Re: time problem

From	Rich <rich@example.invalid>
Newsgroups	comp.os.linux.misc
Subject	Re: time problem
Date	2014-10-27 12:30 +0000
Organization	My Linux Box
Message-ID	<m2ldtc$hv6$2@dont-email.me> (permalink)
References	(6 earlier) <m2i6qd$nta$1@reader1.panix.com> <m2j4r5$n7b$1@dont-email.me> <m2kjjg$cih$1@reader1.panix.com> <m2l6fo$qpq$2@dont-email.me> <m2lcq8$7p7$2@reader1.panix.com>

Show all headers | View raw

JohnF <john@please.see.sig.for.email.com> wrote:
> Rich <rich@example.invalid> wrote:
> > JohnF <john@please.see.sig.for.email.com> wrote:
> >> Rich <rich@example.invalid> wrote:
> >> > <<big snip>>
> >> > Or, setup an ntp client to keep your clocks synced with the ntp
> >> > system on the internet, and just let the Slackware 14.1 rc.d
> >> > scripts save the system time to the hardware clock when you do
> >> > shutdown.
> > 
> >> I think you maybe solved the problem. On a hunch, just noticed that
> >> rc.ntpd had 755 permissions, and indeed ps ax|grep ntp shows
> >>     727 ?  Ss  0:00 /usr/sbin/ntpd -g -p /var/run/ntpd.pid
> >> I must have inadvertantly told slackware's setup to run it during
> >> install (which would also explain why previous releases never
> >> exhibited the problem), and it's somehow been sporadically messing
> >> stuff up ever since. Changed permissions (and kill'ed this
> >> instance), so the problem's hopefully solved. No idea what,
> >> exactly, ntpd might have been doing, but it'll hopefully stop now.

> > That would do it.
> > Having an /etc/hardwareclock file set as "localtime" and the ntpd
> > daemon running would cause exactly what you've been seeing.

> Good to know there's (finally) a rational explanation.

> > Change /etc/localtime to UTC (as you've done, and leave it that
> > way) and you can keep ntpd running and always have the correct
> > time.

> Actually, after all this, I feel more like taking ntpd out back and
> feeding it into the wood chipper (channeling the movie Fargo).

It is not actually ntpd's fault.  Unix systems should always have their
physical clocks storing values that are UTC.  The translation into
human time zones happens when the humans ask "what time is it".

Ntpd is written assuming that the above is true.  That the clocks are
set with physical numbers that are UTC values.  The glitch was that the
actual numbers stored into the hardware by hwclock were physical
numbers after the conversion to local time (because you'd asked it to
do that in your old /etc/hardwareclock setting) while ntpd was setting
physical numbers that are valid UTC time values (as it rightfully should).

So the next time hwclock read in the values, it thought the values were
"local zone" values, and applied a conversion to produce what it
thought should be correct UTC values to put into the system clock.  The
result was a +4 hour discrepancy.

A good analagy is to think of it as imperial vs. metric measurments.

You have a numerical value.  It usually represents feet (localtime).
You want a value in meters (utc).  So you apply the conversion factor
for feet->meters.

But, if the number you read as "feet" is actually a measurment value
that is actually in "meters" already.  Applying the feet->meters
conversion factor will produce the wrong final value.

The above is what was going wrong with your times.  You were applying
"local->UTC" conversion to a numerical value that was already "UTC".

UTC + "local->UTC factor" ==> incorrect value.

Thread

time problem "Bill Cunningham" <nospam@nspam.invalid> - 2014-10-25 18:24 -0400
  Re: time problem Grant <omg@grrr.id.au> - 2014-10-26 11:02 +1100
    Re: time problem "Bill Cunningham" <nospam@nspam.invalid> - 2014-10-25 20:30 -0400
  Re: time problem Baho Utot <baho-utot@columbus.rr.com> - 2014-10-25 20:06 -0400
  Re: time problem philo  <philo@privacy.net> - 2014-10-25 19:53 -0500
    Re: time problem "Bill Cunningham" <nospam@nspam.invalid> - 2014-10-25 21:02 -0400
      Re: time problem Keith Keller <kkeller-usenet@wombat.san-francisco.ca.us> - 2014-10-25 18:20 -0700
        Re: time problem JohnF <john@please.see.sig.for.email.com> - 2014-10-26 03:31 +0000
          Re: time problem Rich <rich@example.invalid> - 2014-10-26 05:36 +0000
            Re: time problem JohnF <john@please.see.sig.for.email.com> - 2014-10-26 07:11 +0000
              Re: time problem Rich <rich@example.invalid> - 2014-10-26 15:43 +0000
                Re: time problem JohnF <john@please.see.sig.for.email.com> - 2014-10-27 04:35 +0000
                Re: time problem Rich <rich@example.invalid> - 2014-10-27 10:22 +0000
                Re: time problem JohnF <john@please.see.sig.for.email.com> - 2014-10-27 12:03 +0000
                Re: time problem Rich <rich@example.invalid> - 2014-10-27 12:21 +0000
                Re: time problem William Unruh <unruh@invalid.ca> - 2014-10-27 13:54 +0000
                Re: time problem JohnF <john@please.see.sig.for.email.com> - 2014-10-27 05:01 +0000
                Re: time problem Rich <rich@example.invalid> - 2014-10-27 10:23 +0000
                Re: time problem JohnF <john@please.see.sig.for.email.com> - 2014-10-27 12:11 +0000
                Re: time problem Rich <rich@example.invalid> - 2014-10-27 12:30 +0000
                Re: time problem William Unruh <unruh@invalid.ca> - 2014-10-27 13:58 +0000
                Re: time problem Rich <rich@example.invalid> - 2014-10-27 18:04 +0000
              Re: time problem William Unruh <unruh@invalid.ca> - 2014-10-26 16:13 +0000
                Re: time problem Chick Tower <c.tower@deadspam.com> - 2014-10-31 03:30 +0000
          Re: time problem William Unruh <unruh@invalid.ca> - 2014-10-26 16:02 +0000
            Re: time problem Andreas Kohlbach <oct14.5.ankman@spamgourmet.com> - 2014-10-26 17:48 -0400
              Re: time problem William Unruh <unruh@invalid.ca> - 2014-10-26 21:58 +0000
              Re: time problem "Nuno J. Silva (aka njsg)" <njsg@invalid.invalid> - 2014-10-27 08:47 +0000
                Re: time problem Andreas Kohlbach <oct14.5.ankman@spamgourmet.com> - 2014-10-27 17:20 -0400
            Re: time problem JohnF <john@please.see.sig.for.email.com> - 2014-10-27 04:45 +0000
    Re: time problem "Bill Cunningham" <nospam@nspam.invalid> - 2014-10-25 21:08 -0400
      Re: time problem philo  <philo@privacy.net> - 2014-10-26 07:15 -0500
  Re: time problem Balwinder S Dheeman <bdheeman.SANSPAM@outlook.com> - 2014-10-26 14:08 +0530
    Re: time problem Charlie Gibbs <cgibbs@kltpzyxm.invalid> - 2014-10-27 03:15 +0000
  Re: time problem William Unruh <unruh@invalid.ca> - 2014-10-26 15:53 +0000
  Re: time problem James Moe <jimoeDESPAM@sohnen-moe.com> - 2014-10-26 14:01 -0700
    Re: time problem Michael Baeuerle <michael.baeuerle@stz-e.de> - 2014-10-27 07:59 +0000
      Re: time problem Marc Haber <mh+usenetspam1118@zugschl.us> - 2014-10-27 13:31 +0100
        "iburst" option of ntpd (was: time problem) Michael Baeuerle <michael.baeuerle@stz-e.de> - 2014-10-28 08:23 +0000
          Re: "iburst" option of ntpd (was: time problem) Marc Haber <mh+usenetspam1118@zugschl.us> - 2014-10-28 10:14 +0100
            Re: "iburst" option of ntpd (was: time problem) William Unruh <unruh@invalid.ca> - 2014-10-28 21:10 +0000
          Re: "iburst" option of ntpd Chris Davies <chris-usenet@roaima.co.uk> - 2014-10-28 14:24 +0000

csiph-web