Re: Monitoring for power outages on a dual-powered system

[DrFranken <midrange@xxxxxxxxxxxx>]

Personally I don't like this setup as it adds exposures and solves none.

1) You will get a message in QSYSOPR (CPF1816 maybe?) *IF* the UPS is 
properly connected to the system. This will happen whether or not both 
power supplies are connected to the UPS. Nothing gained in your setup.
2) You lose the power conditioning capability of the UPS on the non-UPS 
side with your setup.
3) You MUST remember to always keep enough capacity on the UPS to carry 
the entire system when the utility fails or the additional load will 
overload the UPS and 'down she goes'. Since that's the case, why not 
have both power supplies on the UPS to start with?
4) When utility power fails you get a bunch of errors in SAL and WRKPRB 
because one of the Power Supplies loses power and to the system appears 
to have failed. This will probably cause the system to cry wolf with IBM 
about a problem that isn't their fault. So now you get to un-clutter 
these logs and beg IBMs forgiveness.

FYI the 110000AC SRC is logged AFTER the system comes back up so looking 
for that one is like tacking up a 'no smoking' after a gas explosion.

The CPPEA56 is a possibility but that goes back to the system thinking a 
component has failed and you'd then need to look in the problem log for 
SRCs to see what actually failed.  I'd focus on getting the UPS 
interface working properly.

 - Larry "DrFranken" Bolhuis

On 8/2/2010 11:38 AM, James H. H. Lampert wrote:
> We have one side of our production box plugged into the UPS, and the
> other side plugged into the wall. This not only guards against UPS
> malfunctions; it also gives us a way to monitor the power.
>
> This weekend, we had an unattended 80-minute power outage, which (not
> surprisingly) drained the UPS completely dry, and left us down for most
> of a day.
>
> The fact that I found a 110000AC SRC, and a CPPEA56 in the history log
> suggests that there's something I can tie into, so that the system can
> send a distress call as soon as the outage occurs.
>
> But what should I be looking for?
>
> --
> JHHL
>