kafka kraft configuration + dedicated controllers machines

We are preparing Huge Kafka Kraft cluster

Installation is on 35 physical Linux machines and 3 dedicated controllers on 3 VM Linux machines

each broker machine include:

512G RAM memory
64 CORE's

About the brokers machines , each machine include one disk ( RAID10 ) , With 15T size , So total broker storage are ~ 525 TB

About the 3 VM machines that should be used as controllers , of course Machines spec is different and very small comparing to brokers machines

each VM controller machine include:

32G RAM Memory
12 CORE's

But we are not sure about if we need dedicate disk on the controllers Machines or maybe we can use the OS disk ( OS disk is 250G size )

For example we can store the metadata topic under /var/lib/kafka

From our understanding only topic __cluster_metadata should take space on The disk

And this topic take little space

So is it ok to create small volume on OS disk – lets say as 50G to store The controller metadata topic?

Or we need dedicated disk ( as sdb ) on each of the controllers machines?

example:

df -h
Filesystem                       Size  Used Avail Use% Mounted on
devtmpfs                         3.8G     0  3.8G   0% /dev
tmpfs                            3.8G   12K  3.8G   1% /dev/shm
tmpfs                            3.8G  140M  3.6G   4% /run
tmpfs                            3.8G     0  3.8G   0% /sys/fs/cgroup
/dev/mapper/vg-controller-lv_root     50G   17G   34G  34% /
/dev/mapper/vg-controller_lv_var     100G  494M  100G   1% /var
/dev/sda1                        1014M  224M  791M  23% /boot
tmpfs                             764M     0  764M   0% /run/user/0
/dev/sdb                          50G   1M   50G   1% /var/lib/kafka-store_meta_data

enter image description here

enter image description here

Example of default controller configuration: ( from Kraft config )

more controller.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.

#
# This configuration file is intended for use in KRaft mode, where
# Apache ZooKeeper is not present.  See config/kraft/README.md for details.
#

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

# The role of this server. Setting this puts us in KRaft mode
process.roles=controller

# The node id associated with this instance's roles
node.id=1

# The connect string for the controller quorum
controller.quorum.voters=1@localhost:9093

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

# The address the socket server listens on.
# Note that only the controller listeners are allowed here when `process.roles=controller`, and this listener should be consistent with `controller.quorum.voters` value.
#   FORMAT:
#     listeners = listener_name://host_name:port
#   EXAMPLE:
#     listeners = PLAINTEXT://your.host.name:9092
listeners=CONTROLLER://:9093

# A comma-separated list of the names of the listeners used by the controller.
# This is required if running in KRaft mode.
controller.listener.names=CONTROLLER

# Maps listener names to security protocols, the default is for them to be the same. See the config documentation for more details
#listener.security.protocol.map=PLAINTEXT:PLAINTEXT,SSL:SSL,SASL_PLAINTEXT:SASL_PLAINTEXT,SASL_SSL:SASL_SSL

# The number of threads that the server uses for receiving requests from the network and sending responses to the network
num.network.threads=3

# The number of threads that the server uses for processing requests, which may include disk I/O
num.io.threads=8

# The send buffer (SO_SNDBUF) used by the socket server
socket.send.buffer.bytes=102400

# The receive buffer (SO_RCVBUF) used by the socket server
socket.receive.buffer.bytes=102400

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


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

# A comma separated list of directories under which to store log files
log.dirs=/tmp/kraft-controller-logs

# The default number of log partitions per topic. More partitions allow greater
# parallelism for consumption, but this will also result in more files across
# the brokers.
num.partitions=1

# The number of threads per data directory to be used for log recovery at startup and flushing at shutdown.
# This value is recommended to be increased for installations with data dirs located in RAID array.
num.recovery.threads.per.data.dir=1

############################# Internal Topic Settings  #############################
# The replication factor for the group metadata internal topics "__consumer_offsets" and "__transaction_state"
# For anything other than development testing, a value greater than 1 is recommended to ensure availability such as 3.
offsets.topic.replication.factor=1
transaction.state.log.replication.factor=1
transaction.state.log.min.isr=1

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

# Messages are immediately written to the filesystem but by default we only fsync() to sync
# the OS cache lazily. The following configurations control the flush of data to disk.
# There are a few important trade-offs here:
#    1. Durability: Unflushed data may be lost if you are not using replication.
#    2. Latency: Very large flush intervals may lead to latency spikes when the flush does occur as there will be a lot of data to flush.
#    3. Throughput: The flush is generally the most expensive operation, and a small flush interval may lead to excessive seeks.
# 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.

# The number of messages to accept before forcing a flush of data to disk
#log.flush.interval.messages=10000

# The maximum amount of time a message can sit in a log before we force a flush
#log.flush.interval.ms=1000

############################# 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 due to age
log.retention.hours=168

# A size-based retention policy for logs. Segments are pruned from the log unless the remaining
# segments drop below log.retention.bytes. Functions independently of log.retention.hours.
#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=1073741824

# The interval at which log segments are checked to see if they can be deleted according
# to the retention policies
log.retention.check.interval.ms=300000

Best practice would be to store /var/log (server logs) on separate disks (or systems, such as export data to a logging system like Splunk) than Kafka data logs. This way, when either fill-up, the server / OS will not crash, and you can still debug those machines via SSH, for example.

Also, never use /tmp directory for log.dirs

Leave a Comment