Installing the exporter

The module doesn't need to installed on any database. To install, add the module to your module path and set it as a server wide module.

via CLI:

--load web,prometheus,base

or configuration file:

[options]
server_wide_modules = prometheus,web,base

That's it, the exporter is installed.

Configuring the exporter

The exporter has optional configurations. All of them are set on the prometheus section of the configuration file:

[prometheus]
auth = authkey
collectors = python,database,enterprise
db_force = true
metrics_db = /tmp/odoo_prometheus_metrics.db

auth sets the required authentication key. This is useful if Odoo is publicly accessible.

collectors allows you to only enable a subset of collectors. If not set, all of them are active.

db_force is required to be true if you have blocked the list of databases (list_db option)

metrics_db sets a custom path for a sqlite database used to store some metrics across threads/workers. Default is /tmp/odoo_prometheus_metrics.db

Built in collectors

Build custom collectors

The exporter was build in mind to support custom collectors.

Any module can implement a collector, to do that it should extend the class odoo.addons.prometheus.Collector and implement the method collect. This method must return either a single odoo.addons.prometheus.Metric or a list of them.

Example:

# -*- coding: utf-8 -*-
from odoo.addons.prometheus import Collector, Metric

class ExampleCollector(Collector):
    _name = "example"

    def collect(self):
        metric = Metric("odoo_example_metric", "Example metric N1", "gauge")
        metric2 = Metric("odoo_example_metric2", "Example metric N2", "counter")

        # Use self.databases to fetch all available databases
        # Use self.environment with a database name to have a valid environment. This should be used in a context
        for database in self.databases:
            with self.environment(database) as env:
                Partner = env["res.partner"]

                pcount = Partner.search_count([("customer", "=", True)])

                metric.add_label(pcount, database=database) # Named arguments are labels
                metric2.add_label(123) # Metric without labels



        return [metric, metric2]