Mercurial > prosody-modules

view mod_prometheus/README.markdown @ 5285:8e1f1eb00b58

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
mod_sasl2_fast: Fix harmless off-by-one error (invalidates existing tokens!) Problem: This was causing the key to become "<token>--cur" instead of the expected "<token>-cur". As the same key was used by the code to both set and get, it still worked. Rationale for change: Although it worked, it's unintended, inconsistent and messy. It increases the chances of future bugs due to the unexpected format. Side-effects of change: Existing '--cur' entries will not be checked after this change, and therefore existing FAST clients will fail to authenticate until they attempt password auth and obtain a new FAST token. Existing '--cur' entries in storage will not be cleaned up by this commit, but this is considered a minor issue, and okay for the relatively few FAST deployments.
author Matthew Wild <mwild1@gmail.com>
date Wed, 29 Mar 2023 16:12:15 +0100
parents bb66e87a3604
children c90011054335
line wrap: on
line source

---
labels:
- Statistics
summary: Implementation of the Prometheus protocol
...

Description
===========

This module implements the Prometheus reporting protocol, allowing you
to collect statistics directly from Prosody into Prometheus.

See the [Prometheus documentation][prometheusconf] on the format for
more information.

[prometheusconf]: https://prometheus.io/docs/instrumenting/exposition_formats/

**Note:** For use with Prosody trunk (0.12) we recommend the bundled
[mod_http_openmetrics](https://prosody.im/doc/modules/mod_http_openmetrics)
instead. This module (mod_prometheus) will continue to be available in the
community repository for use with older Prosody versions.

Configuration
=============

mod\_prometheus itself doesn’t have any configuration option, but it
requires Prosody’s [internal statistics
provider](https://prosody.im/doc/statistics#built-in_providers) to be
enabled.  You may also want to change the default collection interval
to the one your statistics consumer is using. See below for more information.

```lua
statistics = "internal"
statistics_interval = 15 -- in seconds
```

::: {.alert .alert-warning}
**NOTE:** Make sure to put the statistics variables in the global section of
the configuration, **not** in a `VirtualHost` or `Component` section.  You can
use `prosodyctl check` if you are unsure and want to check your configuration.
:::

See also the documentation of Prosody’s [HTTP
server](https://prosody.im/doc/http), since Prometheus is an HTTP
protocol that is how you can customise its URL.  The default one being
http://localhost:5280/metrics

Scrape interval vs statistics_interval
--------------------------------------

The `statistics_interval` should be set to `"manual"` on trunk if and only
if you have a single Prometheus instance scraping Prosody. This will allow
the internal statistics gathering to run optimally.

If you have multiple instances scraping Prosody, set `statistics_interval`
to the scrape interval of Prometheus to avoid errors in rate calculations
and similar.

Future work will allow the use of `"manual"` with multiple Prometheus
instances and varying scrape intervals (stay tuned).

Compatibility
=============

  ------- -------------
  trunk   Works (but replaced by [mod_http_openmetrics](https://prosody.im/doc/modules/mod_http_openmetrics))
  0.11    Works
  0.10    Works
  0.9     Does not work
  ------- -------------