flume-ng-kafka-sink

Apache FlumeNG sink to push events to Kafka 0.8+

topic selection :

statically via sink configuration
dynamically via event header "topic"

partition key :

from event header "key"

By default ( no "key" header ) event are spread in a round robin fashion across topic's partitions.

With the default kafka partitioner event with the same key will always go to the same partition and therefore to the same consumer. you can override this behaviour by providing a partitioner.class to the kafka server configuration

Build/Install

Build flume-ng-kafka-sink

mvn package

Configuration

agent.sources   = httpSource
agent.channels  = memoryChannel
agent.sinks     = kafkaSink

# Source configuration
agent.sources.httpSource.type = http
agent.sources.httpSource.channels = memoryChannel
agent.sources.httpSource.port = 8026
agent.sources.httpSource.handler = org.apache.flume.source.http.JSONHandler

# Sink configuration
agent.sinks.kafkaSink.type = org.apache.flume.sink.kafka.KafkaSink
agent.sinks.kafkaSink.channel = memoryChannel
    # If omitted topic must be provided by event header topic
agent.sinks.kafkaSink.topic = flume
agent.sinks.kafkaSink.batch.size = 100
    # Producer configuration
agent.sinks.kafkaSink.metadata.broker.list = localhost:9092
agent.sinks.kafkaSink.serializer.class = kafka.serializer.DefaultEncoder
agent.sinks.kafkaSink.producer.type = async
agent.sinks.kafkaSink.batch.num.messages = 100
agent.sinks.kafkaSink.queue.buffering.max.ms = 100

Test

curl -X POST -d "[{"headers":{"topic":"test","key":"key"},"body":'{"foo":"bar"}'}]" http://localhost:8026

Reporting

This sink follow the standard flume reporting interface

To enable json reporting you may start flume with the following options

-Dflume.monitoring.type=http -Dflume.monitoring.port=34545

Just get /metrics to get them

curl http://localhost:34545/metrics 2> /dev/null | python -mjson.tool
{
    "CHANNEL.memoryChannel": {
        "ChannelCapacity": "10000",
        "ChannelFillPercentage": "0.0",
        "ChannelSize": "0",
        "EventPutAttemptCount": "5783",
        "EventPutSuccessCount": "5783",
        "EventTakeAttemptCount": "5805",
        "EventTakeSuccessCount": "5783",
        "StartTime": "1404126896721",
        "StopTime": "0",
        "Type": "CHANNEL"
    },
    "SINK.kafkaSink": {
        "BatchCompleteCount": "57",
        "BatchEmptyCount": "21",
        "BatchUnderflowCount": "1",
        "ConnectionClosedCount": "0",
        "ConnectionCreatedCount": "0",
        "ConnectionFailedCount": "0",
        "EventDrainAttemptCount": "5783",
        "EventDrainSuccessCount": "5783",
        "StartTime": "1404126896722",
        "StopTime": "0",
        "Type": "SINK"
    },
    "SOURCE.httpSource": {
        "AppendAcceptedCount": "0",
        "AppendBatchAcceptedCount": "5783",
        "AppendBatchReceivedCount": "5783",
        "AppendReceivedCount": "0",
        "EventAcceptedCount": "5783",
        "EventReceivedCount": "5783",
        "OpenConnectionCount": "0",
        "StartTime": "1404126896808",
        "StopTime": "0",
        "Type": "SOURCE"
    }
}

heidsoft / flume-ng-kafka-sink Goto Github PK