Re: Weak Battery Options -- MIDRANGE-L

OK I agree that SNMP integration isn't a bad idea because it simplifies cabling, removes the interface cable, and allows multiple systems to get signals from the same UPS.

But what additional information from the UPS could you realistically make productive use of? You might want to query the UPS for estimated remaining run life and assure the i5 shuts down with at least half the battery life still in play for example but you sure wouldn't want to believe the UPSs report of remaining battery run life and keep i5/OS running right to the last second before shutting down! This is a bad idea for several reasons. First the estimate is probably wrong and if it's wrong on the short end, you don't get a clean shutdown. Second even if it's right you now have a UPS with flat or nearly flat batteries. When the power comes back on you'll need to wait for a significant time before brining up your i5 or you risk a second power outage knocking you flat almost immediately since the UPS batteries will still be flat. I have experienced this problem at more than one customer. Run the UPS till it's dead and it drops the system on it's head. Power comes back and both the UPS and the system are brought back immediately. While the system is doing DB Recovery boom boom out go the lights again with the system following right behind. Next IPL is *UGLY.

Pete's issue is that his UPS batteries are so weak that the unit reports weak batteries as soon as it's handed the load. Remember that 'weak battery' is defined by the UPS vendor and generally indicates just a few remaining minutes of power. IBMs response to that signal is a quick shutdown to avoid DB damage and to leave at least a little battery for next time. There IS a difference between PWRDWNSYS *IMMED and pulling the plug!

- Larry

Lukas Beeler wrote:

The UPS integration with i5/OS leaves a lot to be desired anyway.

You can't even attach a network enabled UPS, there's no SNMP UPS
support.

And even with the serial attachement, you can't get much information out
of the UPS. We're mostly using Powerware 9125 UPSes with the System i
deployed to our customers. They can only be attached to a System i
machine using a relay card (making it impossible to attach other IT
systems to the UPS) - when attaching the UPS with a Web/SNMP card, I can
get all the information with a web browser, and it works out of the box
with both Linux und Windows machines.

Of course, this isn't a problem for all those big System i customers IBM
does care about - it's the Small Businesses without a datacenter that
are stuck with no real solution.

-----Original Message-----
From: midrange-l-bounces+l.beeler=dataline.ch@xxxxxxxxxxxx
[mailto:midrange-l-bounces+l.beeler=dataline.ch@xxxxxxxxxxxx] On Behalf
Of Pete Helgren
Sent: Sunday, June 24, 2007 10:23 PM
To: Midrange Systems Technical Discussion
Subject: Re: Weak Battery Options

Thanks Larry. I was leaning that way but it is rather peculiar that IBM

doesn't have some kind of system value to deal with the weak UPS battery

condition. The equivalent being the cache battery on the RAID controller. IBM doesn't automatically degrade the disk performance, they first give you a heads up and later degrade the disks when they calculate that its time to replace the battery. It would be nice if IBM

had a setting that gave you the option of issuing a warning message on the weak battery condition and a time frame of up to 30 seconds of outage before shutting down the system on a loss of power with weak batteries. Just assuming that the weak batteries aren't enough to handle even the most transitory of power flickers seems an bit extreme.

Having JUST replaced my cache batteries last week, I am suffering from battery fatigue I guess. I'll trundle down to my favorite battery place

tomorrow and pick up some new ones....hopefully before Rocky Mountain Power sends me a "flicker" again. Once I take care of that I WILL take a look at a program to be more reasonable about when to shut down the system.

Pete