The Observability Blog

  • Metrics
  • OpenTelemetry

How to monitor Apache Flink with OpenTelemetry

by Jonathan Wamsley on
July 29, 2022

Apache Flink monitoring support is now available in the open source OpenTelemetry collector. You can check out the OpenTelemetry repo here! You can utilize this receiver in conjunction with any OTel collector: including the OpenTelemetry Collector and observIQ’s distribution of the collector.

Below are quick instructions for setting up observIQ’s OpenTelemetry distribution, and shipping Apache Flink telemetry to a popular backend: Google Cloud Ops. You can find out more on observIQ’s GitHub page:

What signals matter?

Apache Flink is an open source, unified batch processing and stream processing framework. The Apache Flink collector records 29 unique metrics, so there is a lot of data to pay attention to. Some specific metrics that users find valuable are:

  • Uptime and restarts
    • Two different metrics that record the duration a job has continued uninterrupted, and the number of full restarts a job has committed, respectively.
  • Checkpoints
    • A number of metrics monitoring checkpoints can tell you the number of active checkpoints, the number of completed and failed checkpoints, and the duration of ongoing and past checkpoints.
  • Memory Usage
    • Memory-related metrics are often relevant to monitor. The Apache Flink collector ships metrics that can tell you about total memory usage, both present and over time, mins and maxes, and how the memory is divided between different processes.

All of the above categories can be gathered with the Apache Flink receiver – so let’s get started.

Before you begin

If you don’t already have an OpenTelemetry collector built with the latest Apache Flink receiver installed, you’ll need to do that first. We suggest using the observIQ OpenTelemetry Collector distro that includes the Apache Flink receiver (and many others) and is simple to install with our one-line installer.

Configuring the Apache Flink receiver

Navigate to your OpenTelemetry configuration file. If you’re using the observIQ Collector, you’ll find it in one of the following location: 

  • /opt/observiq-otel-collector/config.yaml (Linux)

For the observIQ OpenTelemetry Collector, edit the configuration file to include the Apache Flink receiver as shown below:

    endpoint: http://localhost:8081
    collection_interval: 10s

   # Resourcedetection is used to add a unique (
  # to the metric resource(s),...  target_key: namespace

    # Add the exporter for your preferred destination(s)

      receivers: [flinkmetrics]
      processors: [nop]
      exporters: [nop]

If you’re using the Google Ops Agent instead, you can find the relevant config file here.

Viewing the metrics collected

If you followed the steps detailed above, the following Apache Flink metrics will now be delivered to your preferred destination.

The CPU usage of the JVM for a jobmanager or taskmanager.
The CPU time used by the JVM for a jobmanager or taskmanager.
The amount of heap memory currently used.
The amount of heap memory guaranteed to be available to the JVM.
The maximum amount of heap memory that can be used for memory management.
The amount of non-heap memory currently used.
The amount of non-heap memory guaranteed to be available to the JVM.
The maximum amount of non-heap memory that can be used for memory management.
The amount of memory currently used in the Metaspace memory pool.
The amount of memory guaranteed to be available to the JVM in the Metaspace memory pool.
The maximum amount of memory that can be used in the Metaspace memory pool.
The amount of memory used by the JVM for the direct buffer pool.
The total capacity of all buffers in the direct buffer pool.
The amount of memory used by the JVM for the mapped buffer pool.
The number of buffers in the mapped buffer pool.
The amount of managed memory currently used.
The total amount of managed memory.
The total number of live threads.
The total number of collections that have occurred.
The total time spent performing garbage collection.
The total number of classes loaded since the start of the JVM.
The total number of restarts since this job was submitted, including full restarts and fine-grained restarts.
The end to end duration of the last checkpoint.
The total size of the last checkpoint.
The number of checkpoints completed or failed.
The number of checkpoints in progress.
The number of records a task has.
The number of records an operator has.
The last watermark this operator has emitted.

observIQ’s distribution of the OpenTelemetry collector is a game-changer for companies looking to implement OpenTelemetry standards. The single line installer, seamlessly integrated receivers, exporter, and processor pool make working with this collector simple. Follow this space to keep up with all our future posts and simplified configurations for various sources. For questions, requests, and suggestions, reach out to our support team at