0

In our data center we are going to be moving all of our dual power supplied IBM servers (auto sensing 110/220) from an outdated 110 system of dumb power strips to an updated in rack 0u 220 monitored PDUs.

In order to do this we will either need to power down each server one by one to move both power supplies from the 110 plugs to the 220 plugs or move each psu at a time which means having half in 110 and half in 220 for a few seconds while we move the wires. Has anyone ever done this? I have read on a couple data center blogs that it's not a problem, found a support thread on HP about their servers saying it's not a problem, but still want to ask if anyone else has done it. Being able to use different voltages will reduce the time needed for the conversion from a year to instead several months as down time for every single system will not be required.

Again, these servers have dual, some quad, independent hot swappable power supplies that auto sense the incoming voltage and adapt accordingly.

Jason
  • 1

1 Answers1

8

POWER IT DOWN DUDE!!!

Seriously don't even think of running them separately, first of all their phases will be out no matter what you do - basically that's the same as crossing the streams - not advised.

Chopper3
  • 101,299
  • 9
  • 108
  • 239
  • Wrong: the power supplies output DC, which is why they can be redundant. The whole point of redundant power supplies is that they can be plugged into different sources that may be at different phases or voltages. – psusi Jun 16 '11 at 15:29
  • 2
    Output may be DC but input's not yet they'll share a common earth, therefore phase-shifts create spikes >220v - electrics-geek-ahoy! – Chopper3 Jun 16 '11 at 15:30
  • @Chopper3 the whole point of earth is that it is the universal common. It also is not normally carrying any current. While there may be high potential swings between the two AC sources, there is no circuit path between them so it does not matter. – psusi Jun 16 '11 at 15:33
  • Man, two different answers in 3 minutes! What about the phases being screwy do you mean? Would the inside of a server even notice the phases if they are converting power internally at the PSU level? – Jason Jun 16 '11 at 15:33
  • 1
    earth may in general by a universal common but it can be heavily affected 'in-box', I'm not kidding, don't mix two phases, three is fine as they're 120 degrees out of sync - hence the whole need for three phase, there's no such thing as two phase. You have to be very careful with not mixing PDU phases, people die. – Chopper3 Jun 16 '11 at 15:36
  • @Jason no, that is the point. Both PSUs feed the server the same DC power. The server doesn't care which one it comes from. – psusi Jun 16 '11 at 15:37
  • @Jason, have you got an up to date Last Will & Testament? either way shut them down. – Chopper3 Jun 16 '11 at 15:38
  • psusi - you're wrong, and I'm not arguing with you as I have a degree in electronic engineering. – Chopper3 Jun 16 '11 at 15:38
  • @Chopper3 that is just blatantly not true. People plug redundant power supplies into different UPSes and different PDUs all the time; that is the whole point. The two supplies are completely isolated from each other, so the phase of their inputs have nothing to do with each other. – psusi Jun 16 '11 at 15:39
  • Maybe for a test server I'll suit up in arc flash production gear and see what happens. I am not talking about really huge machines, just normal rack mount servers being plugged into an outlet, not directly into a big 75kva floor PDU unit. – Jason Jun 16 '11 at 15:43
  • But they're not isolated, they share a common earth, at that point within the server that supposedly neutral earth ISN'T neutral as two phases are out of step with each other raising the overall current carried levels in the earth - do I have to do a drawing for you. This is a dangerous practice and leads to injuries and deaths every year - seriously go speak to an electrician. If no how about I put you in touch with one of the HP server engineering team I work with? In the meantime can you just admit that you don't know what you're talking about and remove your dangerous comments or I will. – Chopper3 Jun 16 '11 at 15:45
  • 1
    How about this Jason, try it with a multimeter attached to the server chassis. – Chopper3 Jun 16 '11 at 15:46
  • @Chopper3 EVERYTHING shares a common earth. The earth only grounds the chassis for safety; it is not part of the circuit path and does not carry current. You seem to be mixing up common and ground and thinking that both power supplies use a bonded common. They do not have a bonded common; it is isolated from ground. Common may or may not be tied to ground at building transformer, but if they are, both phases will be tied to the same ground. – psusi Jun 16 '11 at 15:50
  • 4
    @Jason - Chopper3 is correct here. I speak from having had the experience of watching someone go stiff when a chassis went hot. You *CAN* do what you're suggesting and probably get away with it, but it's a **BAD IDEA**, potentially a lethal one. Take the outage window on the hardware rather than risk taking one on yourself. – voretaq7 Jun 16 '11 at 15:51
  • @psusi - Not all earths are created equal. http://www.catb.org/jargon/html/magic-story.html is a humorous, non-lethal example of this. – voretaq7 Jun 16 '11 at 15:53
  • Why do you think that whenever elecricians put in multiple phases that are in any proximity to one another they always label "DANGER 415V"? HSE guideline INDG247 says "Only supply a single phase to any one boom" (it's aimed at musicians). I'm not joking here, I've been in this industry for over two decades and have built some fricking enormous data centres and there are rules about mixing phases. Let me spell this out to you, this is about raising the potential in the server chassis, not somehow alteringin the world. – Chopper3 Jun 16 '11 at 15:56
  • So, let me get this straight: Running a server with redundant PSUs on different circuits (even with the same voltage) is a *Bad Idea(TM)*? – Sven Jun 16 '11 at 15:56
  • @Chopper3 yes, there can be a high potential between the two different lines, but since there is no path between them, it does not matter. – psusi Jun 16 '11 at 15:58
  • the chassis is the path by virtual of the shared common earth - that's what I'm saying!! do you know see that they share an earth - they have to, the chassis has to be earthed – Chopper3 Jun 16 '11 at 15:59
  • @SvenW - different circuits/spurs=fine, TWO different phases (out of three)=bad - I thought this was super common knowledge, but then again I did train for this stuff (although it was in the 80's but electricity's not really changed since then has it? ;) ) – Chopper3 Jun 16 '11 at 16:00
  • @Chopper3 and there is no voltage potential between the earth ground in one cable and the earth ground in the other cable! There may be some potential between the common line in each cable, but since common is NOT tied to the chassis, it does not matter. – psusi Jun 16 '11 at 16:01
  • How is the common not tied to the chassis? How is it earthed? – Chopper3 Jun 16 '11 at 16:03
  • @Chopper3 like I said before, it may or may not be earthed at the building transformer. Even if it is, since common is current carrying, it often does have a few volts of potential to true earth ground, which is why it is not tied to the chassis. You can plug the server into an isolation transformer and then common truly is isolated from ground and you will read half the nominal voltage on both hot and common with respect to earth ground, yet the chassis will still read 0v and be safe to touch. – psusi Jun 16 '11 at 16:07
  • @psusi - I've just thought, are you actually just trolling me? that would make more sense - if so you got me – Chopper3 Jun 16 '11 at 16:07
  • @Chopper3 no, I'm trying to explain how you are wrong. – psusi Jun 16 '11 at 16:09
  • And you're qualified how again? – Chopper3 Jun 16 '11 at 16:17
  • How are the ground wires tied to the chassis? Genuine question. I could run both of the PSUs outside of the server sitting on a table, they plug in and out into slots on the back of the server, like a floppy drive disk, while they are plugged in and powered up I can push a button and pop one out still attached to the power cable. – Jason Jun 16 '11 at 16:17
  • @Chopper3: dealing with the engineering I think, is a much better idea than appeals to authority, ad-hominium attacks, or other logical fallacies. – psusi Jun 16 '11 at 16:26
  • Another user on here asked his electrical engineer father about this question, his answer was "Don't even thing about doing this" - I don't know what you need to hear to listen to the advise of qualified people about this matter. Like I say, you're a troll. – Chopper3 Jun 16 '11 at 16:31
  • 1
    ad-hominium - sounds like a new element. – mfinni Jun 16 '11 at 16:54
  • @Chopper3: I presented you with a fact ( common is not bonded to the chassis ) that disproves your hypothesis. If you don't accept that fact, then go test it yourself. Measure the impedance between common and the chassis of any power supply and you will find it is infinite, as is required by UL. If you would rather stick to logical fallacies than reason, then I can't help you. – psusi Jun 16 '11 at 17:06
  • To add a little more background, there once was a time when electrical outlets only had two pins, and so the common WAS used to ground the chassis. That has not been the case for a very long time now. These days UL requires that the chassis be grounded using the third pin: the safety ground, and not tied to common. – psusi Jun 16 '11 at 17:19
  • 1
    Just realised something, I'm in the UK where we have a third ground/earth pin on all our gear - wonder if that's the source of our disagreement... – Chopper3 Jun 16 '11 at 17:25
  • 1
    @Chopper3 that is exactly what I am talking about. All modern gear has the third safety ground and that is the line bonded to the chassis, not the common line. – psusi Jun 16 '11 at 18:22
  • OK guys - now hug it out. – mfinni Jun 16 '11 at 18:36
  • I guess this is something the rest of the world doesnt need to worry about much as you guys have 220 everywhere and dont need to deal with 110? – Jason Jun 16 '11 at 18:39