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
December 02, 2016
Programming
0
820
Kafka Will Get The Message Across, Guaranteed.
Presentation given at SymfonyCon 2016 in Berlin, Germany.
David Zuelke
December 02, 2016
Tweet
Share
More Decks by David Zuelke
See All by David Zuelke
Your next Web server will be written in... PHP
dzuelke
0
160
Getting Things Done
dzuelke
1
420
Your next Web server will be written in... PHP
dzuelke
2
270
Your next Web server will be written in... PHP
dzuelke
3
1.2k
Kafka Will Get The Message Across, Guaranteed.
dzuelke
0
270
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
480
Designing HTTP Interfaces and RESTful Web Services
dzuelke
6
490
Other Decks in Programming
See All in Programming
Microsoft Orleans, Daprのアクターモデルを使い効率的に開発、デプロイを行うためのSekibanの試行錯誤 / Sekiban: Exploring Efficient Development and Deployment with Microsoft Orleans and Dapr Actor Models
tomohisa
0
210
Scale out your Claude Code ~自社専用Agentで10xする開発プロセス~
yukukotani
9
2.6k
Nuances on Kubernetes - RubyConf Taiwan 2025
envek
0
190
TROCCO×dbtで実現する人にもAIにもやさしいデータ基盤
nealle
0
330
なぜ今、Terraformの本を書いたのか? - 著者陣に聞く!『Terraformではじめる実践IaC』登壇資料
fufuhu
4
650
CSC305 Summer Lecture 04
javiergs
PRO
1
110
Flutter로 Gemini와 MCP를 활용한 Agentic App 만들기 - 박제창 2025 I/O Extended Seoul
itsmedreamwalker
0
150
オホーツクでコミュニティを立ち上げた理由―地方出身プログラマの挑戦 / TechRAMEN 2025 Conference
lemonade_37
2
490
学習を成果に繋げるための個人開発の考え方 〜 「学習のための個人開発」のすすめ / personal project for leaning
panda_program
1
110
TanStack DB ~状態管理の新しい考え方~
bmthd
2
330
AWS Serverless Application Model入門_20250708
smatsuzaki
0
130
オープンセミナー2025@広島LT技術ブログを続けるには
satoshi256kbyte
0
120
Featured
See All Featured
GraphQLの誤解/rethinking-graphql
sonatard
71
11k
Designing Dashboards & Data Visualisations in Web Apps
destraynor
231
53k
Balancing Empowerment & Direction
lara
2
590
Documentation Writing (for coders)
carmenintech
73
5k
A Modern Web Designer's Workflow
chriscoyier
695
190k
The Myth of the Modular Monolith - Day 2 Keynote - Rails World 2024
eileencodes
26
3k
How to Think Like a Performance Engineer
csswizardry
25
1.8k
Rebuilding a faster, lazier Slack
samanthasiow
83
9.1k
It's Worth the Effort
3n
187
28k
Let's Do A Bunch of Simple Stuff to Make Websites Faster
chriscoyier
507
140k
Keith and Marios Guide to Fast Websites
keithpitt
411
22k
The Illustrated Children's Guide to Kubernetes
chrisshort
48
50k
Transcript
KAFKA WILL GET THE MESSAGE ACROSS. GUARANTEED. SymfonyCon 2016 Berlin,
Germany
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 a queue
queues are not multi-subscriber :(
"so it's a pubsub thing?"
it's not 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!
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 Twitter firehose and turn it into a pointless demo
;)
None
messaging, of course
track user activity
record runtime metrics
IoT
replicate information between data centers
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]