Dynatrace Managed Q&A
cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Looking to upgrade from Dynatrace Managed to SaaS? See how

Kubernetes application monitoring with namespace-only access

jasonDUT
Newcomer_

We want to monitor applications in Kubernetes but have strict access limitations:

  • Managed Kubernetes cluster

  • No privileged DaemonSets allowed

  • No hostPath / host access

  • Access limited to specific namespaces only

  • Custom ServiceAccount with read-only RBAC (get/list/watch on pods, workloads, services, and metrics APIs)

We cannot deploy Dynatrace OneAgent as a DaemonSet or use cluster-level components.

Question:
What is the officially supported way to monitor applications only in this scenario?
Is application-only monitoring (e.g., webhook or code-level instrumentation) the recommended approach, and what limitations should we expect?

2 REPLIES 2

AntonPineiro
DynaMight Guru
DynaMight Guru

Hi,

Maybe this table might help.

Best regards

❤️ Emacs ❤️ Vim ❤️ Bash ❤️ Perl

t_pawlak
Leader

Hi,
With your constraints (no privileged DaemonSets, no hostPath/host access, namespace-only access, no cluster-level Dynatrace components), you cannot deploy Kubernetes Full-Stack monitoring based on a node-level OneAgent, nor the usual Kubernetes deployment model that relies on cluster-scoped components. The Dynatrace Operator is inherently cluster-scoped (ClusterRole, etc)
https://docs.dynatrace.com/docs/ingest-from/setup-on-k8s/reference/security

You effectively have two supported “application-only” paths (no host-level):

1. Dynatrace Application Monitoring / Application Observability (app-only, no OneAgent DaemonSet on nodes)

  • Monitoring is done by injecting Dynatrace code modules into application Pods (for example via init containers / build-time approaches), without installing OneAgent pods on every node and without host metrics.
  • If the Operator isn’t involved, Dynatrace explicitly notes you lose automatic injection/config/enrichment and must manually adapt your workloads.

Documentation here: https://docs.dynatrace.com/docs/ingest-from/setup-on-k8s/how-it-works/application-monitoring

2. Opentelemetry, I mean export OTLP to Dynatrace.
https://docs.dynatrace.com/docs/ingest-from/dynatrace-oneagent/oneagent-and-opentelemetry/oneagent-o...

 

Is application-only the recommended approach here?

IMO Yes. Given your constraints, it’s essentially the only practical, officially supportable route. Host/network-level features require components/privileges you can’t deploy (for example, CSI-related capabilities can require privileged permissions in typical setups). https://docs.dynatrace.com/docs/ingest-from/setup-on-k8s/how-it-works/components/dynatrace-operator

limitations:
No host-level infrastructure visibility and mabye limited Kubernetes platform/topology view, because without Dynatrace cluster components and broad K8s API access. But here you can do something like here:
https://docs.dynatrace.com/docs/ingest-from/setup-on-k8s/deployment/application-observability

 

 

Featured Posts