node_exporter/docs/TIME.md
Leonid Evdokimov c169b4b1c5 Add metrics from SNTPv4 packet to ntp collector & add ntpd sanity check (#655)
* Add metrics from SNTPv4 packet to ntp collector & add ntpd sanity check

1. Checking local clock against remote NTP daemon is bad idea, local
ntpd acting as a  client should do it better and avoid excessive load on
remote NTP server so the collector is refactored to query local NTP
server.

2. Checking local clock against remote one does not check local ntpd
itself. Local ntpd may be down or out of sync due to network issues, but
clock will be OK.

3. Checking NTP server using sanity of it's response is tricky and
depends on ntpd implementation, that's why common `node_ntp_sanity`
variable is exported.

* `govendor add golang.org/x/net/ipv4`, it is dependency of github.com/beevik/ntp

* Update github.com/beevik/ntp to include boring SNTP fix

* Use variable name from RFC5905

* ntp: move code to make export of raw metrics more explicit

* Move NTP math to `github.com/beevik/ntp`

* Make `golint` happy

* Add some brief docs explaining `ntp` #655 and `timex` #664 modules

* ntp: drop XXX comment that got its decision

* ntp: add `_seconds` suffix to relevant metrics

* Better `node_ntp_leap` comment

* s/node_ntp_reftime/node_ntp_reference_timestamp_seconds/ as requested by @discordianfish

* Extract subsystem name to const as suggested by @SuperQ
2017-09-19 10:36:14 +02:00

3.5 KiB
Raw Blame History

Monitoring time sync with node_exporter

ntp collector

This collector is intended for usage with local NTPD like ntp.org, chrony or OpenNTPD.

Note, some chrony packages have local stratum 10 configuration value making chrony a valid server when it it is unsynchronised. This configuration makes one of node_ntp_sanity heuristics unreliable.

Note, OpenNTPD does not listen for SNTP queries by default, you should add listen on 127.0.0.1 configuration line to use this collector with OpenNTPD.

node_ntp_stratum

This metric shows stratum of local NTPD.

Stratum 16 means that clock are unsynchronised. See also aforementioned note about default local stratum in chrony.

node_ntp_leap

Raw leap flag value. 0 OK, 1 add leap second at UTC midnight, 2 delete leap second at UTC midnight, 3 unsynchronised.

OpenNTPD ignores leap seconds and never sets leap flag to 1 or 2.

node_ntp_rtt

RTT (round-trip time) from node_exporter collector to local NTPD. This value is used in sanity check as part of causality violation estimate.

node_ntp_offset

Clock offset between local time and NTPD time.

ntp.org always sets NTPD time to local clock instead of relaying remote NTP time, so this offset is irrelevant for this NTPD.

This value is used in sanity check as part of causality violation estimate.

node_ntp_reference_timestamp_seconds

Reference Time. This field show time when the last adjustment was made, but implementation details vary from "local wall-clock time" to "Reference Time field in incoming SNTP packet".

time() - node_ntp_reference_timestamp_seconds and node_time - node_ntp_reference_timestamp_seconds represent some estimate of "freshness" of synchronization.

node_ntp_root_delay and node_ntp_root_dispersion

These values are used to calculate synchronization distance that is limited by collector.ntp.max-distance.

ntp.org adds known local offset to announced root dispersion and linearly increases dispersion in case of NTP connectivity problems, OpenNTPD does not account dispersion at all and always reports 0.

node_ntp_sanity

Aggregate NTPD health including stratum, leap flag, sane freshness, root distance being less than collector.ntp.max-distance and causality violation being less than collector.ntp.local-offset-tolerance.

Causality violation is lower bound estimate of clock error done using SNTP, it's calculated as positive portion of abs(node_ntp_offset) - node_ntp_rtt / 2.

timex collector

This collector exports state of kernel time synchronization flag that should be maintained by time-keeping daemon and is eventually raised by Linux kernel if time-keeping daemon does not update it regularly.

Unfortunately some daemons do not handle this flag properly, e.g. chrony-1.30 from Debian/jessie clears STA_UNSYNC flag during daemon initialisation and does not indicate clock synchronization status using this flag. Modern chrony versions should work better. All chrony versions require rtcsync option to maintain this flag. OpenNTPD does not touch this flag at all till OpenNTPD-5.9p1.

On the other hand combination of sync_status and offset exported by timex module is the way to monitor if systemd-timesyncd does its job.