Real-Time Dynatrace Kubernetes Interview Questions (2027)
28 real-time Dynatrace interview questions for Kubernetes: Dynatrace Operator, DynaKube, OneAgent modes, ActiveGate, tokens and troubleshooting.
Real-time Dynatrace interview questions for Kubernetes roles test whether you can deploy and run Dynatrace on a cluster: installing Dynatrace Operator, writing a DynaKube resource, choosing between platform monitoring, application monitoring and cloud-native full-stack, configuring ActiveGate capabilities, and fixing missing data when pods, tokens or the CSI driver misbehave.
This page is about Dynatrace on Kubernetes, troubleshot as it happens. For general platform questions on OneAgent, Davis AI and dashboards, use our Dynatrace interview questions for cloud monitoring, the main Dynatrace interview page on this site. Commands and fields below were checked against Dynatrace's documentation in November 2026, when the manifest install referenced Dynatrace Operator v1.11.0 and the newest DynaKube API version listed was v1beta6.
Key takeaways
- Dynatrace on Kubernetes is deployed and managed by Dynatrace Operator, configured through a DynaKube custom resource in the
dynatracenamespace. - Three documented options: Kubernetes platform monitoring, platform monitoring plus application observability, and platform monitoring plus full-stack observability.
- OneAgent modes in the DynaKube spec are
cloudNativeFullStack(recommended),classicFullStack,applicationMonitoringandhostMonitoring. - Tokens live in a secret named
dynakubewith keysapiToken(Operator token) anddataIngestToken. - ActiveGate capabilities include
routing,kubernetes-monitoring,metrics-ingest,dynatrace-apianddebugging. - Most "no data" incidents come from a wrong
apiUrl, bad tokens, blocked egress or pods injected before the Operator webhook was ready.
Deployment options compared
| Option | What you get | Components | Privilege needs |
|---|---|---|---|
| Kubernetes platform monitoring | Cluster resources, topology, metrics, events, alerts | Operator, webhook, ActiveGate | Kubernetes RBAC only |
| + Application observability | Distributed tracing, code-level insight, profiling | Adds code modules, optional log module, CSI driver by default | CSI driver runs privileged |
| + Full-stack observability | Adds host-level data, host network analysis, disk alerting | Adds host module and log module | Privileged CSI driver and host OS permissions |
Setup questions (1 to 9)
1. How does Dynatrace monitor a Kubernetes cluster?
Dynatrace Operator runs in the cluster and reconciles a DynaKube custom resource. Depending on the configuration it deploys ActiveGate for Kubernetes API monitoring and routing, a webhook that injects code modules into application pods, and OneAgent on nodes for full-stack mode.
2. How do you install the Operator with Helm?
helm install dynatrace-operator oci://public.ecr.aws/dynatrace/dynatrace-operator \
--create-namespace --namespace dynatrace --atomic
Dynatrace's docs note that with Helm 4.0 and later you use --rollback-on-failure in place of --atomic.
3. How do you install it with plain manifests?
kubectl create namespace dynatrace
kubectl apply -f https://github.com/Dynatrace/dynatrace-operator/releases/download/v1.11.0/kubernetes-csi.yaml
On OpenShift the docs use openshift-csi.yaml from the same release with oc apply. Pin a release instead of using "latest" so you can repeat the install.
4. Which tokens are needed and where do they go?
kubectl -n dynatrace create secret generic dynakube \
--from-literal="apiToken=<OPERATOR_TOKEN>" \
--from-literal="dataIngestToken=<DATA_INGEST_TOKEN>"
The Operator token lets the Operator talk to the Dynatrace API; the data ingest token is used for sending metrics and other data. In production, create the secret from your secret manager rather than typing tokens into a shell.
5. Show a minimal DynaKube for cloud-native full-stack.
apiVersion: dynatrace.com/v1beta6
kind: DynaKube
metadata:
name: dynakube
namespace: dynatrace
spec:
apiUrl: https://ENVIRONMENTID.live.dynatrace.com/api
oneAgent:
cloudNativeFullStack: {}
activeGate:
capabilities:
- routing
- kubernetes-monitoring
apiUrl is required, must include /api and cannot be changed later. v1beta6 needs Dynatrace Operator 1.8 or later; older Operators need an older API version, so match the example to your Operator release.
6. How do you check the deployment?
kubectl get dynakube -n dynatrace
kubectl get pods -n dynatrace
The DynaKube should show Running, and you should see Operator, webhook and ActiveGate pods, plus one OneAgent pod per node in full-stack mode.
7. When would you choose application monitoring over full-stack?
When you cannot run privileged host agents, for example on some managed or serverless node types, or when the security team only approves pod-level injection. You keep traces and code-level data and lose host-level metrics and host network analysis.
8. What does ActiveGate do in a cluster?
Its capabilities decide that. kubernetes-monitoring collects data from the Kubernetes API, routing sends OneAgent traffic through it, and metrics-ingest accepts metrics. Dynatrace documents that the kubernetes-monitoring capability and the separate .spec.kubernetesMonitoring section are mutually exclusive.
9. What is classicFullStack, and why avoid it for new installs?
It is the older full-stack mode. Dynatrace marks cloudNativeFullStack as recommended and states that classicFullStack is not supported with platform tokens or in Latest Dynatrace environments.
Day-two operations (questions 10 to 16)
10. How do you limit injection to some namespaces?
Set namespaceSelector under .spec.oneAgent.cloudNativeFullStack (or applicationMonitoring) in the DynaKube so only labelled namespaces get code modules. This keeps system namespaces and batch jobs you do not care about out of monitoring and licensing.
11. Why must application pods restart after the Operator is installed?
Code modules are injected by the webhook when a pod is created. Pods that were already running have no module until they are recreated, so roll the deployments after installing.
12. How do you upgrade the Operator safely?
Read the release notes for DynaKube API changes, upgrade in a non-production cluster first, run helm upgrade with the pinned version, confirm the DynaKube reaches Running, then roll a test workload to confirm injection still works.
13. How do you send traffic through a corporate proxy?
Set the proxy field in the DynaKube spec, and use trustedCAs if the proxy re-signs TLS. Routing OneAgent traffic through an in-cluster ActiveGate also reduces the number of pods that need direct egress.
14. How do you pull images from a private registry?
Mirror the Dynatrace images, then set customPullSecret in the DynaKube for registry credentials. Dynatrace also lists a publicRegistryOverride field in the spec reference.
15. How do Kubernetes events and alerts show up?
With Kubernetes platform monitoring, Dynatrace collects cluster events and metrics such as pod restarts, pending pods and node pressure, and its problem detection can link them to affected services. In an interview, give an example: a service slowing down because its pods are CPU-throttled.
16. How would you use Dynatrace data in a CI/CD pipeline?
Send deployment events to Dynatrace from the pipeline so changes appear on the timeline, and add a quality gate step that checks error rate and response time after a canary release before promoting it. Explain what the gate measures and what happens when it fails.
Real-time troubleshooting scenarios (questions 17 to 26)
17. The DynaKube status is stuck and never reaches Running.
Run kubectl describe dynakube -n dynatrace and read the conditions, then check the Operator logs. Common causes are a wrong apiUrl (missing /api), a token without the required scopes, or no network path from the Operator to the Dynatrace environment.
18. ActiveGate pod is in CrashLoopBackOff.
Check its logs for connection or certificate errors, then resource limits. If a proxy re-signs TLS, add the CA through trustedCAs. If memory is the problem, raise the ActiveGate resources in the DynaKube.
19. Services from one namespace never appear.
Check whether the namespace matches the injection selector, whether the pods were restarted after rollout, and whether the pod has the injection annotations or init container. A missing init container means the webhook did not mutate the pod.
20. The webhook is unavailable and new pods fail to start.
Check the webhook pods and their service. Scale the webhook to more than one replica for production clusters so one node failure does not block pod creation. Explain that a failing admission webhook can affect every new pod, which is why it needs the same care as other cluster components.
21. OneAgent pods do not start on some nodes.
Look for taints the OneAgent DaemonSet does not tolerate, node OS images that block privileged containers, and resource pressure. Add tolerations under .spec.oneAgent.cloudNativeFullStack for the node pools you intend to monitor.
22. After the security team blocked privileged pods, data dropped.
Full-stack needs a privileged CSI driver and host permissions. Either agree an exception for the dynatrace namespace or move to application monitoring and accept losing host-level data. Present it as a trade-off the security team can decide on.
23. A memory leak alert fires for a Java service after a release.
Compare heap and garbage collection metrics before and after the deployment event, check pod restarts for OOMKilled, and use memory profiling if your plan includes it. Roll back if restarts affect users, then hand the developers the evidence.
24. Two clusters show as one in Dynatrace.
Each cluster needs its own DynaKube and Kubernetes monitoring setup, with a distinct name. Check that the ActiveGate in each cluster is reading its own Kubernetes API and not a copied configuration.
25. Monitoring cost jumped after a new team onboarded.
Check which namespaces are injected and whether test or batch namespaces were included by a broad selector. Narrow the selector and agree on which environments need full-stack and which need only platform monitoring.
26. How do you prove Dynatrace is healthy during an incident review?
Show Operator, webhook, ActiveGate and OneAgent pods healthy at the time, the DynaKube status, and data continuity for the affected services. If there was a gap, explain it before anyone draws conclusions from missing charts.
27. How does the setup differ on OpenShift?
The steps are the same, with oc in place of kubectl. Dynatrace's docs create the dynatrace project with oc adm new-project and apply openshift-csi.yaml from the Operator release instead of the Kubernetes manifest. Security context constraints are the usual source of surprises, so check them when OneAgent or CSI driver pods do not start.
28. What would you tell a team that wants to monitor only one critical service?
Start with Kubernetes platform monitoring for the whole cluster, which needs no privileged pods, and add application monitoring for the namespace that runs the critical service. They get cluster context for every incident and code-level traces where it matters, and can widen coverage later. If you need a refresher on why traces matter beyond metrics, see our explainer on observability vs traditional monitoring.
How to prepare
Use a Dynatrace trial and a kind or k3s cluster. Install the Operator with Helm, apply the DynaKube above, deploy a sample app, then break things on purpose: wrong apiUrl, deleted secret, tainted node, pod not restarted. Write down what each failure looks like in kubectl and in the Dynatrace UI. For more question styles, see our most asked Dynatrace questions and the hands-on Dynatrace engineer interview questions. Kubernetes troubleshooting itself is the core of our CKA course in Pune, in Kharadi and live online.
Frequently asked questions
What is a DynaKube?
The custom resource that tells Dynatrace Operator how to monitor a cluster: the environment apiUrl, OneAgent mode, ActiveGate capabilities, tokens secret, proxy and other settings.
Which OneAgent mode should I use on Kubernetes?
Dynatrace recommends cloudNativeFullStack. Use applicationMonitoring when privileged host access is not allowed, and hostMonitoring when you only need host data.
Which namespace does Dynatrace Operator use?
The dynatrace namespace, for Helm, manifest and OpenShift installs.
Which tokens does Dynatrace Operator need?
An Operator token stored as apiToken and a data ingest token stored as dataIngestToken, both in a secret named dynakube in the dynatrace namespace.
Why are my existing pods not monitored after install?
Code modules are injected when pods are created. Restart existing workloads after the Operator and DynaKube are running.
Can I use kubernetes-monitoring and the kubernetesMonitoring section together?
No. Dynatrace documents them as mutually exclusive in the same DynaKube.
Does full-stack mode need privileged access?
Yes. It requires privileged CSI driver access and OS-level permissions for the host module. Platform monitoring alone uses Kubernetes RBAC only.
References
What's Your Reaction?
Like
0
Dislike
0
Love
0
Funny
0
Angry
0
Sad
0
Wow
0