1.8 KiB
date | title | draft | type | menu | ||||
---|---|---|---|---|---|---|---|---|
2019-02-12:00:00+02:00 | Metrics | false | doc |
|
Metrics
Metrics work by exposing a /metrics
endpoint which can then be accessed by prometheus.
To keep the load on the database minimal, metrics are stored and updated in redis.
The metrics
package provides several functions to create and update metrics.
{{< table_of_contents >}}
Exposing New Metrics
First, define a const
with the metric key in redis. This is done in pkg/metrics/metrics.go
.
To expose a new metric, you need to register it in the init
function inside of the metrics
package like so:
{{< highlight golang >}} // Register total user count metric promauto.NewGaugeFunc(prometheus.GaugeOpts{ Name: "vikunja_team_count", // The key of the metric. Must be unique. Help: "The total number of teams on this instance", // A description about the metric itself. }, func() float64 { count, _ := GetCount(TeamCountKey) // TeamCountKey is the const we defined earlier. return float64(count) }) {{< /highlight >}}
Then you'll need to set the metrics initial value on every startup of vikunja.
This is done in pkg/routes/routes.go
to avoid cyclic imports.
If metrics are enabled, it checks if a redis connection is available and then sets the initial values.
A convenience function is available if the metric is based on a database struct.
Because metrics are stored in redis, you are responsible to increase or decrease these based on criteria you define.
To do this, use metrics.UpdateCount(value, key)
where value
is the amount you want to cange it (you can pass
negative values to decrease it) and key
it the redis key used to define the metric.
Using it
A Prometheus config with a Grafana template is available at our git repo.