[KAFKA-14271] Topic recreation fails in KRaft mode when topic contains collidable characters

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Duplicate
• Affects Version/s: 3.3.0
• Component/s: kraft
• Assignee: Unassigned
• Reporter: Jeffrey Tolar
• Created: 30/Sep/22 17:33
• Updated: 21/Dec/22 21:13
• Resolved: 12/Dec/22 23:59

Description

We recently updated one of our clusters from 3.2.0 to 3.3.0 (primarily to get the fix for KAFKA-13909). This cluster is running KRaft mode.

This is a cluster used for some integration tests - each test deletes the topics it uses before the test to ensure a clean slate for the test; the brokers get restarted in-between tests, but the broker data isn't deleted.

With 3.3.0, this semi-crashes Kafka. The brokers stay running, but the topic creation fails:

[2022-09-30 17:17:59,216] WARN [Controller 1] createTopics: failed with unknown server exception NoSuchElementException at epoch 1 in 601 us.  Renouncing leadership and reverting to the last committed offset 18. (org.apache.kafka.controller.QuorumController)
java.util.NoSuchElementException
        at org.apache.kafka.timeline.SnapshottableHashTable$CurrentIterator.next(SnapshottableHashTable.java:167)
        at org.apache.kafka.timeline.SnapshottableHashTable$CurrentIterator.next(SnapshottableHashTable.java:139)
        at org.apache.kafka.timeline.TimelineHashSet$ValueIterator.next(TimelineHashSet.java:120)
        at org.apache.kafka.controller.ReplicationControlManager.validateNewTopicNames(ReplicationControlManager.java:799)
        at org.apache.kafka.controller.ReplicationControlManager.createTopics(ReplicationControlManager.java:567)
        at org.apache.kafka.controller.QuorumController.lambda$createTopics$7(QuorumController.java:1832)
        at org.apache.kafka.controller.QuorumController$ControllerWriteEvent.run(QuorumController.java:767)
        at org.apache.kafka.queue.KafkaEventQueue$EventContext.run(KafkaEventQueue.java:121)
        at org.apache.kafka.queue.KafkaEventQueue$EventHandler.handleEvents(KafkaEventQueue.java:200)
        at org.apache.kafka.queue.KafkaEventQueue$EventHandler.run(KafkaEventQueue.java:173)
        at java.base/java.lang.Thread.run(Thread.java:829)

This appears to be because our topic names contain .'s. Here's a quick reproducer script:

#!/bin/bash

VERSION=3.3.0
TOPIC=$1

set -x

rm -rf -- kafka_2.13-${VERSION} kafka_2.13-${VERSION}.tgz /tmp/kraft-combined-logs
trap 'kill -- "-$$" && wait' EXIT

curl -O https://dlcdn.apache.org/kafka/$VERSION/kafka_2.13-${VERSION}.tgz

tar -xzf kafka_2.13-${VERSION}.tgz

cd kafka_2.13-${VERSION}

id=$(./bin/kafka-storage.sh random-uuid)
./bin/kafka-storage.sh format -t $id -c ./config/kraft/server.properties

./bin/kafka-server-start.sh config/kraft/server.properties > broker.log 2>&1 &
sleep 1

./bin/kafka-topics.sh --create --topic "$TOPIC" --bootstrap-server localhost:9092
./bin/kafka-topics.sh --delete --topic "$TOPIC" --bootstrap-server localhost:9092
./bin/kafka-topics.sh --create --topic "$TOPIC" --bootstrap-server localhost:9092
sleep 1

Running ./test-kafka.sh topic.with.dots exhibits the failure; using topicwithoutdots works as expected.

I'll attach the broker logs from each run.

Issue Links

• duplicates: KAFKA-14337 topic name with "." cannot be created after deletion