This is a Prometheus exporter that, when running on a node, checks the drift of that node's clock against a given NTP server or servers.
These are the metrics supported.
ntp_build_infontp_drift_secondsntp_stratumntp_rtt_secondsntp_reference_timestamp_secondsntp_root_delay_secondsntp_root_dispersion_secondsntp_root_distance_secondsntp_precision_secondsntp_leapntp_scrape_duration_secondsntp_server_reachable
As an alternative to the node-exporter's time module, this exporter does not require an NTP component on localhost that it can talk to. We only look at the system clock and talk to the configured NTP server(s).
Compile make && make install or docker build. The binary can also be
installed with go get:
go install github.com/sapcc/ntp_exporter@latestWe also publish pre-built images on Docker Hub as sapcc/ntp-exporter:
docker pull ghcr.io/sapcc/ntp_exporter:v2.2.0Common command-line options:
-ntp.source string
source of information about ntp server (cli / http). (default "cli")
-version
Print version information.
-web.listen-address string
Address on which to expose metrics and web interface. (default ":9559")
-web.telemetry-path string
Path under which to expose metrics. (default "/metrics")
By default, or when the option -ntp.source cli is specified, the NTP server
and connection options is defined by command-line options:
-ntp.measurement-duration duration
Duration of measurements in case of high (>10ms) drift. (default 30s)
-ntp.high-drift duration
High drift threshold. (default 10ms)
-ntp.protocol-version int
NTP protocol version to use. (default 4)
-ntp.server string
NTP server to use (required).
Command-line usage example:
ntp_exporter -ntp.server ntp.example.com -web.telemetry-path "/probe" -ntp.measurement-duration "5s" -ntp.high-drift "50ms"When the option -ntp.source http is specified, the NTP server and connection
options are obtained from the query parameters on each GET /metrics HTTP
request:
target: NTP server to useprotocol: NTP protocol version (2, 3 or 4)duration: duration of measurements in case of high drifthigh-drift: High drift threshold to trigger multiple probing
For example:
$ curl 'http://localhost:9559/metrics?target=ntp.example.com&protocol=4&duration=10s&high-drift=100ms'Several people have suggested adding a metric like ntp_up that's always 1, so
that people can alert on absent(ntp_up) or something like that. This is not
necessary. Prometheus already generates such a metric by itself during
scraping. A suitable
alert expression could look like
up{job="ntp_exporter",instance="example.com:9559"} != 1 or absent(up{job="ntp_exporter",instance="example.com:9559"})
but the concrete labels will vary depending on your setup and scrape configuration.
![dependabot[bot] avatar](https://avatars.githubusercontent.com/in/29110?v=4 "dependabot[bot]")
![renovate[bot] avatar](https://avatars.githubusercontent.com/in/2740?v=4 "renovate[bot]")
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent