RE: Automatic Reboot, Fsck and a PCI HW-Watchdog

[mgraffam mhv net]

On Sun, 31 May 1998, Brian E.W. Wood wrote:
> On 31-May-98 Gavin wrote:
> > Hi 
> > 
> > How would i best reboot my alpha say .. every 24 hours ? 
> > shutdown command in a cron job .. would that woek .. (and be a good
> > idea)?
> > and could the init script in RH 5.0 be modified so that fsck runs .. no
> > matter what ?
> > eaven when the FS is damaged .. 
> 
> Why on earth would you want to do this? If you have to re-boot every day (or
> even every month) something is seriously wrong. 

I have to admit that I can't figure this out, but running shutdown would
achieve it.

> As for "detecting" crashes, by definition a crash means things are not
> running, so how could the system "detect" this fact. You might have
> another machine test yours, to see if it responds to certain requests
> and thereby "detect" if it is down, but other than taking over its
> tasks I'm not sure what else it could do about it.

A watchdog board gets data written to it periodically by the mainboard.
If this data stream stops (because the mainboard crashes) then the
watchdog board resets the mainboard by tripping its reset line. Simple
idea, and a simple device. 

As I said in my other message, I hacked up a microcontroller to do this
for me via the serial port. 

> Linux systems are designed not to crash, and if they do you should
> discover the reason why, not work around it by re-booting at intervals
> (IMHO).

I agree. My Unix-based machines power down only for hardware upgrades and
reboot only for kernel (or other serious software) upgrades. Once properly
configured, I've never had a machine under Linux, or FreeBSD crash.

I suppose that some really hard-core applications may require absolute
system integrity and warrant rebooting frequently. Such systems would
also be required to do crash detection...

Michael J. Graffam (mgraffam@mhv.net)
http://www.mhv.net/~mgraffam -- Philosophy, Religion, Computers, Crypto, etc
"86% of conspiracy theories have some basis in truth... but, oddly enough,
it's that last 14% that usually gets you killed." 
    --Talas (http://cadvantage.com/~algaeman/conspiracy/public.htm)