ITAPPMONROBOT

December 18, 2007 2:14 PM

"So, at our meeting earlier, you suggested building a robot. Is that something we can really do?"

Somewhat related: Robot cat feeder.

posted by CrunchyFrog (35 comments total)

I heard of a similar trick being used by someone who needed to leave town for a couple of days and couldn't find anyone to feed his fish, so he quickly fashioned a slightly sloping rigid plastic sheet over the tank, on which he laid a bunch of food. He then attached an old pager with the vibration alert active.

Whenever he needed to feed the fish, he just dialed the pager and the vibration shook a bit of the food into the tank.

Like the hack in the article, it is simple, elegant, and will probably work for ages if no one messes with it.
posted by quin at 2:29 PM on December 18, 2007

I weep machine screw tears for ITAPPMONROBOT
posted by CynicalKnight at 2:33 PM on December 18, 2007 [1 favorite]

Somewhat related: Robot cat feeder.

Of course, a nerd overthought it.

Let gravity do the work!

My "Circuit Party Feeders"^* work just fine.

* - For those long weekends when one is away in Miami, Palm Springs, etc.
posted by ericb at 2:34 PM on December 18, 2007

*BIG GRIN*
posted by xorry at 2:35 PM on December 18, 2007

reaching in vain for the restart button that it'd never touch again.

*sniff* it's so sad....
posted by pompomtom at 2:38 PM on December 18, 2007

On a somewhat related note (old robots, cuteness, etc.) New wall-e trailer is up.
posted by mrzarquon at 2:48 PM on December 18, 2007

The Daily WTF used to be funny, but now it's just annoying.

There was one a while ago where someone was complaining that someone had written a bunch of "frameworks" for their code, and so it was hard to figure out. Well, that's basically what you're supposed to do on huge projects, otherwise they become impossible to manage. You create a bunch of APIs that you call in your main program.

I dunno, once funny now dull.
posted by delmoi at 2:57 PM on December 18, 2007

Andy Ihnatko tells a great story about building a telepresence setup with a motion-sensing webcam and an animatronic Darth Vader to scare his cats off the desk.

MacGyver may have diffused bombs, but he had a whole team of scriptwriters behind him, these guys are making things happen FOR REALS...
posted by pupdog at 3:03 PM on December 18, 2007

Robotic Cat Flap Control and 1,000 Megaton, Earth-Orbital, Peace-Keeping Bomb from the quirky Arthur Paul Pedrick, a man with too much time on his hands.
posted by caddis at 3:10 PM on December 18, 2007 [1 favorite]

But did you cry when HAL 9000 died?

Uh, the second time.
posted by Kid Charlemagne at 3:12 PM on December 18, 2007

McGuyver never diffused a bomb. He could well have defused a bomb or two.
posted by everichon at 3:14 PM on December 18, 2007 [2 favorites]

The Daily WTF used to be funny, but now it's just annoying.

Sometimes it's good Delmoi, annoying when it recirculates posts.
posted by mattoxic at 3:14 PM on December 18, 2007

That Daily WTF reminded me of something I've always wanted:

Someone needs to invent a PSU or PC-AT power switch that's robotic and network-aware.

A nice little embedded industrial linux widget. Stable, low power, has a TCP/IP stack and script to check to see if it's host server is actually up and running, awake, and responding to the appropriate pings or server requests. Build it right into the PSU, power switch or motherboard - or, heck, make it an inline plug for your ATX plugs. You could ping it remotely to do a reset, or allow it to reset as programmed.

Such a thing would save so much money, gas, time and shoe leather it's ridiculous.
posted by loquacious at 3:23 PM on December 18, 2007

loquacious, most servers now come with Out of Band Management built in. XServes call it Lights-Out Management, and the utility allows you to remotely start a server as long as it's physically plugged in to the wall. It does need to be configured, of course.
posted by odinsdream at 3:31 PM on December 18, 2007 [1 favorite]

I've used the same CD-eject trick to punch a coworker, who was too close to my computer.

I really should get going on that USB Robotic Boxing Glove. I've been meaning to make on ever since working at AOL (we needed it to better interface with the users).
posted by aubilenon at 3:31 PM on December 18, 2007

Loquacious: Google "phidget" -- robotics kit parts. You could have a servomotor driver with a USB interface.
posted by adamrice at 3:41 PM on December 18, 2007

I had the same problem a few years ago, with a server that would crash, that I couldn't always be there to reset. Finally I came up with a great solution (though less quirky for sure). I got a telephone-control X10 module -- the kind that you can get at Radio Shack, that lets you call your house and turn your lights on and off, etc. using touch tones. Whenever the server would crash, I would call in and send the command to turn it off, then on again. It was set to auto-restart in case of a "power faliure". Just took a few seconds. The system worked perfectly and let me run my company while traveling in China, Thailand, France, Argentina, and so on.
posted by dacoit at 3:54 PM on December 18, 2007 [1 favorite]

Close, but this doesn't quite meet the requirements of a device that will allow us to punch people over the internet.
posted by Kikkoman at 3:55 PM on December 18, 2007 [2 favorites]

Kikkoman: I'm working on it!
posted by aubilenon at 4:03 PM on December 18, 2007

McGuyver never diffused a bomb. He could well have defused a bomb or two.

touché
posted by pupdog at 4:35 PM on December 18, 2007

New wall-e trailer is up.

Hey Astro Zombie .... get the Kleenex ready.
posted by itchylick at 5:01 PM on December 18, 2007

At the local ISP where I used to work, we had a spare web camera pointed at a cheap LCD thermometer in the server room. One of my coworkers hacked together a PHP script to grab a frame from the camera every minute, "read" the display and graph it.

I miss that place.
posted by you at 5:26 PM on December 18, 2007

A good story from wtf, sorta similar lines
posted by mattoxic at 5:40 PM on December 18, 2007 [1 favorite]

Waaaaah. I can even picture it, the CD drive mechanism slowly grinding to a halt, the small lever struggling to perform its now Quixotic^* function.

My reaction to this story reminded me of the IKEA lamp commercial.

* OMG, spelled it right on my first try
posted by Deathalicious at 6:03 PM on December 18, 2007

Metafilter: used to be funny, but now it's just annoying.
posted by MrMoonPie at 6:27 PM on December 18, 2007 [1 favorite]

I once heard of a data center whose servers had chronic memory leaks. They decided the cost of finding the leak was too large -- vs. staffing interns 24/7 to walk around and push the reset button on the servers at prescribed intervals.
posted by RobotVoodooPower at 7:32 PM on December 18, 2007

This is more of a hack than a WTF. You have critical infrastructure which is failing but you can't repair because of a budget freeze (that's not a WTF, that's standard business practice), so you kludge up the simplest possible solution to the problem. (Okay, the simpler solution is to find a different employer.)

odinsdream: Wow, the Wikipedia article you linked is one of the most buzzword-laden wikipedia articles I've ever seen. It's amazing how this industry can make such a big deal out of "you can reboot it from the serial console".
posted by hattifattener at 8:01 PM on December 18, 2007

Somehow I was expecting a story more like this one. But it was pretty cool anyway.
posted by A dead Quaker at 8:21 PM on December 18, 2007

I still like TDWTF. True, they've gotten a lot more preachy of late as Alex tries to tie all the horrible things they dig up into a series of moral lessons about how to do good software development, but the stories serve as an effective check on my own megalomaniacal tendencies when doing architecture. Good software tends to be a happy medium between the lazy hackery and ridiculous over-abstraction that gets featured on TDWTF; I think of their posts as William Hogarth engravings for programmers.
posted by xthlc at 8:29 PM on December 18, 2007

Along RobotVoddooPower's lines, I used to work at a place where the whole monstrous site was written in VB. They brought me in to handle upgrading the network side of things, and we threw a whole bunch of money at hardware to help the load, but even after the big hardware spend, the webserver front-ends would intermittently lock up and need a restart. Rather than fix the root of the problem (the site's code itself), they hired students from the community college to come in and do their studying on 4 hour shifts nights and weekends. While the students were there doing their homework/studies, they'd glance up at a monitor that had one window for each webserver, and if any of the browsers couldn't connect, they'd get up and reboot the corresponding webserver.

This was a "temporary" fix while the site redesign was done. The site was still being "redesigned" 18 months after I left that asylum for better pastures.
posted by barc0001 at 11:29 PM on December 18, 2007

McGuyver never defused a bomb. MacGyver could well have defused a bomb or two.
posted by Spire at 1:31 AM on December 19, 2007

Someone needs to invent a PSU or PC-AT power switch that's robotic and network-aware.

Yeah, they have 'em... Dell DRAC cards and whatever other vendors call 'em. Very useful.
posted by GuyZero at 6:26 AM on December 19, 2007

RobotVoodooPower: I once heard of a data center whose servers had chronic memory leaks. They decided the cost of finding the leak was too large -- vs. staffing interns 24/7 to walk around and push the reset button on the servers at prescribed intervals.

See, that's not a hack, or even inefficient, except for having humans do it. It's actually a smart way to run things, insects vs. mammals. Instead of naming individual servers like we do our pets, and treating them like we do our closest friends and family- individually irreplaceable and invaluable- better to have more cheap servers, basically homogenous, and treat them like insects in a swarm. So long as the swarm survives, the death of a few don't matter- and you can replace physically failed servers easily. Have say 5% more than you need in physical assets, and cycle them regularly. If individual servers are so broken they just don't work, drop them out of use altogether until the next regular weekly swap out. Replace the bad ones with spares, and you have a week to idly investigate the physically failed servers.

If it's not a memory leak, it'll be a bug that only creeps up after time due to a weird counting issue, or race condition. Or failed hardware. Or an OS crash. Or a hotfix problem. Or... hell, you can't anticipate every way a machine will fail, so why try? Just make sure that collectively, the swarm of servers fulfilling a function is up 99.99% of the time and you'll be in good shape. If you automate the removing a server from active rotation, cycling it, sanity testing it, and returning it to rotation before moving on to the next server, you can achieve insanely good uptimes with even the buggiest of old, rickety code and the crashiest of OSes. That's... smart, isn't it?


I work for a large online travel company, and last year there was some memory leak causing servers to fail every ~24 hours. Operations eventually had someone just cycle the servers every day at the same time, whoever was on duty. I suggested they automate that further: cycle the servers regularly in a constant rolling routine, whether they need it or not on the principle that if the code was making it through 72 hour stress tests without major errors, then surely it would keep running fine if no individual server was up more than ~4 hours to even encounter those edge cases/memory leaks/etc. They... didn't listen. I believe the stat I'd also heard was the previous year had some 10 days of unplanned downtime. Not "10 days with incidents", but ~240 hours of unplanned downtime. Eh... some people just don't get what their jobs *really* are...
posted by hincandenza at 1:13 PM on December 19, 2007

odinsdream: Wow, the Wikipedia article you linked is one of the most buzzword-laden wikipedia articles I've ever seen. It's amazing how this industry can make such a big deal out of "you can reboot it from the serial console".

I agree it's got a lot of buzzwords, but it's usually quite a bit more capable than just being able to reboot the system. The internal controller runs by itself as long as the system is plugged in. It has its own IP address and can monitor all of the temperatures, fan speeds, battery and power levels, SMART status, etc., whether the server's main internet connection is functioning or not. It can also send out alerts to an external SMTP server.
posted by odinsdream at 3:17 PM on December 19, 2007

My job would be infinitely more difficult if HP didn't have their iLO tools. There's really no reason to go into the data center when all of your machines are rocking iLO.
posted by esch at 1:05 AM on December 23, 2007