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
700
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
130
Getting Things Done
dzuelke
1
370
Your next Web server will be written in... PHP
dzuelke
2
230
Your next Web server will be written in... PHP
dzuelke
3
970
Kafka Will Get The Message Across, Guaranteed.
dzuelke
0
220
Heroku at BattleHack Venice 2015
dzuelke
0
110
Designing HTTP Interfaces and RESTful Web Services
dzuelke
6
1.2k
The Twelve-Factor App: Best Practices for Modern Web Applications
dzuelke
4
340
Designing HTTP Interfaces and RESTful Web Services
dzuelke
6
450
Other Decks in Programming
See All in Programming
見えないものに着目すると上手くいく、モデリングの勘所 / invisible-driven-design
minodriven
18
3.7k
「正の影響力」を目指して書くこと
honey32
2
110
Adding Security to Microcontroller Ruby
sylph01
2
2k
AIで生成したものをAIでチェックしてる話
ippey
0
140
コンパウンドプロダクト開発の質とスピードを支える Protobuf と Connect #アーキテクチャ_findy / Boosting Compound Product Development Efficiency with Protobuf and Connect
izumin5210
12
640
ERB, ancient and future
m_seki
3
200
入門プロパティベーステスト/learning-property-based-testing
kiway
6
1k
Datadogのmonitorを Terraform管理に爆速で 移行する
kuro_kurorrr
2
280
アイテムレビュー基盤で導入したアーキテクチャとその成果 / Item Review Introduction Architecture Outcome
chichi1091
1
1k
Using Ruby in the browser is wonderful
ledsun
1
1.6k
AFUP Day Lyon 2024 - DDD !== Archi hexagonale
bgalati
0
190
The Frontend Rubyist
largo
0
470
Featured
See All Featured
Design by the Numbers
sachag
274
18k
Helping Users Find Their Own Way: Creating Modern Search Experiences
danielanewman
21
2k
Testing 201, or: Great Expectations
jmmastey
30
6.8k
Rails Girls Zürich Keynote
gr2m
91
13k
The Pragmatic Product Professional
lauravandoore
26
5.9k
Dealing with People You Can't Stand - Big Design 2015
cassininazir
358
22k
Understanding Cognitive Biases in Performance Measurement
bluesmoon
12
1.1k
From Idea to $5000 a Month in 5 Months
shpigford
377
45k
Principles of Awesome APIs and How to Build Them.
keavy
121
16k
Fireside Chat
paigeccino
22
2.7k
Designing on Purpose - Digital PM Summit 2013
jponch
111
6.5k
I Don’t Have Time: Getting Over the Fear to Launch Your Podcast
jcasabona
23
1.7k
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]