This documents the various knowledge needed to add new GitLab self-monitoring Prometheus metrics.
2.6 KiB
Working with Prometheus Metrics
Adding to the library
We strive to support the 2-4 most important metrics for each common system service that supports Prometheus. If you are looking for support for a particular exporter which has not yet been added to the library, additions can be made to the common_metrics.yml
file.
Query identifier
The requirement for adding a new metric is to make each query to have an unique identifier which is used to update the metric later when changed:
- group: Response metrics (NGINX Ingress)
metrics:
- title: "Throughput"
y_label: "Requests / Sec"
queries:
- id: response_metrics_nginx_ingress_throughput_status_code
query_range: 'sum(rate(nginx_upstream_responses_total{upstream=~"%{kube_namespace}-%{ci_environment_slug}-.*"}[2m])) by (status_code)'
unit: req / sec
label: Status Code
Update existing metrics
After you add or change existing common metric you have to create a new database migration that will query and update all existing metrics.
NOTE: Note:
If a query metric (which is identified by id:
) is removed it will not be removed from database by default.
You might want to add additional database migration that makes a decision what to do with removed one.
For example: you might be interested in migrating all dependent data to a different metric.
class ImportCommonMetrics < ActiveRecord::Migration[4.2]
include Gitlab::Database::MigrationHelpers
DOWNTIME = false
def up
::Gitlab::DatabaseImporters::CommonMetrics::Importer.new.execute
end
def down
# no-op
end
end
GitLab Prometheus metrics
GitLab provides Prometheus metrics to monitor itself.
Adding a new metric
This section describes how to add new metrics for self-monitoring (example).
-
Select the type of metric:
Gitlab::Metrics.counter
Gitlab::Metrics.gauge
Gitlab::Metrics.histogram
Gitlab::Metrics.summary
-
Select the appropriate name for your metric. Refer to the guidelines for Prometheus metric names.
-
Update the list of GitLab Prometheus metrics.
-
Trigger the relevant page/code that will record the new metric.
-
Check that the new metric appears at
/-/metrics
.