Linux dies at 6am....

Martin Schwenke martin at
Fri Dec 7 09:38:29 EST 2001

>>>>> "Mark" == Mark Purcell <msp at> writes:

    Mark> On Thu, Dec 06, 2001 at 08:56:25AM +1100, James Macnicol wrote:

    >> Has anyone heard of such a problem or have any ideas how to
    >> debug it?

    Mark> In Debian the default time for running cron.daily is 0625,
    Mark> so it maybe that something which occurs then is causing the
    Mark> problems.

    Mark> Try moving the time for cron.daily in /etc/crontab and see
    Mark> if the problem moves...

Alternatively, when the system comes back, you could do:

  ls -ltru /etc/cron.daily/

This will show you the last access times for the files in
/etc/cron.daily/, which will tell you *approximately* what has been
run and what hasn't.  Only "approximately" because the system will
probably hang before it gets to flush all of the updated timestamps...

Another possibility is to run the cron.daily stuff by hand (as root on
a console, instead of under X) and watch what happens.  You could try:

  run-parts --verbose /etc/cron.daily

If the screen blanks when the system hangs, or for some other reason
you can't see the output, you could just cd to /etc/cron.daily and run
the jobs by and one at a time.

Obligatory war story follows:

At the ANU we had a Solaris box that would reboot every night when it
ran a particular cron job.  We couldn't trace the cause and when we
sent the crash dump to Sun they said it was because of a known bug in
the TCP stack.  A prerelease patch fixed the problem...

So, even though most of the cron jobs are hammering the disk I/O, the
problem might be more obscure than that...  :-)

peace & happiness,

More information about the linux mailing list