? I have not tried this by the way!!

This brings us onto the last point - well that depends on how you have your CA setup and what you do if/when you suffer multiple simultaneous shelf failures !

Assuming your CA is set up correctly and that you have a DR set of all the disks that have just failed on one site then it's my understanding that you should be OK as that is the whole idea of a CA solution. You may lose one or two "transient I/O's".

Hopefully there is some help in there but what you might want to consider is a full test of your CA solution and simulate the loss of 3 shelves and see how your CA solution copes. Only you can answer that one !!!

Regards,
MEA[T]