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
280
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
160
Getting Things Done
dzuelke
1
430
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
820
Heroku at BattleHack Venice 2015
dzuelke
0
130
Designing HTTP Interfaces and RESTful Web Services
dzuelke
6
1.5k
The Twelve-Factor App: Best Practices for Modern Web Applications
dzuelke
4
500
Designing HTTP Interfaces and RESTful Web Services
dzuelke
6
490
Other Decks in Programming
See All in Programming
複雑化したリポジトリをなんとかした話 pipenvからuvによるモノレポ構成への移行
satoshi256kbyte
1
720
Чего вы не знали о строках в Python – Василий Рябов, PythoNN
sobolevn
0
150
Learn CPU architecture with Assembly
akkeylab
1
1.3k
いま中途半端なSwift 6対応をするより、Default ActorやApproachable Concurrencyを有効にしてからでいいんじゃない?
yimajo
2
310
株式会社 Sun terras カンパニーデック
sunterras
0
180
PostgreSQLで手軽にDuckDBを使う!DuckDB&pg_duckdb入門/osk2025-duckdb
takahashiikki
1
230
(Extension DC 2025) Actor境界を越える技術
teamhimeh
1
200
WebエンジニアがSwiftをブラウザで動かすプレイグラウンドを作ってみた
ohmori_yusuke
0
170
CSC509 Lecture 03
javiergs
PRO
0
320
iOSDC.pdf
chronos2500
2
650
大規模アプリにおけるXcode Previews実用化までの道のり
ikesyo
0
990
iOSエンジニア向けの英語学習アプリを作る!
yukawashouhei
0
170
Featured
See All Featured
Fashionably flexible responsive web design (full day workshop)
malarkey
407
66k
Learning to Love Humans: Emotional Interface Design
aarron
274
40k
Building a Scalable Design System with Sketch
lauravandoore
462
33k
Building Better People: How to give real-time feedback that sticks.
wjessup
368
20k
A Modern Web Designer's Workflow
chriscoyier
697
190k
Sharpening the Axe: The Primacy of Toolmaking
bcantrill
45
2.5k
Helping Users Find Their Own Way: Creating Modern Search Experiences
danielanewman
30
2.9k
The Success of Rails: Ensuring Growth for the Next 100 Years
eileencodes
46
7.6k
Designing Experiences People Love
moore
142
24k
Docker and Python
trallard
46
3.6k
[RailsConf 2023] Rails as a piece of cake
palkan
57
5.9k
Faster Mobile Websites
deanohume
310
31k
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]