1.0.66

What's Changed

• Remove waiting for connection to conductor during instrumented client creation by @kehuum in #201

Full Changelog: 1.0.65...1.0.66

Upgrade kafka to 2.0.0.37

kafka 2.0.0.37 Contains the revert of passthrough code

bump kafka to 2.0.0.35

Reverted the passthrough related code

bump kafka to 2.0.0.33

1.0.63

update kafka version to 2.0.0.33 (#194)

Bypass large message processing when passthrough enabled

Bypass large message processing in consumer when passthrough(enable.shallow.iterator) is enabled. This is for raw consumer mirroring.

Make li kafka producer flush behavior to be the same as open source producer

Make li kafka producer flush behavior to be the same as open source producer in terms of the following:

1. Allow flush after producer.close()
2. Throw original flush exception instead of wrapping it into illegal state exception.

bump kafka to 2.0.0.32

contains passthrough fix to test passthrough in BMM cert

Bump kafka to 2.0.0.29

Bump kafka to 2.0.0.29 to pick up oss fix for stale cluster metadata when removing all brokers from one cluster and add the broker(s) to another cluster.

Fix race condition in producer dual close

Fix race condition in producer dual close called from send callback.

Add new LiKafkaConsumerImpl metric - _offsetInvalidOrOutRangeCounter to estimate the potential data loss (#185)

Add new LiKafkaConsumerImpl metric - _offsetInvalidOrOutRangeCounter …

…to estimate the potential data loss (#185)

Added new metric( _offsetInvalidOrOutRangeCounter) into LiKafkaConsumerImpl to count number of OffsetOutOfRange exceptions under following cases with liclosest reset strategy.

1. fetchedOffset < Log Start Offset (LSO)
2. fetchedOffset <= Log End Offset (LEO)

This metric can be used as indicator to estimate the potential data loss upon the LiKafkaConsumer client reseting the offset to "liclosest"