Dynatrace Release Radar 10.26 adds agentic investigations and richer tracing

Agentic AI investigates before a fix

Dynatrace is adding Agentic Operations and the SRE Agent to Workflows, allowing teams to run a problem investigation from a workflow. The agent builds on agentic root cause analysis, gathers evidence, and attaches its findings, root cause and recommended fixes to the problem.

Teams can choose between a ready-made investigation with preconfigured instructions and tools, a custom agent prompt with selected tools and a structured result, or a generative AI prompt for a summary or suggested fix. Dynatrace provides templates for SRE Agent, Alert Reduction, Kubernetes Troubleshooting, Database Operations and Infrastructure Optimization.

The company says agents remain within configured tools and permissions, while approval points remain under the user’s control. An execution log records their steps. Dynatrace also advises users to review the SRE Agent’s findings and recommendations, because “A recommended fix isn’t an applied fix.”

Dynatrace Assist can also answer questions about the page open beside the chat. According to Dynatrace, Assist indicates when it uses screen content and masks personal data with placeholders.

Tracing keeps context across services and queues

Distributed Tracing now carries a problem’s timeframe and filters when users open a trace from a problem. That context remains visible in the chart and carries into Kubernetes, Logs and Services, so investigators do not need to set it again.

For requests crossing a queue or serverless boundary, automatic span-link resolution loads upstream producer traces into the waterfall when a consumer span contains OpenTelemetry span links. Users can open linked traces at the connected span instead of searching for the producer manually.

Dynatrace also says causal root cause analysis now covers OpenTelemetry services. Causal AI follows Smartscape topology on Grail from monitored frontends into OpenTelemetry backends, and from services to monitored databases, combining related service anomalies into a single problem.

More evidence for user, network and cloud investigations

The release adds Apdex scores on Grail for tracking satisfied, tolerating and frustrated experiences across frontends and sessions. Users can filter Experience Vitals, Error Inspector, and Users & Sessions by city, region, country or continent, while frontend events and backend traces from Synthetic results provide additional investigation paths.

Network views expose host-to-host and process-to-process connections collected by OneAgent Network Connection Monitoring. In Kubernetes, teams can inspect pod-to-pod, pod-to-service and pod-to-process connections, including bandwidth, round-trip time, retransmissions, resets and timeouts.

Google Cloud monitoring is also brought into Clouds alongside AWS and Azure, with metrics, logs, metadata, topology, ready-made dashboards and preconfigured health alerts. Dynatrace says Google Cloud projects can be onboarded without deploying ActiveGate for metric polling, while Google Cloud labels remain available across metrics, logs and topology.

For users and incident stakeholders, dashboard owners can create live dashboard links for viewers without a Dynatrace account. Links can use passcodes, expiration dates and IP allowlists, and dashboards can be embedded in an iframe. The owner controls the timeframe, refresh interval and variable values shown to viewers.

Together, these changes give Dynatrace users more ways to investigate problems with preserved context, supporting evidence and controlled sharing. The source presents the updates as tools for the work that comes before a fix, while leaving the decision to approve a recommended action with the user.