Time synchronization disappears from view when it works. When it fails, unrelated systems begin to look broken: TLS certificates appear invalid, login tokens expire immediately, logs cannot be correlated, cluster members disagree, and scheduled jobs run at surprising moments.
Give every machine a reliable path to time. That may mean the router or two internal NTP servers synchronizing from external sources, with clients using those internal addresses. More than one source prevents a maintenance reboot from becoming a network-wide clock problem.
Keep time zones and clocks conceptually separate. Servers can use UTC for clear logs while applications display local time for people. NTP aligns the clock; it does not choose how timestamps are presented.
Monitoring should detect both service failure and excessive offset. A process can be running while the clock remains unsynchronized because its upstream is unreachable. Check new templates, isolated VLANs, and firewall rules for NTP access instead of assuming that DHCP alone solved it.
During troubleshooting, compare several hosts rather than correcting one clock by hand. A manual adjustment hides the shared cause and may introduce a sudden time jump that confuses applications. Fix reachability or the selected sources, then let the time service converge according to its normal policy.
Finally, know what happens after a long power outage. A machine without a useful hardware clock may boot far in the past, before DNS, certificates, or package repositories work correctly. Restoring time early in the boot dependency chain turns many mysterious failures back into ordinary ones.