On Tuesday 03 August 2004 18:21, Mikkel L. Ellertson wrote:
> Frans Ketelaars wrote:
> > On Tuesday 03 August 2004 15:01, Mark Rogers wrote:
> > > Hi Team
> > >
> > > Thanks to all those who have given invaluable help over the last
> > > few weeks. One last item (for the moment) that I need to do is
> > > setup my system (which is always on) to reboot once a day. Im
> > > doing this to ensure there are no problems while absent from the
> > > machine for a couple of weeks. (And will be 10,000km from the
> > > machine). The machine will not be logged on as 'root' and my
> > > security limit is set pretty high (i.e. the reboot command needs
> > > to be looged in as 'root' before it executes). Anyone have any
> > > idea of how I should go about doing this?
> >
> > Slightly OT: you can easyly set up an old computer with an old
> > modem to dial in to (or which has a permanent net connection) and
> > let it powercycle the main machine when it has crashed. See
> > http://www.telenovela-world.com/~spade/linux/howto/Coffee.html for
> > details :-)
> >
> > -Frans
>
> Another way is to pick up a watchdog timer card, and let it reboot
> the machine if it locks up. There are several that are supported
> "out of the box" by the kernel, and it shouldn't be too hard to make
> others work. There is also a software version, but it isn't as
> dependable. It can not recover from a lockup inside a driver that
> has interrupts turned off. The hardware versions do not suffer from
> this limitation.
These watchdog timer cards are expensive though I think. They may be
able to reboot the machine more subtly than a power cycle would, but I
guess this depends on the kind of lock up and the capabilities of the
card.
You could also use an old (>= 386) computer to monitor the main machine
by checking if it continually flips a bit on an unused parallel or
serial port (with the main machine configured to do so) and after a
timeout period power cycling the main machine.
At least it would be cheap (hardware wise) ;-)
> On the other hand, with the normal uptime for a Linux server measured
> in months or years, it usually isn't needed... It has been years
> since I locked up, or crashed a Linux system that wasn't caused by
> hardware failing, loss of power, or by my exparmenting with something
> as root that I knew was risky. (Playing with video modes for DOSumu,
> new device driver, etc...)
>
> Mikkel
-Frans
____________________________________________________
Want to buy your Pack or Services from MandrakeSoft?
Go to http://www.mandrakestore.com
Join the Club : http://www.mandrakeclub.com
____________________________________________________
