We made the unfortunate discovery that our servers in our in-office server room are rusting away. This came to light after the first one had failed.

One obvious candidate is the AC unit, that something is wrong with its humidity regulation. So, I plotted the temperature and humidity. Lots to say, but this illustrates my question nicely:

enter image description here

In relatively sealed room (1.5 by 2.5 meters or so), the fact that humidity stays so high after each thermostat cycle is suspect to me. Also the fact that turning the AC off gave such an obvious suppression of the humidity peaks.

Is this normal AC behavior? I would not have expected the humidity to always return to such a high level. Even at the right side of the graph, it doesn't just stay low, it ‘wants’ to stabilize pretty high.

I also looked into other issues, such as sulfur containing fine particulate matter causing corrosion, but as it is, that's more of an abstract thought. I have no idea how to gauge and/or test that.

I also had an AC maintenance / air quality company look at it, but they seemed to think in terms of office air quality, and couldn't quite follow me on that servers have different requirements than people. For one, their suggestion was to pump fresh air into the room continuously. Seems illogical to me.

Edit: a higher zoom level, too see the correlation between rise/drop in temperature and rise/drop in humidity:

enter image description here

  • Purely out of interest, whereabouts are you in the Netherlands? – Chopper3 Oct 16 '17 at 14:07
  • 2
    We're in Groningen. – Halfgaar Oct 16 '17 at 14:24
  • @Chopper3 why :)? – Halfgaar Oct 16 '17 at 15:22
  • 1
    Just interested why it might be so humid was all, plus I like your country very much too :) – Chopper3 Oct 16 '17 at 19:44

The way an air conditioner dehumidifies is (in brief) the high-pressure refrigerant is forced through a small opening, leading to lower pressure on the other side of the opening. That decompression lowers its temperature; that low-pressure side starts near your evaporator (indoor) coils. Air is blown through the evaporator coils, cooling it. Moisture from the air condenses on the evaporator coils, just like it does the side of a cold drink. That water then drips off the coils, to a tray under them, and then out of the unit's drain (or on some units, into a reservoir which is then pumped out when its full enough).

The moisture has to come from somewhere. Even if somehow the AC evaporated all the condensate, you'd get the same absolute humidity (dew point) you started with.

So there are a couple of likely places it can come from:

  1. Outside air is infiltrating. Looking at the current conditions in Groningen, it appears the outside dew point is 16°C. That is awfully close to the dew point you measured with the A/C off.

  2. How good is the air mixing in the room? Is it possible you're cooling just a corner, and the rapid humidity increase after the AC turns off is just the air mixing? (E.g., is your AC intake next to the output?)

  3. Your drain line could be broken somewhere, and is dumping the water inside your server room, and it's then evaporating back in to the air. Alternatively, the water could be staying in the AC unit (and somehow not building up enough to spill out on the floor) Verify the drain line by inspecting the entire path and confirming water is flowing out.

  4. If you have humidification, it's possible it's running out of control. E.g., it's humidistat may be broken.

  5. You didn't say what test equipment you used to produce that graph, but if it's not accurate...

Also: Your AC appears to cool the room really quickly. This reduces its ability to dehumidify—it's quite possible your AC is simply too big (powerful) for the room.

  • It's a tiny room and a normal office AC unit, so it indeed maybe far too big. But what's the alternative? A doll-house AC unit? – Halfgaar Oct 17 '17 at 7:45
  • @Halfgaar I'm not sure what the smallest split indoor-outdoor unit is... But even with that problem, I'd still expect some dehumidification. That's also going to be the most expensive, so I'd check all the other things first. – derobert Oct 17 '17 at 13:58

Your statistic don't lie, get a new A/C.

In freezing mode the A/C kill the humidity, its just not freezing. Your stat show the inverse, its illogical, please change the unit, or buy a deshumidifer in the short term


There were some good answers, but the details deserve a separate answer, I think.

The comments about the AC unit being too large seemed to be closest: the condensated water never reached sufficient quantities to be drained out because the thermostat cycle was so short, so it got evaporated back. I was advised to turn the temperature up to 24 C, and set the fan to the lowest speed (it was set at highest, arbitrarily). It seems to have made a difference:

AC graph

Not only is the humidity lower, but the spike that normally coincided with rise in temperature (when the AC was in the off-state in the cycle) is gone; There is now a corresponding increase in temperature vs decrease in relative humidity, so probably the absolute humidity stays the same.

Also, the dew point graph is a lot flatter.

  • That's looks much better as the numbers are now moving as expected and values look within tolerance, though not ideal. – DocSalvager Oct 23 '17 at 6:19

Your rel. humidity is quite huge, standart for datacenters is 50% +-15%, you're reaching 80%. Check your AC settings if there are no option for increase humidity enabled, this just looks like not so healthy place for your servers.

Check if there are not any wet places on the walls, ceiling, floor (look under panels if you have floor with two layers). Are there any windows or holes in walls and ceiling (like for cables and pipes)?

  • I indeed found dpcalc.org, saying that I should be around 60% to prevent corrosion. I tried setting the AC to dehumidify, but that only (slightly) made it worse. Hence my question, is this AC unit likely not draining properly? BTW, to the touch everything (walls, etc) seems fine. No condensation on the window, etc. – Halfgaar Oct 16 '17 at 14:40

This may sound crazy but... take a good look at your A/C controls and control wiring.

Is it possible that something is backwards so that when you think you are turning it up, you are really turning it down?

Another angle... It looks like the humidity RISES rapidly when the A/C is on. It should be falling. Is the A/C compressor greatly oversized for the room? The compressor must run a certain amount of time in order to dehumidify properly. An oversized unit doesn't run the compressor enough to dehumidify.

  • If you look closely, you can see that the humidity falls when the AC is on. It's when the thermostat turns off you see that the humidity rises. But the rest of your theory is what I'm moving towards with an AC company. The condensated water in the short cycle is not enough to drip down the drain and is vented back in the room. But, it's "just an AC unit". They don't come much smaller, I don't think. What do other people with normally small server rooms do...? – Halfgaar Oct 21 '17 at 20:48
  • Perhaps I'm misinterpreting the graph? It looks like the humidity (blue line) gradually falls after the A/C is turned off and temperature rises as expected. But the humidity should be rising as well. It's only taking a brief fall immediately after A/C is turned back on. Then there is an large upward trend. Is this right? – DocSalvager Oct 21 '17 at 22:10
  • I added a new picture with higher zoom level, where you can see the humidity falls after the AC turns on, when it gets colder. – Halfgaar Oct 22 '17 at 13:41
  • I just reread your first comment... please don't be offended by this question but... The A/C unit IS mounted in a window or wall so that it exhausts to the outside isn't it? – DocSalvager Oct 22 '17 at 16:42
  • It's a split system, and the radiator with fan is on the roof, yes. But, I also answered my own question; see that for more details. – Halfgaar Oct 23 '17 at 6:01

The other answers all raise good points, but I want to add some more.

Is the server-room in the basement or ground-floor and is the local ground-water table high ? (That wouldn't be strange in many parts of Groningen.)
Could very well be that the floor or walls don't provide a good moisture-seal. With the cooling going on inside this may cause the server-room to continuously draw moisture from the walls/floors which get replenished by moisture migrating in from outside. (Damp surroundings make that even worse.)
If that is the case a coating the floors/walls with a moisture sealant would be a relatively easy fix.

Another point are the connection pipes between inside and outside unit. Some of these get REAL cold and can cause moisture condensation on the pipe. If that happens near the pass-through of the pipes through the wall, condensation buildup on the pipes can migrate alongside the pipe back into the server-room. (Directly on the pipe or through the insulation around the pipe. Rock-wool insulation is notorious for this.) Check the seals around that pass-through to see if you got a moisture leak there.

  • It's on the second floor, so that should be fine. I'll check the pipes, but I wonder. If I set the AC to fan mode, the rise in humidity immediately stops. – Halfgaar Oct 17 '17 at 10:04
  • @Halfgaar if setting the AC to fan only immediately stops the humidity rise, that's hard to explain. Check AC unit documentation to make sure it doesn't have some humidification (or re-evaporation of condensate) feature... – derobert Oct 17 '17 at 14:04
  • @derobert That is what I'm starting to think too... – Tonny Oct 17 '17 at 14:21
  • @derobert I will check, but from memory, the remote control didn't have many options. – Halfgaar Oct 17 '17 at 19:31

80% humidity is much too high. What kind of A/C is that?

A portable, one-piece A/C has a hot air exhaust, so it constantly sucks air out of the room that needs to be replaced from somewhere - usually from the corridor. Is this air is humid or too warm(!) it's the source of your problem. You'd need cool/dry air to flow back.

A two-piece A/C is a much better solution - however, it needs to shed the water from the condenser somewhere; it needs draining. If it doesn't there's something wrong with it. You certainly don't want a model that re-evaporates the condensed water back into the air.

  • It's a split system with proper drain. That is, I suspect the drain may not work properly. What you're saying confirms my suspicion. – Halfgaar Oct 16 '17 at 18:27

Not the answer you're looking for? Browse other questions tagged or ask your own question.