Lesson 5 of 5
Troubleshooting B2B flows
9 min
Listen to this lesson
B2B failures are noisy and expensive. A partner calls before your alert fires, the error text names an XSD you've never seen, and three teams each believe the problem is elsewhere. A runbook that starts from the symptom and walks one layer at a time turns that into a 15-minute exercise.
Where to look
In Cloud Integration you start in Monitor > Integrations and APIs. Monitor Message Processing shows the message processing log (MPL) for each run, with status, steps, errors and attachments; Manage Security holds keystore entries and certificates; Manage Stores holds message queues, data stores and number ranges. For TPM scenarios, Monitor > B2B Scenarios adds the business view of interchanges and links to the MPL of each step.
The AS2 adapters write their own attributes to the MPL, which are often the fastest way to confirm what happened on the wire: MessageDirection, SenderAS2Name, ReceiverAS2Name, AS2MessageID, MDNRequested (synchronous or asynchronous), MDNType and MDNStatus. The AS2 sender also passes headers such as AS2PartnerID, AS2OwnID, AS2Filename and AS2MessageID into the flow, which you can log or use for routing.
Typical failures and their fixes
| Symptom | Where you see it | Usual cause | Fix |
|---|---|---|---|
| Partner gets an HTTP error, no MPL at Contoso | Partner's AS2 log | Wrong URL or address, TLS or client certificate problem, or no sender channel matches the AS2 IDs and subject | Check the endpoint in Manage Integration Content and the uniqueness of the AS2 ID combination |
| Decryption error | MPL of the AS2 sender | Partner encrypted with an old Contoso certificate, or wrong private key alias | Compare certificate serial numbers; confirm the alias in the keystore |
| Signature verification failed | MPL of the AS2 sender | Partner signs with a new certificate | Load the new public certificate and update the public key alias |
| Message rejected as too large | Negative MDN, MPL | Body or attachment above the Conditions limit | Agree file size with the partner or raise the limit |
| Schema or conversion error | MPL of the splitter or converter | XSD file name doesn't match content, wrong version, unsupported repetition character | Match name and content; update step version; agree envelope usage |
| Partner rejects Contoso's 856 but MPL is Completed | Partner's 997 or MDN payload | Negative synchronous MDN without Message Failure on Negative MDN | Enable the option; read the MDN or 997 errors |
| Outbound fails before sending | MPL near the number range step | Number range at maximum without Rotate | Extend the range or enable Rotate after agreeing with the partner |
A starting runbook for AS2 and EDI processing errors in Cloud Integration.
Queues, retries and the dead-letter queue
With Exactly Once, inbound AS2 messages wait in a JMS queue named after the sender channel, which is why SAP recommends giving AS2 sender channels meaningful names. Failed messages are retried with the configured interval and backoff; messages that fail repeatedly can end in the dead-letter queue if you enabled it. Check Manage Stores > Message Queues during an incident: a growing queue for one partner is a sign the receiver or a mapping is failing for all of its messages. Don't delete queued messages to "clear" an incident; each one is a document the partner believes you have received, because you already sent the MDN.
Trace with care
Setting an integration flow's log level to Trace captures payloads at each step for a limited time, which is invaluable for converter and mapping problems. It also writes order prices, bank details and personal data into the log. Contoso's rule, agreed with Nadia: trace only in the test tenant by default; in production only with a ticket, for one flow, and switched back as soon as the evidence is saved.
Building the support runbook
A good B2B runbook is per partner and per flow, and short enough to use at 2 a.m. Contoso's template has one page per partner with: AS2 IDs and endpoints, certificate expiry dates, expected daily volume and cut-off times, the channels and flows involved, number range names, the resend protocol agreed with the partner, the triage path above, and the escalation contacts on both sides.
What would you do?
After Contoso renews its AS2 certificate, Great Lakes' 850s fail at Contoso with decryption errors, while Lakeside's SFTP files are fine. What is the most likely cause and fix?
- Start from the symptom and walk the layers: connection, AS2 adapter, splitter and converter, receiver and MDN, business result.
- AS2 MPL attributes (AS2MessageID, MDNType, MDNStatus) show what happened on the wire.
- Inbound AS2 queues hold documents you already acknowledged; never delete them to clear an incident.
- Trace captures payloads; use it briefly and with approval in production.
- Write one runbook page per partner: IDs, certificates, volumes, resend protocol and contacts.
Check your understanding · not graded
Lesson check
1. A partner posts to Contoso but no message processing log appears. Which cause fits?
2. Which MPL attribute shows whether the partner requested a synchronous or asynchronous MDN?
3. Why shouldn't operators delete messages from an AS2 sender's JMS queue to clear an incident?
4. After a certificate renewal, a partner's inbound messages fail with signature verification errors. What fixes it?