Oracle Integration Cloud: Building Integrations That Survive Production · Module 8 · Error handling, faults and retries
Partial failure in a batch
Lesson 86 of 175 · 2 min
Four hundred records, record 200 fails. Stop, skip, or roll back — and rolling back is usually not available once the first 199 posted. That last clause is the fact people design around too late. A loop is not a transaction. Each record was committed in the target as it was processed, and the target has no notion of a batch that can be undone. "We will roll back on failure" is a plan that requires a capability that does not exist, and discovering that at record 200 is discovering it in production. SO THERE ARE REALLY TWO OPTIONS AND THE THIRD IS COMPENSATION. STOP: 199 posted, 201 not attempted. Clean to reason about IF you know where it stopped, which means the run must record its position — and rerunning has to resume rather than restart, or the 199 are duplicated. SKIP: 399 posted, one in the dead letter.…
The full lesson is part of the course
The video, the complete written lesson and the module quiz are included in Oracle Integration Cloud: Building Integrations That Survive Production, with a certificate on completion and a fourteen-day refund window.
In this module: Module 8 · Error handling, faults and retries
- 1The four kinds of failure
- 2Faults versus errorsFree preview
- 3Scope and fault handlers
- 4What to do with a caught fault
- 5Retry: how many, how far apart
- 6The dead letter, and who reads it
- 7Partial failure in a batch
- 8Compensation, and its limits
- 9Alerting a human
- 10Resubmission: doing it safely
- 11Errors that are not failures
- 12What breaks: five error-handling failures
- 13Lab briefing · Break it five ways