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
モンストを支えるインフラの今とこれから
Search
Isao Shimizu
March 01, 2016
Technology
70
13k
モンストを支えるインフラの今とこれから
dots. Conference Spring 2016
ゲーム開発の裏側
http://eventdots.jp/event/580344
Isao Shimizu
March 01, 2016
Tweet
Share
More Decks by Isao Shimizu
See All by Isao Shimizu
Notion x ポストモーテムで広げる組織の学び / Notion x Postmortem
isaoshimizu
1
160
New Relicを活用したSREの最初のステップ / NRUG OKINAWA VOL.3
isaoshimizu
3
920
「家族アルバム みてね」における運用管理・ オブザーバビリティの全貌 / Overview of Operation Management and Observability in FamilyAlbum
isaoshimizu
5
450
約10年間MIXIのインフラを 支えてきたPagerDutyの活用事例 / PagerDuty on Tour 2024
isaoshimizu
6
1.1k
家族アルバム みてねにおけるGrafana活用術 / Grafana Meetup Japan Vol.1 LT
isaoshimizu
2
1.7k
家族アルバム みてねで直面してきた技術的負債 / MIXI KAG 2024
isaoshimizu
18
9k
今年1年のEKS運用振り返り/3-shake SRE Tech Talk
isaoshimizu
2
370
ポストモーテムの基礎知識と最新事例 / Fundamentals of Postmortem
isaoshimizu
11
3k
全世界1,800万人が利用する「家族アルバム みてね」におけるNew Relic活用法 / FutureStack Tokyo 2023
isaoshimizu
1
560
Other Decks in Technology
See All in Technology
本番環境への影響リスクが低い Real Application Testing (SQL Performance Analyzer) の実施方法の検討と実践
jri_narita
0
230
AIエージェントのオブザーバビリティについて
yunosukey
1
400
株式会社Awarefy(アウェアファイ)会社説明資料 / Awarefy-Company-Deck
awarefy
3
17k
KubeCon EU 2025 Recap - Kubernetes CRD Design for the Long Haul: Tips, Tricks, and Lessons Learned / Kubernetes Meetup Tokyo #70 / k8sjp70-crd-long-haul-recap
everpeace
0
110
インラインRBSコメントに鯛pe checkersもニッコリ
sansantech
PRO
2
200
20250514 1Passwordを使い倒す道場 vol.1
east_takumi
0
160
チェックツールを導入したけど使ってもらえなかった話 #GAADjp
lycorptech_jp
PRO
1
150
DataEngineeringとCloudNativeの今と未来
foursue
0
130
MCP でモノが動くとおもしろい/It is interesting when things move with MCP
bitkey
3
640
Cline&CursorによるAIコーディング徹底活用―Live Vibe Coding付き
pharma_x_tech
2
430
PythonツールであるpygnmiをSONiCのgNMIに対して使ってみた
sonic
0
310
LLMベースAIの基本 / basics of LLM based AI
kishida
9
1.7k
Featured
See All Featured
Making Projects Easy
brettharned
116
6.2k
GraphQLの誤解/rethinking-graphql
sonatard
71
11k
The Art of Delivering Value - GDevCon NA Keynote
reverentgeek
14
1.5k
Chrome DevTools: State of the Union 2024 - Debugging React & Beyond
addyosmani
5
620
Designing Experiences People Love
moore
142
24k
Site-Speed That Sticks
csswizardry
6
560
Docker and Python
trallard
44
3.4k
A Modern Web Designer's Workflow
chriscoyier
693
190k
Reflections from 52 weeks, 52 projects
jeffersonlam
349
20k
Designing for humans not robots
tammielis
253
25k
How STYLIGHT went responsive
nonsquared
100
5.5k
"I'm Feeling Lucky" - Building Great Search Experiences for Today's Users (#IAC19)
danielanewman
227
22k
Transcript
ϞϯετΛࢧ͑ΔΠϯϑϥͷࠓͱ͜Ε͔Β 2016.3.1 dots. Conference Spring 2016 ήʔϜ։ൃͷཪଆ mixi, Inc. XFLAG™
STUDIO @isaoshimizu
About Me • ਗ਼ਫ ܄ SHIMIZU ISAO @isaoshimizu • 20118݄ΑΓגࣜձࣾϛΫγΟ
• ݱࡏΤοΫεϑϥοάελδΦ γεςϜ։ൃ෦ ॴଐ • 2લʹιʔγϟϧɾωοτϫʔΩϯά αʔϏε mixi ͷӡ༻͔Β εϚʔτϑΥϯήʔϜʮϞϯελʔετϥΠΫʯͷӡ༻ 2
ϞϯελʔετϥΠΫ • 201310݄10 ਖ਼ࣜϦϦʔε • iOS൛, Android൛Λఏڙ • શੈքͰ3,000ສਓҎ্͕ϓϨΠʢಉҰͰॏෳμϯϩʔυ͞Εؚͨ·ͣʣ •
ຊɺɺؖࠃɺถɺ߳ߓɾϚΧΦͰల։ • YouTubeͰΞχϝͷ৴ 2,500ສճͷ࠶ੜɺχϯςϯυʔ3DS൛ ग़ՙ100ສຊಥഁ 3
͔͜͜ΒϞϯελʔετϥΠΫ ຊ൛ʹݶఆͨ͠༰ʹͳΓ·͢ 4
ͰݟΔϞϯελʔετϥΠΫͷΠϯϑϥ 5
6 Application 300+ Servers Resque 50 Servers Redis 30 Servers
Memcached 200+ Servers CDN Traffic Max 270Gbps API Traffic Avg 1Gbps TURN 40 Servers MariaDB 250+ Servers Internal Traffic Max 20Gbps
Πϯϑϥߏ 7
8 DC1 DC2 Main Main/Backup Main/Dev
ෳDCͱAmazon VPC • DC 1ʢϝΠϯʣɺDC 2ʢϝΠϯͱόοΫΞοϓʣɺAWSʢϝΠϯͱ։ൃڥʣͰߏ • ֤DCTY2,4ܦ༝ͰAmazon VPCͱDirectConnectͰϓϥΠϕʔτଓ •
֤DCؒยܥো࣌ʹ͓͍ͯ20Gbps·Ͱ͑ΒΕΔઃܭ • ৄ͍͠ղઆ “ڌͨ͠ʢ෦τϥϑΟοΫϘϦϡʔϜฤʣ” http://xflag.com/blog/infradb/internal_traffic_volume.html 9
֤ڌͷׂ 10
11 DC1 DC2 Main Main/Backup Main/Dev Application MariaDB Redis Resque
Memcached MariaDB Redis Memcached TURN Server ։ൃ༻Πϯελϯε܈ Application Resque
֤ڌͷׂ • ຊ൪ͷApplicationɺResqueDC1ɺDC2ͷ྆ํͰՔಇ • ຊ൪ͷRedisɺMariaDBɺMemcached • DC1ʹMaster/Backupͷηοτ • DC2ʹBackup •
BackupDC1ͱDC2߹Θͤͯ2ηοτ͋Δ • AWSͰຊ൪ͷTURN ServerʢϚϧνϓϨΠ༻ʣɺ։ൃ༻Πϯελϯε܈ͳͲ͕Քಇɻ 12
ΫϥΠΞϯτΞϓϦ͔ΒݟͨΠϯϑϥ 13
14 DC1 DC2 AWS TURN Server (EC2) CDN (CloudFront, Akamai)
API API LB LB Internet Client
ΫϥΠΞϯτΞϓϦ͔ΒͷΞΫηεઌ • DC1ͱDC2ʹ͋ΔϩʔυόϥϯαʔʢHAߏʣ • ϩʔυόϥϯαʔͷΞΫηεDNSϥϯυϩϏϯʢ2ͭͷԾIPʣ • ϩʔυόϥϯαʔ͔Β֤ڌͷApplicationαʔόৼΓ͚ • AWSͰϚϧνϓϨΠ༻ʹTURN ServerʢEC2
+ Elastic IPʣ • ήʔϜσʔλɾϦιʔεσʔλ༻ʹCDNʢCloudFront + S3ʣɻAkamaiซ༻ɻ 15
σϓϩΠ 16
17 DC1 DC2 Deploy Server S3 Bucket tarball S3 PUT
GitHub ssh Code Internet
σϓϩΠ • Capistrano 2 • ඞཁͳͱ͖ʹඞཁͳ͚ͩσϓϩΠ͢Δ • capistrano-s3copy-awscli https://github.com/bacter1a/capistrano-s3copy-awscli •
GitHub͔Β࠷৽ͷίʔυΛऔಘͯ͠S3Ξοϓϩʔυ༻ʹtarballΛ࡞ • ApplicationαʔόͰS3ʹΞοϓϩʔυ͞ΕͨtarballΛμϯϩʔυɺల։ • GitHubͷଓ੍ݶରࡦʢgithub.comେྔʹଓ͕ཁٻ͞ΕΔͱڋ൱͞ΕΔͨΊʣ 18
• capistrano-rbenv https://github.com/yyuu/capistrano-rbenv • όʔδϣϯࢦఆͰRubyͷɺΞοϓσʔτָ͕ʹ • Ruby 2.2.4ಋೖࡁΈ 19
ϓϩϏδϣχϯά 20
21 DC1 DC2 Deploy Server S3 Bucket GitHub ssh Chef
Cookbook Ubuntu Package Mirror Internet deb Package deb Package
ϓϩϏδϣχϯά • ChefʹΑΔϓϩϏδϣχϯά͕جຊ • capistrano-paratrooper-chef https://github.com/tk0miya/capistrano-paratrooper-chef • ཧϚγϯͷॳظηοτΞοϓʢ࠷ݶʣAnsibleΛ͏ • OSΫϦʔϯΠϯετʔϧޙͷঢ়ଶͰɺChef͕Πϯετʔϧ͞Ε͍ͯͳ͍ͨΊ
22
• Community CookbookΛincludeͨ͠ಠࣗCookbookΛ࡞ɺར༻͍ͯ͠Δ • ಠࣗCookbookɺXFLAGશମͷΠϯϑϥͰڞ௨ͯ͑͠ΔΑ͏ʹ • ͯ͢ͷαʔό͕ChefͰߏஙՄೳ • Քಇதͷຊ൪DBɺMemcachedʹ࠶ద༻͕ͯ͠ى͖ͳ͍Α͏ʹ 23
• S3ʹMariaDBͷϨϙδτϦͷϛϥʔʢຊՈͷোରࡦʣ • ioDriveͳͲͷυϥΠόஔ͍ͯ͋Δ • aptlyͰߏங http://www.aptly.info/ • ݸਓͷ։ൃϚγϯͰVagrant +
VirtualBox + ChefͰߏͰ͖ΔΑ͏ʹ • Test Kitchenͷ׆༻ https://github.com/test-kitchen/test-kitchen 24
ෛՙରࡦ 25
ࠓʹೖ͔ͬͯΒߴෛՙ͕ଓ͘… 26
ෛՙରࡦ • DBͷγϟʔσΟϯάɺςʔϒϧׂΛܧଓ࣮ࢪ • Applicationαʔόͷ૿ઃ 12-20 core ͳCPU x 2ʢHyper-ThreadingࢉͰ24-40
coreʣͷϚγϯ͕جຊߏ • ΫΤϦվળΩϟογϡରԠ ioDrive(ioMemory)͔ΒSSDͷεέʔϧμϯͷ࣮ݱ • Memcachedͷ૿ઃ 27
Memcachedͷߏ 28
Memcachedͷߏ • Memcached2ϓʔϧߏ • 1ϓʔϧ͋ͨΓ100 • Ϛγϯ1͋ͨΓͷϝϞϦׂΓͯ26GBʢϝϞϦ32GBࡌͷϚγϯʣ • 1ϓʔϧશମͰ2.6TBͷ༰ྔΛͭ •
྆ํͷϓʔϧʹಉ࣌ॻ͖ࠐΈ DoubleWriteCacheStores https://github.com/hirocaster/double_write_cache_stores • ϝϯςφϯεແ͠ͰϓʔϧͷΓସ͕͑ՄೳɻϓʔϧαΠζͷ૿ڧ͕Մೳʹɻ 29
2ϓʔϧߏͷMemcached ೖΕସ͑ํ๏ 30
31 App M App App M M M M M
M M M M M M set get set Pool 1 Pool 2 App App App M M M M M M set get Pool 1 Pool 2 ௨ৗ࣌ ఀࢭɾ૿ઃ M M શΫϦΞ͓ͯ͘͠ M M M M M M Pool2ͷsetΛఀࢭ M Memcachedϗετ
32 App M App App M M M M M
set get set Pool 1 Pool 2 Γସ͑ M M M M M M M M App App App set get Pool 1 Pool 2 Ωϟογϡ͋ͨͨΊ set M M M M M M M M M M M M M M ͠Β͋ͨͨ͘ΊΔ
ϩʔυόϥϯαʔ 33
ϩʔυόϥϯαʔ • ApplicationαʔόͷલஈΞϓϥΠΞϯεͷϩʔυόϥϯαʔΛར༻ • A10 Networks AX2500 • RESTful APIʹΑΔૢ࡞͕Մೳ
• ࣗࣾ։ൃͷGoͷCLIπʔϧʢa10-cliʣ • ApplicationαʔόͷՃɺআ࣌ʹίϚϯυ1ͭͰαʔϏεΠϯɾΞτ͕Ͱ͖Δ 34
35 nginx unicorn nginx unicorn nginx unicorn nginx unicorn MemcachedMariaDBͳͲ
Internet A10 ϩʔυόϥϯαʔ Applicationαʔό
ϞχλϦϯάɾࢹ 36
37 CloudForecast • ͯ͢ͷαʔόͷϝτϦΫεऔಘ • 1,000ͷαʔό͕ܭଌର • ༷ʑͳಠࣗͷϝτϦΫεΛՃ https://github.com/kazeburo/cloudforecast
38 GrowthForecast • DCϥοΫͷిྗͷάϥϑԽ • Fluentd Monitoring AgentͷάϥϑԽ http://kazeburo.github.io/GrowthForecast/
39 • ͯ͢ͷαʔόͷࢹ • 12,000Λ͑Δࢹ߲ • ಠࣗͷϓϥάΠϯΛՃ • PagerDutyͱͷ࿈ܞ •
൪ͷεέδϡʔϦϯά • ΤεΧϨʔγϣϯϙϦγʔ • Slackͷ௨ https://www.nagios.org/
40 • nginx, ApplicationͷϩάΛ FluentdΛͬͯElasticsearchసૹ • APIຖͷॲཧ࣌ؒϨεϙϯελΠϜͷܭଌ • Τϥʔසͷܭଌ •
ಛఆϗετ͔ΒͷΞΫηεͷௐࠪ • Slow QueryͷՄࢹԽ • CloudTrailϩάͷՄࢹԽ https://www.elastic.co/
41 nginx unicorn Amazon S3 GrowthForecast LTSV Log Application Servers
ϩάసૹϑϩʔ Monitoring Agent Response Time Status Code Monitoring Agent Log Log td-agent 2.3.x Max 500Mbps Max 500Mbps
ৗͷπʔϧ 42
43 • ίϛϡχέʔγϣϯSlackͰ౷Ұ • Hubot https://hubot.github.com ར༻ʹΑΔChatOpsͷ࣮ݱ • ։ൃڥͷσϓϩΠͳͲΛSlack͔Β࣮ߦՄೳ •
ৄ͘͠Software Design 20161݄߸ͷChatOpsಛूʹ http://gihyo.jp/magazine/SD/archive/2016/201601 • ΞϓϦέʔγϣϯίʔυΠϯϑϥߏங༻ͷCookbookͳͲ ͯ͢GitHub্Ͱཧ • pull requestsϕʔεͷ։ൃɺϨϏϡʔ • GitHub্Ͱͷ׆ಈSlack௨͞ΕΔ • GitHubͰpull requests͕࡞͞ΕΔͱࣗಈςετ࣮ߦ • Dockerίϯςφͷੜɺςετ࣮ߦɺഁغ • Hubotͱͷ࿈ܞ
͜Ε͔Β͍͖͍ͬͯͨ͜ͱ 44
• ίʔυͷվળɺεέʔϧΞτɺΩϟογϡར༻ͳͲ • εέʔϧΞοϓͰ͖Δ͚ͩΓͨ͘ͳ͍ • ݁Ռతʹεέʔϧμϯʹͭͳ͕Δͱϕετ • ͜ΕΒैདྷ͔Βܧଓ࣮ͯ͠ࢪ͍ͯ͠Δ • ιϑτΣΞͷܧଓతΞοϓσʔτ
• Ubuntu Server 14.04 or 16.04 LTSʢ4/21ϦϦʔε༧ఆʣͷҠߦ • KernelɺϛυϧΣΞͷΞοϓσʔτੵۃతʹ࣮ࢪ 45
• MemcachedͷSlab Rebalancingʢ slab_reassign, slab_automove ʣ • ·ͩௐࠪɺݕূஈ֊ͳͷͰӡ༻ࣄྫ͕͋ΕΓ͍ͨ • ࢹɺϞχλϦϯάڥͷվળ
• ෛՙҟৗʹΑΓૉૣ͘ؾ͚ͮΔӡ༻ʹ • ͏গࣗ͠ಈԽ͍ͨ͠ͱ͜Ζ • Consulͱͷ࿈ܞ 46
• ࣾͷ৽نҊ݅ͰHashicorpࣾπʔϧ Terraformͷར༻͕࢝·͍ͬͯΔ • Ϟϯετʹޙ͔Βಋೖ͢ΔͷίετϦεΫ͕ߴ͍ͷͰಋೖ༧ఆͳ͠ • ෦తʹπʔϧར༻ʹΑΔޮԽ͓͜ͳ͍͖͍ͬͯͨ • ΠϯϑϥςετͷޮԽ •
DockerServerspecͳͲΛ͞Βʹ׆༻͍͖͍ͯͨ͠ • Chef CookbookͷϦϑΝΫλϦϯάɺͬͱγϯϓϧʹ 47
࠷ޙʹ 48
ΤϯδχΞืूதʂ 49 http://xflag.com/recruit/career/engineer.html
͋Γ͕ͱ͏͍͟͝·ͨ͠ 50