OpenTelemetry Collector

OpenTelemetry Collector is a high-performance, scalable, and reliable data collection pipeline for observability data. It receives telemetry data from various sources, performs processing and translation to a common format, and then exports the data to various backends for storage and analysis.

Otel Collector supports multiple data formats, protocols, and platforms, making it a flexible and scalable solution for observability needs.

• How OpenTelemetry Collector works?
• When to use OpenTelemetry Collector?
• otelcol vs otelcol-contrib
• Installation
  • Linux
  • Compiling from sources
• Configuration
  • Troubleshooting
  • Extensions
  • Prometheus integration
• Host Metrics
  • Exporting data to Uptrace
• Resource Detection
• Memory Limiter
• Uptrace

How OpenTelemetry Collector works?

OpenTelemetry Collector is a vendor-agnostic proxy/middleman between your application and distributed tracing tools such as Uptrace or Jaeger.

OpenTelemetry Collector works by receiving telemetry data from various sources, processing and normalizing the data, and then exporting it to various backends for storage and analysis.

Otel Collector provides powerful data processing capabilities, allowing you to perform aggregation, filtering, sampling, and enrichment of telemetry data. You can transform and reshape the data to fit your specific monitoring and analysis requirements before sending it to the backend systems.

Otel Collector is written in Go and licensed under Apache 2.0 license which allows you to change the source code and install custom extensions. That comes at a cost of running and maintaining your own OpenTelemetry Collector instances.

When to use OpenTelemetry Collector?

Most of the time, sending telemetry data directly to a backend is the great way to get started with OpenTelemetry. But you may want to deploy a collector alongside your services to get batching, retries, sensitive data filtering, and more.

The most prominent OpenTelemetry Collector feature is the ability to operate on whole traces instead of individual spans. To achieve that, OpenTelemetry Collector buffers the received spans and groups them by a trace id. That is the key requirement to implement tail-based sampling.

OpenTelemetry Collector can also act as an agent that pulls telemetry data from monitored systems, for example, OpenTelemetry Redis or host metrics.

otelcol vs otelcol-contrib

OpenTelemetry Collector has 2 repositories on GitHub:

• opentelemetry-collector is the core that contains only the most crucial components. It is distributed as otelcol binary.

• opentelemetry-collector-contrib contains the core and all additional available components, for example, Redis and PostgreSQL receivers. It is distributed as otelcol-contrib binary.

You should always install and use the otelcol-contrib, because it is as stable as the core and supports more features.

Installation

OpenTelemetry Collector distributes pre-compiled binaries for Linux, MacOS, and Windows.

Linux

To install otelcol-contrib binary with the associated systemd service, run the following command replacing 0.70.0 with the desired version and amd64 with the desired architecture:

wget https://github.com/open-telemetry/opentelemetry-collector-releases/releases/download/v0.70.0/otelcol-contrib_0.70.0_linux_amd64.deb
sudo dpkg -i otelcol-contrib_0.70.0_linux_amd64.deb

wget https://github.com/open-telemetry/opentelemetry-collector-releases/releases/download/v0.70.0/otelcol-contrib_0.70.0_linux_amd64.rpm
sudo rpm -ivh otelcol-contrib_0.70.0_linux_amd64.rpm

You can check the status of the installed service with:

sudo systemctl status otelcol-contrib

And check the logs with:

sudo journalctl -u otelcol-contrib -f

You can edit the config at /etc/otelcol-contrib/config.yaml and restart OpenTelemetry Collector:

sudo systemctl restart otelcol-contrib

Compiling from sources

You can also compile OpenTelemetry Collector locally:

git clone https://github.com/open-telemetry/opentelemetry-collector-contrib.git
cd opentelemetry-collector-contrib
make install-tools
make otelcontribcol
./bin/otelcontribcol_linux_amd64 --config ./examples/local/otel-config.yaml

Configuration

OpenTelemetry Collector is highly configurable, allowing you to customize its behavior and integrate it into your observability stack. It provides configuration options for specifying inputs, processors, and exporters, enabling you to tailor the agent to your specific needs.

By default, you can find the config file at /etc/otelcol-contrib/config.yaml, for example:

TIP

Don't forget to add the Uptrace exporter to service.pipelines section. Unused receivers and exporters are silently ignored.

# receivers configure how data gets into the Collector.
receivers:
  otlp:
    protocols:
      grpc:
      http:

# processors specify what happens with the received data.
processors:
  resourcedetection:
    detectors: [env, system]
  cumulativetodelta:
  batch:
    send_batch_size: 10000
    timeout: 10s

# exporters configure how to send processed data to one or more backends.
exporters:
  otlp/uptrace:
    endpoint: otlp.uptrace.dev:4317
    headers:
      uptrace-dsn: 'https://<token>@uptrace.dev/<project_id>'

# service.pipelines pull the configured receivers, processors, and exporters together into
# pipelines that process data.
#
# receivers, processors, and exporters that are not used in pipelines are silently ignored.
service:
  pipelines:
    traces:
      receivers: [otlp]
      processors: [batch]
      exporters: [otlp/uptrace]
    metrics:
      receivers: [otlp]
      processors: [cumulativetodelta, batch, resourcedetection]
      exporters: [otlp/uptrace]
    logs:
      receivers: [otlp]
      processors: [batch]
      exporters: [otlp/uptrace]

You can always learn more about Otel Collector using the official documentation.

Troubleshooting

If otelcol is not working as expected, you can check the log output for potential issues. The logging verbosity level defaults to INFO, but you can change it using the configuration file:

service:
  telemetry:
    logs:
      level: 'debug'

To view the logs for potential issues:

sudo journalctl -u otelcol-contrib -f

You can also enable metrics to monitor OpenTelemetry Collector:

receivers:
  prometheus/otelcol:
    config:
      scrape_configs:
        - job_name: 'otelcol'
          scrape_interval: 10s
          static_configs:
            - targets: ['0.0.0.0:8888']

service:
  telemetry:
    metrics:
      address: ':8888'
  pipelines:
    metrics/hostmetrics:
      receivers: [prometheus/otelcol]
      processors: [cumulativetodelta, batch, resourcedetection]
      exporters: [otlp/uptrace]

Extensions

Extensions provide additional capabilities for OpenTelemetry Collector and do not require direct access to telemetry data, for example, Health Check extension responds to health check requests.

extensions:
  # Health Check extension responds to health check requests
  health_check:
  # PProf extension allows fetching Collector's performance profile
  pprof:
  # zPages extension enables in-process diagnostics
  zpages:
  # Memory Ballast extension configures memory ballast for the process
  memory_ballast:
    size_mib: 512

Prometheus integration

See OpenTelemetry Collector Prometheus.

Host Metrics

See OpenTelemetry host metrics.

Exporting data to Uptrace

See Sending data from Otel Collector to Uptrace.

Resource Detection

To detect resource information from the host, Otel Collector comes with resourcedetection processor.

Resource Detection Processor automatically detects and labels metadata about the environment in which the data was generated. Such metadata, known as "resources", provides context to the telemetry data and can include information such as the host, service, container, and cloud provider.

For example, to detect host.name and os.type attributes, you can use system detector:

processors:
  resourcedetection:
    detectors: [env, system]

service:
  pipelines:
    metrics:
      receivers: [otlp, hostmetrics]
      processors: [batch, resourcedetection]
      exporters: [otlp/uptrace]

To add custom attributes such as an IP address, you can use env variables with env detector:

export OTEL_RESOURCE_ATTRIBUTES="instance=127.0.0.1"

To detect more information, you can use more specialized detectors, for example, if you are using Amazon EC2, you can use ec2 detector to also discover cloud.region and cloud.availability_zone attributes:

processors:
  resourcedetection/ec2:
    detectors: [env, ec2]

If you are using Google Cloud:

processors:
  resourcedetection/gcp:
    detectors: [env, gcp]

If you are using Docker:

processors:
  resourcedetection/docker:
    detectors: [env, docker]

You can check the official documentation to learn about available detectors for Heroku, Azure, Consul, and many others.

Memory Limiter

memorylimiterprocessor is a component that allows users to limit the amount of memory consumed by the OpenTelemetry Collector when processing telemetry data. It prevents the collector from using too much memory, which can lead to performance issues or even crashes.

Memory Limiter Processor works by periodically checking the amount of memory consumed by the OpenTelemetry Collector and comparing it to a user-defined memory limit. If the collector is using more memory than the specified limit, the processor will start dropping telemetry data until the memory usage falls below the limit.

To enable memory limiter:

processors:
  memory_limiter:
    check_interval: 1s
    limit_mib: 4000
    spike_limit_mib: 800

service:
  pipelines:
    metrics:
      processors: [memory_limiter]

Uptrace

Uptrace is an OpenTelemetry APM that helps developers pinpoint failures and find performance bottlenecks. Uptrace can process billions of spans on a single server and allows to monitor your software at 10x lower cost.

You can get started with Uptrace by downloading a DEB/RPM package or a pre-compiled Go binary.

• OpenTelemetry backend