CHANGELOG.md

kafka-node CHANGELOG

2018-09-11, Version 3.0.1

• Fixed issue with new topic/partition detection when topic contains dots #1076
• Using double ended queue for message buffer in ConsumerGroupStream and ConsumerStream #1067
• Fixed issue in ConsumerGroupStream where forcing a commit breaks auto commit. #1066

2018-09-05, Version 3.0.0

• Added TS definitions #959 #965 #1022 #965 #1049 #965
• Fixed issue where ConsumerGroup refresh metadata errors weren't being correctly emitted as 'error' #971
• Fixed issue where HighLevelConsumer throw FailedToRebalanceConsumerError: NODE_EXISTS when rebalancing #981
• Fixed Admin.listGroups error handling #985
• Added support to create topics (specifying partitions and replication) using the Admin API (supported in Kafka v0.10 only) #958
• Updated client to remove unnecessary zookeeper reconnect #1011
• Remove noisy debug logs #1009 #1033
• Added SASL/PLAIN Authentication #923 #1033
• Fixed issue where closing a client can still refreshBrokerMetadata #1048
• Fixed empty messages caused by decoding partial messages from 2.0 broker #1050
• Fix consumerGroup removeTopics not able to remove topics #1012
• Added more descriptive errors as to why broker is not available #1023
• Fix Admin API describe groups error handling #1010
• ConsumerGroup leader will now monitors metadata for new topics/partitions. Once change has been detected it will leave and join the group to trigger a rebalance. This interval is controlled by topicPartitionCheckInterval option (defaults to every 30 seconds) #1057

BREAKING CHANGES

• Dropped support for node 4
• KafkaClient createTopics method drops the second async parameter since it uses the Admin API to create Topics

2018-04-27, Version 2.6.1

• Fix issue where duplicate messages are received when connecting to multiple brokers (restored dedicated consumer socket) #956

2018-04-24, Version 2.6.0

• Fix issue during the initial connection phase can end prematurely when metadata request failed #920
• Add addTopics method to the ConsumerGroup #914
• Fix issue where yielding a result in onRebalance in ConsumerGroup leads to an exception being thrown #922
• Add support for Kafka Fetch versions 1 and 2 this enables produced timestamps to be read from the Consumer (Kafka 0.10+). Fetches also now share socket to the broker with other kafka requests (previously fetches were on a dedicated socket) #871
• Add support to auto commit on first join ConsumerGroup configured using commitOffsetsOnFirstJoin #897
• Add support for Buffers as keys, which is useful for Avro encoded keys #932

2018-04-09, Version 2.5.0

• Explicitly cast key to string in hashCode function for KeyedPartitioner #870
• For consumer fetch loop we now clear socket.waiting before invoking callbacks #819
• Add Support for IPv6 #818
• Clear internal topicPayload array if no topic partitions are assigned to the ConsumerGroup #888
• Fix Stale Commit Queue for ConsumerGroupStream #891
• For rebalance case ConsumerGroup will try to commit before joining a group. Added onRebalance callback for manual commit users #889

2018-02-06, Version 2.4.1

• Fix issue where error callbacks for broker requests are not called when connection is closed #863

2018-02-06, Version 2.4.0

• Add compability to Client with broker configurations that separates external and internal traffic #860
• Fix issue where updateMetadata() wipes out entire topic metadata when only updating specific topics #857

2018-01-16, Version 2.3.2

• Fix issue with ConsumerGroupStream where lag stays at one and resuming the consumer re-reads the last read message #850

2018-01-7, Version 2.3.1

• Fix consumer example #842
• Fix issue where ConsumerGroupStream will autoCommit when the stream is explicitly closed #843
• Fix missing highWaterOffset using compression #821
• Fix Snappy buffer.from error using node 4 #827
• Fix fromOffset option not working in ConsumerStream #794
• Fixed issue where producer send failures failed to request refresh of the metadata #810

2017-11-17, Version 2.3.0

• Add support for ListGroups and DescribeGroups protocol. They can be used through the new Admin interface #770
• Fixed missing callback call in commit method of ConsumerGroupStream #776
• Flush queued commits of ConsumerGroupStream on tail end of auto-commit time-out #775

2017-09-11, Version 2.2.3

• Improved recovery of ConsumerGroup from broker down and network issues #758
• Upgrade to snappy 3 #760

2017-08-31, Version 2.2.2

• Fix issue where connections disconnected from being idle will never reinitialize when using KafkaClient #752
• Fix callback of producer send never being called when connection loss occurs with Broker using KafkaClient #751

2017-08-21, Version 2.2.1

• Fix duplicate messages emitted when using KafkaClient and consuming topic/partition that span multiple brokers #747
• Instead of failing silently consumers now emits a error when trying to consume messages that exceeds fetchMaxBytes #744

2017-08-08, Version 2.2.0

• Fix decoding of messages produced using KafkaClient with timestamps #736
• Add Writable stream ProducerStream #734

2017-08-03, Version 2.1.0

• Add two Readable streams ConsumerGroupStream and ConsumerStream #732
• Add V1 and V2 Produce request (adds client timestamps) to KafkaClient only supported in kafka v0.9 and v0.10 #730

2017-07-13, Version 2.0.1

• Fix unreferenced method call in Client/KafkaClient #708

2017-07-08, Version 2.0.0

• Fix out of range error causing idle consumers to crash. This was a performance improvement made in 1.6.1 and reverted in 1.6.2 #672
• Add KafkaClient a version of Client that connects directly to Kafka brokers instead of zookeeper for discovery. #691

Producer/Consumer Key Changes #704

• Fix issue where a specified key attribute in producer payload was missing from the message
• Fix issue where a falsey key value (such as an empty string or 0) was not able to be saved into the message. null or undefined key value will return as null
• Fix issue where key of Buffer value was not able to be saved into the message
• BREAKING CHANGE The key is decoded as a string by default. Previously was a Buffer. The preferred encoding for the key can be defined by the keyEncoding option on any of the consumers and will fallback to encoding if omitted

2017-05-08, Version 1.6.2

• Reverting performance changes using BufferList since it's causing idle consumers to crash #670

2017-05-04, Version 1.6.1

• Fix Offset calling .fetchOffset methods not yielding callback when the topic does not exist #662
• Improved performance of Client using BufferList instead of Buffer.slice #654

2017-03-16, Version 1.6.0

• Add ability to provide your own custom partitioner implementation (see docs for Producer) #625
• Create topics will yield with topics created #618

2017-02-24, Version 1.5.0

• Added highWaterOffset attribute to message and done results to help keep track consumer status #610
• Fixed potential missing commits in HighLevelCosumer and ConsumerGroup #613

2017-02-13, Version 1.4.0

• Logging strategy should be configurable so that kafka-node can be integrated into applications more easily. #597
• Published NPM package should not contain testing and local development artifacts. #598

2017-02-01, Version 1.3.4

• Producers should better recover from brokers going offline and coming back #580

2017-01-31, Version 1.3.3

• Fix issue where fetchEarliestOffsets using Offset could fail #572

2017-01-25, Version 1.3.2

• Fix potential stalled ConsumerGroup when a kafka broker leaves or comes back online #574
• Reduce calls to commit when consumer is idle #568
• Update lodash 4 #565

2017-01-17, Version 1.3.1

• Fix consumer group not reconnecting when a broker comes back online #563
• Removed an non consumer group error from consumer group error list #562

2017-01-12, Version 1.3.0

• Add fetchEarliestOffsets to Offset #544
• Fix issue where consumer heartbeat timeout was not triggering a retry #559

2017-01-12, Version 1.2.1

• Fix potential issue where long running consumers (includes ConsumerGroup, HighLevelConsumer, and Consumer) could throw out of bounds exception during the fetch loop #556

2017-01-11, Version 1.2.0

Consumer Group Changes

• Fix issue where an error in a leaving group during close will prevent a consumer from being closed. #551
• Fix issue with heartbeat where consumer will continue to send heartbeats at regular intervals even when previous ones did not resolve #547
• Add ability for ConsumerGroup to recover from stale offsets. This is configured by the new outOfRangeOffset option which takes same values as fromOffset. #553

2017-01-04, Version 1.1.0

• Fixed issue with unhandled error while using Offset #543
• HighLevelConsumer now allows a configurable retry options used for rebalancing this is found under the rebalanceRetry option key #542
• Fixed issue in HighLevelConsumer where an offset of 0 for new topics was not committed #529
• Upgraded nested-error-stacks dependency for node 7 compatibility #540

2016-11-18, Version 1.0.7

• Fix issue where createTopics using the async set to false was not synchronous. #519

  NOTE: The behavior now is if async is true the callback is not actually called until all the topics are confirmed to have been created by Kafka. The previous behavior the callback would be called after the first request (which does not guarantee the topics have been created). This wasn't consistent with what the doc said.

• Fix issue where messages are lost when sending a batch of keyed messages using the highLevelPartitioner #521

• Upgrade UUID package #520

• Check for in loops using hasOwnProperty to defend against insane libraries that update the prototype of Object #485

• Refactor HighLevelProducer and Producer #508

2016-11-15, Version 1.0.6

• Fix leave group exception that can occur in Consumer Group #513

2016-11-03, Version 1.0.5

• Update doc added how to list all topics #503
• Fix uncaught exceptions that can occur when using ConsumerGroup #505

2016-11-01, Version 1.0.4

• Fix issue where an exception is thrown in client.brokerForLeader when connection with broker is lost in ConsumerGroup it should retry instead #498
• Fix issue where invalid characters could be used in createTopics call #495 #492

2016-10-24, Version 1.0.3

• Fix issue in Consumer Group where using the migrator with no previous HLC offsets will set initial offsets to 0 instead of the offsets provided in "fromOfset" feature #493

2016-10-22, Version 1.0.2

• Fix issue in Consumer Group where using the migrator with no previous HLC offsets will set initial offsets to -1 #490

2016-10-13, Version 1.0.1

• Fix missing support in Consumer Group for fromOffset using earliest and none #483

2016-10-10, Version 1.0.0

Major version change since we're dropping support for Node older than v4.

Added a new Consumer called Consumer Group this is supported in Kafka 0.9 and greater. #477

• Add group membership API protocols
• Add consumerGroup roundrobin assignment
• Add documentation for ConsumerGroup
• Implemented range assignment strategy
• New consumer group should still emit rebalanced and rebalancing events
• Refactor HLC rebalancing test to run against ConsumerGroup as well
• Add migration from existing HLC offsets in ConsumerGroups
• Implement rolling migration of HLC offsets for consumer groups
• Dropping support for node 0.12 and only supporting 4 and above. Updating to version 1.0.0
• Upgrade mocha
• Migrator: cover the cases of slow rebalances by adding verification by checking check four times make sure the zookeeper ownership did not come back

2016-10-03, Version 0.5.9

• Fix issue with highLevelConsumers and how consumer groups react to zookeeper redeploys creating a lot of [NODE_EXISTS] errors #472
• Removed docker-machine support for tests #474
• Minor fixes and additions to doc #475 #471

2016-09-12, Version 0.5.8

• Fix duplicate messages consumed on startup this was triggered by unnecessary rebalance (versions affected: v0.5.4 to v0.5.7) #465

2016-09-07, Version 0.5.7

• Fix regression when calling consumer's setOffset fails to set the given offset #457
• Improved zookeeeper connection loss recovery to verify consumer is still registered #458

2016-08-17, Version 0.5.6

• Fix older version of node issue introduced in last version #447

2016-08-11, Version 0.5.5

• Updated doc #443
• Validate topic's partition value to be a number #442
• Fixed issue where module was relying on deprecated kafka configs (especially broken when SSL is enabled) #427

2016-08-09, Version 0.5.4

• Fix lost client options when creating a Client w/o the new operator #437
• Fix issue rebalances can be missed when an event occurs during rebalance #435
• Fix issue where changes to a topic's number of partitions should trigger a rebalance #430
• Added coverall coverage and additional tests #433 and #432

2016-08-05, Version 0.5.3

• Fix for some long standing high-level consumer rebalance issues: #423
  • Fixed issue where consumers who weren't assigned a partition would never rebalance... ever
  • Fixed issue where calling close consumer did not force the consumer to leave the consumer group and so locks the partition(s) for (the default) of 30000ms
  • Fixed issue where consumers who weren't assigned a partition never emitted a rebalance event after rebalancing
  • Additional cases to be addressed in the next release:
    • If consumer joins or leaves a group during another consumers rebalance then that rebalancing consumer may miss owning partitions
    • Changes to a topic's partition does not trigger a rebalance
• Update doc to warn about possible data loss with way the data is formated when sending with the producer #425
• Added code coverage and additional high-level producer tests #422

2016-07-29, Version 0.5.2

• Fix TypeError: Cannot read property 'sslHost' of undefined #417

2016-07-27, Version 0.5.1

• Prevent fetch before offset's are updated before a rebalance #402
• Add validation to groupId and clientId #405
• Removed unused autoCommitMsgCount config option #406
• Fixed issue where reconnecting brokers emits a connect event instead of reconnect. #413
• Fixed uncaught exception where sslHost or host of undefined is accessed. #413

2016-07-14, Version 0.5.0

• Fix minimatch vulnerability by upgrading snappy to v5.0.5 #400
• Added ESLint to codebase #392
• Added ability to make SSL connections to brokers #383 (Kafka 0.9+ only)

2016-06-27, Version 0.4.0

• Update test to run against docker #387
• Fix missing npm license field warning #386
• Recreate broker sockets to work around nodejs socket issue #4417 #385
• Fixes #319 UnknownTopicOrPartition error when reassigning topic #384
• Add offset#getLatestOffsets function to get all the latest offsets from a group of topics populating those topics partitions #372

2016-05-27, Version 0.3.3

• Fix type error while producing messages #360
• Update README #307 and #352
• Add contributing guidelines #346
• Make Snappy an optional dependency #347

2016-02-21, Version 0.3.2

• Fix client socket when closing and error handling #314
• Make commit() handle case when only callback is passed #306
• Fix typo in offset.js #304

2016-01-09, Version 0.3.1

• Buffer batch for async producers #262

2016-01-08, Version 0.3.0

• Add partitions to producer #260

2015-05-11, Version 0.2.27

• Deps: upgrade snappy to 3.2.0
• Zookeeper#listConsumers: ignore error when there is no such node in zookeeper

2015-04-23, Version 0.2.26

• Fix: add callback to consumer.autoCommit method #198
• Emit an error when there is a problem with the socket connection to the kafka broker #196
• Fix: emit the error instead of silent swallow it #193
• Typo in error message #189

2015-04-01, Version 0.2.25

• Producer support requireAcks option #187
• Update examples #185

2015-03-20, Version 0.2.24

• Bump deps
• Refresh metadata after auto rebalance among brokers #180
• Initialize partition owner with consumerId #178

2015-03-17, Version 0.2.23

• Fix #175: Refresh topic metadata in Producer when broker change
• Refactor Client#refreshMetadata method
• Add the missing semicolons, no offense, just keep style.
• Fix #170: In case of offsetOutOfRange, the consumer should be paused.
• Fix #169: When paused why try to fetch every 1000 ms?
• Ref: remove unused variables.