Container Application Monitoring is a licensing model, not a deployment mode. This guide moves Kubernetes workloads in application monitoring mode from application-only Full-Stack Monitoring (GiB-hours) to Container Application Monitoring (container-hours). Your deployment doesn't change, and no Dynatrace components are added or removed.
For Amazon ECS, see Amazon ECS container workloads.
Container Application Monitoring applies to application monitoring mode only. Change your deployment mode first, and set the opt-in in the same DynaKube edit so your workloads restart only once:
Container Application Monitoring doesn't include Profiling or thread analysis, and you can't add them. If your teams rely on continuous profiling, evaluate this before migrating.
| Capability | Application-only Full-Stack Monitoring | Container Application Monitoring |
|---|---|---|
| Kubernetes platform | by number of Pods | by number of Pods (unchanged) |
| Applications | by container memory (GiB-hours) | by number of containers (container-hours) |
| Included trace volume | Proportional to container memory | 6,000 KiB per container per 15-minute interval |
| OpenTelemetry | Trace data licensed separately as Traces - Ingest & Process | Treated equally with OneAgent |
Many small containers are typically more expensive under container-hours; fewer large containers are typically less expensive.
Choose how much to migrate at once:
| Option | Covers | Restart | Reversible |
|---|---|---|---|
Tenant-wide setting builtin:billing.container-application-monitoring.optin | All eligible OneAgent containers in the environment, on Kubernetes and Amazon ECS | No | No |
Resource attribute dt.has_container_application_monitoring=true | The DynaKube, component, or workload you set it on. Required for OpenTelemetry. | Yes | Yes |
For the steps, DynaKube samples, and setups without Dynatrace Operator, see Kubernetes container workloads.
You can also check for billing usage events with DQL:
fetch dt.system.events| filter event.kind == "BILLING_USAGE_EVENT" and event.type == "Container Application Monitoring"| dedup event.id| summarize container_hours = sum(billed_container_hours), by: { day = bin(timestamp, 24h) }| sort day desc
If you opted in with the resource attribute, remove it and restart the affected workloads. They return to GiB-hour licensing.
The tenant-wide setting can't be disabled once it's enabled. If you want to be able to revert, use the resource attribute instead.