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
202507_ADKで始めるエージェント開発の基本 〜デモを通じて紹介〜(奥田りさ)The Basics of Agent Development with ADK — A Demo-Focused Introduction
risatube
PRO
6
1.4k
0から始めるモジュラーモノリス-クリーンなモノリスを目指して
sushi0120
0
250
QA x AIエコシステム段階構築作戦
osu
0
240
decksh - a little language for decks
ajstarks
4
21k
実践!App Intents対応
yuukiw00w
0
110
Streamlitで実現できるようになったこと、実現してくれたこと
ayumu_yamaguchi
2
270
AIコーディングエージェント全社導入とセキュリティ対策
hikaruegashira
15
9.3k
画像コンペでのベースラインモデルの育て方
tattaka
3
1k
はじめてのWeb API体験 ー 飲食店検索アプリを作ろうー
akinko_0915
0
180
CEDEC 2025 『ゲームにおけるリアルタイム通信への QUIC導入事例の紹介』
segadevtech
2
740
あまり知られていない MCP 仕様たち / MCP specifications that aren’t widely known
ktr_0731
0
220
Quality Gates in the Age of Agentic Coding
helmedeiros
PRO
1
120
Featured
See All Featured
The Power of CSS Pseudo Elements
geoffreycrofte
77
5.9k
The Art of Delivering Value - GDevCon NA Keynote
reverentgeek
15
1.6k
Designing for Performance
lara
610
69k
Bootstrapping a Software Product
garrettdimon
PRO
307
110k
The Art of Programming - Codeland 2020
erikaheidi
54
13k
What’s in a name? Adding method to the madness
productmarketing
PRO
23
3.6k
We Have a Design System, Now What?
morganepeng
53
7.7k
How to Create Impact in a Changing Tech Landscape [PerfNow 2023]
tammyeverts
53
2.9k
[Rails World 2023 - Day 1 Closing Keynote] - The Magic of Rails
eileencodes
35
2.5k
Making the Leap to Tech Lead
cromwellryan
134
9.5k
How STYLIGHT went responsive
nonsquared
100
5.7k
VelocityConf: Rendering Performance Case Studies
addyosmani
332
24k
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]