/migrate-job

/migrate-job — guided full migration

Drive the migrator end-to-end with checkpoints between phases. Asks the user before running each long-running step so they can abort or adjust.

Workflow

1. Confirm prerequisites (invoke aidp-migrator-bootstrap silently — surface any gaps).
2. Ask for the source — either a Databricks workspace path OR a Databricks Job ID. Don't proceed without it.
3. Build the DAG — invoke aidp-build-dag. Show the resulting task count + dep count. Ask "looks right?" before next step.
4. Data check — invoke aidp-check-data. Show the OK/MISSING/EMPTY counts.
   • If any MISSING / EMPTY: ask the user how to handle. Options: (a) migrate catalog first via aidp-migrate-catalog, (b) configure aidp-bucket-mapping, (c) proceed anyway (with a clear warning).
5. Migrate — invoke aidp-migrate-job. Surface live log tail.
6. Status — when the run finishes, invoke /migration-status on the resulting JOB_REPORT.md.

Args (optional — if user provided them upfront)

$ARGUMENTS

If $ARGUMENTS contains a Databricks Job ID or workspace path, jump straight to step 3 (skip the question).

Checkpoints to surface to the user

Between phases, give the user a 1-line summary + the cost estimate of the next step:

[Phase 1/5] DAG built: 7 tasks, 18 dep notebooks, output reports/<MyJob>_manifest.json
[Phase 2/5] About to scan source data on cluster — ~2 min, no token cost. Proceed? (y/N)
...
[Phase 4/5] About to start Pass-2 migration — est. 30-90 min, $10-30 in Claude tokens. Proceed? (y/N)

Do not auto-proceed past the migrate step without explicit user confirmation. Pass-2 is expensive.

When to stop

• User aborts at any checkpoint.
• Data check shows >50% MISSING — likely the catalog hasn't been migrated yet; route to /migrate-catalog instead.
• Cluster is not Active — instruct the user to start it before continuing.