Re: GNU/Linux System Clock Drift

Farley Flud <ff@linux.rocks> wrote:

On 10 May 2025 12:58:18 GMT, Stéphane CARPENTIER wrote:
>

Yes it does.  All PCs will drift.

 
Nonononono. You precisely said you were speaking about systems and not
about hardware.

>
Yesyesyesyesyes.
>
The implication, as any INTELLIGENT person would discern, is that
any PC that is running GNU/Linux will drift in both the hardware
clock and the system clock.

No, you really don't understand. The hardware drift, there is no way
around that. We both agree.

Now, the solution is to check for an atomic clock who doesn't drift
because at the same time it's more precise and there are more than one
clock, so they can average the drift and stay stable. We both agree with
that.

That being said:
- your solution is the stupid one: you put the atomic time instead of
  the old wrong time. On a workstation it's not very important. But it's
  not good because when you check the logs, your system can go backward
  in time or shift as if nothing happened during some time. As long as
  you don't care about your system, you can live with it. But there
  exist better alternatives one of which is, guess what: ntp.
- the ntp solution is the smart one. Instead of changing brutally the
  clock with a nasty shift in the logs, it estimates the shift. And it
  compensate to, at the same time, get the clock on time, avoid
  shifts in the logs and avoid further shift on the system. So, on the
  long run, a system managed by ntp doesn't drift.

How difficult can that be to understand? Even for you, it should be
easy. It works out of the box: just let the system do the default job
and it doesn't drift. Your system drifts because you mess with it: not
because you know better, but because you can't understand the basics.

The hardware drifts, yes, there is no way around. But the system can
estimate its drift and compensate for it. So, once again, unlike your
beliefs, my hardware drift but my system doesn't.

The original question was: "How much drift?"

Yes, and I answered that many times: it doesn't.

But you cannot answer because you don't control your own system.

Unlike you, I answered. My system doesn't drift. It's the right answer.
You don't accept it because it's far above your limited capacities, but
I answered it nevertheless.

And, when you say that all of your computers drift by 0.7s each day, I
don't believe it. At the same time it's huge for a modern clock and
there is no reason for all of your computers to drift that much at the
same level. So you have no clue about how your computers drift.

Your distro does all the work.

And it does it well. Unlike yours because you broke it.

You are just a helpless observer.

You'd love to be a helpless observer because, unlike me, you know your
system drift and you have no clue about it. You don't know how to
estimate it because you wouldn't need to put the same unbelievable
number for all of your computers. And you don't know how to handle it
because you would be using ntp like every sane person.

So, you can't observe anything because you don't know how to do it. And
you don't know how to fix the issue you have because you are too proud
to learn. Good job.

I should also ask: "Does the clock correction include leap seconds?"

Of course, ntp hadn't be created by you: it has been created by people
who know what they do.

Again, you cannot answer.  You could not ever know if your system time
includes leap seconds or not.

Of course my system does include leap seconds. There is nothing
difficult about it. The leap seconds are managed by atomic clocks, so
each time ntp checks the time, if a leap second have been included, it is
taken into account. There is nothing special to do about it. Globally,
there is less than a leap second by year, so there is nothing to say
about it.

Why do you try to make simple things look like they are complicated when
you don't even manage the basics?

--
Si vous avez du temps à perdre :
https://scarpet42.gitlab.io