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
230
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
140
Getting Things Done
dzuelke
1
380
Your next Web server will be written in... PHP
dzuelke
2
250
Your next Web server will be written in... PHP
dzuelke
3
1.1k
Kafka Will Get The Message Across, Guaranteed.
dzuelke
0
760
Heroku at BattleHack Venice 2015
dzuelke
0
120
Designing HTTP Interfaces and RESTful Web Services
dzuelke
6
1.3k
The Twelve-Factor App: Best Practices for Modern Web Applications
dzuelke
4
380
Designing HTTP Interfaces and RESTful Web Services
dzuelke
6
460
Other Decks in Programming
See All in Programming
Universal Linksの実装方法と陥りがちな罠
kaitokudou
1
220
Importmapを使ったJavaScriptの 読み込みとブラウザアドオンの影響
swamp09
4
1.2k
『ドメイン駆動設計をはじめよう』のモデリングアプローチ
masuda220
PRO
8
440
飲食業界向けマルチプロダクトを実現させる開発体制とリアルな現状
hiroya0601
1
390
ECS Service Connectのこれまでのアップデートと今後のRoadmapを見てみる
tkikuc
2
210
C#/.NETのこれまでのふりかえり
tomokusaba
1
160
ECSのサービス間通信 4つの方法を比較する 〜Canary,Blue/Greenも添えて〜
tkikuc
11
2.3k
外部システム連携先が10を超えるシステムでのアーキテクチャ設計・実装事例
kiwasaki
1
230
Piniaの現状と今後
waka292
5
1.5k
シールドクラスをはじめよう / Getting Started with Sealed Classes
mackey0225
3
400
Webの技術スタックで マルチプラットフォームアプリ開発を可能にするElixirDesktopの紹介
thehaigo
2
920
[PyCon Korea 2024 Keynote] 커뮤니티와 파이썬, 그리고 우리
beomi
0
110
Featured
See All Featured
Thoughts on Productivity
jonyablonski
67
4.3k
Making Projects Easy
brettharned
115
5.9k
Product Roadmaps are Hard
iamctodd
PRO
48
10k
Visualization
eitanlees
144
15k
JavaScript: Past, Present, and Future - NDC Porto 2020
reverentgeek
47
5k
Writing Fast Ruby
sferik
626
61k
What's in a price? How to price your products and services
michaelherold
243
12k
Understanding Cognitive Biases in Performance Measurement
bluesmoon
26
1.4k
The Myth of the Modular Monolith - Day 2 Keynote - Rails World 2024
eileencodes
14
1.9k
Typedesign – Prime Four
hannesfritz
39
2.4k
Let's Do A Bunch of Simple Stuff to Make Websites Faster
chriscoyier
504
140k
Build The Right Thing And Hit Your Dates
maggiecrowley
32
2.4k
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]