ProLiant Servers (ML,DL,SL)
cancel
Showing results for 
Search instead for 
Did you mean: 

Rack power subsystem not redundant

 
SOLVED
Go to solution
Arthur Kao
Occasional Advisor

Rack power subsystem not redundant

I've been getting above warning message about "rack power subsystem not redundant" occationally in the past. However recently, as soon as we turn on 1 extra blade on this rack the frequency of the message increases dramaticlly (4-5 email per minute). I looked at each PS under rack view and their usage is all under 50%.

I can see a firmware under Windows 2003 that fixes this issue, but it's not available for Linux (RHEL4) - Server Blade and Power Management Module Firmware.

Is there some other place I can download this firmware or perhaps there's a real problem?

thank you for your insight.
12 REPLIES
Brian Hahne
Regular Advisor
Solution

Re: Rack power subsystem not redundant

We had the same problem.
We looked at this with HP engineering and found a few issues.

1) The original spec of 5 enclosures, powered by 2 3-phase supplies, filled ith blades, was based on G2 power consumption. When G3's came out, power requirements went up. 5 enclosures was dropped, per HP engineering, as a supported configuration. So you should only use 4 enclosures at max.

Also, we had an issue with the enclosures not having enhanced backplanes, which we are in the middle of rectifying.
We had 5 enclosures of mostly G2's, and then 10 G3's. It put us over the capacity threshold of the power supplies by something like 1-2 Watts. It was crazy, but we got emails CONSTANTLY.

The fix was that we first just took G3 blades out until the problem went away, moving them to a new enclosure. Then we reduced each rack to 4 enclosures, on 2 3-phase power supplies.
At this time we are finishing our upgrade to enhanced backplanes. The backplanes aren't a root cause, but engineering did tell us that G3's work better w/ the enhanced backplane, and G4's require it to get dual nics functioning.
Arthur Kao
Occasional Advisor

Re: Rack power subsystem not redundant

thank you for your information.

the rack we are having problem with has 3 enclosures with 1 enclosure of power supply. Therefore from what you saying (4 on 2), we may have to move the 1 enclosure or get 1 more ps.

To me, it's just odd that on the iLo rack view, the power consumption is all < 50% for all the individual power supply. I guess we can't trust that reading then.
Brian Hahne
Regular Advisor

Re: Rack power subsystem not redundant

3 enclosures and 1 power supply?
I think that's your problem right there. I think you need another power supply enclosure.. 3 phase? just get one more.
Oleg Koroz
Honored Contributor

Re: Rack power subsystem not redundant

what you have from hyper terminal for redundancy page ? see attachment
Arthur Kao
Occasional Advisor

Re: Rack power subsystem not redundant

How do you access that page?

thank you.
Oleg Koroz
Honored Contributor

Re: Rack power subsystem not redundant

Arthur Kao
Occasional Advisor

Re: Rack power subsystem not redundant

Here's the power information.

Thank you.
Oleg Koroz
Honored Contributor

Re: Rack power subsystem not redundant


Power available in C box by default use number for side A.
You have 314 and 383 to consume more per side in power enclosure.
With two sides running at same time you use less than 50%, however in case one side goes down you need to divide 314 or 383 in the half, which makes 157 and ~192.
As normally full bay server consume in average 240-325 (depend on the model). Numbers 157 and ~192 won't be able to cover consumption of one blade 240-325 and it might go down or other small production interruptions. To keep redundancy you need to keep numbers 157 and ~192 - above of one server consumption.

How it could be done.
1. add additional 3U 3Phase power enclosure
2. you can add 3U 1Phase in combination with 3Phase enclosure
3. you can add 1U power enclosure to supply power for one of Blade enclosures
Or last but not the best options are:
4. pull one server out
5. Remove some components like fiber card if not used or one - two processors from (dual box) to reduce power consumption.
Oleg Koroz
Honored Contributor

Re: Rack power subsystem not redundant

Whole game is only about to get 157W and ~192W number over higher number that you can see in server consumption from Hyper Terminal in Blade enclosure
Example added

Also one of notes you had not mentioned firmware - latest one 2.32 you can update from firmware CD it has option to update blade Components
And it could be done with command line from Linux
Se pages for Updates

http://h18023.www1.hp.com/support/files/server/us/download/26082.html

http://h18023.www1.hp.com/support/files/server/us/download/25831.html
Oleg Koroz
Honored Contributor

Re: Rack power subsystem not redundant

Sorry Arthur for getting back on it, I'm positively you need to know that too
With standard enclosures you have all power shared, with enhanced enclosure you have shared only for power provided to signalling Backplanes and to Interconnects, where power to servers depend strictly from side A or B.
I added Image with enhanced enclosure where it show real number for shared power 720W (which is usually for one enclosure 240W) remaining power is consumed by Servers. Number in standard enclosure would be in same range +/- and depend on the Interconnects.

Shared power is constant and not depends on the number of power enclosures.

So, to clarify 50% of redundancy you need to divide shared power in case one side down. Which make for us 360W per side and options to have more from TO GO is 314W or 383W. So, side A is main factor of Errors that you have for today. You can apply same solution as it listed before to get numbers up.
I guess both of us learned something today ...
;o)
Oleg Koroz
Honored Contributor

Re: Rack power subsystem not redundant

Sorry Arthur for getting back on it, I'm positively you need to know that too
With standard enclosures you have all power shared, with enhanced enclosure you have shared only for power provided to signalling Backplanes and to Interconnects, where power to servers depend strictly from side A or B.
I added Image with enhanced enclosure where it show real number for shared power 720W (which is usually for one enclosure 240W) remaining power is consumed by Servers. Number in standard enclosure would be in same range +/- and depend on the Interconnects.

Shared power is constant and not depends on the number of power enclosures.

So, to clarify 50% of redundancy you need to divide shared power in case one side down. Which make for us 360W per side and options to have more from TO GO is 314W or 383W. So, side A is main factor of Errors that you have for today. You can apply same solution as it listed before to get numbers up.
I guess both of us learned something today ...
;o)
Arthur Kao
Occasional Advisor

Re: Rack power subsystem not redundant

thanks again for your detailed reply.