[SOLVED] ntp drift file will not update

ewaller · 2015-04-04 01:58:15

I am sorry I asked you to post to the wiki. I did not mean to subject you to this.
I am not involved with the wiki; I must admit I've not edited it my self. This is starting to sound like Bismark's take on Laws.

As I said, I have no influence over the wiki. But, were I trying to get ntpd to run, a note that systemd interferes with that and takes it upon itself is a reasonable thing. I note that the OP did not even have reason to consider he was being stepped on by systemd. Why would they look in the documentation about the init system?

<Taking of my moderator hat>
That said. I use systemd. I am sick to tears about the ongoing bitching on both sides of the argument. I am not a big fan of systemd; but I don't want to lose any sleep over an init system.
I like the idea that systemd is working towards solving the secure boot problem. That is init related stuff. Why is systemd screwing with ntp? At least let's warn people in an appropriate place.
<moderator hat back in place>

Last edited by ewaller (2015-04-04 01:58:45)

Alad · 2015-04-04 02:03:46

I am sorry I asked you to post to the wiki. I did not mean to subject you to this. (...)

Not trying to subject anything to anyone. It's simply tiresome to see all kinds of "workarounds" on the wiki where nobody has any idea what's the relation to the original problem, and how/why they (do or do not) solve it.

As this is drifting off-topic, I've opened a discussion on the article talk page. OP: You should still post your pstree.

Last edited by Alad (2015-04-04 02:07:48)

ewaller · 2015-04-04 02:18:00

Okay, but...
I have been around the track a few times. I know because of my tenure here that systemd tentacles reached out and embraced time synchronization. For more three decades, ntp has been the canonical way to manage time synchronization. Why then is it unreasonable to add a note in the ntp wiki page that something has usurped that? Why would one expect one who is debugging a problem with ntp to decide to go read the documentation on a controversial init system that presumably has nothing to do with time synchronization?

What is worse: What if I am trying to run a time server? Systemd is great to keep a client synchronized. It breaks the ability to be a mid-stratum reference machine unless you reconfigure the init system. It might be nice to know.

Alad · 2015-04-04 03:53:21

ewaller, I appreciate your concern. systemd time-related functionality is mentioned in Time, and conflicts with existing daemons do belong in their respective articles (when properly shown).

What I find strange is how "enabling" ntp via timedatectl (when it already is through ntpd) supposedly solves the "format error frequency file" in the first post, if ntpd indeed stays running. Perhaps a file was deleted (adjtime?) in the process, and timedated regenerated it. I'm sure wiki readers would be interested in these matters as well.

Last edited by Alad (2015-04-04 04:10:28)

rg_arc · 2015-04-04 03:55:41

https://ptpb.pw/Qq8F

Alad · 2015-04-04 04:09:31

There's some threads on systemd-networkd relating to timesyncd. Not sure if related. Anyway, if I find out more I'll post it to the wiki discussion page.

Last edited by Alad (2015-04-04 04:19:25)

rg_arc · 2015-04-04 04:50:17

ewaller wrote:

Okay, but...
I have been around the track a few times. I know because of my tenure here that systemd tentacles reached out and embraced time synchronization. For more three decades, ntp has been the canonical way to manage time synchronization. Why then is it unreasonable to add a note in the ntp wiki page that something has usurped that? Why would one expect one who is debugging a problem with ntp to decide to go read the documentation on a controversial init system that presumably has nothing to do with time synchronization?
What is worse: What if I am trying to run a time server? Systemd is great to keep a client synchronized. It breaks the ability to be a mid-stratum reference machine unless you reconfigure the init system. It might be nice to know.

Yeah, my note in the wiki is really just a small reference to other issues with the wiki in general. Basically everything needs to be completely revised and certain references to other services need to be made so there is no confusion.

Furthermore, I understand people are polarized on the systemd/sysinit issue. Ultimately, who cares, if Arch Linux is going in the systemd direction I'll follow because linux is not about keeping certain methods around and doing things a specific way but about an ever changing community.

Last edited by rg_arc (2015-04-04 05:17:12)

Raynman · 2015-04-04 09:36:18

I had been following this thread yesterday, but how did it suddenly turn into another systemd discussion...

I see two issues (driftfile creation and being recognized or not by timedatectl) that are both related to ntpd, but there's not enough diagnostic information from before/after the various things rg_arc tried to really determine what was going on.

It is also unclear how exactly timedatectl interacts with the various NTP implementations, so we can't (yet) add the correct information to the wiki until somebody figures it out (read source, ask upstream as suggested on the wiki talk page).

rg_arc, you added a general description about enabling NTP-based time synchronization using timedatectl to the page about ntpd. You also said in this thread

rg_arc wrote:

Timesync will use NTP ultimately, not the other way around.

and

rg_arc wrote:

I don't believe removing my edit is appropriate as this is still related to using NTP.

But as Alad said, that page is specific to one implementation of NTP, namely ntpd (to quote Lahwaac from the wiki talk page: "ntpd != NTP", systemd-timesyncd uses NTP, not ntpd), so the information should also be specific to NTP and for that we need to at least understand how exactly timedatectl handles the set-ntp command (probably also depends on which NTP implementations are available/installed on the system).

Alad now added a clear warning, maybe that's a good compromise until the issue is clarified.

rg_arc · 2015-04-04 10:18:46

Raynman wrote:

I had been following this thread yesterday, but how did it suddenly turn into another systemd discussion...
I see two issues (driftfile creation and being recognized or not by timedatectl) that are both related to ntpd, but there's not enough diagnostic information from before/after the various things rg_arc tried to really determine what was going on.
It is also unclear how exactly timedatectl interacts with the various NTP implementations, so we can't (yet) add the correct information to the wiki until somebody figures it out (read source, ask upstream as suggested on the wiki talk page).
rg_arc, you added a general description about enabling NTP-based time synchronization using timedatectl to the page about ntpd. You also said in this thread
rg_arc wrote:
Timesync will use NTP ultimately, not the other way around.
and
rg_arc wrote:
I don't believe removing my edit is appropriate as this is still related to using NTP.
But as Alad said, that page is specific to one implementation of NTP, namely ntpd (to quote Lahwaac from the wiki talk page: "ntpd != NTP", systemd-timesyncd uses NTP, not ntpd), so the information should also be specific to NTP and for that we need to at least understand how exactly timedatectl handles the set-ntp command (probably also depends on which NTP implementations are available/installed on the system).
Alad now added a clear warning, maybe that's a good compromise until the issue is clarified.

I think it would be best for me to try and be more detailed with how my system is tuned. I have never used timesyncd before, so any information about that service will be appreciated.

What I think may be unclear with this wiki page is the distinction between systemd's ntpd.service and ntpd itself. The systemd service executes "/usr/bin/ntpd -g -u ntp:ntp" when it starts up, at least on my system it does. What I put on the wiki was very simple and may need more clarification on the subject, of which I don't currently have enough information. The timesyncd wiki page and the ntpd wiki page both are in need of expansion and revision. I don't really have any polarizing opinions on what .service should be used over the other, but I do believe there should be more clarification as to what is going on between timesyncd and ntpd.

lahwaacz · 2015-04-04 11:15:09

rg_arc wrote:

What I think may be unclear with this wiki page is the distinction between systemd's ntpd.service and ntpd itself. The systemd service executes "/usr/bin/ntpd -g -u ntp:ntp" when it starts up, at least on my system it does.

The ntpd.service is not part of systemd:

$ pacman -Qo /usr/lib/systemd/system/ntpd.service 
/usr/lib/systemd/system/ntpd.service is owned by ntp 4.2.8.p1-1

It is not part of upstream ntpd either, it is provided by Arch packagers as part of the extra/ntp package and the options used are the recommended way to start ntpd on Arch.

rg_arc wrote:

The timesyncd wiki page and the ntpd wiki page both are in need of expansion and revision.

This is meaningless without stating what is missing or why it should be improved. Feel free to start a discussion on the talk page on the wiki, or use the article status templates and mark the page directly (as Alad did).

Finally, please let's keep the wiki matters on the wiki, after all this thread is marked as [SOLVED]. Please contribute to the linked talk page on the wiki.

ewaller · 2015-04-04 15:16:48

lahwaacz wrote:

Finally, please let's keep the wiki matters on the wiki, after all this thread is marked as [SOLVED]. Please contribute to the linked talk page on the wiki.

I agree. Both as a moderator and the one who inadvertently got this started. That is why I kind of bowed out of the conversation last night.
I'll check back in a bit, and if there are no further comments, I'll close this thread so that it does not reanimate in the future.

rg_arc · 2015-04-04 18:31:40

Well it isn't meaningless as we did start a discussion and that _is_ meaningful if something constructive comes out of it instead of constant bickering over small changes to an open wiki. Systemd does initiate ntpd at system start that is what I have been saying this whole time, don't mix things up.

Last edited by rg_arc (2015-04-04 18:33:33)

Arch Linux

#26 2015-04-04 01:58:15

Re: [SOLVED] ntp drift file will not update

#27 2015-04-04 02:03:46

Re: [SOLVED] ntp drift file will not update

#28 2015-04-04 02:18:00

Re: [SOLVED] ntp drift file will not update

#29 2015-04-04 03:53:21

Re: [SOLVED] ntp drift file will not update

#30 2015-04-04 03:55:41

Re: [SOLVED] ntp drift file will not update

#31 2015-04-04 04:09:31

Re: [SOLVED] ntp drift file will not update

#32 2015-04-04 04:50:17

Re: [SOLVED] ntp drift file will not update

#33 2015-04-04 09:36:18

Re: [SOLVED] ntp drift file will not update

#34 2015-04-04 10:18:46

Re: [SOLVED] ntp drift file will not update

#35 2015-04-04 11:15:09

Re: [SOLVED] ntp drift file will not update

#36 2015-04-04 15:16:48

Re: [SOLVED] ntp drift file will not update

#37 2015-04-04 18:31:40

Re: [SOLVED] ntp drift file will not update

Board footer