¶ In this lesson you will
- Recognise each failure from what the user sees
- Confirm it in Kibana
- Fix it and recover the application
¶ How these were made
We broke the demo workflow on purpose, one fault at a time, ran real applications, captured what the applicant, the officer and Kibana showed, and restored the good workflow. Try to diagnose each drill before reading the answer.
¶ Drill A · The applicant cannot submit
Create application does nothing; DevTools shows 403
-
- Travel details: summary of the entered data
- Edit: go back to the form
- Send to email: notification channel
- Confirmation checkbox: information is true and accurate
- Note: information cannot be changed after this action
- Create application: submits the application
Kibana: UNAUTHORIZED_ROLE_ON_TRANSITION on event APPLY
- KQL query: message:"UNAUTHORIZED_ROLE_ON_TRANSITION"
- Hit count: 8 hits
- PROCESSING_FAILURE_TRACKER: event=APPLY startState=NEW
- WARN: Forbidden Exception: UNAUTHORIZED_ROLE_ON_TRANSITION
- EXCEPTION_OCCURRED: ApplicationServiceImpl / workflow_engine
The fault: only agents may APPLY
{
"startState": "NEW",
"event": "APPLY",
"authorisedRoles": ["ROLE_AGENT"],
"endStateOne": { "stateCode": "PAYMENT_PENDING", ... }
}
- Citizens are refused. Authorise the applicant roles the service is published for
¶ Drill B · Paid, but no officer sees it
The applicant sees a normal "Submitted for Processing"
- Status looks normal: Submitted for Processing
- Last updated one minute after submission, then nothing
- The timeline stops here: no officer ever receives it
Kibana: ASSIGN_OFFICE → WORKFLOW_ENGINE_ERROR, OFFICE_NOT_FOUND, four attempts
- KQL query: message:"C260930160938VLLWAM" and message:"OFFICE_NOT_FOUND"
- Time range: Last 1 hour
- Hit count: 12 hits
- PROCESSING_FAILURE_TRACKER: event=ASSIGN_OFFICE
- ASSIGN_OFFICE -> WORKFLOW_ENGINE_ERROR
The tracker line: event, start state, source INTERNAL
- KQL query: message:"PROCESSING_FAILURE_TRACKER" and message:"C260930160938VLLWAM"
- Hit count: 8 hits
- Expanded document panel (1 of 8)
- message: ASSIGN_OFFICE 404 from officer-engine /officer/v1/office?officeId=...
- traceId: (empty) for internal processing
The OFFICE_ASSIGNMENT action pointed to an office id that does not exist. The engine tried four times (the first attempt and 3 retries, 60 seconds apart), then gave up. The application C260930160938VLLWAM is still in PENDING_ASSIGN_OFFICE. Fix the office in the workflow (a Service Improvement on a live service), then ask support to re-trigger the event.
¶ Drill C1 · The Approve button is missing
Only Request for Action and Reject: Approve is hidden
- Application No: C260930161617RK5OKX
- Hide details: collapse the summary
- Traveller: full names and phone number
- Trip: location, date and purpose of travel
- Request for Action button
- Reject button
No error anywhere: the portal simply hides transitions the signed-in officer may not fire. The APPROVE transition had authorisedUsers set to another user. Kibana has nothing to show, so read the workflow JSON.
¶ Drill C2 · An empty red toast on Reject
The officer rejects with a reason
- Reject form
- Reason: rejection reason input
- Comment: rejection comment
- Cancel
- Reject: submits the rejection
An empty error toast; the dialog stays open
- Error toast: the Reject transition failed
- Reject confirmation dialog still open
- Reject button in the dialog
- Reject action still available in the panel
Kibana with the traceId from DevTools: 42 lines
- KQL query: traceId:"6abd1baceb7eeb2030e42443c2391c10"
- Time range: Last 20 minutes
- Hit count: 42 hits
- EXCEPTION_OCCURRED: ApplicationServiceImpl
- TOKEN_UTIL => Token Data: officer identity for the request
The cause: getApplicationState → NotFoundException
- KQL query: traceId:"6abd1baceb7eeb2030e42443c2391c10" and message:"getApplicationState"
- Hit count: 2 hits
- exception_class: NotFoundException
- message: EXCEPTION_OCCURRED ... methodName=getApplicationState
- root_cause_class: NotFoundException
The REJECT transition's end state was COURSE_DEMO_REJECTED_TYPO, a state that does not exist. The portal saved it without complaint. After fixing the code, the officer's next click works (user actions are synchronous; nothing to re-trigger).
The four drills
| Drill | User sees | Kibana | Fix |
|---|---|---|---|
| A | 403 on submit-application | UNAUTHORIZED_ROLE_ON_TRANSITION, event=APPLY | Authorise applicant roles on APPLY |
| B | Normal status, empty inbox | WORKFLOW_ENGINE_ERROR OFFICE_NOT_FOUND; tracker source=INTERNAL | Fix office; re-trigger ASSIGN_OFFICE |
| C1 | Approve button missing | Nothing | Fix authorisedRoles / authorisedUsers |
| C2 | Empty red toast, 500 | CACHE_VALUE_NOT_FOUND → NotFoundException in getApplicationState | Fix the state code; retry |
¶ Check yourself: match the pairs
Show the matches
| Item | Matches |
|---|---|
| 403 when the citizen clicks Create application | UNAUTHORIZED_ROLE_ON_TRANSITION |
| Paid, stuck in PENDING_ASSIGN_OFFICE | OFFICE_NOT_FOUND |
| Empty toast on Reject, 500 | NotFoundException in getApplicationState |
| Approve button hidden | No log: authorisedUsers |
← D5 · Kibana: find an application's story · D7 · Cheat sheet: states, events, actions, go-live →