Kubernetes CRD Extended Collection¶

Introduction¶

This document describes how to create a DataKit resouce in a Kubernetes cluster and configure an extension collector.

Add Authentication¶

If it is an upgraded version of DataKit, you need to add authentication in the apiVersion: rbac.authorization.k8s.io/v1 entry of datakit.yaml, that is, copy the following lines and add them to the end:

- apiGroups:
  - guance.com
  resources:
  - datakits
  verbs:
  - get
  - list

Creat v1beta1 DataKit Instance, Create DataKit Object¶

Write the following to the yaml configuration, such as datakit-crd.yaml, where each field has the following meaning:

k8sNamespace: Specify namespace, locates a collection's Pod with deployment, required
k8sDaemonSet: Specify the daemonset name to locate a collection's Pod with namespace
k8sDeployment: Specify the deployment name, and locates the Pod of a collection with namespace
inputConf: Collector configuration file, find the corresponding Pod according to namespace and deployment, replace the wildcard information of Pod, and then run the collector according to inputConf content. The following wildcard characters are supported.
$IP: Pod's intranet IP
$NAMESPACE: Pod Namespace
$PODNAME: Pod Name
$NODENAME: The name of the current node

Execute the kubectl apply -f datakit-crd.yaml command.

Attention

DaemonSet and Deployment are two different Kubernetes resources, but here k8s DaemonSet and k8s Deployment can exist at the same time. That is, under the same Namespace, the Pod created by DaemonSet and the Pod created by Deployment share the same CRD configuration. This is not recommended, however, because fields like source are used to identify data sources in specific configurations, and mixing them leads to unclear data boundaries. It is recommended that only one k8s DaemonSet and k8s Deployment exist in the same CRD configuration.
Datakit only collects Pod in the same node as it, which belongs to nearby collection and will not be collected across nodes.

Example¶

A complete example is as follows, including:

Create CRD Datakit
Namespace and Datakit instance objects used for testing
Configure the Prom collector (inputConf)

apiVersion: apiextensions.k8s.io/v1
kind: CustomResourceDefinition
metadata:
  name: datakits.guance.com
spec:
  group: guance.com
  versions:
  - name: v1beta1
    served: true
    storage: true
    schema:
      openAPIV3Schema:
        type: object
        properties:
          spec:
            type: object
            properties:
              instances:
                type: array
                items:
                  type: object
                  properties:
                    k8sNamespace:
                      type: string
                    k8sDaemonSet:
                      type: string
                    k8sDeployment:
                      type: string
                    datakit/logs:
                      type: string
                    inputConf:
                      type: string
  scope: Namespaced
  names:
    plural: datakits
    singular: datakit
    kind: Datakit
    shortNames:
    - dk
---
apiVersion: v1
kind: Namespace
metadata:
  name: datakit-crd
---
apiVersion: "guance.com/v1beta1"
kind: Datakit
metadata:
  name: my-test-crd-object
  namespace: datakit-crd
spec:
  instances:
    - k8sNamespace: "testing-namespace"
      k8sDaemonSet: "testing-daemonset"
      inputConf: |
        [inputs.prom]
          url="http://prom"

Ngxin Ingress Configuration Sample¶

Here, we use DataKit CRD extension to collect Ingress metrics, that is, we collect Ingress metrics through prom collector.

Prerequirements¶

Deployed DaemonSet DataKit
If the Deployment is called ingress-nginx-controller, the yaml configuration over there is as follows:

...
spec:
  selector:
    matchLabels:
      app.kubernetes.io/component: controller
  template:
    metadata:
      creationTimestamp: null
      labels:
        app: ingress-nginx-controller  # 这里只是一个示例名称
...

Configuration Step¶

Create Datakit CustomResourceDefinition

Execute the following create command:

cat <<EOF | kubectl apply -f -
apiVersion: apiextensions.k8s.io/v1
kind: CustomResourceDefinition
metadata:
  name: datakits.guance.com
spec:
  group: guance.com
  versions:
    - name: v1beta1
      served: true
      storage: true
      schema:
        openAPIV3Schema:
          type: object
          properties:
            spec:
              type: object
              properties:
                instances:
                  type: array
                  items:
                    type: object
                    properties:
                      k8sNamespace:
                        type: string
                      k8sDeployment:
                        type: string
                      datakit/logs:
                        type: string
                      inputConf:
                        type: string
  scope: Namespaced
  names:
    plural: datakits
    singular: datakit
    kind: Datakit
    shortNames:
    - dk
EOF

View deployment:

$ kubectl get crds | grep guance.com
datakits.guance.com   2022-08-18T10:44:09Z

Create a Datakit resource

Prometheus configuration can be found in link

Execute the following yaml:

apiVersion: guance.com/v1beta1
kind: DataKit
metadata:
  name: prom-ingress
  namespace: datakit
spec:
  instances:
    - k8sNamespace: ingress-nginx
      k8sDeployment: ingress-nginx-controller
      inputConf: |-
        [[inputs.prom]]
          url = "http://$IP:10254/metrics"
          source = "prom-ingress"
          metric_types = ["counter", "gauge", "histogram"]
          measurement_name = "prom_ingress"
          interval = "60s"
          tags_ignore = ["build","le","method","release","repository"]
          metric_name_filter = ["nginx_process_cpu_seconds_total","nginx_process_resident_memory_bytes","request_size_sum","response_size_sum","requests","success","config_last_reload_successful"]
        [[inputs.prom.measurements]]
          prefix = "nginx_ingress_controller_"
          name = "prom_ingress"
        [inputs.prom.tags]
          namespace = "$NAMESPACE"

!!! Note that namespace is customizable, while k8sDeployment and k8sNamespace must be accurate.

View deployment:

$ kubectl get dk -n datakit
NAME           AGE
prom-ingress   18m

View Metric Collection

Log in to Datakit pod and execute the following command:

$ datakit monitor

You can also log in to Guance Cloud Platform, "Indicator"-"Viewer" to view metric data