The Problem Nobody Talks About
Every our Dynatrace tenant is a snowflake.When the Demoability team operates six Dynatrace environments — from playground-dev on the sprint ring to the customer-facing live tenant — each one carries a different set of active Extensions 2.0 monitoring configurations. One tenant monitors IBM MQ and Oracle SQL. Another runs three SNMP auto-discovery scanners across WAN and datacenter segments. A third tracks Salesforce event streams. Some extensions need nothing more than an AG group name; others require OAuth tokens, API keys, database passwords, and private RSA keys stored safely in AWS Secrets Manager.
Managing this by hand in the Dynatrace UI works — until it doesn’t. A configuration drifts silently. A credential rotates and the extension stops collecting. Someone enables a feature set in one environment and forgets to replicate it to the others. There is no audit trail, no review process, no rollback.
The answer was obvious: manage it as code. The harder question was how.
Terraform Is the Right Tool — and a Daunting One
The dynatrace-oss/dynatrace Terraform provider supports dynatrace_hub_extension_v2_config, which can declaratively manage monitoring configurations for every installed extension. In theory, this is clean. In practice, it surfaces a wall of complexity immediately.
The Dynatrace API replaces every stored credential in its GET responses with a 16-character hex placeholder. This means every terraform plan shows a perpetual diff on password fields — unless you engineer around it with ignore_changes and a dual-hash fingerprint that can distinguish real credential changes from API noise. Then there is version management: each extension has an active installed version per tenant, and version mismatches between state and the API cause silent re-pushes. Six environments × 20+ extension types × per-environment variation in feature sets, AG groups, device IP lists, and secret keys is not a configuration problem. It is a code generation problem.
Where AI Becomes the Force Multiplier
Rather than manually inspecting each tenant’s Dynatrace Hub and transcribing configs by hand, the team uses an AI agent equipped with a purpose-built dt-export skill. The skill knows how to locate the provider binary in the Terragrunt cache, set the required environment variables, and invoke the provider’s built-in export CLI against a live tenant:
# Export all hub extension v2 configs from the tenant<br>export DYNATRACE_ENV_URL="https://{environmentid}.apps.dynatrace.com"<br>export DYNATRACE_CLIENT_ID="dt0s02...."<br>export DYNATRACE_CLIENT_SECRET="..."<br>export DYNATRACE_TOKEN_URL="https://{environmentid}.apps.dynatrace.com needs onboarding.<br><em>Export all existing extension configs, map them to deployment variables, </em><br><em>create environments/xyz/terragrunt.hcl, and update SECRETS.md with the required secret keys.</em>
The export produces raw .tf files reflecting exactly what is deployed. The agent then reads these files, maps the field structure to the repo’s module interface, and outputs ready-to-commit terragrunt.hcl configuration blocks — with credentials extracted into AWS Secrets Manager key references rather than plain values.
Example prompts that capture the full workflow:
“Export the current extension monitoring configurations from the live3gen environment. Find any configs for SNMP auto-discovery that are not yet in Terraform state and add them to environments/live3gen/terragrunt.hcl following the existing patterns.”
Or
“A new Dynatrace environment https://{environmentid}.apps.dynatrace.com needs onboarding. Export all existing extension configs, map them to deployment variables, create environments/xyz/terragrunt.hcl, and update SECRETS.md with the required secret keys.”
The dynatrace-extensions skill provides the agent with the full module interface, variable schema, and credential-scrubbing rules — so it produces code that passes terragrunt validate on the first attempt rather than requiring manual correction.
The Outcome
What previously required hours of UI inspection, manual JSON transcription, and trial-and-error Terraform authoring now takes minutes.
Features:
- New environments are onboarded with a single prompt.
- Existing configs are versioned, reviewed, and deployed through a pull request.
- Six tenants, fully reconciled, with terragrunt plan returning valid information what will be changes or not.