Traffic Metrics

A user can enable traffic metrics by editing a Mesh resource and providing the desired Mesh-wide configuration. If necessary, metrics configuration can be customized for each Dataplane individually, e.g. to override the default metrics port that might be already in use on that particular machine.

Out-of-the-box, Kuma provides full integration with Prometheus:

• if enabled, every dataplane will expose its metrics in Prometheus format
• furthemore, Kuma will make sure that Prometheus can automatically find every dataplane in the mesh

To collect metrics from Kuma, you need to first expose metrics from Dataplanes and then configure Prometheus to collect them.

To expose Prometheus metrics from every dataplane in the mesh, configure a Mesh resource as follows:

which is a convenient shortcut for

apiVersion: kuma.io/v1alpha1
kind: Mesh
metadata:
  name: default
spec:
  metrics:
    enabledBackend: prometheus-1
    backends:
    - name: prometheus-1
      type: prometheus
      conf:
        skipMTLS: false
        port: 5670
        path: /metrics
        tags: # tags that can be referred in Traffic Permission when metrics are secured by mTLS  
          service: dataplane-metrics

type: Mesh
name: default
metrics:
  enabledBackend: prometheus-1
  backends:
  - name: prometheus-1
    type: prometheus
    conf:
      skipMTLS: true # by default mTLS metrics are also protected by mTLS. Scraping metrics with mTLS without transparent proxy is not supported at the moment.

which is a convenient shortcut for

type: Mesh
name: default
metrics:
  enabledBackend: prometheus-1
  backends:
  - name: prometheus-1
    type: prometheus
    conf:
      skipMTLS: true
      port: 5670
      path: /metrics
        service: dataplane-metrics

Both snippets from above instruct Kuma to configure every dataplane in the mesh default to expose an HTTP endpoint with Prometheus metrics on port 5670 and URI path /metrics.

Override Prometheus settings per Dataplane

To override Mesh-wide defaults for a particular Pod, use Kuma-specific annotations:

• prometheus.metrics.kuma.io/port - to override Mesh-wide default port
• prometheus.metrics.kuma.io/path - to override Mesh-wide default path

E.g.,

As a result, dataplane for this particular Pod will expose an HTTP endpoint with metrics on port 1234 and URI path /non-standard-path.

To override Mesh-wide defaults on a particular machine, configure Dataplane resource as follows:

type: Dataplane
mesh: default
name: example
metrics:
  enabledBackend: prometheus-1
  backends:
  - name: prometheus-1
    type: prometheus
    conf:
      skipMTLS: true
      port: 1234
      path: /non-standard-path

Although dataplane metrics are now exposed, Prometheus doesn’t know anything about it just yet.

To help Prometheus to automatically discover dataplanes, Kuma provides a tool - kuma-prometheus-sd. kuma-prometheus-sd is meant to run alongside Prometheus instance. It knows location of Kuma Control Plane is and can fetch an up-to-date list of dataplanes from it. It then transforms that information into a format that Prometheus can understand, and saves it into a file on disk. Prometheus watches for changes to that file and updates its scraping configuration accordingly.

Use kumactl install metrics | kubectl apply -f - to deploy configured Prometheus with Grafana.

If you’ve got Prometheus deployment already, you can use to bring Kuma metrics to your main Prometheus cluster.

First, you need to run kuma-prometheus-sd, e.g. by using the following command:

kuma-prometheus-sd run \
  --cp-address=http://kuma-control-plane.internal:5681 \
  --output-file=/var/run/kuma-prometheus-sd/kuma.file_sd.json

The above configuration tells kuma-prometheus-sd to talk to Kuma Control Plane at http://kuma-control-plane.internal:5681 and save the list of dataplanes to /var/run/kuma-prometheus-sd/kuma.file_sd.json.

Then, you need to set up Prometheus to read from that file, e.g. by using prometheus.yml config with the following contents:

scrape_configs:
- job_name: 'kuma-dataplanes'
  scrape_interval: 15s
  file_sd_configs:
  - files:
    - /var/run/kuma-prometheus-sd/kuma.file_sd.json

and running

Now, if you check Targets page on Prometheus UI, you should see a list of dataplanes from your mesh, e.g.

Kuma lets you expose Dataplane metrics in a secure way by leveraging mTLS. Prometheus needs to be a part of the Mesh for this feature to work, which is the default deployment model when kumactl install metrics is used on Kubernetes.

apiVersion: kuma.io/v1alpha1
kind: Mesh
metadata:
  name: default
spec:
  mtls:
    enabledBackend: ca-1
    backends:
    - name: ca-1
      type: builtin
  metrics:
    enabledBackend: prometheus-1
    backends:
    - name: prometheus-1
      conf:
        skipMTLS: false
        port: 5670
        path: /metrics
        tags: # tags that can be referred in Traffic Permission  
          service: dataplane-metrics

Allow the traffic from Grafana to Prometheus Server and from Prometheus Server to Dataplane metrics and for other Prometheus components:

apiVersion: kuma.io/v1alpha1
kind: TrafficPermission
metadata:
  namespace: default
  name: metrics-permissions
spec:
  sources:
    - match:
        service: prometheus-server.kuma-metrics.svc:80
  destinations:
    - match:
        service: dataplane-metrics
    - match:
       service: "prometheus-alertmanager.kuma-metrics.svc:80"
    - match:
       service: "prometheus-kube-state-metrics.kuma-metrics.svc:80"
    - match:
       service: "prometheus-kube-state-metrics.kuma-metrics.svc:81"
    - match:
       service: "prometheus-pushgateway.kuma-metrics.svc:9091"
---
apiVersion: kuma.io/v1alpha1
kind: TrafficPermission
mesh: default
metadata:
  namespace: default
  name: grafana-to-prometheus
spec:
   sources:
   - match:
      service: "grafana.kuma-metrics.svc:80"
   destinations:
   - match:
      service: "prometheus-server.kuma-metrics.svc:80"

This feature requires transparent proxy, therefore for now it’s not available in Universal for now.

In addition to exposing metrics from Dataplane, you may want to expose metrics from application next to Kuma DP.

Use standard prometheus.io annotations either on Pod or Service

apiVersion: apps/v1
kind: Deployment
metadata:
  namespace: kuma-example
  name: kuma-tcp-echo
spec:
  ...
  template:
    metadata:
      ...
      annotations:
        prometheus.io/scrape: "true"
        prometheus.io/port: "1234"
        prometheus.io/path: "/non-standard-path"
    spec:
      ...

Use Discovery Service of . In the future Kuma will help to expose metrics in more native way.

Remember that in order to consume paths protected by mTLS, you need Traffic Permission that lets Prometheus consume applications.

Kuma ships with 3 default dashboards that are available to import from Grafana Labs repository.

This dashboards lets you investigate the status of a single dataplane in the mesh.

This dashboard lets you investigate the aggregated statistics of a single mesh.