Using custom metrics
For this lab Grafana and Alertmanager need to be enabled for the team.
To be able to collect custom metrics you will first need to configure your app to expose metrics. This is called instrumentation of code and can include annotating the code with metadata, or adding in logic to calculate and expose data.
Instrumenting code means you write code to expose information about the technical, business, and customer context. This information can then be collected and analyzed using Prometheus and Grafana.
In this lab we'll use a container that exposes custom metrics and then show how the metrics can be collected and analyzed.
Create a Workload
-
Go to the Catalog and click on the
k8s-deployment
template. -
Click on the Values tab.
-
Fill in the name
custom-metrics
. -
Use the following values:
image: repository: quay.io/rbaumgar/monitor-demo-app-jvm pullPolicy: IfNotPresent tag: latest containerPorts: - name: web containerPort: 8080 protocol: TCP servicePorts: - port: 8080 targetPort: 8080 protocol: TCP name: web replicaCount: 1 serviceMonitor: create: true endpoints: - interval: 30s port: web scheme: http path: /q/metrics
-
Click the Submit button.
Our metrics will now be scraped by the Platform Prometheus. Before we continue, let's first generate some load:
-
Expose the
custom-metrics
service (see lab Expose services). -
Run the following command in your terminal:
for i in {1..1000}; do curl https://custom-metrics-labs.<your-domain>/hello; sleep 10; done
-
Wait for approximately 10 minutes.
Create a dashboard in Grafana to use the metrics
-
In the left menu, click on Apps and open Grafana.
-
Go to the Grafana dashboard homepage.
-
In the top right, click New and select New Dashboard.
-
Click Add visualization.
-
In the Data source tab, select Prometheus Platform.
-
In the A collapsible section, select a metric from the Metric drop-down list. In our example, we use the
application_greetings_total
metric. -
Click Run queries.
-
You should now see a Time series graph like this:
-
Save the dashboard in Grafana.
Make the dashboard persistent
Now you know how to create a dashboard in Grafana for your custom metrics. You could now save the dashboard, but if Grafana would get re-started, the dashboard will be gone. To make the dashboard persistent we need to add it to a configmap.
-
Go to Apps and open Gitea.
-
In the list of repositories there is a repository called
otomi/team-<team-name>-argocd
. Go to this repository. -
Click Add File and then New File.
-
Name the file
my-custom-dashboard.yaml
. -
Add the following manifest to the file:
apiVersion: v1 kind: ConfigMap metadata: name: my-custom-dashboard labels: grafana_dashboard: "1" release: grafana-dashboards-labs # change labs to the name of your team data: my-dashboard.json: |- # paste your dashboard json here
-
Before committing changes, go back to Grafana.
-
Click on Dashboard settings (in the top right).
-
Go to JSON model tab.
-
Copy the JSON model and paste it into the ConfigMap. Make sure to indent with 4.
-
Delete the dashboard created in Grafana.
-
Commit changes in Gitea.
The dashboard will now automatically be loaded into the team's Grafana instance.
Create custom rules
Now we are exporting metrics, these metrics can also be used to generate alerts. To generate alerts, we first need to create a Prometheus Rule:
-
Go to Apps and open Gitea.
-
In the list of repositories there is a repository called
otomi/team-<team-name>-argocd
. Go to this repository. -
Click Add File and then New File.
-
Name the file
my-custom-rules.yaml
. -
Add the following manifest to the file:
apiVersion: monitoring.coreos.com/v1 kind: PrometheusRule metadata: labels: prometheus: system name: labs-custom-rules spec: groups: - name: custom.rules rules: - alert: 50GreetingsReached annotations: description: We reached 50 greetings! summary: The number of greetings has reached more than 50. expr: application_greetings_total > 50 for: 1m labels: severity: warning
-
Commit changes in Gitea.
See alerts based on the rule in Alertmanager
Go to Apps and open Alertmanager. You will see Alertmanager has received the alerts from Prometheus:
If a receiver has been configured for the team, like Slack, then you would also have received a message with the alert.
Updated about 7 hours ago