Desktop restarting at midnight

[ivan]

I have a very strange problem.  One of our computers restarts the desktop at midnight.  I only noticed it because I have been on the night shift this week.

It is only this box - all the others stay stable with no hint of a restart at midnight, just as this box used to.  The only difference is that we had to replace the boot drive hard disk a couple of weeks ago and restore the image from backup.  Everything else on the machine works as it always did with all programs appearing where they were before the disk change.

We can live with it if we have to but I would be interested to know if anyone has seen this and if there is a way it can be fixed.

[Martin Iturbide]

Hi Ivan

Just curious, which OS/2-eCS version and release (ver /r) is having this issue? is that a full restart of the computer, or just the WPS desktop restarts?

Regards

[ivan]

Hi Martin,

It is WSeB with FP6 and any other fixes we could find applied.

It is just the WPS desktop restart - very like the xWorkplace restart desktop.  No it is not xWorkplace causing it because I uninstalled it on Tuesday and it still did it that night.

As I said we can live with it but it is a bit disconcerting when it happens when you are in the middle of doing something.  Another thing to note is that the image file was made two weeks before we had to replace the hard disk and it wasn't doing it up to then - it has only appeared after the new hard disk was installed and reimaged.

[Martin Iturbide]

I have no idea, but I have some questions about the restart periodicity.

Does it always happens at midnight? Everyday? Is it exactly when the clock marks 12:00 everyday, or it is just at night from time to time? Can you pinpoint the exact moment when it happens?

[ivan]

Exactly at midnight (00:00:00) every day - I just watched it happen again.

This has been consistent all this week because I have watched it.  I can't say anything about the previous week because the guy on duty that week didn't use this box but I don't think it started just to annoy me.

Something has to be triggering it and I think I have looked at everything that might.  It can't be the backup timing, that doesn't kick in until 02:00 and it hasn't missed that at all.

I don't think it is triggered by the new hard disk - all the SMART test come up clean and the firmware passes the manufacturers tests.

I can see this remaining a puzzle until we retire the box.
 

[Doug Bissett]

I heard about something like that, many years ago. It turned out to be a clock synchronizing signal, sent over the power lines (to sync all of the wall clocks). Why it would affect only one machine would be a puzzle, but looking for something like that might solve the riddle.

[Alex Taylor]

It's quite likely caused by the WPS process trapping. POPUPLOG.OS2 may contain some clues, or the system error log might.

[ivan]

Thanks Doug, I've got the recording oscilloscopes set to start recording just before midnight and we'll see what transpires.

Nothing in popuplog Alex, apart from the normal complaints re soffice.bin throwing a hissy fit when oo4.1 is closed.

An Edit well after midnight:

Your point is well taken Doug - it was a control pulse sent down one of the phases of the three phase supply we have.  At midnight alternate street lights are turned off by a control signal and those that are turned off are on one phase.  The reason the computer started acting up was because when we changed the drive it was plugged back into a different socket, one that was powered by the phase that carried the switching signal.  Why the signal should do what it did requires more investigation but we will be looking at installing signal filters on all incoming power lines.

[Doug Bissett]

Your point is well taken Doug - it was a control pulse sent down one of the phases of the three phase supply we have.

Interesting. I would think that a switching power supply (all of them for desktop computers, AFAIK) would filter something like that, but stranger things have been known to happen (and not all power supplies are equal).

You may want to check the ground on that outlet. It the ground is missing, it may amplify such a signal, and cause a problem.

Perhaps it generates a power on reset, which would cause a boot.

I don't know about your power, but here we use 115 volts, and most power supplies have a switch to select either 115 volts, or 230 volts. Here, if the switch is set for 230 volts, the power supply output tends to be low, and small power bumps (or "noise") will cause problems.

we will be looking at installing signal filters on all incoming power lines.

It may be more effective to get a UPS, although the cost would be more, overall. The cheapest option would be to determine which outlets are on that phase, and paste on a notice to not use those outlets with computer equipment.

Of course, you may have the problem again, if they decide to add the switching signal to another phase.

[ivan]

Hi Doug,

Here we have 415v three phase which gives 230v between phase and neutral.  What they are doing is superimposing a 175Hz signal on the 50Hz supply.  The signal is like Morse Code with the first 'dash' being longer (I assume it is a wake up signal).

I don't see any way that a UPS can filter out that type of signal - the big 5Kw on the servers didn't and I assume the smaller ones on essential equipment don't either.

Regarding filters on incoming power lines, we can buy off the shelf filters designed to block the signals sent out by the 'home plug' type of network extenders (network using mains cables) so it shouldn't be too difficult to adapt them to block 175Hz as well.

My first thoughts as to what happens is that the 175HZ signal interacts with one or more of the crystal controlled frequencies in the computer which is seen by the software as a reset but that is only a thought.  On Monday I will give the problem to my service department to kick about and see if they come up with other ideas.

[Doug Bissett]

What they are doing is superimposing a 175Hz signal on the 50Hz supply.

175 HZ could be difficult to filter, since it is so close to the base 50 HZ.

I don't see any way that a UPS can filter out that type of signal

A good UPS would detect a frequency change, or over/under voltage, and switch to battery until it goes back to normal (a few seconds in this case).

On Monday I will give the problem to my service department to kick about and see if they come up with other ideas.

As I said earlier, the cheapest way is to mark the outlets that are on that phase, so they don't get used by computer equipment, but first I would try another computer (or two), to see if it works, or fails. You may have a marginal power supply in the one that fails. A switching power supply should be able to handle something like that, but if the signal is big enough, or the power supply is marginal, it may slip through and cause problems.

Have fun, and good luck...

[Andi B.]

Do you mean with 'restart desktop' restart WPS only or the whole system? If restart WPS I can not see how this can be related to some power disturbances. If you mean restart whole system and it is somehow related to the 175Hz control signal you describe than go and buy a new power supply unit as yours is badly broken. Even if this 175Hz signal has amplitude high as the normal voltage (which I doubt) every PSU even the low quality ones have to handle such situations without any disturbance on the secondary (the mobo) side. AFAIR 20ms power supply interruptions while full load on the secondary side have to be handled by the PSU too. Usually a few 100ms interruptions are handled as hardly no one runs the PSU at full load. Thinking of surge and burst tests a PSU has to withstand I do not want not believe this 175Hz signal can cause troubles even for very low quality PSUs. Your mobo and your drives do not have any chance to notify such problems with any but a badly broken PSU.

Btw. SQR (3) * 230V = 398V. Usually called 3 x 230V ~ / 400V system  ;-)

[ivan]

Hi Andy,

I will admit it is a mystery.  There is nothing wrong with the power supply, all the voltages are normal as read from the bios health screen as well as the fact it is an Antec 600W supplying about 300W.

I do think the main problem is the mother board - it is 2001 vintage as are the CPU, memory and other cards with the exception of the video card which is 2004 vintage.

The initial 175Hz 'tone' lasts for 1.3 seconds, presumably because they know the rather indifferent nature of rural French power lines

Yes I do know single and three phase power formulae having one of my degrees as an electrical engineer.  Twenty years ago the nominal single phase voltage was 220v giving a 380v three phase this has since been increased to a nominal 230v (actually it is higher) with something approaching 415v three phase.

In the new year I have updating all the computers as a project.  What motherboards we will use I haven't decided yet but they will be AMD CPUs and graphics cards.  If anyone wants to throw in a recommendation I would be pleased to have a look at it.
   

[Doug Bissett]

In the new year I have updating all the computers as a project.  What motherboards we will use I haven't decided yet but they will be AMD CPUs and graphics cards.  If anyone wants to throw in a recommendation I would be pleased to have a look at it.

I don't have a specific suggestion, but I will comment that I have always had good luck with AMD processors, on Asus motherboards. My newest one is an Asus M3A78-EM board (built in ATI video, and really bad Realtek NIC - but it works with Multimac), with quad core Phenom processor. That is now pretty old, so things may have changed. I have thought about updating it, but my sister, and one brother, have the same setup, and that makes it easier for me to support them.

Please keep us informed. Many people would like to know what works.