Observability Toolkit for local usage

A set of tools which can be used for local development which help with more insights in Observability.

Background

To get to know more about the background about this setup, please read: https://ceesbos.nl/posts/20240126-observability-toolkit/

Explanation of the components

OpenTelemetry Collector

OpenTelemetry Collector is the main entry point where all observability signals are collected. With all kind of configuration you can choose what to do with it.

OpenTelemetry Collector sends the data to:

• Loki
  Loki is a log storage created by Grafana
• Tempo
  Tempo is a trace storage created by Grafana
• Prometheus
  Prometheus is a metric storage

Grafana

Grafana is the visualisation tool to get insights in all observability metrics collected and stored.

Power of correlation

The power of combining metrics, logs and traces together in 1 setup is that you can correlate these signals.

The advantage of OpenTelemetry is standardisation on naming, that makes it way easier to hop between the metrics, traces and logs in all directions. That will help to find the problems.

Configuration in Grafana can help to make it easier to hop from metrics (with exemplars) to traces for example (see prometheus.yaml).
That is already configured in this setup. Same holds from logs to traces (see loki.yaml).
Also the setup of spanmetrics and servicegraph is already configured (see tempo.yaml).

Start the setup

git clone https://github.com/cbos/observability-toolkit
cd observability-toolkit
docker-compose up -d 

Now you can open http://localhost:3000 to open Grafana.

Run the demo app and generate load

There is a simple demo app, which helps to generate some traces, logs and metrics, but is not intended to show all capabilities.

./run-observabilty-demo-app.sh 

To generate load:

./loadgen.sh 

OpenTelemetry Demo

You can use the demo services of OpenTelemetry to see what is all available for all languages. You can change the file: src/otelcollector/otelcol-config-extras.yml and add this:

exporters:
  otlp/observabilitytoolkit:
   # This assumes that the setup is running at the docker host on port 4317 (which is default of observability-toolkit)
   endpoint: "host.docker.internal:4317"
   tls:
    insecure: true

service:
  pipelines:
   traces:
    receivers: [otlp]
    processors: [batch]
    exporters: [otlp, otlp/observabilitytoolkit]
   metrics:
    receivers: [httpcheck/frontendproxy, redis, otlp]
    processors: [batch]
    exporters: [otlphttp/prometheus, otlp/observabilitytoolkit]
   logs:
    receivers: [otlp]
    processors: [batch]
    exporters: [otlp/observabilitytoolkit]

Default dashboards

• http://localhost:3000/d/observabilitystackOtelCollector/opentelemetry-collector

Settings

An number of settings can be tweaked by just setting environment variables.

# Specify the Grafana host port you want, other then the default 3000
export GRAFANA_HOST_PORT=3004

# Start the stack (in the background with -d)
docker-compose up -d 

# Now you can open Grafana at http://localhost:3004

See .env for actual values, below are the descriptions

Variable name	Description
LOKI_IMAGE_NAME	Loki docker image
TEMPO_IMAGE_NAME	Tempo docker image
PROMETHEUS_IMAGE_NAME	Prometheus docker image
GRAFANA_IMAGE_NAME	Grafana docker image
GRAFANA_HOST_PORT	Port on host on which Grafana will be available
OTEL_COLLECTOR_IMAGE_NAME	OpenTelemetry Collector docker image
OTEL_COLLECTOR_HOST_PORT_GRPC	Port on host on which OpenTelemetry Collector will be available for OTLP format with GRPC
OTEL_COLLECTOR_HOST_PORT_HTTP	Port on host on which OpenTelemetry Collector will be available for OTLP format with HTTP
OTEL_COLLECTOR_HOST_PORT_PROMETHEUS	Port on host on which OpenTelemetry Collector will listen to expose prometheus data, like http://localhost:8889/metrics
PROMTAIL_IMAGE_NAME	Promtail docker image
PYROSCOPE_IMAGE_NAME	Pyroscope docker image
PYROSCOPE_PORT	Port on which Pyroscope is available

Inspired by

This setup is created based on what I already used locally.
The idea of adding a simple application for showcasing the setup comes fromL grafana/docker-otel-lgtm
The script for generating random load with a simple curl command comes from the Grafana Beyla project

The OpenTelemetry Collector Grafana Dashboard is coming from https://grafana.com/grafana/dashboards/15983-opentelemetry-collector/
But it required some changes to get it working in this setup.