Xbusd exposes prometheus metrics to the wanted host:port by adding the following options to the configuration file:
- The tcp host on which the metrics are served
--metrics-http-hoston the cli)
- The tcp port on which the metrics are served
--metrics-http-porton the cli)
The global metrics are bus-related and should be collected only once per bus no matter how many xbusd instances share the load.
These metrics are served only if the “metrics” service is started (which is the default).
They can be served on a separated url (
/metrics/global) is needed, typically
when several xbusd servers share the load, by enabling the setting
--metrics-merge-global on the cli).
New in version 3.2.6.
Some metrics are costly to retrieve and xbusd anticipates the metrics access so it replies quickly.
To get accurate process counts in the metrics, it is recommended to let xbusd know the typical delay between two queries, by settings the following setting:
Approximate delay between two queries by prometheus (
--metrics-delay-hinton the cli). Default is: “15s”.
Possible units are: “h”, “m” or “s”.
It should never be smaller than the actual prometheus delay between checks. A larger value will only lower the refresh rate of the information, which can be interesting to lower the load on the database.