• Remi (unregistered)

    So you mean you have a computer running only to reboot the other computer ? Sounds like a big waste of energy... :/ But I like the macgyver wire stuff

  • StDoodle (unregistered) in reply to Andir
    Andir:
    I have a much better solution to this, that would allow multiple servers to be rebooted with minimal hardware.

    Rig up one of those 12" clocks (you know.. the ones with hour/minute hands made of metal) so that one wire from every PC rubs against the center bolt of the clock and the other wire is positioned to trigger a reset at a specific time each day. You'd have to run the trigger wire from the back in a spot so it just barely touched the hour hand. (maybe a fine wire on the end of the hour hand?)

    Just be careful not to run 1.21 gigawatts through the thing, or you might send all of your equipment back to 1985 (which, from the sound of things, may be 'when' it is from).

  • screwdriver (unregistered) in reply to kastein

    Parallel ports are sold as General-Purpose-IO-Ports (GPIO) with industrial PCs these days. After all they're just as good as any of the GPIO-ports of any micro-controller. Together with using sound-cards as analog outputs they provide the base to recycle any old desktop or server into a very, very powerful toy for any real-time application.

  • (cs) in reply to jimlangrunner
    jimlangrunner:
    now THAT is enterprisey! You should wins some sort of enter-prize. heh.

    Ow. That hurt.

  • (cs) in reply to screwdriver
    screwdriver:
    Together with using sound-cards as analog outputs
    *facepalm* You just let us know how those output decoupling caps work out for you, yeh?
  • (cs) in reply to Remi
    Remi:
    So you mean you have a computer running _only_ to reboot the other computer ? Sounds like a big waste of energy... :/
    Right. That's why all the effin' Einsteins in this thread keep on suggesting running a second task on this crashed-hard-locked-up-solid-with-a-nonresponsive-PCI-bus computer and just having it reboot itself any time it notices it has crashed. To save power.

    despairs at the state of humanity

  • Anonymouse (unregistered) in reply to kastein

    The sad part is, we actually used a setup similar to this where I work. We plugged four external modems into a mostly-dead UPS (it would no longer charge a battery) and wired up a serial cable to a solid-state relay that would flip the power switch on the UPS whenever DTR was enabled on the serial line. The modems sometimes would fail and need to be power-cycled, and it was a lot easier to VPN into the network and click "Cycle Modems" than to drive all the way to the office and cycle them by hand.

  • (cs) in reply to DaveK
    DaveK:
    TRWTF is the URL to the "Meaninglessness" article. I mean, how does that article link come out as

    http://thedailywtf.com/Articles/MeaningleBneB.aspx

    unless someone's written it out by hand with german betas and then OCR'd it off a wooden table?

    Version 1 of the "Friendly URL" function. Guess what the following T-SQL prints out: PRINT REPLACE( N'Meaninglessness', N'ß', N'B')

    Obviously, lesson learned.

  • Matt A (unregistered) in reply to Nick
    Nick:
    I get the idea that the f*cking card in question is a proprietary piece of crap and that it's not an isolated defect but a design flaw present in all instances of this f*cking card.

    So replacing the fcking network card would leave you out some fcking money with the same fcking problem. Better to make a fcking simple solution that gives you some f*cking fun at the same time.

    f*ckin'-A man
  • (cs) in reply to Alex Papadimoulis
    Alex Papadimoulis:
    Version 1 of the "Friendly URL" function. Guess what the following T-SQL prints out: PRINT REPLACE( N'Meaninglessness', N'ß', N'B')

    Obviously, lesson learned.

    Ah, but will it be prepared for the horrors of a capital 'ß'?

    np: The Notwist - Hands On Us (The Devil, You + Me)

  • Some farmer... (unregistered)

    You are missing the required bailing wire! Basically, just replace the contacts with some bailing wire and you'll have a true solution!

  • (cs) in reply to Alex Papadimoulis
    Alex Papadimoulis:
    DaveK:
    TRWTF is the URL to the "Meaninglessness" article. I mean, how does that article link come out as

    http://thedailywtf.com/Articles/MeaningleBneB.aspx

    unless someone's written it out by hand with german betas and then OCR'd it off a wooden table?

    Version 1 of the "Friendly URL" function. Guess what the following T-SQL prints out: PRINT REPLACE( N'Meaninglessness', N'ß', N'B')

    So you mean that first it sees the double S and replaces it with a 'ß' for you because it thinks it knows what you meant, and then does the replacement you asked for?

    That's a bit microsoft-word-y, isn't it?

    Alex Papadimoulis:
    Obviously, lesson learned.
    What, "Computers suck"? ;-)
  • daqq (unregistered) in reply to anonymous coward

    You really should consult some hardware designers - inductors have the bad habbit of "charging" up to a lot of volts after they are switched off - this could kill your parralel board or your computer - if nothing else you should add another transistor to do the switching and a clamp diode as well as some power supply other than the parallel port.

  • (cs)

    These ingenious solutions make me think of going all the way through, and set up a PLC controlling a solenoid put in front of the "reset" button. When the ping fails, send the appropiate signal to your PLC, and voila! Reset server!

    Also, I'd wrap it all up inside a nice humanoid figure and put a USR logo on it. ;)

  • (cs) in reply to danixdefcon5
    danixdefcon5:
    Also, I'd wrap it all up inside a nice humanoid figure and put a USR logo on it. ;)
    I'm pretty sure that acronym refers only to the modem manufacturer, not USR&MM...
  • hi (unregistered) in reply to anonymous coward

    uhm, 8? try 12. and no card needed most of the time. plus usb.

  • (cs) in reply to daqq
    daqq:
    You really should consult some hardware designers - inductors have the bad habbit of "charging" up to a lot of volts after they are switched off - this could kill your parralel board or your computer - if nothing else you should add another transistor to do the switching and a clamp diode as well as some power supply other than the parallel port.
    Yup, gotta love inductive back-EMF. V = L*di/dt will destroy things rather quickly if you don't do that... luckily, most N-channel MOSFETs and power BJTs come with an integral clamp diode. Other than that, dropping one across the relay coil with the anode at the top usually does the trick.

    (I'm a hardware engineer, I just do software/firmware as my day job...)

    PS: my snide comment replying to some guy who suggested a direct connection above was mostly because that's pretty much forcing 5 or 12 volts (depending on design) from the CDROM drive motor controller straight into a logic input + pullup resistor that's expecting an open/closed connection. Not cool at all, which is why the relay and MOSFET are needed.

  • (cs) in reply to minkey

    Yeah, I can see this failing when the wires get bent too far out of shape from being forced past eachother like that, or by being brushed against by a random piece of clothing passing by.

  • (cs) in reply to minkey

    That was supposed to be a reply to this:

    minkey:
    Maybe it's just the 6 years of Air Force Aircraft maintenance speaking, but two bare wires touching scares me. You couldn't have bought a 2 dollar button at radio shack that the drive hits? so you could place the button where you want AND cover everything?

  • Wolfgang (unregistered) in reply to anonymous coward

    The safest method would be a optocoupling transistor. You can drive the LED with either a serial port's DTS line (add a 510 Ohms resistor) or one of the parallel port's lines (280 Ohms resistor). Use the transistor output for a open collector on the reset switch.

    Using a NE555, three resistors and a capacitor you can build a simple watchdog timer. Set it in a way, that after 5 seconds without reseting the watchdog timer, the system reset is triggered. Then a little app (could be even a shell script) that sends a watchdog reset to /dev/lp0 every second. Add some capacitor, discharge resistor and rectifier diode into the line from the port line, so that a static line level wouldn't keep the watchdog reseted.

  • Bush (unregistered)

    Did none of you guys ever hear of a power strip that runs linux? A hosting company I was with was using it, you could telnet to it and power cycle any port.

    I bet if you tried hard enough you could write a script on it that would cycle the pc if it couldn't ping it.

    I mean seriously...

  • Herby (unregistered)

    Look, the operative word here is CHEAP. All the components here were obtained at NO cost to the maker. The wire is obviously some left over telco wire, and the computer is some rag-tag thing left over from an old upgrade. The total budget for all of this is NOTHING (zero, nada). That is what makes it so good. If there were a larger budget, the whole thing wouldn't be necessary. Also being no cost, it could be done without any approvals, probably after hours.

    THAT is the beauty of this!

  • Michael Ensly (unregistered)

    This website is the main reason I am doing a degree in astrophysics

  • Mike D. (unregistered) in reply to anonymous coward

    Pedantry ahead...

    anonymous coward:
    If you're really in the business of abusing obsolete technology to get some general purpose IO lines, old parallel port cards give you 8 open collector outputs capable of switching about 10mA each.

    Attach these lines to a relay and you could reset 8 machines per card.

    As someone else pointed out, don't forget the diode across the relay coil. 1N4148 should work fine, and connect the striped end to the supply voltage.

    anonymous coward:
    Also, the flow control lines of a serial port can be used to similar effect.
    You're not using them to parasitically power a serial device? Bleh.
    kastein:
    luckily, most N-channel MOSFETs and power BJTs come with an integral clamp diode. Other than that, dropping one across the relay coil with the anode at the top usually does the trick.
    Cathode's the end with the stripe. If you connect the stripe end to the lower voltage (probably the parallel port pin), the diode will conduct instead of the relay and the smoke comes out.

    Also, the body diode in a MOSFET is a side effect of the way they're built, and isn't really that good for protection. They're generally not very fast-recovery.

    Blackice:
    So you took the time to connect your wire to the problem computer's reset switch, why not just connect the wire to the robot cd drive's "eject signal" wire? No exposed wires, no moving parts (at least none that affect your "circuit")
    What wire is this? The CD drives I've messed with were pretty well sealed up due to the laser. No easily-accessible eject signal.
    DaveK:
    screwdriver:
    Together with using sound-cards as analog outputs
    *facepalm* You just let us know how those output decoupling caps work out for you, yeh?
    DC-blocking caps. Decoupling caps are the ones connected across the power supply so the supply voltage stays put when all those transistors yank on it.
    Wolfgang:
    The safest method would be a optocoupling transistor. You can drive the LED with either a serial port's DTS line (add a 510 Ohms resistor) or one of the parallel port's lines (280 Ohms resistor). Use the transistor output for a open collector on the reset switch.
    This is the best idea in the bunch. No moving parts and you don't have to share grounds.
    Wolfgang:
    Using a NE555, three resistors and a capacitor you can build a simple watchdog timer. Set it in a way, that after 5 seconds without reseting the watchdog timer, the system reset is triggered. Then a little app (could be even a shell script) that sends a watchdog reset to /dev/lp0 every second. Add some capacitor, discharge resistor and rectifier diode into the line from the port line, so that a static line level wouldn't keep the watchdog reseted.
    I like the spirit of this, but a 555 circuit is complicated enough to be a project in itself, and your system will have to start hitting that watchdog before the 5 seconds is up after boot. I haven't seen anything with a BIOS boot that fast that wasn't some semi-custom embedded hardware.
  • (cs) in reply to Mike D.
    Mike D.:
    Pedantry ahead...
    Them's fightin' words!
    Mike D.:
    DaveK:
    screwdriver:
    Together with using sound-cards as analog outputs
    *facepalm* You just let us know how those output decoupling caps work out for you, yeh?
    DC-blocking caps. Decoupling caps are the ones connected across the power supply so the supply voltage stays put when all those transistors yank on it.
    They're both "decoupling" in the most literal sense of the word: isolating one subsystem from the influences of another. In the case of the audio card, there's an output circuit on one side of the cap, and an input circuit on the other side, and the cap simultaneously both couples those two circuits at AC frequencies - and decouples them at DC. Wikipedia thinks both usages are valid: http://en.wikipedia.org/wiki/Capacitive_coupling http://en.wikipedia.org/wiki/Decoupling_capacitor
  • (cs) in reply to JamesQMurphy
    JamesQMurphy:
    jimlangrunner:
    now THAT is enterprisey! You should wins some sort of enter-prize. heh.

    Ow. That hurt.

    Shouldn't the "enter-prize" go to the door opening robot?

  • Bob (unregistered) in reply to Remi
    Remi:
    So you mean you have a computer running _only_ to reboot the other computer ? Sounds like a big waste of energy... :/ But I like the macgyver wire stuff

    Run different server apps on both computers and have each computer reboot the other if it fails.

  • Mike D. (unregistered) in reply to DaveK
    DaveK:
    Mike D.:
    Pedantry ahead...
    Them's fightin' words!
    Great, I'm debating Stephen Colbert. "Pick a side, we're at war." ^_^;;;
    DaveK:
    They're both "decoupling" in the most literal sense of the word: isolating one subsystem from the influences of another. In the case of the audio card, there's an output circuit on one side of the cap, and an input circuit on the other side, and the cap simultaneously both couples those two circuits at AC frequencies - and decouples them at DC. Wikipedia thinks both usages are valid: http://en.wikipedia.org/wiki/Capacitive_coupling http://en.wikipedia.org/wiki/Decoupling_capacitor
    Well, if you're going by Wikipedia standards, then it sounds like you're set. But if you're going to spend any significant time reading electrical engineering trade literature, like data sheets, you're going to find that a decoupling cap means a capacitor connected across some DC supply, used to decouple the current transients of the load from the supply and from other loads. A series cap in any other line is referred to as either a blocking capacitor, a DC blocking capacitor, or an isolation capacitor.

    We use the terms in this manner to disambiguate these specific uses. There's already enough stuff in electrical engineering to make you go mad; creating ambiguity isn't going to help.

    CAPTCHA: "refoveo"... where do you get these?

  • methinks (unregistered) in reply to Florian Junker
    Florian Junker:
    There's still another problem: What happens when the restart robot dies? Does it have a backup?

    I see an infinite row of old-pc-with-wired-cd-tray robots, each guarding its predecessor...

  • Wasz (unregistered) in reply to Blackice
    Blackice:
    So you took the time to connect your wire to the problem computer's reset switch, why not just connect the wire to the robot cd drive's "eject signal" wire? No exposed wires, no moving parts (at least none that affect your "circuit")

    The old fashioned way won't make it to dailywtf

  • Watson (unregistered)

    "A little more complex and ... a little more elegant".

    No, we don't do any oxymorons here, nosir.

  • anonymous (unregistered) in reply to Neil
    Neil:
    XioPod:
    ping something.. if no return, shutdown -r -f

    done..

    Did you consider that the box that needs resetting may be having hard crashes from the haywire PCI card?

    reset button is similar to a soft reset, in that it doesn't necessarily reset peripheral cards. i had a coprocessor card that needed a power cycle when the POS locked up. it laughed at a software reboot, didn't even blink at the reset button. your kneejerk response could have been reworded to be informative, instead of just raining on the guy's idea. did you ever consider the merit in trying it first?

    :loop ping wireless.address | find "100%" if errorlevel 1 goto pinged shutdown -r -f :pinged ping -n 10 127.0.0.1 goto loop

  • (cs) in reply to Mike D.
    Mike D.:
    We use the terms in this manner to disambiguate these specific uses.
    I admit it, I was being ambiguous. I thought the context would serve to disambiguate.
    Mike D.:
    There's already enough stuff in electrical engineering to make you go mad; creating ambiguity isn't going to help.
    On the contrary, madness is the only thing that will save us now! :-P
    Mike D.:
    CAPTCHA: "refoveo"... where do you get these?
    Lorem ipsum.
  • (cs) in reply to Watson
    Watson:
    "A little more complex and ... a little more elegant".

    No, we don't do any oxymorons here, nosir.

    We do, but site rules only allow only self-affirmatory oxymorons.
  • (cs) in reply to anonymous
    anonymous:
    reset button is similar to a soft reset

    On most x86 machines, reset immediately power cycles the machine. Power off can be customized through the bios (depending on the bios) to either hard shutdown or soft shutdown.

  • OldTechSupport (unregistered) in reply to Some farmer...
    Some farmer...:
    You are missing the required bailing wire! Basically, just replace the contacts with some bailing wire and you'll have a true solution!
    Not to be confused with *BALING* wire. See, they're hay bales, not bucket handles...

    (Yes, pet peeve!)

  • James Cape (unregistered)

    Virtually all new servers come with with out-of-band management console that accepts SNMP or IPMI commands to restart the box.

  • Are Three (unregistered)

    Doesn't it reset it twice? Once when the tray opens and once when it closes? Or do you keep the tray open until the next time its services are called upon?

    Regardless... a very clever solution. I wish I had thought of something similar several years ago when I couldn't get that remote power switch approved. At least I got paid while driving in.

  • Bob (unregistered) in reply to Neil
    XioPod:
    ping something.. if no return, shutdown -r -f 
    done..
    Neil:
    Did you consider that the box that needs resetting may be having hard crashes from the haywire PCI card?
    I was thinking the same thing as Neil. Surely if the card was causing the computer to crash, it would have been mentioned in the original story? But yeah, either the author left out an important detail (crashing computer), or he MacGuyvered an unnecessarily complicated solution.

    I had a similar problem a while ago. I had a homemade Debian Linux router/general server with wired and wireless network. Every once in a while, the wireless card would cause the system to kernel panic, and I had to manually reset. This CD-ROM solution would have worked fine, but I ended up just replacing the official drivers with an open source version (which had its own set of problems, but none of them included kernel panic).

  • Are Three (unregistered) in reply to anonymous

    that should be find "100%%"

    otherwise the batch file only issues a 'find "100"' command.

    a single % evaluates to a null command line argument when run in a batch file.

  • Random832 (unregistered) in reply to Alex Papadimoulis
    Alex Papadimoulis:
    Version 1 of the "Friendly URL" function. Guess what the following T-SQL prints out: PRINT REPLACE( N'Meaninglessness', N'ß', N'B')

    Obviously, lesson learned.

    The other question is... why?

    No, seriously, why?

    Why not PRINT REPLACE( N'Meaninglessness', N'ß', N'ss') ?

  • Metal Lord (unregistered)

    seems like you're not THAT short of time over there, champ

  • (cs) in reply to Mike D.
    Mike D.:
    kastein:
    luckily, most N-channel MOSFETs and power BJTs come with an integral clamp diode. Other than that, dropping one across the relay coil with the anode at the top usually does the trick.
    Cathode's the end with the stripe. If you connect the stripe end to the lower voltage (probably the parallel port pin), the diode will conduct instead of the relay and the smoke comes out.
    Eep, you got me. I knead too proofreed my poasts beter.
    methinks:
    Florian Junker:
    There's still another problem: What happens when the restart robot dies? Does it have a backup?

    I see an infinite row of old-pc-with-wired-cd-tray robots, each guarding its predecessor...

    You're very clever, but it's turtles right the way down.
  • iToad (unregistered) in reply to minkey
    minkey:
    Maybe it's just the 6 years of Air Force Aircraft maintenance speaking, but two bare wires touching scares me. You couldn't have bought a 2 dollar button at radio shack that the drive hits? so you could place the button where you want AND cover everything?

    Actually, if you use duct tape to hold this thing together, then you have to use bare wires. If you get more advanced and use tie wraps, then you are allowed to use a button.

  • Real-modo (unregistered) in reply to jimlangrunner
    jimlangrunner:
    minkey:
    Maybe it's just the 6 years of Air Force Aircraft maintenance speaking, but two bare wires touching scares me. You couldn't have bought a 2 dollar button at radio shack that the drive hits? so you could place the button where you want AND cover everything?

    Seems that when I was working the the shop, we used micro switches (really about .5x1x2) with a roller bearing on the lever-end for limit switches. Depending on what we wanted to happen, we wired the NO or NC (normally open or normally closed) to good effect.

    So you COULD wire the NO side and place the switch on the (wooden) table leg at about the limit of the CD tray's travel and get the same effect. And you could cover those bare wires. You could also use the NC side for an indicator lite. or something. Light's on means everything is okay (or soon to be). Lights out means a reboot is coming.

    Enterprisey enough?

    Nope. No XML.

  • Real-modo (unregistered) in reply to Bob
    Bob:
    Remi:
    So you mean you have a computer running _only_ to reboot the other computer ? Sounds like a big waste of energy... :/

    Run different server apps on both computers and have each computer reboot the other if it fails.

    Yeah!

    Take that, VMWare Corporation!

  • (cs) in reply to DaveK
    DaveK:
    Watson:
    "A little more complex and ... a little more elegant".

    No, we don't do any oxymorons here, nosir.

    We do, but site rules only allow only self-affirmatory oxymorons.
    What the heck would one (or both, assuming the transitive property) of those be?

    "I'm good enough, I'm smart enough, and doggone, people aren't really people; just a conglomeration of individuals."

    Apologies to Mr Smalley, but I don't see this one flying off the shelves any time soon.

  • cph (unregistered)

    That "old desktop PC" that "has performed flawlessly for 9 months" probably draws around 100W when idle... That's almost $100 wasted in electricity costs over 9 months assuming average SoCal prices of $.14/kWh.

    Buying a remote-enabled power switch would have been cheaper overall !!

  • fragalot (unregistered) in reply to anonymous coward
    anonymous coward:
    Attach these lines to a relay and you could reset 8 machines per card.

    Right. so your plan is to power a relay directly from the parallel port? why not just use a simple transistor.. cheaper too

  • quux (unregistered) in reply to Andir

    ...and cron v0.001 was born

Leave a comment on “The Son of ITAPPMONROBOT”

Log In or post as a guest

Replying to comment #:

Log Out

« Return to Article