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
250
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
400
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
780
Heroku at BattleHack Venice 2015
dzuelke
0
120
Designing HTTP Interfaces and RESTful Web Services
dzuelke
6
1.4k
The Twelve-Factor App: Best Practices for Modern Web Applications
dzuelke
4
430
Designing HTTP Interfaces and RESTful Web Services
dzuelke
6
470
Other Decks in Programming
See All in Programming
もう少しテストを書きたいんじゃ〜 #phpstudy
o0h
PRO
17
4k
SwiftUI Viewの責務分離
elmetal
PRO
2
270
Rubyと自由とAIと
yotii23
6
1.7k
Rails アプリ地図考 Flush Cut
makicamel
1
130
CloudNativePGを布教したい
nnaka2992
0
110
Boos Performance and Developer Productivity with Jakarta EE 11
ivargrimstad
0
310
Formの複雑さに立ち向かう
bmthd
1
930
「個人開発マネタイズ大全」が教えてくれたこと
bani24884
1
190
Unity Android XR入門
sakutama_11
0
180
たのしいSocketのしくみ / Socket Under a Microscope
coe401_
8
1.3k
5分で理解する SOLID 原則 #phpcon_nagoya
shogogg
1
300
Datadog DBMでなにができる? JDDUG Meetup#7
nealle
0
140
Featured
See All Featured
Code Review Best Practice
trishagee
67
18k
Faster Mobile Websites
deanohume
306
31k
Optimizing for Happiness
mojombo
376
70k
Principles of Awesome APIs and How to Build Them.
keavy
126
17k
A better future with KSS
kneath
238
17k
Building a Modern Day E-commerce SEO Strategy
aleyda
38
7.1k
A designer walks into a library…
pauljervisheath
205
24k
Agile that works and the tools we love
rasmusluckow
328
21k
Adopting Sorbet at Scale
ufuk
74
9.2k
StorybookのUI Testing Handbookを読んだ
zakiyama
28
5.5k
Raft: Consensus for Rubyists
vanstee
137
6.8k
The World Runs on Bad Software
bkeepers
PRO
67
11k
Transcript
KAFKA WILL GET THE MESSAGE ACROSS. GUARANTEED. PHP Benelux 2017
Belgium
David Zuelke
None
dz@heroku.com
@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 & dz@heroku.com