Resolves https://github.com/spring-cloud/spring-cloud-stream-binder-kafka/issues/1081
The wrong consequence from the absense of scheduler.shutdown():
1) First of all we created the pool with 1 thread.
2) After we lost the reference on it and created the pool with 2 threads.
3) But, the first pool is not yet collected by GC and now you have 3 threads together. And so on.
Each thread does nothing, but it takes system memory and takes part in the scheduling process.
After 30 topics, for example, we potentially have (30+1)*15=465 threads.
It is already serious additional load on the switching contexts and the native memory.
Removed waiting after scheduler stop request.
checkstyle fixes
Going forward, binder will use the default ack on error settings
in Spring Kafka, which is true by default. If the applicaitons
need to change this behavior, then a custom error handler must
be provided through ListenerContainerCustomizer.
Resolves https://github.com/spring-cloud/spring-cloud-stream-binder-kafka/issues/1079
This comnmit effectively removes changes introduced by a4ad9e2c0b.
Removing spring.factories mechanism of registering binders in preference to using spring.binders.
This commit effectively reverts 4e6881830a
Remove spring.factories changes introduced in the commit mentiond above.
Revert the binder to use spring.binders mechanism to register binder configurations.
When destination-is-pattern property is enabled and native deserialization
is used, then Kafka Streams binder does not use the correct Serde types.
Fixing this issue by providing the proper Serdes to the StreamsBuilder.
Resolves https://github.com/spring-cloud/spring-cloud-stream-binder-kafka/issues/1051
When functional beans are discovered from libraries in the classpath,
it causes issues when Kafka Streams functions are scanned and bootstrapped.
Binder expects the users to provide function definition property although the
application does not directly include these functional beans or is aware of it.
Fixing this issue by excluding non kafka streams function from scanning.
Null check around adding micrometer listener to the StreamsBuilder.
Addressing issues raised by the comments here:
https://github.com/spring-cloud/spring-cloud-stream-binder-kafka/issues/1030#issuecomment-804039087
Resolves https://github.com/spring-cloud/spring-cloud-stream-binder-kafka/issues/1046
When using ack mode `MANUAL_IMMEDIATE`, we must acknowledge the delivery on
the container thread, otherwise the commit will be queued for later processing,
possibly causing out of order commits, incorrectly reducing the committed offset.
Wait for the send to complete before acknowledging; use a timeout slightly larger
than the configured producer delivery timeout to avoid premature timeouts.
Tested with user-provided test case.
Fix missing timeout buffer.
More context for this issue: https://gitter.im/spring-cloud/spring-cloud-stream?at=6050f43595e23446e43cacd1
StreamsListener (for Micrometer) is now directly available
in Spring Kafka starting from 2.5.3 (as KafkaStreamsMicrometerListener).
Removing the temporory interface added in the binder.
Addressing PR review comments.
Modifying tests to verify.
Resolves https://github.com/spring-cloud/spring-cloud-stream-binder-kafka/issues/1040
* Kafka Streams function detection improvements
Allow Kafka Streams functions defined as Component beans
to be candidates for establishing bindings. Currently, Kafka Streams
functions need to be written as functional beans using @Bean.
Adding this improvement so that if applications prefer to write
the business logic using @Component, then it is possible to do so.
Adding test cases to verify the behavior.
Resolves https://github.com/spring-cloud/spring-cloud-stream-binder-kafka/issues/1030
* Kafka Streams functions and bean name overriding
Whn Kafka Streams function bean names are overridden,
there is an issue with scanning it properly for binding.
Addressing this issue.
* Adding docs for Component based model
* Addressing PR review comments
When using multi binder setup in Kafka Streams binder, there is an issue
in which the binder health indicator is not getting bootstapped due to a
ConditionalOnBean is unable to find a match for KafkaStreamsRegistry bean.
Fixing this issue by using an ObjectProvider instead of ConditionalOnBean.
Resolves https://github.com/spring-cloud/spring-cloud-stream-binder-kafka/issues/1006
