Dynatrace 1.349 adds SRE agents, OpenTelemetry root cause analysis and Azure cost data

Autonomous agents in Workflows

The headline addition is Agentic Operations and the SRE Agent. Dynatrace says users can now delegate SRE tasks to autonomous agents for automated, reliable and trustworthy execution. According to the company, SRE Agents conclude a problem investigation even before a human looks at it, and the report includes the root cause, key findings and a proposed fix written onto the problem.

Agentic workflows can also handle scheduled work such as right-sizing, cost cleanup or a weekly report, and hand the owning team a ready-to-use result. Dynatrace states that the agents use only the tools a customer allows and stay within the permissions that customer sets. Users decide where a person must approve before a change lands, and the execution log records every step.

Four new items are available in Workflows:

• Run agent: runs a ready-made agent as a workflow step without writing a prompt. The first is the SRE Agent, pre-configured by Dynatrace with its instructions and tool selection. More agents will follow.
• Prompt agentic AI: users define the prompt, context, instructions, permitted tools and an output schema. The agent decides which tools to call, runs its own queries and returns a structured result.
• Prompt generative AI: summarizes events and activity in any language and suggests fixes for open problems and code changes, optionally grounded in Dynatrace Documentation, Community and Developer Portal content.
• Agentic workflow templates: starting points for SRE Agent, Alert Reduction, Kubernetes Troubleshooting, Database Operations, Infrastructure Optimization and more.

Causal AI for OpenTelemetry and a new JSM connector

Dynatrace root cause analysis now covers OpenTelemetry services. When several services in a chain degrade together, their anomalies are merged into one problem with a causal explanation that names the root cause service, rather than one alert per affected service. According to the notes, this works from a monitored web or mobile frontend down into an OpenTelemetry backend, and from a service to the databases it calls, including cloud databases and those monitored with a Dynatrace extension. Causal AI follows the Smartscape topology on Grail data, so no tags or correlation rules need to be maintained.

A new Jira Service Management (JSM) connector for Workflows creates, updates or closes a JSM alert whenever Dynatrace detects a problem, security vulnerability or attack. A stable alias ties the three actions to the same JSM alert, which Dynatrace says keeps the on-call view free from duplicate notifications. Alert mapping and routing are configured directly in Workflows.

Azure cost data, tracing controls and access changes

Cost & Carbon Optimization now supports ingestion of Azure cloud cost and usage data, alongside AWS. Data is imported daily from Azure Cost and Usage data exports using the FOCUS 1.2 specification, an open billing standard from the FinOps Foundation. Up to 12 months of historical data can also be ingested. The data is stored as business events in Grail and can be queried with DQL, and OpenPipeline processors can enrich it, for example by mapping Azure tags to Dynatrace entities. Cost Explorer now shows Azure and AWS data, and because both use FOCUS data sets, the same queries work for either provider.

Other feature updates include:

• Group management: a redesigned experience lets admins create groups and invite members in one flow, and preview effective policies before saving.
• Request naming: permissions can now follow IAM policies instead of management zones. This is opt-in, and users without a matching policy lose access.
• Adaptive Traffic Management: controls per scope, such as a host group, Kubernetes cluster or namespace, are now available in Settings under Trace ingest control. Previously they were available only through the Settings API.

What it means for users

The release notes list 27 feature updates, 6 breaking changes and 28 fixes and maintenance items. Teams planning to adopt the IAM-based request naming should note that access is lost for users without a matching policy. For operations teams, the combination of agent-written investigations, merged causal problems and automatic JSM alert handling is aimed at reducing alert noise and manual handoffs, as Dynatrace describes it.