sarama-cluster-tester produce writes 1000 messages to the configured topic.
sarama-cluster-tester stream prints all of the messsages.
sarama-clsuter-tester consume GROUP_ID consumes the messages as a consumer group.
The producer writes 1000 messages to the topic and then exits.
This seeds the topic for use by consumer tests.
Each consumer process starts 5 consumer groups.
Four of the consumers are well-behaved, meaning they:
- consume messages until they're instructed to shut down
- deliver messages to the reporting routine
- acknowledge messages after reporting
The other consumer is like the well-behaved ones, except:
- it consumes 10 messages and then shuts down
- it does not acknowledge messages after they're reported
A normal run of the consumer at a scale of 1, then, will process around 800 messages (4/5 consumers), then pause for a bit while rebalancing is triggered due to the death of the bad consumer, and then process the remaining messages.
At the end, it reports:
2017/09/08 19:51:17 at=result found=1000 seen=1010
Meaning that all 1000 messages were observed, and 10 were re-transmitted after the bad consumer died.
Running the consumer at a scale of 2 behaves similarly. At the end, the two consumers report results like:
app[consumer.1]: 2017/09/08 19:54:18 at=result found=536 seen=536
app[consumer.2]: 2017/09/08 19:54:18 at=result found=484 seen=484
---
app[consumer.2]: 2017/09/08 19:56:00 at=result found=652 seen=662
app[consumer.1]: 2017/09/08 19:56:00 at=result found=352 seen=358
The number of results found is always >= 1000.