Mailinglist Archive: opensuse (3349 mails)

< Previous Next >
Re: [SLE] 9.3 cron.daily run time
  • From: Randall R Schulz <rschulz@xxxxxxxxx>
  • Date: Sat, 7 May 2005 12:40:24 -0700
  • Message-id: <200505071240.24871.rschulz@xxxxxxxxx>
Ken,

On Saturday 07 May 2005 12:19, Ken Schneider wrote:
> On Sat, 2005-05-07 at 10:08 -0700, Randall R Schulz wrote:
> > Ken,
> >
> > On Saturday 07 May 2005 09:12, Ken Schneider wrote:
> > > ...
> > >
> > > What ever happened to using the system clock to schedule cron
> > > jobs? Seemed to work correctly for how many years? Now you
> > > schedule based on how long ago a file was created which is rather
> > > short sighted to me.
> >
> > I believe the shortsightedness is your own.
> >
> > Many users don't leave their systems running 24/7 and if the daily
> > processing scheduling was done in the manner you suggest, it would
> > simply be skipped if the system did not happen to be up at the
> > selected time.
>
> This is true. What is the possibility of time creep on the files so
> that after a while the time at which they run starts becoming later
> in the day? Or is that something that actually runs in cron from the
> system clock. Also what about the people that just can't seem to set
> their clocks properly?

What about them?

The magnitude of the discrepancy between the sanctioned time standard
and the clock on any given machine (which is never zero) is not the
issue. If the system is running 24 hours a day, daily cron jobs will
get run even if they're scheduled using the naive means you suggest.
But if the system is not up 24 hours per day, then they may not. In
fact, if you pick an overnight time for the daily cron jobs and the
system is only running during, say, business hours, then the daily cron
will probably never get run.

And I don't understand what you're saying about "time creep." Richard
already acknowledged that it was a simple bug and he supplied the fix.
I applied it to my system, for what it's worth. I also messed up my
clock in my clumsy effort to immediately force the correct ctime upon
the time-stamp file, as you noticed.


> > > Like I have said in the past
> > > If it ain't broke, don't fix it (break it)
> >
> > More like "If you don't like it, lump it" or, perhaps, "My way or
> > the highway."
>
> I see no need for you to get offensive. I merely point out that
> something has worked well for many years, they change it and now
> people have problems. Perhaps better testing is in order then. And
> better documentation.

The point was not to give offense, but clearly the fixed-scheduling
scheme _is_ broken and _does not_ work well for any installation that's
not running on a continuous basis, and that's a lot of systems. Glib
statements like yours don't rise to the level of engineering
principles, they're just insults hurled by people who disagree with a
decision that someone else made. And I believe in calling things as I
see them.


> --
> Ken Schneider


Randall Schulz

< Previous Next >
Follow Ups