[PVE-User] NTP on PVE8...

[Alwin Antreich]

Hi Marco,



August 19, 2024 at 11:00 AM, "Marco Gaiarin" <gaio at lilliput.linux.it> wrote:



> 
> Running 'pve7to8' i've catch:
> 
>  INFO: Checking for supported & active NTP service..
>  WARN: systemd-timesyncd is not the best choice for time-keeping on servers, due to only applying updates on boot.
>  While not necessary for the upgrade it's recommended to use one of:
>  * chrony (Default in new Proxmox VE installations)
>  * ntpsec
>  * openntpd
> 
> This is indeed mostly true, and in a cluster using a NTP daemon is a must.
> No doubt on that.
> 
> But for a standalone server, systemd-timesyncd seems sufficient to me; and
> surely NOT 'only applying updates on boot': systemd-timesyncd sync time at
> regilar interval!
> 
>  PollIntervalMinSec=, PollIntervalMaxSec=
>  The minimum and maximum poll intervals for NTP messages. Polling starts at the minimum poll interval, and is adjusted within the specified limits in response to received
>  packets.
> 
>  Each setting takes a time span value. The default unit is seconds, but other units may be specified, see systemd.time(5). PollIntervalMinSec= defaults to 32 seconds and
>  must not be smaller than 16 seconds. PollIntervalMaxSec= defaults to 34 min 8 s (2048 seconds) and must be larger than PollIntervalMinSec=.
> 
systemd-timesyncd only uses one server to update its time, which can lead to time jumps when this NTP can't be accessed reliably. NTPs (eg. ntpd/chrony) are usually using 3+ servers to calculate a mean time and compensate for jitter and other delays. This allows to keep the time more stable and especially helps when Ceph or HA is enabled. ;)

I do recommend chrony for servers (physical or virtual).

Cheers,
Alwin