[CAMEL-18985] camel-kafka: messages are getting lost with "breakOnFirstError"

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.18.5, 3.19.0, 3.20.0, 3.20.1, 4.0-M2
• Fix Version/s: 3.18.7, 3.20.5, 3.21.0, 4.0-M3
• Component/s: camel-kafka
• Estimated Complexity: Unknown
• Assignee: Otavio Rodolfo Piske
• Reporter: Oleg Pershai
• Created: 30/Jan/23 01:59
• Updated: 26/Apr/23 15:31
• Resolved: 26/Apr/23 15:02

Description

Description:

Messages are getting lost with "breakOnFirstError=true" when processing of a particular message failed several times in a raw.

 

Configuration:

• autoCommitEnable=false
• allowManualCommit=true
• breakOnFirstError=true
• autoOffsetReset=earliest
• maxPollRecords is greater than one (e.g. 5 in this test)

 

Test Scenario:

1. inbound-topic contains 10 messages: 0,1,2,3,4,5,6,7,8,9
2. camel-route polls 5 messages: 0,1,2,3,4
3. camel-route successfully processes message=0 and manually commits offset
4. camel-route successfully processes message=1 and manually commits offset
5. camel-route fails with processing message=2 first time
6. breakOnFirstError=true causes camel-route to reconnect and poll from the committed offset
7. camel-route polls 5 messages: 2,3,4,5,6
8. camel-route fails with processing message=2 second time
9. ACTUAL: camel-route reconnects and polls messages 7,8,9 - as result messages 3,4,5,6 are never processed

EXPECTED: camel-route should do the following on the step 9:

• reconnect and poll the same 5 messages again: 2,3,4,5,6
• process message=2 third time (the test is configured to succeed on the 3rd attempt)
• continue with processing messages 3,4,5,6...

 

GitHub project reproducing the issue:

https://github.com/opershai/camel-kafka-lost-messages-demo 

Test-runs:

• 3.18.5 (failing): https://github.com/opershai/camel-kafka-lost-messages-demo/actions/runs/4039702564 
• 3.19.0 (failing): https://github.com/opershai/camel-kafka-lost-messages-demo/actions/runs/4039815226 
• 3.20.0 (failing): https://github.com/opershai/camel-kafka-lost-messages-demo/actions/runs/4039709753 
• 3.20.1 (failing): https://github.com/opershai/camel-kafka-lost-messages-demo/actions/runs/4039673222 
• 2.25.4 (passing): https://github.com/opershai/camel-kafka-lost-messages-demo/actions/runs/4057277373 

 

Impact:

It seems "breakOnFirstError" is not fully reliable in LTS versions of Camel Kafka component at the moment:

• 3.20.x - due to this bug
• 3.18.x - due to this bug
• 3.14.x - due to another bug https://issues.apache.org/jira/browse/CAMEL-17925 as fix was not back-ported to the 3.14.x

Issue Links

• links to: GitHub Pull Request #9932
• links to: GitHub Pull Request #9936
• links to: GitHub Pull Request #9947