Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

For the heck of it, what's the fastest way to kill a server, other than just pulling the plug?

Kill as in a(n) (un)safe shutdown, you could theoretically reboot it and have it survive the incident.

Edit: Tangentially related to serverfault.com, my apologies. IMHO, the question is related in that these are possible things users could be attempting, as well as in the case of adverse conditions there might be reasons server admins want to take down multiple boxes lest undefined "damage" might occur.

Edit2: opened to community wiki.

share

locked by sysadmin1138 Nov 4 '11 at 19:06

This question exists because it has historical significance, but it is not considered a good, on-topic question for this site, so please do not use it as evidence that you can ask similar questions here. This question and its answers are frozen and cannot be changed. More info: help center.

closed as off topic by Zoredache, John Gardeniers, ThatGraemeGuy, Chris Thorpe, RobM Jul 15 '10 at 8:00

Questions on Server Fault are expected to relate to professional server, networking, or related infrastructure administration within the scope defined by the community. Consider editing the question or leaving comments for improvement if you believe the question can be reworded to fit within the scope. Read more about reopening questions here.If this question can be reworded to fit the rules in the help center, please edit the question.

1  
How is this related to system administration? –  Zoredache Jul 15 '10 at 0:16
4  
C4 ;) kills the server with 8040 meters per second which IS fast ;) –  TomTom Jul 15 '10 at 1:27
2  
Bazookas always work well. –  John Gardeniers Jul 15 '10 at 2:06
2  
I would say that exposing the server to a naked singularity (i.e. a black hole) at zero range would make it go offline quicker. –  Chopper3 Jul 15 '10 at 8:09
1  
I strongly suggest reopening it as a community wiki. It's a really fun question, and quite technically interesting, too. –  Massimo Jul 15 '10 at 9:27

13 Answers 13

up vote 10 down vote accepted

On a Sun server, Stop-A suspends the kernel immediately. You can press the key combination or send an equivalent signal on the serial console.

On Linux, Alt-SysRq-O crashes the kernel immediately. Through a remote shell, you can do echo o > /proc/sysrq-trigger. There are other magic SysRq keys that will cause an immediate unclean reboot. Alt-SysRq-S is handy to sync to disk first.

share
    
Bingo, cheers mate. –  EricR Jul 15 '10 at 0:22
    
I'd like to add that "echo b > /proc/sysrq-trigger" is useful when an NFS client is wedged so badly that a clean reboot is likely to hang forever. In that case, I would rather take the chance of local filesystem corruption (which is unlikely to be a problem with journaling) than to be left with a remote server that is unreachable because SSH has shut down and NFS hangs while trying to unmount. –  200_success Jul 15 '10 at 0:35
    
I shall have to try that next time our VM running Zimbra decides it neither wants to work or properly shut down. echo S > /proc/sysrq-trigger followed by echo 0 > /proc/sysrq-trigger should be slightly safer than just hitting the virtual bit red button. –  David Spillett Jul 15 '10 at 1:11

You're not scaling. If it doesn't scale, it's not a solution.

EMP pulse.

Theoretically, if it didn't hard-fry the electronics, you could reinstall the OS and things would survive the incident. Fastest way to turn off an entire datacenter of machines -- just use a larger pulse -- and will work better than just pulling the power, as those UPSes/batteries won't get in your way.

Of course, it's not ideal scaling, as you're dealing inverse-square law for power vs coverage area and any decent computer scientist will wince at such an O constraint and look for a better algorithm. Perhaps a bunch of EMP pulse emitters, each triggerable by being subjected to a pulse, would give you an easy scaling cascade system, at the cost of small delays in the effect as the pulses ripple out.

share
    
Wow, I like that. Wire them all together and have one box shielded (or not for that matter) set off localised EMPs throughout your data centre. Scalable and fun. –  EricR Jul 15 '10 at 15:41

give an oracle dba the root password.

share
1  
You'd really have to hate that poor server to do that!!! –  Massimo Jul 15 '10 at 16:47

For Windows, this one is simply delicious:

http://msdn.microsoft.com/en-us/library/ff545499(v=VS.85).aspx

Just a little Registry edit, and you'll be able to trigger an instant BSOD on any Windows system by pressing CTRL + SCROLL LOCK twice.

share
    
Oohh, nice - I didn't know about this –  Mark Henderson Jul 15 '10 at 5:53
4  
Well, there should be some way to force a system crash in order to test and debug it, I suppose... MS developers couldn't just sit there waiting for it to randomly happen. Even if in good old times this wouldn't have meant waiting so much :-p –  Massimo Jul 15 '10 at 5:57

For Windows:

shutdown -s -f -t 0 \\ServerName

This forces an immediate shutdown and causes all the applications to close (whether they'd like to or not).

share

This is handy for shutting down a Windows computer remotely:

C:\WINDOWS\system32\shutdown.exe -r -m \\MYCOMPUTERNAME01
C:\WINDOWS\system32\shutdown.exe -r -m \\192.168.1.186

-r means you are completely shutting it down and rebooting
-m means it's a remote machine
share
1  
The -f and -t flags could be of some assistance here. –  jscott Jul 15 '10 at 11:20

Two fun ones that may not give you 'comes back up' (data corruption is vicious):

dd if=/dev/zero of=/dev/mem
dd if=/dev/urandom of=/dev/kmem

Of course, the other way around would be interesting too.

share
1  
What OSs have a /dev/mem? I run ubuntu, and there wasn't a /dev/mem last I checked –  EricR Jul 15 '10 at 17:52
1  
Linux xxx 2.6.32-xx-generic xxx-Ubuntu SMP Mon Jul 5 09:20:59 UTC 2010 x86_64 GNU/Linux \\ $ ls -l /dev/mem crw-r----- 1 root kmem 1, 1 2010-07-14 00:02 /dev/mem –  Slartibartfast Jul 16 '10 at 3:05

Cross the pins in a USB port with something metal (paper clip works well), and you'll have a really quick power down (ie, instant). I've only ever seen this happen in a laptop, but I'd imagine it'd happen on a desktop as well. Amazingly, the computer has booted back up normally after it was done.

share
    
Interesting, hadn't heard of that before, something to try next time. –  EricR Jul 15 '10 at 3:19
    
P.S. - Only try it if you're ok with frying the whole computer in case something goes wrong. :) –  Paul Kroon Jul 15 '10 at 16:50
1  
On a similar note, get a usb cable and plug both ends into the same computer. You'll certainly fry some components that way. –  kaerast Jul 15 '10 at 17:34
    
According to this blog post, USB 2.0 devices are now required to withstand short-circuits. –  grawity Jul 18 '10 at 21:29

Run a fork bomb as root? (or any user if no user resource limits are in place)

Or cat hair, dust, and other general household detritus gumming up all the cooling fans?

share
    
Ah, good old fork bombs... –  Massimo Jul 15 '10 at 5:53

C4, or other explosive

share
4  
Question stated "have it reboot and survive the incident". C4 would not let it do that. –  Kevin M Jul 15 '10 at 1:21
4  
It would with remote detonator. User would survive ;) –  TomTom Jul 15 '10 at 3:11
    
@TomTom, excellent. :D –  EricR Jul 15 '10 at 3:19
1  
thermite. youtube.com/watch?v=jhfF2Et77aw –  petrus Jul 15 '10 at 10:43

Universal: Find a paperclip and jam it in the "Reset" hole

share
1  
...or in the power supply. But be sure to have good insulation... –  Massimo Jul 15 '10 at 5:52

Depending on your system

halt --force

might get the job done (as far as quickness goes).

share
    
I'll look in to it, thanks. I've never used the force option of halt. –  EricR Jul 15 '10 at 0:23

shutdown -t now has always been fast enough for me

share
    
sure, but that's no fun! That's the nice, pleasant, clean way of killing a server where it shuts down all running processes, kills the daemons, and stops. –  EricR Jul 15 '10 at 0:10
    
@Eric: I took your parentheses in "a(n) (un)safe" to indicate both alternatives. –  Dennis Williamson Jul 15 '10 at 0:36
    
@Dennis, fair enough. I should've worded that more clearly. –  EricR Jul 15 '10 at 3:19

Not the answer you're looking for? Browse other questions tagged or ask your own question.