DEV Community

Cover image for Dynatrace SaaS 1.349: what’s new and what to check in production
Span Observer
Span Observer

Posted on Originally published at span.observer

Dynatrace SaaS 1.349: what’s new and what to check in production

Dynatrace SaaS 1.349 began rolling out on 27 October 2026, according to the official release notes page (last updated 5 November 2026). Your tenant may already have it, without anyone asking. The useful question is not « what is new? » but « what could break quietly on Monday morning? »

This article sorts the release by operational risk: breaking changes first, then settings that move, then new features worth testing before you switch them on. It applies to Dynatrace SaaS. Managed follows a different schedule and is out of scope.

A note on honesty: span.observer has not tested 1.349 on a tenant. The facts come from the release notes, which give only a rollout start date and no date per feature. The risk ranking is my own analysis.

Dynatrace SaaS 1.349: the breaking changes to handle first

The page lists seven breaking changes. The summary below pairs each with the expected action.

YAML summary of the Dynatrace SaaS 1.349 breaking changes: log ordinal attribute, ingest enrichment rules, Settings scoping, the 250-permission limit, notebooks and classic metric limit rules

The first two touch the data itself. In logs, the ordinal attribute only orders a record when it holds an integer. A value such as "15" moves to overwrittenN.ordinal. Any query, processor or alert that reads the old field needs a review, or the field should be renamed before ingest.

For enrichment, you can no longer create a rule on a host scope in builtin:ingest.enrichment.config. Existing rules stay editable « until further notice » but will be removed later. Duplicate rules within one scope are now rejected by the Settings API. If a configuration pipeline pushes the same rule twice, it will fail.

Access control: the quietest risk

One change can open or close access with no error message: Settings object scoping, in Group 3 environments only. Per the page, management zone conditions in IAM policies never match any more, and security context conditions derived from management zones stop working. Host groups must be addressed by a string identifier (mygroup), not an entity ID.

In practice, an ALLOW statement can stop granting a permission and a DENY statement can stop blocking one. Audit your policies first, separating what relies on management zones from what does not. For background, see Grail versus Classic: permissions, tags and architecture.

The page adds that security context enrichment initially applies only to values set at the edge, for example by OneAgent. Do not rely on that context alone to protect settings written elsewhere.

On the API side, the call /iam/v1/accounts/{accountUuid}/groups/{groupUuid}/permissions accepts at most 250 permissions per request. Above that, POST and PUT return 400. A script that pushes every permission of a group in one go must now split its payload.

Terminal output: jq length on a sample file of 620 permissions returns 620, then splitting by 250 gives three batches of 250, 250 and 120

The example above uses a dummy file and jq’s _nwise function. Adapt it to the real body format of your requests, which I have not verified. A PUT that would exceed the limit is replaced by a PUT followed by POSTs.

Settings that change place or source

  • Classic metric limit alert rules: the four built-in rules under builtin:internal.metric.limits-alerts are disabled in every SaaS environment, and active ones were deactivated automatically. Replace them with Metric events, or move to Latest Dynatrace and Metrics on Grail.
  • Request naming: access now follows IAM policies and security context instead of management zones. If you opt in, a user without a matching policy loses access.
  • Browser Monitor: in Group 3, ActiveGate 1.347 or later is needed to schedule and run monitors. The dt.synthetic.browser(.step).duration metrics are then computed from RUM JavaScript.
  • Notebooks: invalid JSON is flagged and invalid sections are left out. Uploading a corrected file creates a new notebook, so sharing does not carry over.
  • Multi-environment ActiveGate: deprecated. Existing deployments keep working, with no new features. An end-of-life date will be announced.

New features to test before enabling

Three features headline the release. None forces action, which leaves time to test them.

  1. Agents in Workflows: three actions, Run agent (the SRE Agent writes root cause, key findings and a proposed fix onto the problem), Prompt agentic AI and Prompt generative AI. Per the page, agents use only permitted tools, can require human approval and log every step. Start with a workflow that only comments on a problem. See also our Workflow examples and guardrails for agents in production.
  2. Causal AI for OpenTelemetry services: anomalies along one chain merge into a single problem naming the root-cause service, with no tags or correlation rules. Enabled by default only in Group 3, not in Group 1 or Group 2.
  3. Jira Service Management connector: it creates, updates and closes JSM alerts as problems change state, using a stable alias to avoid duplicates.

Among the smaller items, note Azure cost import in FOCUS 1.2 format (up to 12 months of history), custom buckets for metrics (the default default_metrics bucket keeps 462 days) and a Snowflake connection wizard that will replace the extension, itself due for deprecation. For setup, see Snowflake with Dynatrace.

How to check it works

  • Permissions: test with an ordinary service account and user account, before and after. An access that changes result points to a policy to fix.
  • Logs: look for records where overwrittenN.ordinal is set; they reveal non-integer values. I have not run that query.
  • Automation: rerun your configuration pipelines; a duplicate enrichment rule fails immediately.
  • Alerts: list those that relied on the four disabled rules.

Best practices

  • Find out which group (1, 2 or 3) your environment is in: several changes depend on it.
  • Give each breaking change a named owner and a date.
  • Reread the release notes before each wave: the page is updated after publication.

Frequently asked questions

When was Dynatrace SaaS 1.349 released?

The page states a rollout start on 27 October 2026. It gives no date for individual features.

Which environments are hit by the scoping change?

According to the notes, only Group 3. Check your group before auditing IAM policies.

Should I enable the agents right away?

No. Existing environments keep their settings; test first on a low-stakes workflow.

Conclusion

For 1.349, the priorities are the IAM policy audit, the log ordinal attribute and enrichment rules. The agents can wait. Sources: Dynatrace SaaS 1.349 release notes and the vendor’s blog post.

span.observer is an independent community blog, not affiliated with Dynatrace LLC. Dynatrace is a trademark of Dynatrace LLC. Versions and dates change: check them in the official documentation for your environment.


Originally published on span.observer. French version: https://span.observer/fr/dynatrace-saas-1-349-verifier-production/.

Top comments (0)