dslreports logo
site
 
    All Forums Hot Topics Gallery
spc

spacer




how-to block ads


Search Topic:
uniqs
922
share rss forum feed


advgNY

@verizon.net

Server is Killing Power Supplies

I have a rack of 4 HP DL350's and 1 HP DL320. All of the servers are connected to an APC Smart-UPS 3000XL. The UPS logs show no issues with the power.

The DL320 server (the only one w/ 1 power supply) - has been through 3 of them this year. The server works fine for a few months, and then the PS dies and needs to be replaced. Today, I am now installing a 4th one.

Any ideas what the cause may be? I have a UPS hooked up....power is clean....is it heat? I have all 5 servers together in the rack...should I try moving the DL320 so it has better airflow? I have never seen this before...

Thanks!


yaplej
Premium
join:2001-02-10
White City, OR
I am no electrical engineer but don't the power supplies just plug into a common voltage control board? With one power supply it pulls 100% power from it and with 2+ the power is distributed over them.

They should output DC so its not like you would have some odd AC distortion/sync problem.

I assume no amount of pulling out/inserting the same power supply causes it to come back online? Seems like it would have to be something wrong with the voltage controller though. Its not probable that someone would get 3 bad power supplies in a row unless your supplier is giving you junk.

Looks like HP has that voltage module/power supply back-plane embedded into the power supply cage.
HP Part# 532473-001
--
sk_buff what?

Open Source Network Accelerators
»www.trafficsqueezer.org
»www.opennop.org


JoelC707
Premium
join:2002-07-09
Lanett, AL
kudos:5
reply to advgNY
I'm not familiar with HP servers but it looks like the DL320 is a single PSU only model. Do any of the DL350's experience this problem or is it just the DL320? How loaded is that server, and what size PSU is in there (presuming there are options for different wattage PSUs)?

Is this rack in a datacenter or a server room? Mainly I'm curious how extensive the cooling is. I've always heard it's wise to not have gaps in rack units, and if you have servers spaced apart you should use blanks to prevent mixing of hot/cold air (probably other reasons too, never worked in a large scale deployment to have to plan for that).

In my small scale deployments, (just a few servers in a single rack with a single AC unit cooling it) I've always found it best to space out servers so they don't "share" heat. With all of them stacked, they are all thermally connected and will (to an extent) heat up as a whole and cause others that may run cooler to run hotter. With them spaced apart (1U is enough) you can get airflow between them and in my experience, they will run much cooler. Granted I've never had failures like you're having but I know excessive heat will cause failures.


bryandj23

@bayvillewireless.com
Just a thought, but what kind of hardware (HDD's, PCI cards, etc) is on the DL320?

We witnessed power supply failures on Dell Poweredge 1900's (single PS) that were stocked with 3.5" 15k HDD's. At (if I remember correctly) 800 watts max, the PS over time couldn't handle spinning 8 drives efficiently.


Billy Brethr

join:2005-04-01
San Antonio, TX
reply to advgNY
Post manufacturer & model number(s) of the power supplies for review here.

My experience with desktops is that there is a wide variation in quality in PSU's, and that not only do the cheap ones die quickly, but they frequently take motherboards, CPUs, memory and hard drives as they go.


jimk
Premium
join:2006-04-15
Raleigh, NC
Reviews:
·Time Warner Cable
·voip.ms
said by Billy Brethr:

Post manufacturer & model number(s) of the power supplies for review here.

My experience with desktops is that there is a wide variation in quality in PSU's, and that not only do the cheap ones die quickly, but they frequently take motherboards, CPUs, memory and hard drives as they go.

HP tends to brand their server power supplies as their own (especially the hot swap models used in just about everything other than some low end tower servers), so it can be tricky to determine the real OEM just by looking at the label. If you can find the UL file number, that might reference back to the real OEM. They don't often cheap out on server parts, so it will probably be from a good manufacturer.

Still, it would be useful to post the part number in case anyone else here has had experience with this particular power supply, since HP uses the same power supplies in a lot of their servers. It would also be useful to find out more about the server they are being used in (new Gen8 server, or something older).

I would like to hear what symptoms are being experienced to lead to power supply replacement. If the power supply runs too hot or is too heavily loaded, that could shorten the life. I have seen some PSUs that don't like being on a line interactive UPS and prefer a double-conversion online UPS (an indication of questionable power supply design). There are a lot of possibilities, so it can be tricky to troubleshoot without more information.


Billy Brethr

join:2005-04-01
San Antonio, TX
reply to advgNY
Another thing to do is look at the voltages as reported in BIOS. I've never seen any hard technical data as far as tolerances go, but in general I regard +/- 5% as good, +/- 10% as bad, and in between as highly suspect. I use this rule is for diagnosing non-specific problems that might be explained by data corruption as a result of bad power. Someone more knowledgeable might have a "tighter" standard, but certainly no looser than that.

Also, check the log of your UPS. Do you have frequent outages, spikes, lightening strikes, etc..., also known as "power events"? What are the settings for your UPS? I set mine as close to 110 VAC as possible, and IMO the settings for APC are way too loose. I think the tightest I can get it is something like +/- 10% which I regard as completely unacceptable. Meaning that, even if your UPS thinks your power is good, it might be bad. Consider the consequences of running 5 computers from power that is consistently 101 VAC when specs require 110.

Finally, check for "droop". Run a voltage reporting software and then run something like Prime95 to push the system and see what happens to the voltages when the machine is under heavy load. It might look good in BIOS, but once you start drawing amperage the voltage drops outside spec.


advgNY

@comcast.net
reply to advgNY
The power supply in the DL320 G5 that keeps dieing is 400 watts. (single PSU) It only has x2 500GB 10K SATA drives, 8GB RAM, and an Intel Xeon 3050 2.13GHz Dual-Core CPU. The PCI slots are empty.

All other servers have been running without issue.

It was situated in the rack below a lot of other "longer" servers (the DL320 is short), so I moved it to rule out heat as the issue. It now runs cool to the touch.

UPS logs were checked and show 100% normal.

The only thing I can possibly think of is that the drives are 10K, and they *may* have been 7.5K originally when we bought the server. I am going to check the specs just to be sure 10K is ok. But it doesn't seem overburdened even with 10K drives...

I looked at the voltage module. I think that is only for dual PSU HP servers?

I will check the BIOS and check out Prime95.

JoelC707
Premium
join:2002-07-09
Lanett, AL
kudos:5
A 10K vs 7.2K drive would really only present a measurable different load on initial spin-up. Sure it will likely use more power while spinning too but it's not going to be much different. The specs you listed are honestly quite tame and should not present that high of a load.

I used APC's calculator and selected the DL320 G5 and configured it like you listed. It gave me 154 watts. It's been my experience that APC's calculator tends to estimate high (wise choice when sizing for a UPS) but 154 watts seems about right honestly. HP has a power calculator and I downloaded the one for the G5 and older (it has G5 and G6 servers listed in it) but it doesn't seem to include the DL320.

I'm gonna say either there is something wrong with the power source (is the power cable the same across all dead PSUs?) or the server itself. Heat could certainly have been an issue. Higher heat means the fans have to spin harder and as such consume more power. If your PSU were borderline already, extra heat to dissipate could push it over the edge (though given the calculator and listed specs, I doubt it was/is overloaded).


Billy Brethr

join:2005-04-01
San Antonio, TX
reply to advgNY
said by advgNY :

The power supply in the DL320 G5 that keeps dieing is 400 watts. (single PSU)

»h20566.www2.hp.com/portal/site/h···chetoken

Power Supply: Auto-sensing 450-Watt Power Factor Correcting (PFC) Power Supply, CE Mark Compliant
Hp specs list the PSU as 450 watts. Also, they claim it's has PFC, which to me is the primary delineator between "good" and "not good" Power Supplies. It's possible that if your 400 Watt PSU is also not PFC, it may only be producing 60% of that 400 Watts, or even less. You might want to check your Vendor and find the reason for this discrepancy. They may be selling you cheap-crap PSU's and claiming they are authentic Hp OEM.


dennismurphy
Put me on hold? I'll put YOU on hold
Premium
join:2002-11-19
Parsippany, NJ
kudos:3
Reviews:
·Verizon FiOS
reply to advgNY
Do you have an HP service contract?? If you do, when you open a case, tell them it's been multiple failures of the same part and you want the failing part CAPTURED. Let them do a failure diagnosis on it and figure out what's happening.

HP is one of, if not the only, OEM that still designs their own power supplies. I've been in the Houston lab where the PSUs are engineered - they take an incredible amount of pride in the effort that goes into their PSUs. They really are best-of-class... Cleaner, more efficient output than anyone else on the market. Good stuff.

The Houston folks are the ones that would diagnose a captured PSU. Let them at it!