Kafka-node is a Node.js client with Zookeeper integration for Apache Kafka 0.8.1 and later.
- Features
- Install Kafka
- API
- Troubleshooting / FAQ
- HighLevelProducer with KeyedPartitioner errors on first send
- How do I debug an issue?
- How do I get a list of all topics?
- For a new consumer how do I start consuming from the latest message in a partition?
- FailedToRebalanceConsumerError: Exception: NODE_EXISTS[-110]
- HighLevelConsumer does not consume on all partitions
- How to throttle messages / control the concurrency of processing messages
- How do I produce and consume binary data?
- What are these node-gyp and snappy errors?
- How do I configure the log output?
- Running Tests
- LICENSE - "MIT"
- Consumer and High Level Consumer
- Producer and High Level Producer
- Manage topic Offsets
- SSL connections to brokers (Kafka 0.9+)
- Consumer Groups managed by Kafka coordinator (Kafka 0.9+)
Follow the instructions on the Kafka wiki to build Kafka 0.8 and get a test broker up and running.
connectionString
: Zookeeper connection string, defaultlocalhost:2181/
clientId
: This is a user-supplied identifier for the client application, defaultkafka-node-client
zkOptions
: Object, Zookeeper options, see node-zookeeper-clientnoAckBatchOptions
: Object, when requireAcks is disabled on Producer side we can define the batch properties, 'noAckBatchSize' in bytes and 'noAckBatchAge' in milliseconds. The default value is{ noAckBatchSize: null, noAckBatchAge: null }
and it acts as if there was no batchsslOptions
: Object, options to be passed to the tls broker sockets, ex. { rejectUnauthorized: false } (Kafka +0.9)
Closes the connection to Zookeeper and the brokers so that the node process can exit gracefully.
cb
: Function, the callback
client
: client which keeps a connection with the Kafka server.options
: options for producer,
{
// Configuration for when to consider a message as acknowledged, default 1
requireAcks: 1,
// The amount of time in milliseconds to wait for all acks before considered, default 100ms
ackTimeoutMs: 100,
// Partitioner type (default = 0, random = 1, cyclic = 2, keyed = 3, custom = 4), default 0
partitionerType: 2
}
var kafka = require('kafka-node'),
Producer = kafka.Producer,
client = new kafka.Client(),
producer = new Producer(client);
ready
: this event is emitted when producer is ready to send messages.error
: this is the error event propagates from internal client, producer should always listen it.
payloads
: Array,array ofProduceRequest
,ProduceRequest
is a JSON object like:
{
topic: 'topicName',
messages: ['message body'], // multi messages should be a array, single message can be just a string or a KeyedMessage instance
key: 'theKey', // only needed when using keyed partitioner
partition: 0, // default 0
attributes: 2 // default: 0
}
cb
: Function, the callback
attributes
controls compression of the message set. It supports the following values:
0
: No compression1
: Compress using GZip2
: Compress using snappy
Example:
var kafka = require('kafka-node'),
Producer = kafka.Producer,
KeyedMessage = kafka.KeyedMessage,
client = new kafka.Client(),
producer = new Producer(client),
km = new KeyedMessage('key', 'message'),
payloads = [
{ topic: 'topic1', messages: 'hi', partition: 0 },
{ topic: 'topic2', messages: ['hello', 'world', km] }
];
producer.on('ready', function () {
producer.send(payloads, function (err, data) {
console.log(data);
});
});
producer.on('error', function (err) {})
⚠️ WARNING: Batch multiple messages of the same topic/partition together as an array on themessages
attribute otherwise you may lose messages!
This method is used to create topics on the Kafka server. It only works when auto.create.topics.enable
, on the Kafka server, is set to true. Our client simply sends a metadata request to the server which will auto create topics. When async
is set to false, this method does not return until all topics are created, otherwise it returns immediately.
topics
: Array, array of topicsasync
: Boolean, async or synccb
: Function, the callback
Example:
var kafka = require('kafka-node'),
Producer = kafka.Producer,
client = new kafka.Client(),
producer = new Producer(client);
// Create topics sync
producer.createTopics(['t','t1'], false, function (err, data) {
console.log(data);
});
// Create topics async
producer.createTopics(['t'], true, function (err, data) {});
producer.createTopics(['t'], function (err, data) {});// Simply omit 2nd arg
client
: client which keeps a connection with the Kafka server. Round-robins produce requests to the available topic partitionsoptions
: options for producer,
{
// Configuration for when to consider a message as acknowledged, default 1
requireAcks: 1,
// The amount of time in milliseconds to wait for all acks before considered, default 100ms
ackTimeoutMs: 100,
// Partitioner type (default = 0, random = 1, cyclic = 2, keyed = 3, custom = 4), default 2
partitionerType: 3
}
var kafka = require('kafka-node'),
HighLevelProducer = kafka.HighLevelProducer,
client = new kafka.Client(),
producer = new HighLevelProducer(client);
ready
: this event is emitted when producer is ready to send messages.error
: this is the error event propagates from internal client, producer should always listen it.
payloads
: Array,array ofProduceRequest
,ProduceRequest
is a JSON object like:
{
topic: 'topicName',
messages: ['message body'], // multi messages should be a array, single message can be just a string,
key: 'theKey', // only needed when using keyed partitioner
attributes: 1
}
cb
: Function, the callback
Example:
var kafka = require('kafka-node'),
HighLevelProducer = kafka.HighLevelProducer,
client = new kafka.Client(),
producer = new HighLevelProducer(client),
payloads = [
{ topic: 'topic1', messages: 'hi' },
{ topic: 'topic2', messages: ['hello', 'world'] }
];
producer.on('ready', function () {
producer.send(payloads, function (err, data) {
console.log(data);
});
});
⚠️ WARNING: Batch multiple messages of the same topic/partition together as an array on themessages
attribute otherwise you may lose messages!
This method is used to create topics on the Kafka server. It only work when auto.create.topics.enable
, on the Kafka server, is set to true. Our client simply sends a metadata request to the server which will auto create topics. When async
is set to false, this method does not return until all topics are created, otherwise it returns immediately.
topics
: Array,array of topicsasync
: Boolean,async or synccb
: Function,the callback
Example:
var kafka = require('kafka-node'),
HighLevelProducer = kafka.HighLevelProducer,
client = new kafka.Client(),
producer = new HighLevelProducer(client);
// Create topics sync
producer.createTopics(['t','t1'], false, function (err, data) {
console.log(data);
});
// Create topics async
producer.createTopics(['t'], true, function (err, data) {});
producer.createTopics(['t'], function (err, data) {});// Simply omit 2nd arg
client
: client which keeps a connection with the Kafka server. Note: it's recommend that create new client for different consumers.payloads
: Array,array ofFetchRequest
,FetchRequest
is a JSON object like:
{
topic: 'topicName',
offset: 0, //default 0
}
options
: options for consumer,
{
groupId: 'kafka-node-group',//consumer group id, default `kafka-node-group`
// Auto commit config
autoCommit: true,
autoCommitIntervalMs: 5000,
// The max wait time is the maximum amount of time in milliseconds to block waiting if insufficient data is available at the time the request is issued, default 100ms
fetchMaxWaitMs: 100,
// This is the minimum number of bytes of messages that must be available to give a response, default 1 byte
fetchMinBytes: 1,
// The maximum bytes to include in the message set for this partition. This helps bound the size of the response.
fetchMaxBytes: 1024 * 1024,
// If set true, consumer will fetch message from the given offset in the payloads
fromOffset: false,
// If set to 'buffer', values will be returned as raw buffer objects.
encoding: 'utf8'
}
Example:
var kafka = require('kafka-node'),
Consumer = kafka.Consumer,
client = new kafka.Client(),
consumer = new Consumer(
client,
[
{ topic: 't', partition: 0 }, { topic: 't1', partition: 1 }
],
{
autoCommit: false
}
);
By default, we will consume messages from the last committed offset of the current group
onMessage
: Function, callback when new message comes
Example:
consumer.on('message', function (message) {
console.log(message);
});
Add topics to current consumer, if any topic to be added not exists, return error
topics
: Array, array of topics to addcb
: Function,the callbackfromOffset
: Boolean, if true, the consumer will fetch message from the specified offset, otherwise it will fetch message from the last commited offset of the topic.
Example:
consumer.addTopics(['t1', 't2'], function (err, added) {
});
or
consumer.addTopics([{ topic: 't1', offset: 10 }], function (err, added) {
}, true);
topics
: Array, array of topics to removecb
: Function, the callback
Example:
consumer.removeTopics(['t1', 't2'], function (err, removed) {
});
Commit offset of the current topics manually, this method should be called when a consumer leaves
cb
: Function, the callback
Example:
consumer.commit(function(err, data) {
});
Set offset of the given topic
-
topic
: String -
partition
: Number -
offset
: Number
Example:
consumer.setOffset('topic', 0, 0);
Pause the consumer. Calling pause
does not automatically stop messages from being emitted. This is because pause just stops the kafka consumer fetch loop. Each iteration of the fetch loop can obtain a batch of messages (limited by fetchMaxBytes
).
Resume the consumer. Resumes the fetch loop.
Pause specify topics
consumer.pauseTopics([
'topic1',
{ topic: 'topic2', partition: 0 }
]);
Resume specify topics
consumer.resumeTopics([
'topic1',
{ topic: 'topic2', partition: 0 }
]);
force
: Boolean, if set to true, it forces the consumer to commit the current offset before closing, defaultfalse
Example
consumer.close(true, cb);
consumer.close(cb); //force is disabled
client
: client which keeps a connection with the Kafka server.payloads
: Array,array ofFetchRequest
,FetchRequest
is a JSON object like:
{
topic: 'topicName'
}
options
: options for consumer,
{
// Consumer group id, default `kafka-node-group`
groupId: 'kafka-node-group',
// Optional consumer id, defaults to groupId + uuid
id: 'my-consumer-id',
// Auto commit config
autoCommit: true,
autoCommitIntervalMs: 5000,
// The max wait time is the maximum amount of time in milliseconds to block waiting if insufficient data is available at the time the request is issued, default 100ms
fetchMaxWaitMs: 100,
// This is the minimum number of bytes of messages that must be available to give a response, default 1 byte
fetchMinBytes: 1,
// The maximum bytes to include in the message set for this partition. This helps bound the size of the response.
fetchMaxBytes: 1024 * 1024,
// If set true, consumer will fetch message from the given offset in the payloads
fromOffset: false,
// If set to 'buffer', values will be returned as raw buffer objects.
encoding: 'utf8'
}
Example:
var kafka = require('kafka-node'),
HighLevelConsumer = kafka.HighLevelConsumer,
client = new kafka.Client(),
consumer = new HighLevelConsumer(
client,
[
{ topic: 't' }, { topic: 't1' }
],
{
groupId: 'my-group'
}
);
By default, we will consume messages from the last committed offset of the current group
onMessage
: Function, callback when new message comes
Example:
consumer.on('message', function (message) {
console.log(message);
});
Add topics to current consumer, if any topic to be added not exists, return error
topics
: Array, array of topics to addcb
: Function,the callbackfromOffset
: Boolean, if true, the consumer will fetch message from the specified offset, otherwise it will fetch message from the last commited offset of the topic.
Example:
consumer.addTopics(['t1', 't2'], function (err, added) {
});
or
consumer.addTopics([{ topic: 't1', offset: 10 }], function (err, added) {
}, true);
topics
: Array, array of topics to removecb
: Function, the callback
Example:
consumer.removeTopics(['t1', 't2'], function (err, removed) {
});
Commit offset of the current topics manually, this method should be called when a consumer leaves
cb
: Function, the callback
Example:
consumer.commit(function(err, data) {
});
Set offset of the given topic
-
topic
: String -
partition
: Number -
offset
: Number
Example:
consumer.setOffset('topic', 0, 0);
Pause the consumer. Calling pause
does not automatically stop messages from being emitted. This is because pause just stops the kafka consumer fetch loop. Each iteration of the fetch loop can obtain a batch of messages (limited by fetchMaxBytes
).
Resume the consumer. Resumes the fetch loop.
force
: Boolean, if set to true, it forces the consumer to commit the current offset before closing, defaultfalse
Example:
consumer.close(true, cb);
consumer.close(cb); //force is disabled
The new consumer group uses Kafka broker coordinators instead of Zookeeper to manage consumer groups. This is supported in Kafka version 0.9 and above only.
API is very similar to HighLevelConsumer
since it extends directly from HLC so many of the same options will apply with some exceptions noted below:
- In an effort to make the API simpler you no longer need to create a
client
this is done inside theConsumerGroup
- consumer ID do not need to be defined. There's a new ID to represent consumers called member ID and this is assigned to consumer after joining the group
- Offsets, group members, and ownership details are not stored in Zookeeper
ConsumerGroup
does not emit aregistered
event
var options = {
host: 'zookeeper:2181',
zk : undefined, // put client zk settings if you need them (see Client)
batch: undefined, // put client batch settings if you need them (see Client)
ssl: true, // optional (defaults to false) or tls options hash
groupId: 'ExampleTestGroup',
sessionTimeout: 15000,
// An array of partition assignment protocols ordered by preference.
// 'roundrobin' or 'range' string for built ins (see below to pass in custom assignment protocol)
protocol: ['roundrobin'],
// Offsets to use for new groups other options could be 'earliest' or 'none' (none will emit an error if no offsets were saved)
// equivalent to Java client's auto.offset.reset
fromOffset: 'latest', // default
// how to recover from OutOfRangeOffset error (where save offset is past server retention) accepts same value as fromOffset
outOfRangeOffset: 'earliest', // default
migrateHLC: false, // for details please see Migration section below
migrateRolling: true
};
var consumerGroup = new ConsumerGroup(options, ['RebalanceTopic', 'RebalanceTest']);
// Or for a single topic pass in a string
var consumerGroup = new ConsumerGroup(options, 'RebalanceTopic');
You can pass a custom assignment strategy to the protocol
array with the interface:
topicPartition
{
"RebalanceTopic": [
"0",
"1",
"2"
],
"RebalanceTest": [
"0",
"1",
"2"
]
}
groupMembers
[
{
"subscription": [
"RebalanceTopic",
"RebalanceTest"
],
"version": 0,
"id": "consumer1-8db1b117-61c6-4f91-867d-20ccd1ad8b3d"
},
{
"subscription": [
"RebalanceTopic",
"RebalanceTest"
],
"version": 0,
"id": "consumer3-bf2d11f4-1c73-4a39-b498-cfe76eb65bea"
},
{
"subscription": [
"RebalanceTopic",
"RebalanceTest"
],
"version": 0,
"id": "consumer2-9781058e-fad4-40e8-a69c-69afbae05184"
}
]
callback(error, result)
result
[
{
"memberId": "consumer3-bf2d11f4-1c73-4a39-b498-cfe76eb65bea",
"topicPartitions": {
"RebalanceTopic": [
"2"
],
"RebalanceTest": [
"2"
]
},
"version": 0
},
{
"memberId": "consumer2-9781058e-fad4-40e8-a69c-69afbae05184",
"topicPartitions": {
"RebalanceTopic": [
"1"
],
"RebalanceTest": [
"1"
]
},
"version": 0
},
{
"memberId": "consumer1-8db1b117-61c6-4f91-867d-20ccd1ad8b3d",
"topicPartitions": {
"RebalanceTopic": [
"0"
],
"RebalanceTest": [
"0"
]
},
"version": 0
}
]
We have two options for automatic migration from existing highLevelConsumer
group. This is useful to preserve the previous committed offsets for your group.
We support two use cases:
- You have downtime and your old HLC consumers are no longer available
- Where the old HLC group is still up and working and you are doing a rolling deploy with zero downtime
For case 1 use below setting:
{
migrateHLC: true, // default is false
migrateRolling: false // default is true
}
For case 2 setting migrateRolling
to true
will allow the ConsumerGroup to start monitoring zk
nodes for when topic ownership are relinquished by the old HLC consumer. Once this is done the ConsumerGroup will connect and the previous HLC offsets from zookeeper will be migrated automatically to the new Kafka broker based coordinator.
- Group name should be consistent with old highLevelConsumer
- Should never overwrite existing offsets
- Only offsets for Topics that were once in the highLevelConsumer will be migrated over offsets for new topics will follow the
fromOffset
setting
client
: client which keeps a connection with the Kafka server.
ready
: when zookeeper is readyconnect
when broker is ready
Fetch the available offset of a specific topic-partition
payloads
: Array,array ofOffsetRequest
,OffsetRequest
is a JSON object like:
{
topic: 'topicName',
partition: 0, //default 0
// time:
// Used to ask for all messages before a certain time (ms), default Date.now(),
// Specify -1 to receive the latest offsets and -2 to receive the earliest available offset.
time: Date.now(),
maxNum: 1 //default 1
}
cb
: Function, the callback
Example
var kafka = require('kafka-node'),
client = new kafka.Client(),
offset = new kafka.Offset(client);
offset.fetch([
{ topic: 't', partition: 0, time: Date.now(), maxNum: 1 }
], function (err, data) {
// data
// { 't': { '0': [999] } }
});
groupId
: consumer grouppayloads
: Array,array ofOffsetCommitRequest
,OffsetCommitRequest
is a JSON object like:
{
topic: 'topicName',
partition: 0, //default 0
offset: 1,
metadata: 'm', //default 'm'
}
Example
var kafka = require('kafka-node'),
client = new kafka.Client(),
offset = new kafka.Offset(client);
offset.commit('groupId', [
{ topic: 't', partition: 0, offset: 10 }
], function (err, data) {
});
Fetch the last committed offset in a topic of a specific consumer group
groupId
: consumer grouppayloads
: Array,array ofOffsetFetchRequest
,OffsetFetchRequest
is a JSON object like:
{
topic: 'topicName',
partition: 0 //default 0
}
Example
var kafka = require('kafka-node'),
client = new kafka.Client(),
offset = new kafka.Offset(client);
offset.fetchCommits('groupId', [
{ topic: 't', partition: 0 }
], function (err, data) {
});
Example
var partition = 0;
var topic = 't';
offset.fetchLatestOffsets([topic], function (error, offsets) {
if (error)
return handleError(error);
console.log(offsets[topic][partition]);
});
Example
var partition = 0;
var topic = 't';
offset.fetchEarliestOffsets([topic], function (error, offsets) {
if (error)
return handleError(error);
console.log(offsets[topic][partition]);
});
Error:
BrokerNotAvailableError: Could not find the leader
Call client.refreshMetadata()
before sending the first message. Reference issue #354
This module uses the debug module so you can just run below before starting your app.
export DEBUG=kafka-node:*
Call client.loadMetadataForTopics
with a blank topic array to get the entire list of available topics (and available brokers).
client.once('connect', function () {
client.loadMetadataForTopics([], function (error, results) {
if (error) {
return console.error(error);
}
console.log('%j', _.get(results, '1.metadata'));
});
});
If you are using the new ConsumerGroup
simply set 'latest'
to fromOffset
option.
Otherwise:
- Call
offset.fetchLatestOffsets
to get fetch the latest offset - Consume from returned offset
Reference issue #342
This error can occur when a HLC is killed and restarted quickly. The ephemeral nodes linked to the previous session are not relinquished in zookeeper when SIGINT
is sent and instead relinquished when zookeeper session timeout is reached. The timeout can be adjusted using the sessionTimeout
zookeeper option when the Client
is created (the default is 30000ms).
Example handler:
process.on('SIGINT', function () {
highLevelConsumer.close(true, function () {
process.exit();
});
});
Alternatively, you can avoid this issue entirely by omitting the HLC's id
and a unique one will be generated for you.
Reference issue #90
Your partition will be stuck if the fetchMaxBytes
is smaller than the message produced. Increase fetchMaxBytes
value should resolve this issue.
Reference to issue #339
- Create a
async.queue
with message processor and concurrency of one (the message processor itself is wrapped withsetImmediate
so it will not freeze up the event loop) - Set the
queue.drain
to resume the consumer - The handler for consumer's
message
event pauses the consumer and pushes the message to the queue.
In the consumer set the encoding
option to buffer
.
Set the messages
attribute in the payload
to a Buffer
. TypedArrays
such as Uint8Array
are not supported and need to be converted to a Buffer
.
{
messages: Buffer.from(data.buffer)
}
Snappy is a optional compression library. Windows users have reported issues with installing it while running npm install
. It's optional in kafka-node and can be skipped by using the --no-optional
flag (though errors from it should not fail the install).
npm install kafka-node --no-optional --save
Keep in mind if you try to use snappy without installing it kafka-node
will throw a runtime exception.
By default, kafka-node
uses debug to log important information. To integrate kafka-node
's log output into an application, it is possible to set a logger provider. This enables filtering of log levels and easy redirection of output streams.
A logger provider is a function which takes the name of a logger and returns a logger implementation. For instance, the following code snippet shows how a logger provider for the global console
object could be written:
function consoleLoggerProvider (name) {
// do something with the name
return {
debug: console.debug.bind(console),
info: console.info.bind(console),
warn: console.warn.bind(console),
error: console.error.bind(console)
};
}
The logger interface with its debug
, info
, warn
and error
methods expects format string support as seen in debug
or the JavaScript console
object. Many commonly used logging implementations cover this API, e.g. bunyan, pino or winston.
For performance reasons, initialization of the kafka-node
module creates all necessary loggers. This means that custom logger providers need to be set before requiring the kafka-node
module. The following example shows how this can be done:
// first configure the logger provider
const kafkaLogging = require('kafka-node/logging');
kafkaLogging.setLoggerProvider(consoleLoggerProvider);
// then require kafka-node and continue as normal
const kafka = require('kafka-node');
On the Mac install Docker for Mac.
npm test
Achieved using the KAFKA_VERSION
environment variable.
# Runs "latest" kafka on docker hub*
npm test
KAFKA_VERSION=0.8 npm test
KAFKA_VERSION=0.9 npm test
KAFKA_VERSION=0.10 npm test
*See Docker hub tags entry for which version is considered latest
.
npm run stopDocker
Copyright (c) 2015 Sohu.com
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.