Re: Root partition on ext3 totally damaged (total frustration)

From: Floyd L. Davidson (floyd_at_barrow.com)
Date: 10/22/04 

Date: Fri, 22 Oct 2004 04:02:48 -0800

Jean-David Beyer <jdbeyer@exit109.com> wrote:
>Floyd L. Davidson wrote:
>>>Here is a one-week event log from one of my UPSs:
>>>
>>>100401 05/03/04 07:07:09 Scheduled UPS self-test passed
>>>200006 05/09/04 07:21:57 UPS on battery: Deep momentary sag 075.4 V
>>>100300 05/09/04 07:21:57 Normal power restored: UPS on line
>>>200007 05/09/04 08:28:17 UPS on battery: Large momentary spike 124.8 V
>>>100300 05/09/04 08:28:17 Normal power restored: UPS on line
>>>200007 05/10/04 06:18:21 UPS on battery: Large momentary spike 125.4 V
>>>100300 05/10/04 06:18:21 Normal power restored: UPS on line
>>>100401 05/10/04 07:07:13 Scheduled UPS self-test passed
>> I doubt that you'd even know that was happening except for the
>> UPS being able to log it.
>
>The ones at night blink the incandescent lights. That is fairly long.

Yep. If it blinks an incandescent light, that's /long/, and would
probably cause most computers to reset.

>> Assuming the voltages are accurate,
>> that is an hilarious way to define "large momentary spike"! I
>> lived in one place for years where the *normal* voltage was
>> 124V. (Light bulbs don't last long...)
>
>Mine is like that...
>Date, Time, Min, Max, Curr. BattV Freq. %load TempC
>10/21/04,21:00:26,122.8,124.1,123.5,55.05,60.00,023.4,037.3, ,
>10/21/04,21:10:25,122.8,124.1,123.5,55.05,60.00,023.9,037.3, ,
>10/21/04,21:20:25,122.8,124.1,123.5,55.05,60.00,023.9,037.3, ,
>10/21/04,21:30:25,121.5,124.8,124.1,55.05,60.00,023.9,037.3, ,
>10/21/04,21:40:26,123.5,124.8,124.1,55.05,60.00,023.4,037.3, ,
>10/21/04,21:50:25,122.8,124.8,124.1,55.05,60.00,023.4,037.3, ,
>10/21/04,22:00:25,123.5,124.8,124.1,55.05,60.00,023.4,037.3, ,
>
>and I wish I could get 125volt bulbs around here.

"Rough duty" bulbs do last longer. Also any bulb claimed to
be "long life" will probably last longer. I ended up using
130V bulbs meant for radio towers and traffic signals...

>I think they are correct in saying there was a large momentary
>spike, but their A/D convertor providing the readings is
>probably an averaging meter and averages over an entire second
>or two (the computer polls the UPS every second or two).

The UPS need a *fast* response to be able to start up and switch
to the UPS in less than some small fraction of a second,
otherwise the switch itself would cause the computers to reset.
If they averaged over anything greater than a couple tenths of a
second it would be far too slow. I doubt that the poll
determines what is reported or the timing for it. It almost
certainly just picks up already generated packets of
information. (More discusson on what the packet might contain
is farther down.)

>> 75 volts is a genuine
>> sag, but the power supply might even be able to keep the
>> computer running on 75 volts for a long period of time!
>
>Well, the SSI12V specification says a compliant power supply
>whall not cause a reset if the power misses one complete
>cycle. It states, further, that it shall provide the specified
>output voltages when the line voltage is in the range of 90V to
>140V and 180V to 264V with the frequency between 47Hz and
>63Hz. Furthermore, power supplies intended to meet this
>specification shall be tested with 85V input to ensure
>compliance.

So they will *definitely* last at 85V (which would be at full
load), and my bet is that in most instances (all except for
perhaps a fully loaded power supply) they will function just as
well at 75V... and probably at 50V too!

In fact the actual range is probably *not* 90V to 140V, but
rather 90V to 264V, with no change required to switch between
115 VAC and 230 VAC power.

>Now maybe power supplies are all designed to runat 75 volts, but
>they are not required to be. This specification is for server
>systems, not the usual home systems.

Yup, that's the case. Note that they must run at *full load*
for the specification, and virtually no well designed system
ever actually puts a full load on the power supply. So while
it is tested at 85V, in practice it will work at voltages far
lower (in almost all cases, but that is *not* guaranteed).

>> 100 V, like 125 V, is definitely _not_ going to reset the PC.
>> (Not that it is bad to have a UPS that can switch that fast!
>> It's *great*. Just that you don't want to get too excited about
>> trivial power fluctuations.)
>>
>As I said earlier, I do not think those voltage readings are to
>be taken seriously. A flick down to 100 volts for less than a
>second would not affect an incandescent lamp's brightness that

Depends... You might well be able to see flickers that involve
only a few cycles of the power (especially if you are /not/
looking directly at the lamp, and see it in your peripheral
vision).

>much, but I can see them flicker, so the flick must be more than
>that. And the blackout below did not go to 9.7v, it went to 0.0
>volts.

That can mean either that the voltage is averaged and/or that
the report is based on a sample at some later interval (almost
certainly the case). The interval might be regular, not not
synchronized to anything, or it might be reported at some
specific time delay after a fault condition is triggered. That
would make sense as a way to provide more useful information.
And yes it could also be averaged over a longer period of time
than the fault trigger circuits, which might be consider as a
more useful indication of what happened.

The reason for doing that is because the UPS has to trigger a
fault and switch to backup power very quickly, but the actual
fluctuation causing the switch might well be far more severe than
just barely matching the fault trigger. For example, if a fault
is triggered at 109V or 127V (118VAC +/- 9V), for example, it
would be useless to report the exact voltage at that exact
moment. They might instead either log the voltage averaged over
the next 10 to 30 cycles (.12 to .5 seconds), or they might just
wait .1 seconds and log the instantaneous voltage at the point.

Either way, you are right that the logged voltage is an
"indicator", not an exact measure that means much about what
maximum or minimum voltage was actually experienced. (My only
complaint is with calling 125 volts a "large spike", which it
isn't.)

>>>100300 09/18/04 08:38:25 Normal power restored: UPS on line
>>>200000 09/18/04 10:10:22 UPS on battery
>>>100300 09/18/04 10:10:22 Normal power restored: UPS on line
>>>200003 09/19/04 01:01:17 UPS on battery: Blackout 009.7 V

...

>> My problem is that I have multiple SMP systems, and software
>> shutdown doesn't work well with SMP kernels.
>
>Both my systems are SMP systems. One is 4.5 years old, and the
>new one was put in service in March 2004. The software shutdown
>works perfectly well with both of them. The older machine has
>shut itself down quite a few times in the last 4 years, and for
>the shorter (15-minute) outages, coasts right through them..

I've got one fairly old Asus P2B-DS and two Tyan S2462 boards.
I did try to make the Asus board do shutdowns, but it doesn't
work. (It's been a long time, but it also seems to me that
the kernel was saying power management was disabled because it
is an SMP system???) I didn't really make any effort at it
with the Tyan S2462.

Could be I need to revisit that!

>> I've been going on ext3 and faith (and backups).
>>
>Not me. I sure do the backups, and maybe ext3 is better than
>ext2, but I do not use either with my dbms stuff; I use raw file
>systems for that.

Yes, it appears we have entirely different needs, which
definitely points out that there simply are no simple answers,
and each situation has to be evaluated on its own merits. Even
home networks... 

-- 
FloydL. Davidson           <http://web.newsguy.com/floyd_davidson>
Ukpeagvik (Barrow, Alaska)                         floyd@barrow.com

Relevant Pages

  • Re: line conditioner
    ... For example, if AC voltage is 110 volts, ... > I suspect that this line conditioner outputs something ... > First learn what a computer power supply must do. ... >>> so it's better to get a UPS. ...
    (alt.comp.hardware.pc-homebuilt)
  • Re: PC doesnt start
    ... With the switch pressed the voltage are within the specs but, ... I measured the voltage at the Power switch plug, when connected to the MoBo, ... Why then would a power switch or connections be ...
    (microsoft.public.windowsxp.hardware)
  • Re: PC doesnt start
    ... when the front switch was preessed the voltage went ... If the reset circuit does not work on power up, ... When you pressed the front power switch, then green wire voltage ...
    (microsoft.public.windowsxp.hardware)
  • Re: Thunderstorms
    ... the cable modem belongs to the phone ... We just switch it off at the wall (all wall sockets here have switches, ... Most UPS boxes are surge protectors. ... wall switch is nothing at all to a power surge from lightning, ...
    (comp.sys.mac.apps)
  • Re: Back again with some questions...(long - yep, you know it)
    ... I have on hand a three-pole 30A switch, ... so the full 50A load (plus power supply capacitors!) is never switched on by any contacts. ... I'll use seperate drive transformers for each IGBT so I can generate a gate waveform of positive and negative voltage, adequate to sink the current to get the devices turning on and off promptly. ... ZCS for IGBTs) in which case the datasheet is less than helpful. ...
    (sci.electronics.design)