Add a Kubernetes Distribution

[TylerHelmuth]

This PR adds a new distribution specifically for monitoring Kubernetes. For full details see #357.

Tested in https://github.com/TylerHelmuth/opentelemetry-collector-releases:

• action: https://github.com/TylerHelmuth/opentelemetry-collector-releases/actions/runs/8470931109
• release: https://github.com/TylerHelmuth/opentelemetry-collector-releases/releases/tag/v0.97.7
• packages:
  • https://github.com/TylerHelmuth/opentelemetry-collector-releases/pkgs/container/opentelemetry-collector-releases%2Fopentelemetry-collector-k8s
  • https://hub.docker.com/r/thelmuth34/opentelemetry-collector-k8s/tags

Example helm chart values.yaml using this distro:

mode: daemonset

clusterRole:
  create: true
  rules:
    - apiGroups: 
        - ""
      resources:
        - nodes/proxy
      verbs:
        - get

resources:
  limits:
    cpu: 200m
    memory: 512Mi
  requests:
    cpu: 100m
    memory: 256Mi

image:
  repository: thelmuth34/opentelemetry-collector-k8s
  tag: 0.97.7

command:
  name: otelcol-k8s

presets:
  logsCollection:
    enabled: true
  kubernetesAttributes:
    enabled: true
    extractAllPodLabels: true
    extractAllPodAnnotations: true
  kubeletMetrics:
    enabled: true

config:
  receivers:
    jaeger: null
    zipkin: null
    kubeletstats:
      collection_interval: 30s
      metric_groups:
        - node
        - pod
      metrics:
        k8s.node.uptime:
          enabled: true
        k8s.pod.uptime:
          enabled: true
        k8s.pod.cpu_limit_utilization:
          enabled: true
        k8s.pod.cpu_request_utilization:
          enabled: true
        k8s.pod.memory_limit_utilization:
          enabled: true
        k8s.pod.memory_request_utilization:
          enabled: true

  service:
    pipelines:
      traces:
        receivers: [otlp]
        exporters: [debug]
      metrics:
        exporters: [debug]
      logs:
        exporters: [debug]

ports:
  jaeger-compact:
    enabled: false
  jaeger-thrift:
    enabled: false
  jaeger-grpc:
    enabled: false
  zipkin:
    enabled: false

mode: deployment

image:
  repository: thelmuth34/opentelemetry-collector-k8s
  tag: 0.97.7

command:
  name: otelcol-k8s

resources:
  limits:
    cpu: 200m
    memory: 512Mi
  requests:
    cpu: 100m
    memory: 256Mi

# We only want one of these collectors - any more and we'd produce duplicate data
replicaCount: 1

presets:
  # enables the k8sclusterreceiver and adds it to the metrics pipelines
  clusterMetrics:
    enabled: true
  # enables the k8sobjectsreceiver to collect events only and adds it to the logs pipelines
  kubernetesEvents:
    enabled: true

config:
  receivers:
    k8s_cluster:
      collection_interval: 30s
    jaeger: null
    zipkin: null
  processors:
    transform/events:
      error_mode: ignore
      log_statements:
        - context: log
          statements:
            # adds a new watch-type attribute from the body if it exists
            - set(attributes["watch-type"], body["type"]) where IsMap(body) and body["type"] != nil

            # create new attributes from the body if the body is an object
            - merge_maps(attributes, body, "upsert") where IsMap(body) and body["object"] == nil
            - merge_maps(attributes, body["object"], "upsert") where IsMap(body) and body["object"] != nil

            # Transform the attributes so that the log events use the k8s.* semantic conventions
            - merge_maps(attributes, attributes[ "metadata"], "upsert") where IsMap(attributes[ "metadata"])
            - set(attributes["k8s.pod.name"], attributes["regarding"]["name"]) where attributes["regarding"]["kind"] == "Pod"
            - set(attributes["k8s.node.name"], attributes["regarding"]["name"]) where attributes["regarding"]["kind"] == "Node"
            - set(attributes["k8s.job.name"], attributes["regarding"]["name"]) where attributes["regarding"]["kind"] == "Job"
            - set(attributes["k8s.cronjob.name"], attributes["regarding"]["name"]) where attributes["regarding"]["kind"] == "CronJob"
            - set(attributes["k8s.namespace.name"], attributes["regarding"]["namespace"]) where attributes["regarding"]["kind"] == "Pod" or attributes["regarding"]["kind"] == "Job" or attributes["regarding"]["kind"] == "CronJob"

            # Transform the type attribtes into OpenTelemetry Severity types.
            - set(severity_text, attributes["type"]) where attributes["type"] == "Normal" or attributes["type"] == "Warning"
            - set(severity_number, SEVERITY_NUMBER_INFO) where attributes["type"] == "Normal"
            - set(severity_number, SEVERITY_NUMBER_WARN) where attributes["type"] == "Warning"


  service:
    pipelines:
      traces: null
      metrics:
        exporters: [ debug ]
      logs:
        processors: [ memory_limiter, transform/events, batch ]
        exporters: [ debug ]
        
ports:
  jaeger-compact:
    enabled: false
  jaeger-thrift:
    enabled: false
  jaeger-grpc:
    enabled: false
  zipkin:
    enabled: false

Closes #357

[TylerHelmuth]

Depends on open-telemetry/opentelemetry-helm-charts#1092

[dmitryax]

The set of components LGTM assuming we remove the deprecated logging exporter

[TylerHelmuth]

/cc @open-telemetry/operator-approvers

[jpkrohling]

It would be great to have the operator folks review this as well.

@avillela , you had a list of components during your talk yesterday and while I believe they are all here already, I'd appreciate it if you could double check.

[jpkrohling]

Thank you for making this work, @TylerHelmuth !

[swiatekm]

🚀

[avillela]

It would be great to have the operator folks review this as well.

@avillela , you had a list of components during your talk yesterday and while I believe they are all here already, I'd appreciate it if you could double check.

Are you referring to the Collector components for k8s @jpkrohling?

[mx-psi]

My general comment is "let's start smaller", by releasing for fewer OSs, architectures and types of artifacts

[TylerHelmuth]

@mx-psi @swiatekm-sumo you're comments on reducing the produced artifacts is valid, and annoying lol I will play around with what changes are necessary in cmd/goreleaser to reduce the artifact scope for this, and future, distros.

[pavolloffay]

Shouldn't here be as well the operator maintainers?

[swiatekm]

Doesn't @open-telemetry/operator-approvers include the maintainers?

[TylerHelmuth]

Correct. I went with -approvers instead of maintainers because thats how we do our own code owners file and its how the otel site handles adding other teams as code owners. In my opinion helm and operator Approvers are totally qualified to review PRs for this distro.

[jaronoff97]

@TylerHelmuth is there anything we need to do to resolve the error message github is giving us here?

[TylerHelmuth]

Ya, we need to give those groups access to this repo. I'll open a community issue for that.

[TylerHelmuth]

Created open-telemetry/community#2039

[TylerHelmuth]

@open-telemetry/helm-approvers please review

[dmitryax]

LGTM

[andrzej-stencel]

Shouldn't we add this distro to the main README.md file?

Current list of distributions:

• OpenTelemetry Collector (also known as "otelcol")
• OpenTelemetry Collector Contrib (also known as "otelcol-contrib")

[TylerHelmuth]

@astencel-sumo updated the readme and removed some statements that are no longer universally true.

[TylerHelmuth]

Ran another test against the latest commits in this branch (and 1 extra commit to update the references to my repo/dockerhub for the artifacts:

• https://github.com/TylerHelmuth/opentelemetry-collector-releases/actions/runs/8633831138
• https://github.com/TylerHelmuth/opentelemetry-collector-releases/actions/runs/8633831142
• https://github.com/TylerHelmuth/opentelemetry-collector-releases/actions/runs/8633831160

Artifacts: https://github.com/TylerHelmuth/opentelemetry-collector-releases/releases/tag/v0.97.9

Tested locally using the values.yamls from the description with the updated tag and can see the collector running as expected.

[jaronoff97]

🎉