Kafka Benchmark Commands

benchmark-commands.txt

Producer

Setup
bin/kafka-topics.sh --zookeeper esv4-hcl197.grid.linkedin.com:2181 --create --topic test-rep-one --partitions 6 --replication-factor 1
bin/kafka-topics.sh --zookeeper esv4-hcl197.grid.linkedin.com:2181 --create --topic test --partitions 6 --replication-factor 3

Single thread, no replication

bin/kafka-run-class.sh org.apache.kafka.clients.tools.ProducerPerformance test7 50000000 100 -1 acks=1 bootstrap.servers=esv4-hcl198.grid.linkedin.com:9092 buffer.memory=67108864 batch.size=8196

Single-thread, async 3x replication

bin/kafktopics.sh --zookeeper esv4-hcl197.grid.linkedin.com:2181 --create --topic test --partitions 6 --replication-factor 3
bin/kafka-run-class.sh org.apache.kafka.clients.tools.ProducerPerformance test6 50000000 100 -1 acks=1 bootstrap.servers=esv4-hcl198.grid.linkedin.com:9092 buffer.memory=67108864 batch.size=8196

Single-thread, sync 3x replication

bin/kafka-run-class.sh org.apache.kafka.clients.tools.ProducerPerformance test 50000000 100 -1 acks=-1 bootstrap.servers=esv4-hcl198.grid.linkedin.com:9092 buffer.memory=67108864 batch.size=64000

Three Producers, 3x async replication
bin/kafka-run-class.sh org.apache.kafka.clients.tools.ProducerPerformance test 50000000 100 -1 acks=1 bootstrap.servers=esv4-hcl198.grid.linkedin.com:9092 buffer.memory=67108864 batch.size=8196

Throughput Versus Stored Data

bin/kafka-run-class.sh org.apache.kafka.clients.tools.ProducerPerformance test 50000000000 100 -1 acks=1 bootstrap.servers=esv4-hcl198.grid.linkedin.com:9092 buffer.memory=67108864 batch.size=8196

Effect of message size

for i in 10 100 1000 10000 100000;
do
echo ""
echo $i
bin/kafka-run-class.sh org.apache.kafka.clients.tools.ProducerPerformance test $((1000*1024*1024/$i)) $i -1 acks=1 bootstrap.servers=esv4-hcl198.grid.linkedin.com:9092 buffer.memory=67108864 batch.size=128000
done;

Consumer
Consumer throughput

bin/kafka-consumer-perf-test.sh --zookeeper esv4-hcl197.grid.linkedin.com:2181 --messages 50000000 --topic test --threads 1

3 Consumers

On three servers, run:
bin/kafka-consumer-perf-test.sh --zookeeper esv4-hcl197.grid.linkedin.com:2181 --messages 50000000 --topic test --threads 1

End-to-end Latency

bin/kafka-run-class.sh kafka.tools.TestEndToEndLatency esv4-hcl198.grid.linkedin.com:9092 esv4-hcl197.grid.linkedin.com:2181 test 5000

Producer and consumer

bin/kafka-run-class.sh org.apache.kafka.clients.tools.ProducerPerformance test 50000000 100 -1 acks=1 bootstrap.servers=esv4-hcl198.grid.linkedin.com:9092 buffer.memory=67108864 batch.size=8196

bin/kafka-consumer-perf-test.sh --zookeeper esv4-hcl197.grid.linkedin.com:2181 --messages 50000000 --topic test --threads 1

server-config.properties

# Licensed to the Apache Software Foundation (ASF) under one or more
# contributor license agreements.  See the NOTICE file distributed with
# this work for additional information regarding copyright ownership.
# The ASF licenses this file to You under the Apache License, Version 2.0
# (the "License"); you may not use this file except in compliance with
# the License.  You may obtain a copy of the License at
# 
#    http://www.apache.org/licenses/LICENSE-2.0
# 
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# see kafka.server.KafkaConfig for additional details and defaults

############################# Server Basics #############################

# The id of the broker. This must be set to a unique integer for each broker.
broker.id=0

############################# Socket Server Settings #############################

# The port the socket server listens on
port=9092

# Hostname the broker will bind to and advertise to producers and consumers.
# If not set, the server will bind to all interfaces and advertise the value returned from
# from java.net.InetAddress.getCanonicalHostName().
#host.name=localhost

# The number of threads handling network requests
num.network.threads=4
 
# The number of threads doing disk I/O
num.io.threads=8

# The send buffer (SO_SNDBUF) used by the socket server
socket.send.buffer.bytes=1048576

# The receive buffer (SO_RCVBUF) used by the socket server
socket.receive.buffer.bytes=1048576

# The maximum size of a request that the socket server will accept (protection against OOM)
socket.request.max.bytes=104857600


############################# Log Basics #############################

# The directory under which to store log files
log.dirs=/grid/a/dfs-data/kafka-logs,/grid/b/dfs-data/kafka-logs,/grid/c/dfs-data/kafka-logs,/grid/d/dfs-data/kafka-logs,/grid/e/dfs-data/kafka-logs,/grid/f/dfs-data/kafka-logs

# The number of logical partitions per topic per server. More partitions allow greater parallelism
# for consumption, but also mean more files.
num.partitions=8

############################# Log Flush Policy #############################

# The following configurations control the flush of data to disk. This is the most
# important performance knob in kafka.
# There are a few important trade-offs here:
#    1. Durability: Unflushed data is at greater risk of loss in the event of a crash.
#    2. Latency: Data is not made available to consumers until it is flushed (which adds latency).
#    3. Throughput: The flush is generally the most expensive operation. 
# The settings below allow one to configure the flush policy to flush data after a period of time or
# every N messages (or both). This can be done globally and overridden on a per-topic basis.

# Per-topic overrides for log.flush.interval.ms
#log.flush.intervals.ms.per.topic=topic1:1000, topic2:3000

############################# Log Retention Policy #############################

# The following configurations control the disposal of log segments. The policy can
# be set to delete segments after a period of time, or after a given size has accumulated.
# A segment will be deleted whenever *either* of these criteria are met. Deletion always happens
# from the end of the log.

# The minimum age of a log file to be eligible for deletion
log.retention.hours=168

# A size-based retention policy for logs. Segments are pruned from the log as long as the remaining
# segments don't drop below log.retention.bytes.
#log.retention.bytes=1073741824

# The maximum size of a log segment file. When this size is reached a new log segment will be created.
log.segment.bytes=536870912

# The interval at which log segments are checked to see if they can be deleted according 
# to the retention policies
log.cleanup.interval.mins=1

############################# Zookeeper #############################

# Zookeeper connection string (see zookeeper docs for details).
# This is a comma separated host:port pairs, each corresponding to a zk
# server. e.g. "127.0.0.1:3000,127.0.0.1:3001,127.0.0.1:3002".
# You can also append an optional chroot string to the urls to specify the
# root directory for all kafka znodes.
zookeeper.connect=esv4-hcl197.grid.linkedin.com:2181

# Timeout in ms for connecting to zookeeper
zookeeper.connection.timeout.ms=1000000

# metrics reporter properties
kafka.metrics.polling.interval.secs=5
kafka.metrics.reporters=kafka.metrics.KafkaCSVMetricsReporter
kafka.csv.metrics.dir=/tmp/kafka_metrics
# Disable csv reporting by default.
kafka.csv.metrics.reporter.enabled=false

replica.lag.max.messages=10000000

What is the advantage of using zookeeper in different machine?
I used my zk in the same machine with kafka, although I'm afraid it's a dangerous decision.
I did this because I could save three machines.
some key point is:
1.kafka data transferring on the NIC will make zk connection timeout due to hit the limits of NIC.
2.if zk use the same data disk with kafka,zk will have a IO blocking while kafka busy reading and writing.
so when I use zk within the three machines same with kafka. I set zk data dir to an independent disk. such as os disk. usually ssd.

For recent version(test with 2.3.0):

1. clone kafka source code, then run

./gradlew jarAll -x signArchives -x test -x javadoc -x scaladoc

2. run test

 bin/kafka-run-class.sh org.apache.kafka.tools.ProducerPerformance --topic test-rep-one --num-records 50000000 --record-size 100 --throughput=-1 --producer.config ./test.conf 

test.conf:

[root@dx-app2 kafka]# cat test.conf 

# list of brokers used for bootstrapping knowledge about the rest of the cluster
# format: host1:port1,host2:port2 ...
bootstrap.servers=localhost:9092

# the default batch size in bytes when batching multiple records sent to a partition
batch.size=8196

# the total bytes of memory the producer can use to buffer records waiting to be sent to the server
buffer.memory=67108864

@daixiang0
Hi, I can't understand the first step. What do you mean by saying "clone kafka source code"?

I think the answer is simple, for example, run 'git clone ...location' command where ...location is Kafka source on github.

Is this test setup running zookeeper on 3 nodes or just 1? It's not clear from this file.

I believe the zookeeper switch on 'kafka-consumer-perf-test' has deprecated for 'kafka-consumer-perf-test.bat' . (I am running remotely from kafka path CLI)
Running the below from CLI I am getting the error.

kafka-consumer-perf-test --topic test --bootstrap-server test:XXXX --messages 10 --threads 1 --consumer.config C:*****\consumer.properties --group test --timeout 100000 --print-metrics

Exception in thread "main" java.util.IllegalFormatConversionException: f != java.lang.Integer
at java.util.Formatter$FormatSpecifier.failConversion(Formatter.java:4302)
at java.util.Formatter$FormatSpecifier.printFloat(Formatter.java:2806)
at java.util.Formatter$FormatSpecifier.print(Formatter.java:2753)
at java.util.Formatter.format(Formatter.java:2520)
at java.util.Formatter.format(Formatter.java:2455)
at java.lang.String.format(String.java:2940)
at scala.collection.immutable.StringLike.format(StringLike.scala:354)
at scala.collection.immutable.StringLike.format$(StringLike.scala:353)
at scala.collection.immutable.StringOps.format(StringOps.scala:33)
at kafka.utils.ToolsUtils$.$anonfun$printMetrics$3(ToolsUtils.scala:60)
at kafka.utils.ToolsUtils$.$anonfun$printMetrics$3$adapted(ToolsUtils.scala:58)
at scala.collection.mutable.ResizableArray.foreach(ResizableArray.scala:62)
at scala.collection.mutable.ResizableArray.foreach$(ResizableArray.scala:55)
at scala.collection.mutable.ArrayBuffer.foreach(ArrayBuffer.scala:49)
at kafka.utils.ToolsUtils$.printMetrics(ToolsUtils.scala:58)
at kafka.tools.ConsumerPerformance$.main(ConsumerPerformance.scala:82)
at kafka.tools.ConsumerPerformance.main(ConsumerPerformance.scala)

@daixiang0
Hi, I can't understand the first step. What do you mean by saying "clone kafka source code"?

I think to clone source code is to build Kafka from source. If you've already installed, just go run the shell script. :)

Hello

Single-thread, sync 3x replication

bin/kafka-run-class.sh org.apache.kafka.clients.tools.ProducerPerformance test 50000000 100 -1 acks=-1 bootstrap.servers=esv4-hcl198.grid.linkedin.com:9092 buffer.memory=67108864 batch.size=64000

Three Producers, 3x async replication
bin/kafka-run-class.sh org.apache.kafka.clients.tools.ProducerPerformance test 50000000 100 -1 acks=1 bootstrap.servers=esv4-hcl198.grid.linkedin.com:9092 buffer.memory=67108864 batch.size=8196

Sorry but what is the difference between sync and async replication ? i was thinking synch mode mode corresponds to acks=1 and async correspond to acks=all but the command lines let me think something different ?

Thanks