[KAFKA-10353] Trogdor - Fix RoundTripWorker to not fail when the topic it's trying to create already exists

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Duplicate
• Assignee: Unassigned
• Reporter: Stanislav Kozlovski
• Created: 04/Aug/20 06:58
• Updated: 04/Aug/20 14:10
• Resolved: 04/Aug/20 14:09

Description

Trogdor's RoundTripWorker calls WorkerUtils#createTopics with a failOnCreate flag equal to true, making the code throw an exception if the topic already exists.

https://github.com/apache/kafka/blob/28b7d8e21656649fb09b09f9bacfe865b0ca133c/tools/src/main/java/org/apache/kafka/trogdor/workload/RoundTripWorker.java#L149

This is prone to race conditions when scheduling multiple workers to start at the same time - only one will succeed in creating a topic and running the test, while the rest will end up with a fatal error

This has also been seen to happen in the RoundTripFaultTest system test where a network exception can cause the CreateTopics request to reach Kafka but Trogdor retry it and hit a TopicAlreadyExists exception on the retry, failing the test.

Issue Links

• duplicates: KAFKA-10308 fix flaky core/round_trip_fault_test.py