Power-Cycling Windows 7 PC: Pitfalls?

(PeteCresswell) wrote:
I have a remotely-located Windows 7 PC acting as a server for a bunch of
IP cams at a second remote location.

Long story short: once or twice a month things get weird and the only
fix seems to be power-cycling the PC, switch, and radio link to the
cams.

During summer months I can usually get the owner of the property to
toggle the switch on the power strip that serves all the devices.

But other times of the year, the owner is away and getting somebody to
get in to the property and do the deed is a big deal.


Two options come to mind:

- For about $100, a Raspberry Pi hooked into an AC relay
that I can get to via the Internet and power cycle the
devices whenever I need to - basically duplicating the
current situation except that the homeowner or
neighbor is out of the picture.

- For about $10, a simple mechanical timer switch that
power cycles the PC and all once every 24 hours - so
when the cams go down, the situation is remedied
by the next day.


I like #2 for it's low cost and simplicity.

But are there pitfalls around power-cycling the
PC over-and-over again day-after-day?...

Some of the less expensive equipment that we install and monitor are
essentially mini ATX computers hooked to specialized control devices.
They tend to lock up when there are nearby lightning strikes or other
odd events.
We have them set to reboot every 24 hours, thousands of them spanning
years, and we have never had a problem with daily rebooting. Some of
the equipment in really bad locations we have hooked to Power Stones and
via Ethernet we can reboot them when ever necessary.
A mechanical timer would work but power outages will vary the reboot
time considerably. Instead of rebooting at 1 AM it will end up
rebooting a 9 PM, right in the middle of a critical time period.
Electronic timers go bad. We will not use either.

Per Paul in Houston TX:
We have them set to reboot every 24 hours, thousands of them spanning
years, and we have never had a problem with daily rebooting.

By "rebooting" do you mean having the power cut and re-applied without
doing a Windows Shutdown?

.... as opposed to something like Windows Scheduler event to gracefully
shut down and a BIOS clock to start up again...
--
Pete Cresswell

(PeteCresswell) wrote:
Per Paul in Houston TX:
We have them set to reboot every 24 hours, thousands of them spanning
years, and we have never had a problem with daily rebooting.

By "rebooting" do you mean having the power cut and re-applied without
doing a Windows Shutdown?

... as opposed to something like Windows Scheduler event to gracefully
shut down and a BIOS clock to start up again...

Most of the machines use Task Scheduler and a *.bat file in C\root to
reboot the units. The bat contains the reboot function:
%windir%\system32\SHUTDOWN.exe -r -t 00
We put Power Stones only on the really dysfunctional devices, mostly the
ones that get hit by lightning every few days and lock up.
Eventually we have to send techs to sites but we try to postpone that as
long as possible.
Most of the machines are on cell modems and we use RDP or Telnet over
VPN to make changes. The ones that need to be secure use Abierto gateways.

PeteCresswell wrote:

Per CRNG:

That's a controlled shutdown rather than "pulling the plug".

That's what I am talking about: "Pulling the plug".... Wasn't thinking
so much about hardware exposure as System exposu how well Windows 7
deals with just having the plug yanked.

Using NTFS for the file system with its journaling? You could also
disable the write cache buffering in your drives.

On Thu, 9 Feb 2017 18:16:14 -0500, Stan Brown
wrote in


On Thu, 9 Feb 2017 13:17:04 -0600, VanguardLH wrote:
Many users power off their computer when they are done using it. That
means, at least, one power cycle per day.


But that's not what the OP is proposing. He's proposing routinely
cutting the power, not routinely shutting Windows down.

+1
--
Web based forums are like subscribing to 10 different newspapers
and having to visit 10 different news stands to pickup each one.
Email list-server groups and USENET are like having all of those
newspapers delivered to your door every morning.

David E. Ross wrote:

PeteCresswell wrote:

Per CRNG:

That's a controlled shutdown rather than "pulling the plug".

That's what I am talking about: "Pulling the plug".... Wasn't thinking
so much about hardware exposure as System exposu how well Windows 7
deals with just having the plug yanked.

I strongly suspect that "pulling the plug" does no damage to Windows but
can cause extreme damage to the hardware. Actually, the damage can be
caused if there is a voltage spike when power is restored. The hardware
damage can appear to be software damage.

Um, with the PC powered off (whether it was properly shutdown or the
power was abruptly removed), power UP the computer would be the same as
EVERYONE does it: push a switch, power is immediately there.

However, I did have another thought: lots of computers will not power up
just because their A/C source came up. ATX mobos and PSUs don't power
up when you plug the PSU into a live wall outlet or turn on the power
strip. ATX uses soft power. When the input A/C power is available, the
PSU-on logic on the mobo gets power (+5VDC standby) from the PSU. The
user would then have to press the Power button on the case to pull PS-ON
to ground to get the PSU to actually bring up the PC.

So you could use shutdown.exe to properly shutdown the computer or you
could use a web switch to power off the PC but telling the web switch to
turn back on will not power up the PC.

Maybe there's a BIOS/UEFI setting to make the mobo power up with the PSU
gets power and somehow have the BIOS/UEFI drop the PS-ON line (on the
20-14 PSU header on the mobo) to tell the PSU to come full on. I have
not tried this but maybe an alarm event could be defined in the
BIOS/UEFI that makes the PC power up. So, say, 5 minutes after shutting
down the PC would come back up. If that works, no web switch is needed,
just the shutdown.exe scheduled event.

http://www.inspectmygadget.com/2009/...on-a-schedule/

So have a scheduled event that shutdowns the PC at 7:25 AM and use the
BIOS-level alarm to power up the PC at 7:30 AM.

Not all BIOS/UEFI mobos will have this feature. My buddy's brother had
a gaming PC that went bad and he didn't want to fix it and his brother
eventually bought a newer gaming PC. So he had one he didn't want but
didn't want to repair. I took it (free) and replaced the PSU with a
much better quality one and with more reserve capacity. A fan speed
sensor on the mobo was bad but I managed to software control the RPM of
the CPU fan (so it would be more quiet until temperature went up past a
threshold when the software [Speedfan] would increase the RPM). The
video card was nominal, there was no onboard video, the video card died
in a month, so I had to replace that. Alas, there is almost NOTHING
available for BIOS settings. So I couldn't even test the BIOS alarm
function to see if the PC would come back up at a scheduled time. Tis a
far cry from the mobos that I use in my own builds with a lot more
BIOS/UEFI tweaks.

Ken Blake wrote:

On Thu, 9 Feb 2017 13:17:04 -0600, VanguardLH wrote:

Many users power off their computer when they are done using it.

Yes, but I would probably say "most," rather than "many."

"many" covers "most". So saying "many" includes "most". On the other
hand, "most" exceeds or equals "many". Neither of us know the numbers.

That means, at least, one power cycle per day.

Yes, and for many people, they use it several times a day, and that
means multiple power cycles.

Users power off and on their TVs, too; however, from what I've seen and
do myself is to leave the TV even when I leave it when I know or expect
to be reusing the TV a short time later. So more likely a scenario is a
power cycle in the morning (if they have time to compute before they
commute) and again upon return to home from work. Again, and only from
personal experience, those that work from home, unemployed, or retired
end up leaving the PC powered up all the time or might let it cycle into
standby or hibernate modes (depending on how it was pre-configured).

The only time my computer is ever powered off is when I'm away from
home for a week or two, for example when I'm away on a vacation.

Yep, one of the last things I do before leaving the house to go to the
airport for a trip is to power off the PC, cable modem/router, other
electronics, check the stove, room lights, etc. Switch off when go off.

Char Jackson wrote:

PeteCresswell wrote:

I have a remotely-located Windows 7 PC acting as a server for a bunch
of IP cams at a second remote location.

Long story short: once or twice a month things get weird and the
only fix seems to be power-cycling the PC, switch, and radio link to
the cams.

If it were me, I'd want to know what it means to "get weird" and fix
that issue properly. Rebooting, especially by cycling the power,
doesn't seem like a good solution to me.

I used to use a magicJack USB dongle which required installing their
software (softphone) for VOIP communications. After a while, and
usually after the warranty expired (plus they only replace the unit just
ONCE and not every time for a failed unit), the damn thing would stop
working. I managed to figure out how to use a substitute softphone (I
didn't like the ad banner in theirs) but changing the software didn't
help. There was a problem in the hardware. Researching the chip they
used and which version discovered it was a known problem but no firmware
upgrades available (and they later replaced it with a newer chip). The
only way to get it working was to unplug the USB dongle and plug it back
in. Well, if you are trying to do this remotely then power cycling the
PC is your only option to power cycle the USB dongle.

I remember back when I used to have analog data modem cards in my PC
(full-chip modems and then Winmodems). Sometimes the card got into an
unusable state. The software could not get the card working again.
There was no way to issue a reset to the card to have it alone do a
power cycle or refresh its configuration. It was a card in a slot in
the mobo inside the case. Not safe to open the case while power was
still on and yank and reseat the card. So I had to power cycle the PC.
This was not specifically to power cycle the modem. Instead it was to
get the CPU on power up to send a Reset signal to all devices to force
them into a known starting state. Actually a power cycle was not
required, just a reboot which got the CPU to issue a Reset to the
internal devices. Getting the hardware reset got the modem back to
working. Yeah, the hardware had a problem but it was easier to
occasionally reboot the PC rather than see if buying and trying another
analog data/fax modem eliminated the problem. The problem did not occur
often enough to warrant the cost of replacement. Besides, back then,
the standing mantra of tech support when there were Windows problems was
to perform a reboot. In this case, the problem was with hardware and
not the OS but then I was having to reboot Windows often enough back
then that a couple more per month was not a significant increase in
nuisance.

PeteCresswell wrote:

Per Paul in Houston TX:

We have them set to reboot every 24 hours, thousands of them spanning
years, and we have never had a problem with daily rebooting.

By "rebooting" do you mean having the power cut and re-applied
without doing a Windows Shutdown?

... as opposed to something like Windows Scheduler event to
gracefully shut down and a BIOS clock to start up again...

See my reply to Ross. Although it is possible to use a web switch to
abruptly yank power from the PC, I'm not sure turning the web switch
back on will also turn on the PC. ATX uses soft power which means
bringing up the input A/C to the PSU will not turn on the PC. You have
to be there to press the Power button.

A Paul mentioned, you could try using an alarm event in your BIOS/UEFI
to schedule when to power up the PC. You would use shutdown.exe to
properly shutdown the PC and turn off power (although the +5VDC standby
from the PSU still goes to the mobo to power the PS-ON login on the
mobo) and then have a BIOS alarm turn the PC back on. If you wanted a
daily power cycle at 3 AM, schedule a shutdown.exe event to power off at
2:55 AM (or a little earlier if shutdown of the OS and powering off
takes longer) and schedule a BIOS power-on alarm at 3 AM.

If you use *only* the Power button on the computer case to shutdown the
OS and power off the PC, does a following press of the Power button to
power up the PC and load a fresh copy of Windows and all startup
programs and services result in your problem resolved? When you do a
cold boot (shutdown & power off followed by pressing Power button to
power on), do you see the LEDs flash on your keyboard? That means the
CPU issued a Reset signal to all internal devices to put them into a
known starting state. I believe "shutdown.exe /r" performs a shutdown
of the OS, powers off, and then uses a BIOS feature (alarm) to get the
PC powered back on. You could test. If the power bounce is too short
(when using the /r argument to reboot the PC) then just use shutdown
with /s instead of /r to shutdown the OS and power off the PC, and
following that scheduled shutdown with a BIOS scheduled alarm to power
up the PC.

You do not want to go into Standby or Hibernate mode. You want to
shutdown the OS and also power off the system. In case you have some
stubborn processes that refuse to unload, you will need to add the /f
argument to shutdown.exe to kill the unresponsive processes.

Also, on power up (whether using shutdown /r or a BIOS alarm or a web
switch), and if the camera software does not run as a service, you will
need to configure Windows to auto-login. You won't be there to login so
you need Windows to skip the login dialog to allow startup programs for
your Windows account to load. That also means anyone getting physical
access to that PC can use it without logging in. I suppose you could
configure the screen saver to be password protected but the intruder
could simply power cycle the PC to get Windows loaded and into your
auto-login Windows account.

CRNG wrote:

Stan Brown wrote:

VanguardLH wrote:

Many users power off their computer when they are done using it. That
means, at least, one power cycle per day.

But that's not what the OP is proposing. He's proposing routinely
cutting the power, not routinely shutting Windows down.

+1

Um, what did I say again that you cited? That users *power down* their
PCs. Yes, those users probably do shutdown the OS. That does not
preclude that they also POWER OFF their computer. Unlike the OP, most
users want to properly shutdown the OS *before* the power gets dropped.
Maybe the OP doesn't but that doesn't obviate my statement. As yet, the
OP has not declared that his OS gets into a state where it refuses to
shutdown. So far, the OP has not mentioned criteria regarding the state
of the OS, service, and apps, if any, that their state cannot be
changed, and whether that state is allowed to change or not before
dropping power.

shutdown.exe DOES both shutting down the OS *and* powering off the PC.
ATX mobos can be powered off using OS commands or using the Power
button. "shutdown.exe /s" shuts down the OS and powers off the PC. I
would have to test if "shutdown.exe /r" for a reboot both shutdowns the
OS, powers off the PC, and then use a BIOS alarm to restore power. If
/r does a reboot without removing power then /s would have to get used
followed by using a BIOS alarm to power up: schedule shutdown.exe with
/s 5 minutes, or longer if required, before the BIOS alarm is scheduled
to power up the PC.

What happens to your PC when you run "shutdown.exe /s" in an elevated
command shell? If your PC refuses to shutdown the OS and power off
perhaps you need to add the /f argument to force unresponsive processes
to unload so the OS will shutdown.

On Fri, 10 Feb 2017 01:31:20 -0600, VanguardLH wrote:
Ken Blake wrote:

[quoted text muted]

Many users power off their computer when they are done using it.

Yes, but I would probably say "most," rather than "many."

"many" covers "most". So saying "many" includes "most". On the other
hand, "most" exceeds or equals "many". Neither of us know the numbers.


Geeze. He was agreeing with you, with emphasis. At least that's how
it looked to me.

People are so touchy on the Internet.

--
Stan Brown, Oak Road Systems, Tompkins County, New York, USA
http://BrownMath.com/
http://OakRoadSystems.com/
Shikata ga nai...

"(PeteCresswell)" wrote in message
...
I have a remotely-located Windows 7 PC acting as a server for a bunch of
IP cams at a second remote location.

Long story short: once or twice a month things get weird and the only
fix seems to be power-cycling the PC, switch, and radio link to the
cams.

Do you mean that if you were physically present at the house, you'd just
turn the PC's mains supply off and on? Wouldn't you do a restart from within
Windows, to make sure that everything was closed down tidily. Power-cycling
a Windows PC sounds very risky and prone to filesystem corruption.

I'd say the best thing when accessing remotely is to power-cycle the switch
and radio link, but first access the PC remotely (eg via Teamviewer) and
restart it, either from the start menu or else from a DOS prompt with the
"shutdown" command with whatever arguments *restart* it.

On 2/9/2017 11:00 AM, (PeteCresswell) wrote:
I have a remotely-located Windows 7 PC acting as a server for a bunch of
IP cams at a second remote location.

Long story short: once or twice a month things get weird and the only
fix seems to be power-cycling the PC, switch, and radio link to the
cams.

During summer months I can usually get the owner of the property to
toggle the switch on the power strip that serves all the devices.

But other times of the year, the owner is away and getting somebody to
get in to the property and do the deed is a big deal.


Two options come to mind:

- For about $100, a Raspberry Pi hooked into an AC relay
that I can get to via the Internet and power cycle the
devices whenever I need to - basically duplicating the
current situation except that the homeowner or
neighbor is out of the picture.

- For about $10, a simple mechanical timer switch that
power cycles the PC and all once every 24 hours - so
when the cams go down, the situation is remedied
by the next day.


I like #2 for it's low cost and simplicity.

But are there pitfalls around power-cycling the
PC over-and-over again day-after-day?...

Here is a power switch that might meet your requirements.
It automatically pings a location on the internet and if the preset
remote site quits responding to pings it will power reset the outlet.
If the internet is up to the switch you can remote into it using either
a web browser or a smart phone application and order a power reset.
This unit is selling for around $70 and I believe they offer what they
call a two-outlet unit that also allows the remote ping site to be set
to something other than their built in site for around $100.

https://3gstore.com/product/4186_sin...FdgKgQodVdAB5g

No need to build a Raspberry Pi box when you can purchase something like
this ready to go. Rated for 10A Max load which should handle your
modem, router, cams, and possibly a single PC/Monitor depending on size.

On Fri, 10 Feb 2017 11:20:12 -0000, "NY" wrote in


"(PeteCresswell)" wrote in message
.. .
I have a remotely-located Windows 7 PC acting as a server for a bunch of
IP cams at a second remote location.

Long story short: once or twice a month things get weird and the only
fix seems to be power-cycling the PC, switch, and radio link to the
cams.

Do you mean that if you were physically present at the house, you'd just
turn the PC's mains supply off and on? Wouldn't you do a restart from within
Windows, to make sure that everything was closed down tidily. Power-cycling
a Windows PC sounds very risky and prone to filesystem corruption.

+1
--
Web based forums are like subscribing to 10 different newspapers
and having to visit 10 different news stands to pickup each one.
Email list-server groups and USENET are like having all of those
newspapers delivered to your door every morning.

VanguardLH wrote:


However, I did have another thought: lots of computers will not power up
just because their A/C source came up.

Actually, there is a BIOS setting for that, and two implementations.

In the BIOS, in the Power section,

Power options:
shut off on power restored [do not start because power has come back]
power up immediately [good for servers...]
previous state [when the power comes on, put the PC in the
soft ON or soft OFF state it was previously in]

Motherboards and chipsets have two levels of support.
Some seem to have all these options in hardware. If it's
supposed to stay off, it continues in soft OFF state.

I have at least two Asus motherboards here, that
have the BIOS settings, but the hardware support has
to be "emulated". That means, each time the power is
restored, the ATX soft ON triggers for about 1 to 1.5
seconds. The CPU runs, does a quick check of that BIOS
setting. If the PC is supposed to be OFF, it does
a soft OFF and doesn't complete POST. And it does
that, because there isn't the necessary state saving
info in SouthBridge/SuperIO.

It's a similar kind of monkey business that some PCs do,
so they can support overclocking. The motherboard does
a double start. On the first start up, nominal clock is used.
The BIOS loads the overclock controller with the BIOS
settings the user has selected, then the BIOS asserts
RESET and comes up a second time. They don't seem to
like clock ramping for some reason (which does seem to
work).

Even my current PC has some of this behavior. If the
mains power has been off and +5VSB drops, I'll catch it
doing a double start the next time. However, if +5VSB
is present day after day, each start is just a single start.

So you will catch these machines in some odd behaviors.

Paul