Why is ntpd crashing?

Question

CentOS

Good afternoon, please help me figure out why ntpd is crashing.
The situation is, there is a virtual machine on Centos, an ESXi virtualization host. On this machine, for some reason, time runs away by +3 hours every morning. At it there is a configured ntpd. But for some reason it falls in the evening. After rebooting ntpd
, the time becomes correct.
Here is a snippet of the log.

Mar 29 10:40:03 r ntpd[21873]: Listen normally on 3 ens192 10.0.0.11 UDP 123
Mar 29 10:40:03 r ntpd[21873]: Listening on routing socket on fd #20 for interface updates
Mar 29 10:40:03 r ntpd[21873]: 0.0.0.0 c016 06 restart
Mar 29 10:40:03 r ntpd[21873]: 0.0.0.0 c012 02 freq_set kernel 1.443 PPM
Mar 29 10:40:11 r ntpd[21873]: 0.0.0.0 c615 05 clock_sync
Mar 30 19:39:35 r ntpd[21873]: 0.0.0.0 0618 08 no_sys_peer
Mar 30 21:29:49 r ntpd[21873]: 0.0.0.0 0617 07 panic_stop -10734 s; set clock manually within 1000 s.
Mar 30 21:29:49 r systemd[1]: ntpd.service: main process exited, code=exited, status=255/n/a
Mar 30 21:29:49 r systemd[1]: Unit ntpd.service entered failed state.
Mar 30 21:29:49 r systemd[1]: ntpd.service failed.

Answer 1

- is the time normal on the host machine?
- there are also virtual machines for ESXi, how are they over time?
- what does "ntpq -p" show?

Answer 2

NTP refuses to work if the difference is more than 15 minutes. At boot time, just before ntpd starts, ntpdate is started, which sets the correct time one time. So a reboot solves the problem.
The main problem is that linux hardware clock is actually emulated by vmware programmatically. When ESX is loaded, it can be very late and is always inaccurate.
Another problem appears when migrating between ESXs with different processor frequencies. Linux suddenly starts to hurry too much - visually it is very clearly visible if you run top.
And yet, maybe you have not disabled power saving in the ESX BIOS. The dynamic frequency of the processor can lead to the same effect.