v0.17.0

Implemented:

• #167 Add support for SSL/TLS connections in both gRPC and HTTP services. Set both a certificate and key file in the configuration to take advantage of this feature. (Requested in #160).
• Default client ID was changed to kp_<hostname>_<contain-id> if running in a docker container, otherwise kp_<hostname>_<pid>. If hostname cannot be retrieved, then the client ID is kp_<random-token>.

Fixed:

• #168 Configured addresses for HTTP, gRPC and Unix were not being respected.

v0.16.0

Implemented:

• #156 #151 Added formal support for Kafka versions up to v2.1.0.
• #155 When the last group member leaves a consumer group, all group records are removed from ZooKeeper. Group records were never deleted from ZooKeeper before, that could cause issues with disposable groups.
• #144 Added support for specifying and receiving Kafka record headers over both the HTTP and gRPC produce and consume interfaces.
• #146 Added new configuration flags for the producer partitioner, broker-enforced produce timeout, and network-level timeouts.

Fixed:

• #54 Rebalancing fails due to non existent topic.

v0.15.0

Implemented:

• #135 Added support for Kafka v0.11.0.0 - v1.0.0
• consumer.rebalance_timeout was removed, so rebalancing is triggered as soon as membership status of a consumer group or subscription of a consumer group member changes.
• #138 Pending offsets are committed to Kafka faster on rebalancing. That could take up to consumer.offsets_commit_interval before, but now it happens as soon as possible.

Fixed:

• #120 Consumption from atopic stopped for a group.
• #123 Inexplicable offset manager timeouts.
• #124 Subscription to a topic fails indefinitely after ZooKeeper connection loss.
• #140 Offset manager keeps losing connection with a broker.

v0.14.0

Implemented:

• Added HTTP API endpoints to list topics optionally with partitions and
  configuration.
• Ack timeout can now be greater than subscription timeout. So in absence of
  incoming consume requests for a topic, rebalancing may be delayed until
  ack timeout expires for all offered messages. If a new request comes while
  waiting for ack timeout to expire, then the subscription timer is reset for
  the topic.
• Posts can now be performed with content type x-www-form-urlencoded, in that
  case message should be passed in the msg form parameter.
• Structural logging with sirupsen/logrus and mailgun/logrus-hooks/kafkahook.
• Support for Kafka version 0.10.2.0.

Fixed:

• Explicit acks were fixed for HTTP API. Turned out values of noAck,
  ackPartition, and ackOffset parameters had been ignored.
• A race was found in tests that if a request comes while a topic expiration
  is in progress consumption from the topic may never resume.
• #100 Consumption from a
  partition stops if the segment that we read from expires.

v0.13.0

Major update brings At-Least-One delivery guarantee via synch production and explicit acknowledgement of consumed messages.

• Changelog

v0.12.0

Support for gRPC producer/consumer API and more

• Getting Started
• Changelog

v0.11.1

Bug fix release.

• Getting Started
• Changelog

v0.11.0

Major implementation refactoring to make internal components more prominent and connections between them easier to track. Besides several bugs have been detected and fixed during the refactoring.

• Getting Started
• Changelog

v0.10.1

This release fixes a critical issue that existed ever since the first public release (v0.7.0).

• Getting Started
• Changelog

v0.10.0

This release adds support for Kafka 0.9.0.x. Note that consumer group management is still performed via ZooKeeper.

• Getting Started
• Changelog