Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Kafka Will Get The Message Across, Guaranteed.
Search
David Zuelke
January 28, 2017
Programming
0
260
Kafka Will Get The Message Across, Guaranteed.
Presentation given at PHP Benelux 2017 near Antwerp, Belgium.
David Zuelke
January 28, 2017
Tweet
Share
More Decks by David Zuelke
See All by David Zuelke
Your next Web server will be written in... PHP
dzuelke
0
150
Getting Things Done
dzuelke
1
410
Your next Web server will be written in... PHP
dzuelke
2
260
Your next Web server will be written in... PHP
dzuelke
3
1.1k
Kafka Will Get The Message Across, Guaranteed.
dzuelke
0
800
Heroku at BattleHack Venice 2015
dzuelke
0
130
Designing HTTP Interfaces and RESTful Web Services
dzuelke
6
1.4k
The Twelve-Factor App: Best Practices for Modern Web Applications
dzuelke
4
460
Designing HTTP Interfaces and RESTful Web Services
dzuelke
6
480
Other Decks in Programming
See All in Programming
クラシルリワードにおける iOSアプリ開発の取り組み
funzin
1
810
イベントソーシングとAIの親和性ー物語とLLMに理解できるデータ
tomohisa
1
160
TSConfig Solution Style & subpath imports to switch types on a per-file basis
maminami373
1
180
DevDay2025-OracleDatabase-kernel-addressing-history
oracle4engineer
PRO
7
1.6k
Babylon.js 8.0のアプデ情報を 軽率にキャッチアップ / catch-up-babylonjs-8
drumath2237
0
110
Devinで実践する!AIエージェントと協働する開発組織の作り方
masahiro_nishimi
6
2.6k
Building an Application with TDD, DDD and Hexagonal Architecture - Isn't it a bit too much?
mufrid
0
370
Passkeys for Java Developers
ynojima
1
200
CRUD から CQRS へ ~ 分離が可能にする柔軟性
tkawae
0
230
AIにコードを生成するコードを作らせて、再現性を担保しよう! / Let AI generate code to ensure reproducibility
yamachu
7
6.1k
Zennの運営完全に理解した #完全に理解したTalk
wadayusuke
1
140
MLOps Japan 勉強会 #52 - 特徴量を言語を越えて一貫して管理する, 『特徴量ドリブン』な MLOps の実現への試み
taniiicom
2
570
Featured
See All Featured
Connecting the Dots Between Site Speed, User Experience & Your Business [WebExpo 2025]
tammyeverts
1
82
Navigating Team Friction
lara
186
15k
Distributed Sagas: A Protocol for Coordinating Microservices
caitiem20
331
21k
What's in a price? How to price your products and services
michaelherold
245
12k
Rails Girls Zürich Keynote
gr2m
94
13k
CoffeeScript is Beautiful & I Never Want to Write Plain JavaScript Again
sstephenson
160
15k
KATA
mclloyd
29
14k
The Straight Up "How To Draw Better" Workshop
denniskardys
233
140k
How to Ace a Technical Interview
jacobian
276
23k
Responsive Adventures: Dirty Tricks From The Dark Corners of Front-End
smashingmag
252
21k
Building Flexible Design Systems
yeseniaperezcruz
329
39k
Reflections from 52 weeks, 52 projects
jeffersonlam
349
20k
Transcript
KAFKA WILL GET THE MESSAGE ACROSS. GUARANTEED. PHP Benelux 2017
Belgium
David Zuelke
None
[email protected]
@dzuelke
KAFKA
LinkedIn
APACHE KAFKA
"uh oh, another Apache project?!"
None
KEEP CALM AND LOOK AT THE WEBSITE
None
"Basically it is a massively scalable pub/sub message queue. architected
as a distributed transaction log."
"so it's a queue?"
it's not just a queue
queues are not multi-subscriber :(
"so it's a pubsub thing?"
it's not just a pubsub thing
pubsub broadcasts to all subscribers :(
it's a log
None
not that kind of log
WAL
Write-Ahead Log
WRITE-AHEAD LOG
None
1 foo 2 bar 3 baz 4 hi
1 create document: "foo", data: "…" 2 update document: "foo",
data: "…" 3 create document: "bar", data: "…" 4 remove document: "foo"
None
never corrupts
sequential I/O
None
sequential I/O
every message will be read at least once, no random
access
FileChannel.transferTo (shovels data straight from e.g. disk cache to network
interface, no copying via RAM)
"HI, I AM KAFKA" "Buckle up while we process (m|b|tr)illions
of messages/s."
TOPICS
streams of records
1 2 3 4 5 6 7 …
1 2 3 4 5 6 7 8 … producer
writes consumer reads
can have many subscribers
1 2 3 4 5 6 7 8 … producer
writes consumerB reads consumerA reads
can be partitioned
P0 1 2 3 4 5 6 7 … P1
1 2 3 4 … P2 1 2 3 4 5 6 7 8 … P3 1 2 3 4 5 6 …
partitions let you scale storage!
partitions let you scale consuming!
None
all records are retained, whether consumed or not, up to
a configurable limit
PRODUCERS
byte[]
(typically JSON, XML, Avro, Thrift, Protobufs)
(typically not funny GIFs)
can choose explicit partition, or a key (which is used
for auto-partitioning)
https://github.com/edenhill/librdkafka & https://arnaud-lb.github.io/php-rdkafka/
BASIC PRODUCER $rk = new RdKafka\Producer(); $rk->addBrokers("127.0.0.1"); $topic = $rk->newTopic("test");
$topic->produce(RD_KAFKA_PARTITION_UA, 0, "Unassigned partition, let Kafka choose"); $topic->produce(RD_KAFKA_PARTITION_UA, 0, "Yay consistent hashing", $user->getId()); $topic->produce(1, 0, "This will always be sent to partition 1");
CONSUMERS
cheap
only metadata stored per consumer: offset
guaranteed to always have messages in right order (within a
partition)
can themselves produce new messages! (but there is also a
Streams API for pure transformations)
None
BASIC CONSUMER $conf = new RdKafka\Conf(); $conf->set('group.id', 'myConsumerGroup'); $rk =
new RdKafka\Consumer($conf); $rk->addBrokers("127.0.0.1"); $topicConf = new RdKafka\TopicConf(); $topicConf->set('auto.commit.interval.ms', 100); $topic = $rk->newTopic("test", $topicConf); $topic->consumeStart(0, RD_KAFKA_OFFSET_STORED); while (true) { $msg = $topic->consume(0, 120*10000); do_something($msg); }
AT-MOST ONCE DELIVERY $conf = new RdKafka\Conf(); $conf->set('group.id', 'myConsumerGroup'); $rk
= new RdKafka\Consumer($conf); $rk->addBrokers("127.0.0.1"); $topicConf = new RdKafka\TopicConf(); $topicConf->set('auto.commit.enable', false); $topic = $rk->newTopic("test", $topicConf); $topic->consumeStart(0, RD_KAFKA_OFFSET_STORED); while (true) { $msg = $topic->consume(0, 120*10000); $topic->offsetStore($msg->partition, $msg->offset); do_something($msg); }
AT-LEAST ONCE DELIVERY $conf = new RdKafka\Conf(); $conf->set('group.id', 'myConsumerGroup'); $rk
= new RdKafka\Consumer($conf); $rk->addBrokers("127.0.0.1"); $topicConf = new RdKafka\TopicConf(); $topicConf->set('auto.commit.enable', false); $topic = $rk->newTopic("test", $topicConf); $topic->consumeStart(0, RD_KAFKA_OFFSET_STORED); while (true) { $msg = $topic->consume(0, 120*10000); do_something($msg); $topic->offsetStore($msg->partition, $msg->offset); }
EXACTLY-ONCE DELIVERY
you cannot have exactly-once delivery
THE BYZANTINE GENERALS "together we can beat the monsters. let's
both attack at 07:00?" "confirm, we attack at 07:00" ☠
USE CASES
• LinkedIn • Yahoo • Twitter • Netflix • Square
• Spotify • Pinterest • Uber • Goldman Sachs • Tumblr • PayPal • Airbnb • Mozilla • Cisco • Etsy • Foursquare • Shopify • CloudFlare
ingest the Twitter firehose and turn it into a pointless
demo ;)
None
messaging, of course
track user activity
record runtime metrics
aggregate logs
IoT (you could still e.g. use MQTT over the wire,
and bridge to Kafka)
replicate information between data centers (also see Connector API)
Event Sourcing broker :)
WAL / Commit Log for another system
billing!
"shock absorber" between systems to avoid overload of DBs, APIs,
etc.
in PHP: mostly producing messages; better languages exist for consuming
The End
THANK YOU FOR LISTENING! Questions? Ask me: @dzuelke &
[email protected]