github
/
node_exporter
mirror of https://github.com/prometheus/node_exporter
1
0
Fork 0
Code Issues Projects Releases Wiki Activity
Exporter for machine metrics
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
1036 Commits
34 Branches
53 Tags
95 MiB
Tree: f00f3db08b
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from 'f00f3db08b'
${ noResults }
node_exporter/docs/TIME.md

80 lines
3.5 KiB
Raw Normal View History Unescape

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
7 years ago
# Monitoring time sync with node_exporter
## `ntp` collector
This collector is intended for usage with local NTPD like [ntp.org](http://ntp.org/), [chrony](https://chrony.tuxfamily.org/comparison.html) or [OpenNTPD](http://www.openntpd.org/).
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](https://en.wikipedia.org/wiki/Network_Time_Protocol#Clock_strata) 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](https://en.wikipedia.org/wiki/Network_Time_Protocol#Clock_synchronization_algorithm) 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
Add _seconds suffix to node_time. (#823)
7 years ago
`node_time_seconds - node_ntp_reference_timestamp_seconds` represent some estimate of
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
7 years ago
"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.