Agent v7 is the latest major version of the Datadog Agent. The only change from Agent v6 is that this version only includes support for Python 3 for integrations and custom checks.

See the Upgrade to Agent v7 documentation to learn how to upgrade your Agent to version 7. All official integrations support Python 3 out-of-the-box. Follow the Python 3 Custom Check Migration guide to migrate your custom checks to Python 3.

Note: You can test this migration with Agent v6, by Using Python 3 with Datadog Agent v6.

Agent version 6 main changes:

The big difference between Agent 5 and Agent 6 is that Agent 6 is a complete rewrite of the core Agent in Golang. Golang has allowed the Agent to take advantage of concurrency. In place of the three processes the Agent v5 used to run—the Forwarder, the Collector, and DogStatsD—there is now only one process: the Agent. It also comes with a number of other core improvements:

• Agent v6 has significantly improved resource usage over Agent v5:

  • Decreased CPU usage
  • Decreased memory usage
  • Fewer file descriptors
  • All around decreased footprint

• Agent 6 uses two additional ports:

  • 5000 to expose its runtime metrics.

  • 5001 for the Agent CLI/GUI commands.

    Note: You can specify different ports for expvar_port and cmd_port in the datadog.yaml file.

• Custom build your Agent v6 and DogStatsD much easier and with many more configuration options, to include or exclude almost anything.

Agent v6 new functionalities:

To see all changes between Agent v5 and v6, consult the Datadog Agent dedicated changes documentation. The following are key differentiators:

• Distributions metrics can be performed on the server directly to calculate real, effective global percentiles.
• DogStatsD can be used over a Unix socket instead of over UDP.
• Live Process monitoring is available for Windows.
• Prometheus OpenMetrics is supported natively.
• All your logs can be sent to Datadog for alerting, analysis, and correlation with metrics.