PDA

View Full Version : Several dead Xserve G5 PSUs- What's going on?




Projekt
Mar 24, 2013, 11:59 PM
Of our cluster of three Xserve G5's, TWO have had dead power supplies recently. As the dead PSUs start to pile up, I'm wondering if there is a way to remedy this myself. These servers are behind a top of the line power conditioner and haven't even been subjected to so much as an outage. Is there a known issue with these power supplies? Where would I even start if I wanted to begin replacing capacitors in an attempt to get these working again?

Any help is appreciated.



AlphaDogg
Mar 25, 2013, 12:09 AM
2 is a coincidence, 3 is a trend.

Lil Chillbil
Mar 25, 2013, 12:41 AM
Why are you guys still using g5 xserves, great and all but for as a professional should you still be using 10 year old tech for pro use

justperry
Mar 25, 2013, 12:46 AM
Why are you guys still using g5 xserves, great and all but for as a professional should you still be using 10 year old tech for pro use

Why discard them if they "serve" them well?

Hrududu
Mar 25, 2013, 01:27 AM
I wish I knew a little more about the Xserve, but I would highly suggest against opening up the PSU and trying to work on them. Aside from hurting yourself, if you make a mistake, you could fry a logicboard or CPU when you try and use it again. If its a late enough model G5, you may still be able to get the PSUs from Apple. If they are all similar build dates and the PSU's failed a few months apart, that could just be the lifespan. How many have you replaced in the last couple years, or were these originals?

Projekt
Mar 25, 2013, 09:53 AM
I haven't been managing this particular building for the lifespan of these machines, but I assume the PSUs are original to the machine. That being said, the servers themselves have been incredibly robust and it seemed odd to me to have a rash of PSU failures in such a short amount of time. I've put two more on order, but I'll give apple a call and see if they have any suggestions. I'd be surprised if anything comes of that, though, as I think anything PPC is now legacy hardware in Apple's eyes.

And to respond to Chillbil– the days when these ran mission-critical systems is long gone, but they still handle our archives very well. The strange thing is, they are in the same room as our newer equipment and have all the standard protections against surges and such, which made me especially curious about these failures.

Nameci
Mar 25, 2013, 11:07 AM
I haven't been managing this particular building for the lifespan of these machines, but I assume the PSUs are original to the machine. That being said, the servers themselves have been incredibly robust and it seemed odd to me to have a rash of PSU failures in such a short amount of time. I've put two more on order, but I'll give apple a call and see if they have any suggestions. I'd be surprised if anything comes of that, though, as I think anything PPC is now legacy hardware in Apple's eyes.

And to respond to Chillbil– the days when these ran mission-critical systems is long gone, but they still handle our archives very well. The strange thing is, they are in the same room as our newer equipment and have all the standard protections against surges and such, which made me especially curious about these failures.

Are you an engineer? Because you seem to think like one.

Engineers are more interested on the cause of the failure rather than the replacement process itself. Are you afraid it will also affect your new equipments in due time?

4JNA
Mar 25, 2013, 11:33 AM
i'll add my two cents. 10+ years give/take was the middle of the 'Chinese Capacitor Disaster' and most of the PSUs and mobo/logic boards from that era have failed or will anytime now. think 'iMac G5' for example...

EDIT: Dell (http://www.guardian.co.uk/technology/blog/2010/jun/29/dell-problems-capacitors) was hard hit by this, tons more links, this one summed it up nicely.

if you are curious, take one of the covers off a failed supply that has been disconnected for more than a couple days and look at the tops of the capacitors. the ones that are marked on the top with what looks like a 'Y' or that the top is divided into 3rds are the most likely suspect. the shape on the top of the cap is based on the factory or manufacturer, and the 'Y' caps were the hardest hit by the bad design. the second in line were the '+' marked caps.

even stranger yet, i've got new old stock (new, never used/installed) mobos from the same time frame that are now starting to show defective caps as well, so even replacement PSUs might have an issue unless they have been recapped or refurbed. best of luck.

Projekt
Mar 25, 2013, 11:39 AM
Great information, thank you. Do you know of anywhere that sells either recapped PSUs or does PSU repair? If its just a couple caps, couldn't I do it myself with some time and a multimeter?

justperry
Mar 25, 2013, 11:45 AM
Great information, thank you. Do you know of anywhere that sells either recapped PSUs or does PSU repair? If its just a couple caps, couldn't I do it myself with some time and a multimeter?

Yes, but you have to desolder them to get rests, switch the multimeter to Ohms and if the meter goes up it's OK, if nothing happens it's a bad one, but switch the ohm setting from lower to higher, so 10 Ohm-100 Ohm-1000-Ohm and so on.

4JNA
Mar 25, 2013, 11:49 AM
Do you know of anywhere that sells either recapped PSUs or does PSU repair? If its just a couple caps, couldn't I do it myself with some time and a multimeter?

sorry, no good links for you. with company money to spend, i'd buy from a reputable vendor with a decent warranty and forget about it until they go 'poof'.

usually recapping is no big deal, but the problem i've found with PSUs is that after they manufacture and test, they 'goop' or 'glue' everything together to prevent vibration/shipping damage. take the cover off any PSU and you'll see what i'm talking about. tough to get that crap off to get to the couple failed caps, if you can even find them with all that stuff on them.

life or death or can't spend any money, sure as it can't get any more dead and caps are cheap. time worth more than the replacement cost of the PSU, probably not then and there might be more damage caused by the fail that can't be seen. best of luck.

Lil Chillbil
Mar 25, 2013, 01:19 PM
Why discard them if they "serve" them well?

because they are pretty much imac g5s when you get right down to it. and imac g5s just arn't professional anymore

Projekt
Mar 25, 2013, 01:40 PM
As I said, they just handle archives and are no longer running critical infrastructure. I've ordered the new PSUs already, and its not so much the cost as much as it is my own curiosity.

skateny
Mar 25, 2013, 03:24 PM
Dell (http://www.guardian.co.uk/technology/blog/2010/jun/29/dell-problems-capacitors) was hard hit by this, tons more links, this one summed it up nicely.

Good read.

"Even the firm defending Dell in the lawsuit was affected when Dell balked at fixing 1,000 suspect computers, according to e-mail messages revealed in the dispute."

Wildy
Mar 26, 2013, 05:01 AM
because they are pretty much imac g5s when you get right down to it. and imac g5s just arn't professional anymore

Unfortunately no, the dual processors, ECC RAM, larger memory capacity, multiple NICs and high fault tolerances put this is a completely different class to the iMac G5s.

A server is designed to handle a couple of jobs, and do them really well. If it can still run those jobs, why replace it? As stated, they're no longer mission-critical, so next-day spares and servicing is no longer a requirement - why not use these machines? The only thing which makes them no longer relevant in the data centres of today is that because of the architecture, they don't support virtualization.


In response to the OP:
You already know, but I will repeat: be careful when handling PSUs - the caps are high capacity and will store charge for quite a long time.
http://www.pcstats.com/articleimages/200302/capblown_3.jpg

This is the sort of thing you want to be looking for (an extreme case) - but sometimes the bulge may occur on the underside of the cap which makes it a bit harder to spot. It shouldn't be hard to find and resolder a replacement once (if) you find the offending caps.

Are you sure they were the original PSUs and not some third-party offering?

Projekt
Mar 26, 2013, 10:35 AM
Yes, I looked into the issue and these are definitely original to the machine. If I can find some time this week I'll try my hand at fixing them, if only to measure the output voltages and see if I've succeeded (I'm not inclined to put my DIY-PSU back into a company machine)

seveej
Mar 27, 2013, 03:06 AM
I'm no engineer, but I have some of the mindset. in fact I often characterize myself (or more precisely my mindset) as an orphan having been raised by an engineer and an empath.

That said, I have had my share of experiences with computer hardware and PSU failures. PSU failures are (outside of fire breaking out in the server room) the worst thing that can happen to a computer - partially because spares can be hard to come by, but mostly because a PSU failure may lead to other parts failing as well (most probably power spikes). I've once had a PSU fail on me in such a way, that not a single component (CPU, RAM, MoBo, drives, PCI cards) was salvageable. I sincerely hope this is not the OP's case...

But, to get to the issue, and I do not expect y'all to agree, I am of the opinion, that no computer component which does not have moving parts SHOULD be able to break. That they frequently do so is, IMHO, attributable to a combination of design flaws or component QA issues.

I think the QA-aspect does not necessitate elaboration, on the other hand what I define as a design flaw implies an (IMO) skewed prioritization between power (capability), price and durability.

What I mean is that (and I think anyone with 10+ years working with computers can relate to) the whole concept of the lifespan of computer hardware being 3-5 years is absurd. We worked with that (then) state-of-the-art hardware in 1993 or 2003 and while we don't expect (or maybe "should not expect") it to run 1080p video, we should be able to expect it to do the same tasks today as it did back then.

Fact is, there are a lot of people who would be well served by 10-year old hardware, and I'm not thinking about sub-saharan Africa here. My personal everyday computing environment is based on machines from 2011, 2008, 2003, 2002 and 2001 (first two and fourth mentioned in sig), I've outfitted a number of modest computer users (with needs ranging from word processing via light DTP to web design) with :apple:-hardware from 1998 through 2008 (mainly 2003-2006). One of the central reasons why I've transitioned back to apple-hardware is that I've found the hardware to be more durable, but I'm starting to worry...

Notwithstanding physical wear and tear (and knocks) and components with moving parts (HDD, opticals, fans and hinges), I think there is no reason why machines could not be designed to reliably function for 10+ years.
In fact I just resurrected my old Sinclair (anno 1982) and it still works...

I acknowledge that the computer business -model is largely based on designed (forced) obsolescence, but I seriously question the sensibility of it - not only from an environmental standpoint.

RGDS,