Skip to main content
Requires the complete reservation and settlement workflow from steps 1–3. Test with synthetic data in a test environment.

Test the failure paths

1

Validate the happy path

Seed one PENDING order and a matching balance. Confirm that available funds decrease, reserved funds increase, one ledger row appears, and the response exposes all three identifiers.
2

Replay the same request

Send the same requestKey. Confirm that no second reservation or ledger entry is created and the prior transaction can be identified.
3

Race two requests

Submit two different requests against the same balance. Confirm that row locking and serializable isolation prevent overspending.
4

Reject insufficient funds

Confirm a 409 response, no partial balance update, no reserved order state, and no ledger insert.
5

Interrupt settlement

Simulate a provider timeout. Confirm that the Job retry preserves transactionId, operationId, retry lineage, and provider idempotency metadata.
6

Verify least privilege

Run with the real service account, then remove one required Table or Secret permission and confirm the workflow fails closed.

Investigate one operation

Start with the identifier closest to the report: In Jobs History, inspect status, duration, attempt, selected template and Component version, timestamps, metadata, and actor. In Databases → Audit, verify the balance and order transition. In Trace, follow the operationId across Endpoint execution and Job dispatch.
A successful Endpoint response proves that reservation committed and settlement was scheduled. It does not prove that the provider settled the order. Customer notification should follow a terminal Job state plus a confirmed transaction/provider state.

Production extensions

This guide keeps the architecture visible rather than pretending to be a complete payment system. A production implementation commonly adds:
  • a reconciliation Job for transactions stuck in PENDING_SETTLEMENT;
  • an append-only settlement/reversal ledger policy;
  • provider webhook verification and replay protection;
  • currency- and tenant-specific limits;
  • dual-control approval for high-value or manual operations;
  • dashboards and alerts for age, retries, failure classes, and balance invariants;
  • retention and redaction rules aligned with financial and privacy obligations.
Continue with first production workflow. Previous: step 3. Return to the workflow overview.
Last modified on November 8, 2026