[jira] [Created] (KAFKA-15771) ProduceRequest#partitionSizes() is not an atomic operation
Xiaobing Fang created KAFKA-15771: - Summary: ProduceRequest#partitionSizes() is not an atomic operation Key: KAFKA-15771 URL: https://issues.apache.org/jira/browse/KAFKA-15771 Project: Kafka Issue Type: Bug Components: producer Reporter: Xiaobing Fang Encountered a concurrency issue in method ProduceRequest#partitionSizes() while developing with Kafka. When both Thread 1 and Thread 2 concurrently call method ProduceRequest#partitionSizes(), Thread 2 may receive an incomplete or empty result if Thread 1 is still in the process of initializing partitionSizes. This is an incorrect state. the code to ensure that Thread 2 obtains the final state rather than an intermediate one. -- This message was sent by Atlassian Jira (v8.20.10#820010)
[jira] [Created] (KAFKA-15136) The inspection of field allowAutoTopicCreation field in MetadataRequest is unreasonable
Xiaobing Fang created KAFKA-15136:
-
Summary: The inspection of field allowAutoTopicCreation field in
MetadataRequest is unreasonable
Key: KAFKA-15136
URL: https://issues.apache.org/jira/browse/KAFKA-15136
Project: Kafka
Issue Type: Improvement
Components: clients
Reporter: Xiaobing Fang
```java
if (!data.allowAutoTopicCreation() && version < 4)
throw new UnsupportedVersionException("MetadataRequest versions older than 4
don't support the " +
"allowAutoTopicCreation field");
```
Background:
Based on my understanding, the code in MetadataRequest is intended to ignore
the allowAutoTopicCreation field when version≤4. However, if kafka server is
configured with "auto.create.topics.enable"=false and the client sets
allowAutoTopicCreation=false, UnsupportedVersionExceptionwill be thrown, which
is not expected.
Issues:
# `KafkaAdminClient#handleDescribeTopicsByNames()` throws an exception when
sending MetadataRequest to a lower version server, making it unusable.
# MetadataRequestTest avoids this issue by setting
`allowAutoTopicCreation=true` in tests for version≤4, but this is not
reasonable. And the comments in
[testAutoTopicCreation|https://github.com/apache/kafka/blob/1f4cbc5d53259031123b6e9e6bb9a5bbe1e084e8/core/src/test/scala/unit/kafka/server/MetadataRequestTest.scala#L146]
may also be problematic.
Solution:
# Remove the checking code in MetadataRequest.
# Add a field `hasSetAllowAutoTopicCreation` in MetadataRequest. Only throw an
exception when `version≤4` and `hasSetAllowAutoTopicCreation=true`, without
considering the value of allowAutoTopicCreation field.
If there is indeed an issue, I can work on fixing it. Looking forward to your
reply.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
[jira] [Created] (KAFKA-13752) Using `equals` instead of `==` when Uuid compare in Java
Xiaobing Fang created KAFKA-13752: - Summary: Using `equals` instead of `==` when Uuid compare in Java Key: KAFKA-13752 URL: https://issues.apache.org/jira/browse/KAFKA-13752 Project: Kafka Issue Type: Improvement Components: clients Reporter: Xiaobing Fang `Uuid.ZERO_UUID == new Uuid(0L, 0L)` is true in scala, but in java is false. -- This message was sent by Atlassian Jira (v8.20.1#820001)
[jira] [Created] (KAFKA-13424) Redundant cleanup operations in TopicChangeHandler or TopicDeletionManager
Xiaobing Fang created KAFKA-13424: - Summary: Redundant cleanup operations in TopicChangeHandler or TopicDeletionManager Key: KAFKA-13424 URL: https://issues.apache.org/jira/browse/KAFKA-13424 Project: Kafka Issue Type: Improvement Components: controller Affects Versions: 2.8.0 Reporter: Xiaobing Fang In `TopicDeletionManager.completeDeleteTopic()`, `controllerContext.removeTopic(topic)` will be done after deleting topic znode in `client.deleteTopic(topic, controllerContext.epochZkVersion)`. This will trigger `TopicChangeHandler`, and process `controllerContext.removeTopic` aggine in `processTopicChange`. It is equivalent to deleting Topic once and executing `controllerContext.removeTopic` twice. Is this the expected situation？ -- This message was sent by Atlassian Jira (v8.3.4#803005)
