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
trinity で Cloud Composer に ワークフローを簡単デプロイ / Easy...
Search
Hiroka Zaitsu
October 25, 2019
Technology
0
880
trinity で Cloud Composer に ワークフローを簡単デプロイ / Easy workflow deployment to Cloud Composer with trinity
2019.10.25 Fukuoka.go#14+Umeda.go
https://fukuokago.connpass.com/event/146447/
Hiroka Zaitsu
October 25, 2019
Tweet
Share
More Decks by Hiroka Zaitsu
See All by Hiroka Zaitsu
GMOペパボのデータ基盤とデータ活用の現在地 / Current State of GMO Pepabo's Data Infrastructure and Data Utilization
zaimy
2
130
ビジネス職が分析も担う事業部制組織でのデータ活用の仕組みづくり / Enabling Data Analytics in Business-Led Divisional Organizations
zaimy
1
500
Vertex AI Matching Engine と CLIP を使って EC サービスの類似画像検索機能を作る / Development of similar image search function for EC services using Vertex AI Matching Engine and CLIP
zaimy
0
740
BigQuery の日本語データを Dataflow と Vertex AI でトピックモデリング / Topic modeling of Japanese data in BigQuery with Dataflow and Vertex AI
zaimy
1
5.9k
データサイエンティストの仕事紹介 / Data Scientist Job Introduction
zaimy
1
600
GMOペパボのサービスと研究開発を支えるデータ基盤の裏側 / Inside Story of Data Infrastructure Supporting GMO Pepabo's Services and R&D
zaimy
1
1.8k
正則化とロジスティック回帰/machine-learning-lecture-regularization-and-logistic-regression
zaimy
0
8.7k
ECサイトにおける閲覧履歴を用いた購買に繋がる行動の変化検出 / Change Detection in Behavior Followed by Possible Purchase Using Electronic Commerce Site Browsing History
zaimy
1
930
ハンドメイド作品を対象としたECサイトにおける大量生産品の検出 / Detection of Mass-produced Goods at EC Site to Trade Handmade Goods
zaimy
3
4.8k
Other Decks in Technology
See All in Technology
人と生成AIの協調意思決定/Co‑decision making by people and generative AI
moriyuya
0
230
激動の時代、新卒エンジニアはAIツールにどう向き合うか。 [LayerX Bet AI Day Countdown LT Day1 ツールの選択]
tak848
0
630
【CEDEC2025】ブランド力アップのためのコンテンツマーケティング~ゲーム会社における情報資産の活かし方~
cygames
PRO
0
160
興味の胞子を育て 業務と技術に広がる”きのこ力”
fumiyasac0921
0
450
[MIRU25] NaiLIA: Multimodal Retrieval of Nail Designs Based on Dense Intent Descriptions
keio_smilab
PRO
1
160
OpenTelemetry の Log を使いこなそう
biwashi
5
1.2k
20250728 MCP, A2A and Multi-Agents in the future
yoshidashingo
1
160
私とAWSとの関わりの歩み~意志あるところに道は開けるかも?~
nagisa53
1
140
Bet "Bet AI" - Accelerating Our AI Journey #BetAIDay
layerx
PRO
1
480
Vision Language Modelと自動運転AIの最前線_20250730
yuyamaguchi
2
900
2025-07-25 NOT A HOTEL TECH TALK ━ スマートホーム開発の最前線 ━ SOFTWARE
wakinchan
0
180
AI駆動開発 with MixLeap Study【大阪支部 #3】
lycorptech_jp
PRO
0
280
Featured
See All Featured
Faster Mobile Websites
deanohume
308
31k
RailsConf 2023
tenderlove
30
1.2k
Exploring the Power of Turbo Streams & Action Cable | RailsConf2023
kevinliebholz
34
5.9k
Product Roadmaps are Hard
iamctodd
PRO
54
11k
Typedesign – Prime Four
hannesfritz
42
2.7k
Design and Strategy: How to Deal with People Who Don’t "Get" Design
morganepeng
130
19k
Improving Core Web Vitals using Speculation Rules API
sergeychernyshev
18
1k
Building Applications with DynamoDB
mza
95
6.5k
Refactoring Trust on Your Teams (GOTO; Chicago 2020)
rmw
34
3.1k
Learning to Love Humans: Emotional Interface Design
aarron
273
40k
Dealing with People You Can't Stand - Big Design 2015
cassininazir
367
26k
The Cost Of JavaScript in 2023
addyosmani
51
8.7k
Transcript
ࡒେՆ / Pepabo R&D Institute, GMO Pepabo, Inc. 2019.10.25 Fukuoka.go#14+Umeda.go
trinity Ͱ Cloud Composer ʹ ϫʔΫϑϩʔΛ؆୯σϓϩΠ
σʔλαΠΤϯςΟετ ࡒ େՆ / @zaimy 2 Hiroka Zaitsu ϖύϘݚڀॴ ݚڀһ
1. Cloud Composer ͱ 2. Cloud Composer ͷσϓϩΠ࣌ͷࠔΓ͝ͱ 3. trinity
ʹΑΔղܾͷࢼΈ 4. ࠓޙΔ͜ͱ 3 ࣍
1. Cloud Composer ͱ
• GCP ͷ "ϑϧϚωʔδυͷϫʔΫϑϩʔ ΦʔέετϨʔγϣϯ αʔϏε" • Apache Airflow Λ
GCP ্ʹߏங͢Δ • ϖύϘͷϩάج൫ʢDWHʣΛ Treasure Data ͔Β GCP Ҡߦத • ϫʔΫϑϩʔαʔϏε Treasure Workflow (Ϛωʔδυ Digdag) ͔Β Cloud Composer Ҡߦத 5 Cloud Composer ͷ֓ཁ
ϫʔΫϑϩʔͷίʔυϕʔε repository └ dags ɹ ├ workflowA ɹ │ ├
main.py ɹ │ └ hoge.sql ɹ └ workflowB ɹ ɹ ├ main.py ɹ ɹ └ piyo.sql 6 • dags σΟϨΫτϦԼʹϫʔΫϑϩʔ୯ҐͰ αϒσΟϨΫτϦΛΔ • ϫʔΫϑϩʔຊମʢDAGʣͷ python ίʔυ • ϫʔΫϑϩʔͰར༻͢ΔΫΤϦ • ઃఆϑΝΠϧɹͳͲ ※σΟϨΫτϦߏΛ Cloud Storage ͱ߹ΘͤΔ߹
ϫʔΫϑϩʔͷσϓϩΠʢՃͱߋ৽ʣ $ gcloud composer environments storage dags import \ --environment
ENVIRONMENT_NAME \ --location LOCATION \ --source LOCAL_FILE_TO_UPLOAD 7 ίʔυϕʔε $MPVE4UPSBHF "JSqPX HDMPVEDPNQPTFSJNQPSU
ϫʔΫϑϩʔͷআ ͦͷ1 - Cloud Storage ͔Βআ $ gcloud composer environments
storage dags delete \ --environment ENVIRONMENT_NAME \ --location LOCATION \ DAG_NAME.py 8 ίʔυϕʔε $MPVE4UPSBHF "JSqPX HDMPVEDPNQPTFSEFMFUF
ϫʔΫϑϩʔͷআ ͦͷ2 - Airflow ͔Βআ $ gcloud composer environments run
--location LOCATION \ ENVIRONMENT_NAME delete_dag -- DAG_NAME 9 ίʔυϕʔε $MPVE4UPSBHF "JSqPX HDMPVEDPNQPTFSEFMFUF@EBH
2. Cloud Composer ͷ σϓϩΠ࣌ͷࠔΓ͝ͱ
• ϫʔΫϑϩʔͷՃͱߋ৽ • import ϫʔΫϑϩʔ୯ҐͰͷ࣮ߦ • ࠩͷ͋ΔϫʔΫϑϩʔʹରͯ͠ݸผʹ࣮ߦ͢Δඞཁ͕͋Δ • import
Cloud Storage ͷϑΝΠϧΛ্ॻ͖͢Δ • ίʔυϕʔεͰআͨ͠ϑΝΠϧ ݸผʹআ͠ͳ͍ݶΓ Cloud Storage ʹΔ 11 gcloud ίϚϯυΛͦͷ··ӡ༻ʹ͏ͱେม
• ϫʔΫϑϩʔͷআ • delete ͱ Airflow ͷ dag_delete ͷ2ճίϚϯυΛ࣮ߦ͢Δඞཁ͕͋Δ •
delete ϑΝΠϧ୯Ґ, dag_delete ϫʔΫϑϩʔ୯ҐͰͷ࣮ߦ • ࠩͷ͋ΔϑΝΠϧ/ϫʔΫϑϩʔʹରͯ͠ݸผʹ࣮ߦ͢Δඞཁ͕͋Δ • ։ൃʹΑΓेݸͷϫʔΫϑϩʔʹʑ͕ࠩੜ·Ε͍ͯ͘ • ࠩΛػցతʹݕग़ͯ͠ Cloud Composer ʹಉظ͍ͨ͠ 12 gcloud ίϚϯυΛͦͷ··ӡ༻ʹ͏ͱେม
• όέοτ/σΟϨΫτϦؒͰϑΝΠϧΛಉظ͢Δ Cloud Storage ͷίϚϯυ • ϑΝΠϧͷߋ৽࣌ࠁʹࠩҟ͕͋Εಉظରͱఆ͞ΕΔ • ༰͕มߋ͞Ε͍ͯͳͯ͘ॲཧରʹͳͬͯ͠·͏ •
Cloud Storage ʹґଘ͢Δ • Airflow GCP Ҏ֎ͰߏஙͰ͖ΔͷͰଞͷετϨʔδʹରԠ͍ͨ͠ 13 gsutil rsync Ͳ͏͔ͳ
• ಛఆͷ git ϦϙδτϦͱಉظ͢Δ Airflow ͷػೳ • ୯ҰͷϒϥϯνͷΈࢦఆՄೳ • ຊ൪ڥʹ
master ͷίʔυΛಉظ͢Δʹྑͦ͞͏ • ςετڥ CI Ͱ feature branch ͷίʔυΛσϓϩΠ͍ͨ͠ 14 Airflow sync Ͳ͏͔ͳ
3. trinity ʹΑΔղܾͷࢼΈ
• ίʔυϕʔεͱ Cloud Storage ͱ Airflow ͷ3ͭΛಉظ͢Δ • ϫʔΫϑϩʔ୯ҐͰɺσΟϨΫτϦߏͱϑΝΠϧ༰͔ΒϋογϡΛܭࢉ •
͋Δ࣌ͷϫʔΫϑϩʔఆٛΛද͢ϋογϡ • ίʔυϕʔε͔Βܭࢉͨ͠ϋογϡͱ Cloud Storage ʹอଘ͞Ε͍ͯΔ ϋογϡ͕ҟͳΔϫʔΫϑϩʔΛಉظૢ࡞ͷରʹ͢Δ 16 trinity ͷํ
• https://github.com/zaimy/trinity • A tool to synchronize workflows between Codebase,
Cloud Storage and Airflow metadata. • ͳͥ Goʁ • ΫϩείϯύΠϧͰ Mac, Linux, Windows ʹରԠͰ͖Δ • ϫʔΫϑϩʔ୯ҐͰॲཧ͕ՄೳͳͷͰฒྻԽ͍ͨ͠ 17 trinity ͷ࣮ $ trinity --bucket=BUCKET_NAME \ --composer-env=COMPOSER_ENV_NAME
1. ίʔυϕʔεͰϋογϡΛܭࢉͯ͠ϫʔΫϑϩʔ͝ͱʹอଘ 2. ίʔυϕʔεͱ Cloud Storage ͷϫʔΫϑϩʔΛϦετͯ͠ൺֱ i. ίʔυϕʔεʹ͔͠ͳ͚Ε Cloud
Storage ʹΞοϓϩʔυʢՃʣ ii. Cloud Storage ʹ͔͠ͳ͚Ε Cloud Storage ͱ Airflow ͔Βআ iii. ྆ํʹ͋Είʔυϕʔεͱ Cloud Storage ͷϋογϡΛൺֱ a. ࠩҟ͕͋Ε Cloud Storage ͷϫʔΫϑϩʔΛஔʢߋ৽ʣ 18 ॲཧͷྲྀΕ
؆୯ʹಉظతͳσϓϩΠ͕ Ͱ͖ΔΑ͏ʹͳͬͨ !
• ςετՃͱϦϑΝΫλϦϯά • Go ͷ࡞๏ߟ͑ํʹԊ͍͖͍ͬͯͨ • ػೳՃ • Airflow ʹ
dags Ҏ֎ʹ plugins ͋ΔͷͰରԠ͢Δ • dry-run 20 ࠓޙΔ͜ͱ
None