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

[Vincent Blut]

On 2020-02-01T22:46+0100, Michael Biebl wrote:
>Am 01.02.20 um 22:37 schrieb Santiago Vila:
>> On Sun, Jan 12, 2020 at 08:41:19PM +0100, Michael Biebl wrote:
>>
>>> I guess at this point it is best to ask chrony, ntp, openntpd, ntpsec
>>> and virtualbox [1] to drop the Conflicts= line again.
>>> Maybe we should even do that for buster via a stable upload?
>>>
>>> Thoughts?
>>
>> Hi. I definitely think this should be fixed in stable, in whatever way
>> it's considered best for stable.
>>
>> The last thing a system admin would expect from Debian stable is that
>> the clock is off by several minutes in a system where a time-keeping
>> package like ntp or chrony is present. This was completely unexpected
>> for me. (In fact, I would have reported this as serious but I prefer
>> to concentrate on finding a good fix).
>>
>> Regarding the proper fix: Anything which makes chrony and ntp work
>> again (without surprises) would do. I agree that the less intrusive
>> the change, the better.
>>
>> In the Debian 10 instances at GCE where I found this I just did this:
>>
>> systemctl disable systemd-timesyncd
>>
>> Would it make sense to ship systemd-timesyncd disabled by default in
>> buster and add a README to enable it only if the user really decides
>> to enable it? (Maybe also documenting this in the release notes).
>
>I think this would break more setups then it fixes.
>The default behaviour of systemd-timesyncd has been since two releases
>to be enabled by default. We can't easily change that.
>
>> That would be the most simple solution for stable that I can think,
>> as it would reduce the number of packages to change to just one.
>
>Unfortunately I think that disabling systemd-timesyncd by default is one
>of the most intrusive changes. After all, systemd is installed by
>default (and thus systemd-timesyncd enabled by default). I fear this is
>a no-go.

I wholeheartedly agree with you, Michael. To me having an {S}NTP 
implementation by default is a must. Disabling systemd-timesyncd at this 
point would probably seen as a serious regression.

Despite what Balint thinks, I think that we, maintainers of alternative 
NTP implementations, should just stop conflicting with 
systemd-timesyncd.service on stable and oldstable. Regarding Bulleye, 
that should not be an issue if Balint and Michael’s work is pushed in 
the archive.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://alioth-lists.debian.net/pipermail/pkg-systemd-maintainers/attachments/20200201/2048ebe4/attachment-0001.sig>