Skip to main content

Installation

Install via Helm​

You can install production-ready KubeSkoop instance by Helm.

Execute the following command to add the KubeSkoop repository to Helm and install KubeSkoop:

# Add kubeskoop repo
helm repo add kubeskoop https://kubeskoop.io/
# Upgrade helm repo
helm repo update
# Install kubeskoop
helm install -n kubeskoop kubeskoop kubeskoop/kubeskoop

The command will install KubeSkoop into the kubeskoop namespace and turn on the agent, controller and webconsole components of KubeSkoop by default.

Customize Configuration​

You can view the values that can be configured by executing the following command.

helm show values kubeskoop/kubeskoop

You can save the values to the values.yaml file, make changes to the parameters you need, and then install KubeSkoop with theses values.

# If you have already installed KubeSkoop, you can upgrade it with `helm upgrade`.
helm install -f values.yaml -n kubeskoop kubeskoop kubeskoop/kubeskoop

You can also use the --set parameter to customize the configuration.

helm install --set controller.config.prometheusEndpoint=http://prometheus -n kubeskoop kubeskoop kubeskoop/kubeskoop

Configure Probes​

In the config section of the values, you can configure the metrics/event probes that need to be turned on.

The default configuration is as follows:

config:
  metricProbes:
    - name: conntrack
    - name: qdisc
    - name: netdev
    - name: io
    - name: sock
    - name: tcpsummary
    - name: tcp
    - name: tcpext
    - name: udp
    - name: kernellatency
    - name: packetloss
    - name: flow
      args:
        enablePortInLabel: false
    - name: tcpretrans
  eventProbes:
    - name: biolatency
    - name: kernellatency
    - name: packetloss
      args:
        enableStack: false
    - name: tcpreset
    - name: tcpretrans
  eventSinks:
    - name: stderr

For more information about the probes, please see the documentation.

Configure Prometheus and Loki Endpoint​

Some features (e.g., network graph, anomaly events) of Web Console relies on Prometheus and Loki. You need to provide the endpoint for the Prometheus and Loki services during installation.

tip

If there is no ready-to-use Prometheus or Loki instances, you can refer to theses documentations:

Configure prometheusEndpoint and lokiEndpoint in the controller.config section.

Assuming that prometheus and loki are installed in the monitoring namespace, and the service names are prometheus and loki, a sample configuration is as follows:

controller:
 enabled: true
 config:
   # configure endpoint address for prometheus.
   prometheusEndpoint: 'http://prometheus.monitoring:9090'
   # configure endpoint address for loki.
   lokiEndpoint: 'http://loki.monitoring:3100'

At the same time, the event sinks needs to be configured in config.eventSink so that events can be collected via Loki.

config:
  # ...some probe configurations
  eventSinks:
    - name: stderr
    # configure loki event sink and specify the address as the service address of loki.
    - name: loki
      args:
        addr: 'http://loki.monitoring:3100'

Finally, install/update KubeSkoop via helm install or helm upgrade.