Skip to content

Latest commit

 

History

History

docker

README.md

Overview

This directory contains a Dockerfile and an example-instrumentation.yaml file for creating a Docker init container that can be used to auto-instrumetation your Python app via the kubernetes OTel Operator.

Installation

Install cert manager into your k8s cluster unless already installed.

Install the OTel Operator Custom Resource Definitions:

kubectl apply -f https://github.com/open-telemetry/opentelemetry-operator/releases/latest/download/opentelemetry-operator.yaml

Install an OpenTelemetryCollector definition into the k8s cluster. The following defines a Collector sidecar that just prints debug statements.

apiVersion: opentelemetry.io/v1beta1
kind: OpenTelemetryCollector
metadata:
  name: my-sidecar
spec:
  mode: sidecar
  config:
    receivers:
      otlp:
        protocols:
          grpc:
            endpoint: 0.0.0.0:4317
          http:
            endpoint: 0.0.0.0:4318
    exporters:
      debug: {}
    service:
      pipelines:
        traces:
          receivers: [otlp]
          exporters: [debug]

Install an Instrumentation definition. It will be activated when you create an app with an inject-python annotation, running an init container before your Python app. The init container makes OTel libraries (including a sitecustomize.py) available at /otel-auto-instrumentation-python and sets the PYTHONPATH for the starting application.

apiVersion: opentelemetry.io/v1alpha1
kind: Instrumentation
metadata:
  name: splunk-otel-python
spec:
  exporter:
    endpoint: http://localhost:4318
  sampler:
    type: always_on
  python:
    env:
      - name: OTEL_EXPORTER_OTLP_PROTOCOL
        value: http/protobuf
    image: "splunk-otel-python-init:v2.0.0"

Create an application image and run it via something like this Deployment. Note the annotations, indicating that we want both Python auto instrumentation installed and our sidecar Collector to run.

apiVersion: apps/v1
kind: Deployment
metadata:
  name: ticker
spec:
  selector:
    matchLabels:
      app: ticker
  replicas: 1
  template:
    metadata:
      labels:
        app: ticker
      annotations:
        sidecar.opentelemetry.io/inject: "true"
        instrumentation.opentelemetry.io/inject-python: "true"
    spec:
      containers:
      - name: ticker
        image: ticker:v1
        env:
        - name: PYTHONUNBUFFERED
          value: "1"

When you run this application, k8s should create a pod with a Collector sidecar and auto instrumentation enabled. You can check the Collector logs to make sure it's receiving data.