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
720
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
380
Your next Web server will be written in... PHP
dzuelke
2
240
Your next Web server will be written in... PHP
dzuelke
3
1k
Kafka Will Get The Message Across, Guaranteed.
dzuelke
0
220
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
370
Designing HTTP Interfaces and RESTful Web Services
dzuelke
6
460
Other Decks in Programming
See All in Programming
Cloudflare Workers x AWS Lambdaの組み合わせユースケース / Cloudflare Workers x AWS Lambda Combination Use Case
seike460
PRO
2
310
はしめてのプログラミングとロボット制御
watawatavoltage
0
290
The rollercoaster of releasing an Android, iOS, and macOS app with Kotlin Multiplatform | droidcon Berlin
prof18
0
110
Temporalを取り巻く仕様を整理する
sajikix
0
110
なぜ宣言的 UI は壊れにくいのか / Why declarative UI is less fragile
uenitty
29
13k
feature環境をGitHub ActionsとCloudFormationでいい感じに管理する
nealle
2
310
初心者がおさえておきたいAWS CDKのベストプラクティス 2024
konokenj
15
7.3k
HMSコンペ 11th Solution (team : kansai-kaggler)
t88
1
680
DynamoDB コスト最適化っぽいことの基本 with Terraform
kuro_kurorrr
2
250
開発部に不満を持っていたCSがエンジニアにジョブチェンしてわかった「勝手に諦めない」ことの大切さ
sakuraikotone
28
16k
Javaの現状2024夏 / Java current status 2024 summer
kishida
4
1.4k
最古の関数型言語「Lisp」ことはじめ / lisp_in_kamiyama
uhooi
1
190
Featured
See All Featured
Git: the NoSQL Database
bkeepers
PRO
423
64k
Building Effective Engineering Teams - LeadDev
addyosmani
47
2.2k
Build your cross-platform service in a week with App Engine
jlugia
227
17k
Principles of Awesome APIs and How to Build Them.
keavy
124
16k
ParisWeb 2013: Learning to Love: Crash Course in Emotional UX Design
dotmariusz
105
6.8k
I Don’t Have Time: Getting Over the Fear to Launch Your Podcast
jcasabona
26
1.8k
Refactoring Trust on Your Teams (GOTO; Chicago 2020)
rmw
29
2.5k
Reflections from 52 weeks, 52 projects
jeffersonlam
346
19k
No one is an island. Learnings from fostering a developers community.
thoeni
17
2.8k
StorybookのUI Testing Handbookを読んだ
zakiyama
15
4.9k
Building Applications with DynamoDB
mza
89
5.8k
How To Stay Up To Date on Web Technology
chriscoyier
784
250k
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]