96

Is there a way to reboot a Linux system (Debian in particular) without rebooting the hardware?

I have a RAID controller that takes a bit to get itself running before the OS starts up, and I would like it if there was a way to quickly reboot the Linux OS without having to go through the whole reboot process of restarting the RAID controller, etc.

Improve this question
1
  • 6
    All UEFI based systems are able to switch OS whtout rebooting. You also havekexec()which is likeexec(), but for Unix kernels.
    – user2284570
    Mar 25, 2015 at 3:08

3 Answers 3

Reset to default
107

I use kexec-reboot on nearly all of my production systems.

It works incredibly well, allowing me to bypass the long POST time on HP ProLiant servers and reduce the boot cycle from 5 minutes to ~45 seconds.

See: https://github.com/error10/kexec-reboot

The only caveat is that it doesn't seem to work on RHEL/CentOS 6.x systems booting UEFI. But most sane OS/hardware combinations work.

Improve this answer
11
  • 66
    I wrote that! But it still works pretty well...
    – Michael Hampton
    Mar 25, 2015 at 0:27
  • 1
    Fascinating. So how does the actual process look? According to an article by Hariprasad Nellitheertha (linked to from the kexec wikipedia page), "Unlike the normal reboot process, kexec does not perform a clean shutdown of the system before rebooting. It is left to you to kill all applications and unmount file systems before attempting a kexec reboot." I notice some questionable things, for example, Wim Coekaerts says that "sync; umount -a; kexec -e" will do a very fast reboot. But you can't umount a busy filesystem. Perusing the web, it seems that many ignore the 'kill all applications' part.
    – Mike S
    Mar 25, 2015 at 14:12
  • 2
    @MikeS This isn't kexec. It's the kexec-reboot utility. The Github description outlines the difference and what the utility adds to the process.
    – ewwhite
    Mar 25, 2015 at 16:42
  • 4
    @MikeS After staging kexec, you simply do a shutdown -r. Your system scripts will notice that kexec has been staged and do the right thing. How does it look? Three or four or fifteen minutes shorter. I have a very horrible quality video I took off the iDRAC console, but you really don't want to suffer through that... The kexec-reboot script is meant to automate the process of getting the right command line arguments to kexec, which can be tricky.
    – Michael Hampton
    Mar 25, 2015 at 16:46
  • 1
    @MichaelHampton - cool. I see that on CentOS 6.5 near the end of /etc/rc3d/rc6.d/S01reboot there is: [ -n "$kexec_command" ] && $kexec_command -e -x >& /dev/null . So, iff a kernel is loaded by kexec -l, the system will reboot into it without a complete /sbin/reboot... kexec is baked in! I didn't know that, thanks. I suppose more modern distributions have it covered as well. BTW, if kexec wasn't previously run with -l, the reboot script will say "Nothing has been loaded!" which is why the output is sent to /dev/null, and why I was never aware of this facility.
    – Mike S
    Mar 25, 2015 at 18:01
18

Yes, it is possible. kexec will allow a Linux kernel to be booted directly from Linux without going through the BIOS boot process.

Improve this answer
2
  • Also, the UEFI allow to switch the OS without rebooting.
    – user2284570
    Mar 25, 2015 at 3:09
  • 2
    @user2284570 As far as I'm aware, UEFI provides no such mechanism.
    – Jonathon Reinhart
    Mar 3, 2019 at 17:43
17

Yes. You must use kexec-tools. To make the life easier, I use kexec-reboot.

$sudo apt-get install kexec-tools
$wget https://raw.githubusercontent.com/vadmium/kexec-reboot/master/kexec-reboot
$chmod +x kexec-reboot
$sudo mv kexec-reboot /usr/local/sbin/kexec-reboot
$sudo /usr/local/sbin/kexec-reboot
Improve this answer
3
  • 13
    So that there is no confusion, I did not write this.
    – Michael Hampton
    Mar 26, 2015 at 4:06
  • This is likely Ubuntu/Debian-only, right?
    – ewwhite
    Mar 26, 2015 at 22:09
  • 1
    @ewwhite It would probably work on RHEL/CentOS, but on a reading of the code I found several problem spots where edge cases might cause it to fail. And the source seems to acknowledge this with a comment: # The error checking is very basic.
    – Michael Hampton
    Mar 26, 2015 at 23:23

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .