Integrating k3s with Dynatrace: differences between installing OneAgent on each node vs Dynatrace Operator

[NicoleMT]

Hello,
I am planning to integrate a single k3s cluster with Dynatrace and I noticed there are two installation options:

1. Installing OneAgent on each cluster node.

2. Using the Dynatrace Operator for instrumentation.

I would like to better understand the differences between these two approaches, especially regarding:

• The level of observability provided.

• Licensing/consumption impact.

• Recommendations depending on the environment (lab vs production).

Has anyone in the community already worked with this setup and could share experiences or best practices?

Thank you in advance!

[Julius_Loman]

The supported method of deploying and observing k8s clusters is Dynatrace Operator.  You shall not attempt to use any other approach unless you have a really good reason not to use the operator in any deployment modes.  Typically you want to use cloud native full stack mode, but it really depends on the details.

While deploying OneAgent directly on the nodes might work in some deployments, it's not a supported way of getting deep monitoring of k8s workloads.