Bug#947936: chrony: Does (still) not start properly on boot on buster

[Vincent Blut]

Control: tags -1 pending




https://salsa.debian.org/debian/chrony/-/blob/master/debian/changelog


On 2020-01-02T13:38+0100, Santiago Vila wrote:

Santiago, Michael, does that changelog entry about this issue feels clear to you?

Package: chrony
Version: 3.4-4
Severity: important

Dear maintainer:

Apparently, installing chrony does not ensure at all that it will work.

Google has moved from ntp in Debian 9 to chrony in Debian 10 for their
default Debian GCE images, and I discovered this on a lot of GCE
instances having a clock several minutes off.

The problem I found is very similar to the one described here:

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=933370

I believe the best summary of the problem was made by Michael Biebl
here:

https://github.com/systemd/systemd/issues/7104#issuecomment-471329392

Quoting Michael:

As it stands, the current practice of having systemd-timesyncd.service
enabled by default (in Debian) and alternative implementations like
chrony or ntpd declare Conflicts=systemd-timesyncd.service in their
service file does not work reliably.

AFAIK, this has been fixed on the systemd side in version 241-3 by
dropping the "Conflicts" systemd had on chrony or ntpd.

Unfortunately, AFAIK, conflicts are bi-directional, so apparently the
problem will persist in buster as far as chrony still has conflicts
in the systemd unit file.

I've noticed this problem happens randomly (it happens in some
instances, it does not happens in some others), so I don't have a
"recipe" as such to reproduce it.

However, I have a particular instance at GCE showing this behaviour
which I could try to clone to give you ssh access if required (please
contact me privately for details).

The behaviour is the following:

Both systemd-timesyncd and chrony are enabled (which is the default
on GCE instances). Just after a reboot, "systemctl status chrony" shows this:

● chrony.service - chrony, an NTP client/server
  Loaded: loaded (/lib/systemd/system/chrony.service; enabled; vendor preset: 
enabled)
  Active: inactive (dead)
  Docs: man:chronyd(8)
  man:chronyc(1)
  man:chrony.conf(5)

and I see this in the boot log:

systemd[1]: Condition check resulted in Network Time Synchronization being 
skipped.

If I do "systemctl disable systemd-timesyncd" and reboot, chrony is
properly loaded and it runs.

If I do "systemctl enable systemd-timesyncd" again and reboot, chrony
is shown as "inactive (dead)" again.

At this point, if I edit 
/etc/systemd/system/multi-user.target.wants/chrony.service
to remove the Conflicts line and reboot, chrony is properly loaded and it runs.

[ I'm reporting this as "important" because I believe it to be the kind
 of bug that should be fixed in a point release of Debian 10 ].

[ Cc to Michael Biebl in case he would like to comment on the issue ].

Thanks.

signature.asc
Description: PGP signature