AMQP contract

• work-inbound — produced tasks (Producer → Worker)
• work-outbound — processed / discarded / dead-letter tasks (Worker → Producer/Audit)
• certified-result — certified tasks (Producer → Audit)

⚠️ Dead-letter convention

If a message fails processing and is dead-lettered, it is re-published to the work-outbound exchange with routing key:

task.produced.<color>

In the context of the Audit service, such messages are treated as DISCARDED

Scenarios Reproducing

Scenario 3

What are the results when you change the deployment setup to: 3 instances of A, 3 instance of B, 1 instance of C AND when the processing time on B is changed to 8 seconds?

Reproducing

1. Change in Worker application.yml setting worker.handling-ms: 8000 and worker.fail-every-second-task: false
2. Clean & Build in gradle
3. Kill all workers

pkill -f worker-0.0.1-SNAPSHOT.jar

4. Up 3 workers:

for p in 8020 8021 8022; do java -jar build/libs/worker-0.0.1-SNAPSHOT.jar --server.port=$p > worker-$p.log 2>&1 & done

Scenario 4

1. Create modified version of B that immediately throws and error on every second task it receives.

2. Ensure that all of the tasks are processed by Workers, including those that initially failed (you can increase the number of deployed Workers if needed)

Reproducing

1. Change in Worker application.yml setting worker.handling-ms: 8000 and worker.fail-every-second-task: true
2. Clean & Build in gradle
3. Kill all workers

pkill -f worker-0.0.1-SNAPSHOT.jar

4. Up 3 workers:

for p in 8021 8022 8023; do java -jar build/libs/worker-0.0.1-SNAPSHOT.jar --server.port=$p > worker-$p.log 2>&1 & done