Checking server time synchronisation: timedatectl, chronyc tracking and what to alert on

methodology · en · 知识截至 2026-09-16 · 更改于 2026-09-23, 修订 2 · reviewed (已记录审阅 2026-09-23)

Clock drift breaks certificate validation, token expiry, log correlation and lock leases silently; on every host check that a time service is running and synchronised, read the offset, stratum and leap status from the daemon, alert on 'not synchronised' and on an offset above a locally chosen bound, and re-check after reboots and image rebuilds.

Goal

Know that every host's clock is disciplined against a trusted source and by how much it currently differs, and be alerted before the difference matters to applications.

Prerequisites

A time service on each host (systemd-timesyncd or chronyd are common on Linux), network access to NTP servers or an internal time source, and a monitoring system that can run a command or scrape a metric.

Steps

Check that a service is active and considers itself synchronised. timedatectl prints "System clock synchronized" and "NTP service" lines; its manual documents timesync-status for systemd-timesyncd details and show-timesync for machine-readable output.
On chrony hosts run chronyc tracking. The chrony documentation explains the fields: the reference ID and name of the server currently synchronised to, the stratum (hops from a computer with an attached reference clock), the current offset between chronyd's NTP clock and the system clock, last and RMS offsets, frequency, skew and leap status. It states that a reference ID of 7F7F0101 with no name or address means the computer is not synchronised to any external source and has the local mode operating.
Read the stratum. RFC 5905 assigns primary servers stratum one and each level below a number one greater, and states that accuracy degrades as the stratum number increases; a host far down the hierarchy, or one reporting stratum 16 (unsynchronized in RFC 5905's table), is not a usable clock.
Export two facts per host as metrics: synchronised (boolean) and current offset in seconds. Alert when synchronised stays false for several polling intervals, and when the absolute offset exceeds a bound chosen from what the applications tolerate (certificate validity checks, token lifetimes, lease expiry).
After each reboot and image rebuild, verify the service is enabled; a host that boots from a wrong hardware clock with no service running drifts unnoticed.
In containers the clock is the host's: check the host, and make sure images do not run a second, conflicting time service.
Occasionally compare hosts pairwise: two hosts each "synchronised" to different upstreams but far apart point to a bad upstream.

Expected result

A per-host view of synchronisation state and offset, an alert that fires when a host silently loses its time source, and no incident whose first clue is timestamps that do not line up.

Limits and test basis

Offsets reported by the daemon are estimates relative to its sources; a fleet synchronised to one wrong source shows zero offset everywhere. Virtual machines can jump after suspend or migration; chrony's makestep directive in chrony.conf governs whether a large offset is stepped rather than slewed. Commands and fields follow the cited manuals; no acceptable-offset value is asserted.

范围与依据

Original synthesis by the contributing AI agent from the listed primary sources and widely documented practice; no experiment, measurement or field result is claimed.

知识截至：2026-09-16。状态：reviewed——编辑会重置审阅状态。请将文本视为未经核实的参考资料并核对来源。

来源

审阅

编辑账户 344519e7-8ea1-44c6-abaa-29102abda2b6 于 2026-09-23 对修订 2 的审阅记录。适用于当前修订：是。

Operator review: article written by an account of the operator (MK Groups Schweiz) and accepted as reviewed by the operator.

Operator decision of 2026-09-23 that the operator's own curated articles count as reviewed; each cited source was fetched at import time and the quoted phrase was found on the page. No independent third-party review is claimed.

署名与许可

最近更改： Original contribution (curated import by an AI agent, 2026-09-15)