Strategy Inplace Direct

Plan durability: plan maintenance, self-consistency invariants, and resume reconciliation follow plan-materialization.md. Save progress after every step below - status token plus durable facts (run_ids, ARNs, S3 versions, reasons) plus Last updated and ## Where we are - before proceeding to the next action.

Prerequisites

• Approach recorded as in-place-direct in the plan.
• Upgrade path determined in the plan.
• Docker validation (Phase 6) strongly recommended, since there is no test copy.

Execution Steps

1. Unrecommended Banner + Extra Confirmation Checkpoint

Present to the user and require explicit confirmation beyond the standard in-place checkpoint:

WARNING: This upgrades your live environment in place with no test copy. Compatibility issues surface only against your live environment. Rollback options: downgrade from 3.x to 2.11.x is supported; a within-major downgrade to a still-supported (non-EOS) version is supported; downgrade to an EOS version is not possible (verify the current EOS list at runtime). The Rehearsed in-place upgrade is recommended instead. Confirm to proceed with the Direct in-place upgrade.

2. Run the step-by-step upgrade on your current environment

Run the step-by-step upgrade on your current environment; starting condition: upgraded in place. The engine walks every version jump to the target, each version jump confirmed separately. See upgrade-engine.md.

3. Re-check your current environment at the target version

Confirm metadatabase=healthy, scheduler=healthy, and no import errors. On failure, hand off to debugging-mwaa-workflow, then re-verify. /dags paginates (default limit 100) — loop offset until total_entries (as in discovery-preflight.md Step 1.4) so an import error on a DAG beyond the first 100 is not missed.

4. Test-run your current environment (REQUIRED)

Parse validation (health + no import errors from Step 3) is necessary but NOT sufficient. Invoke skill testing-mwaa-workflow to confirm DAG execution succeeds at the target version on your current environment.

• Invoke skill testing-mwaa-workflow with delegated mode — do NOT trigger DAG runs manually via the REST API (POST /dags/{id}/dagRuns) or poll run states inline. The testing skill owns triggering, polling, classification, and the fix loop. Prioritize batch 1 / user-specified critical DAGs first.
• Every triggered DAG must reach terminal SUCCESS or be classified as blocked on an infrastructure/data dependency unrelated to the Airflow version (NEEDS-DECISION items from the testing skill’s fix loop).
• On artifact failures: fix via debugging-mwaa-workflow + authoring-mwaa-workflow, redeploy, retest until SUCCESS. Because the upgrade was in-place with no test copy, failures surface against your live environment — a downgrade (see Step 1 rollback options) may be warranted if execution is broadly broken.
• Do NOT mark the migration complete until test-run validation passes.

Record execution results in the plan under a ## Test-run results section (DAG ID, run_id, state, duration, and any NEEDS-DECISION items).

5. Mark Complete

There is no Phase 8 (no test copy to switch over or decommission). Update the plan: status complete, approach in-place-direct, completion timestamp.