[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: question(feature request..?)

From: Sorin Serban
Subject: Re: question(feature request..?)
Date: Sun, 23 May 2004 18:45:14 +0300
User-agent: Mutt/1.5.6i

On Sun, May 23, 2004 at 04:20:01PM +0800, Uwe Dippel wrote:
> address@hidden wrote:
> >hello. i have come to play with grub a bit and i am really excited about
> >it. however, i wonder if there is any function for rebooting in case of
> >kernel panic and booting a different entry in the menu (something
> >similar to what lilo can do: boot once + reboot on panic..)
> >
> >
> > 
> >
> Sorin,
> not being so familiar with lilo, this sounds phantastic (in all meanings 
> of the word). It implies that the boot-loader doesn't cede control to 
> the kernel; rather works like a shell around it; controlling its ups and 
> downs.
> Earlier I was working on a concept of a 'standing-up' OS and so find 
> this a welcome support: If a boot-loader, as a very small, simple and 
> reliable program, keeps in overall control during kernel load; that is 
> through an initrd, mount, etc. and could intervene at any moment of 
> failure, we might want to keep it as an outer shell around that kernel 
> forever.
> If that boot-loader can throw out the kernel during boot, why should it 
> not be able to throw out the kernel after boot, whenever the system 
> comes to a grinding halt; be it after two weeks of uptime ?
> You might argue that the admin can restart the system; or the system 
> reboot automatically. Right, but at times this won't work; the admin be 
> unavailable or the system got corrupted and reboots over and over. One 
> could even envisage a rudimentary message passing from the dying kernel 
> to the boot-loader to inform it that reboot won't work; another 
> partition ought to be started or even that a re-install respectively 
> replacing the current install with the latest good image is the thing to 
> do. Then the boot-loader could automatically boot to another partition 
> with another install (back-up install) or quite another partition that 
> simply regenerates the main install / partitions from a good image. With 
> this boot in turn passing a message to that boot-loader in case of 
> successfully writing back the image and auto-finishes with /sbin/reboot 
> or similar; getting the boot-loader to go back to running the by now 
> repaired / reinstalled system.
> Or just boot to a live-CD set up to return basic functionality to a 
> system while waiting for the repair. Think of a compromised webserver 
> that comes back automatically from live-CD.
> The point should be clear by now. Can the experts in here say something 
> about the feasability of this; and if lilo is really so close to being a 
> URM (Uptime Reference Monitor) ?
> Then I could only hope for grub to catch up ... !

well, i seemed to have phrased that wrong, i was veru tired to. hte
previous reply seems to be more or less what i meant. the boot on panic
option is a kernel option, that u can append (panic=[timeout_in_seconds]
) . for what u were talking about, there are utilityes like watchdogs
(dun really know if they can do _all_ u want there, though). it is
really not the bootloaders job to watch for what happens to the system
after it booted (i.e. the kernel is loaded and it has taken control..)

10x a lot for ur input.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]