Production deployment of the Docker container with Marathon

【ヒカ☆ラボ】Dockerを利用した開発事例～Docker導入から運用まで～
https://atnd.org/events/86945

ref1. Docker / Apache Mesos / Marathon による3倍速いIQONクローラーの構築
http://tech.vasily.jp/entry/iqon-crawler-by-docker-and-mesos-and-marathon

Transcript

1. Production deployment of the
   Docker container with Marathon
   ʲώΧˑϥϘʳDockerΛར༻ͨ͠։ൃࣄྫ
   2017/04/13 Thu. Tatsuro Mitsuno VASILY, inc.
   

   View Slide

2. WHO AM I
   @kotatsu360
   ▸ Tatsuro Mitsuno / ޫ໺ ୡ࿕
   ▸ 2012/04 Yahoo Japan Corporation
   ▸ 2016/04 VASILY, Inc
   ▸ Infrastructure Engineer
   ▸ Twitter GitHub Qiita: @kotatsu360
   Icon illustrated by YOSHI https://ja-jp.facebook.com/yoshi.yone.7
   

   View Slide

3. OUR COMPANY
   ▸ Company: 

   VASILY, Inc. / גࣜձࣾVASILY
   ▸ Address:

   ౦ژ౎඼઒۠੢ޒ൓ా2-23-5 δχΞεϏϧ 6F
   ▸ Employee:

   35
   ▸ Engineer:

   16
   

   View Slide

4. ▸ over 2.5M users
   ▸ over 1.6 M fashion items
   ▸ crawl from hundreds of EC sites
   Θͨ͠ͷʮ޷͖ʯ͕͜͜ʹ͋Δ
   

   View Slide

5. TOC
   ▸ Introduction
   ▸ VASILY Crawler Overview
   ▸ Crawler Application
   ▸ Crawler Infrastructure
   ▸ Container Production Deployment
   ▸ deploy application
   ▸ update container image
   

   View Slide

6. Containerized Application
   Introduction
   

   View Slide

7. CONTAINERIZED APPLICATION
   Containerized Application is ...
   ▸ Cool
   ▸ Modern
   ▸ Difficult
   ▸ The most difficult point I think is production deployment.
   but ...
   

   View Slide

8. >
   CONTAINERIZED APPLICATION
   in Development in Production
   
   run
   simple & easy complexity & difficulty
   EC2
   INSTANCE
   EC2
   INSTANCE
   CI/CD
   service
   container
   image
   registory
   
   ?
   ?
   ?
   ?
   ?
   

   View Slide

9. >
   CONTAINERIZED APPLICATION
   in Development in Production
   
   run
   simple & easy complexity & difficulty
   EC2
   INSTANCE
   EC2
   INSTANCE
   CI/CD
   service
   container
   image
   registory
   
   ?
   ?
   ?
   ?
   ?
   Today's theme
   

   View Slide

10. Overview
    VASILY Crawler
    

    View Slide

11. OVERVIEW
    Crawler Application
    Ruby 2.3.3 Amazon Simple Queue Service (SQS)
    

    View Slide

12. OVERVIEW
    Crawler Application
    ▸ Containerized Application
    ▸ Parallel Distributed Processing Application ( called `worker` )
    XPSLFS
    %
    XPSLFS
    %
    XPSLFS
    $
    XPSLFS
    $
    XPSLFS
    #
    XPSLFS
    #
    XPSLFS
    "
    XPSLFS
    "
    Independent

    from each other via SQS
    1 worker, 1 container
    

    View Slide

13. OVERVIEW
    Crawler Infrastructure
    

    View Slide

14. OVERVIEW
    Apache MESOS
    ▸ A distributed systems kernel
    ▸ Apache Mesos abstracts CPU, memory, storage,
    and other compute resources away from
    machines.
    ▸ Native support for launching containers with
    Docker and AppC images.
    ▸ Mesos Frameworks
    ▸ Hadoop / Cassandra / Marathon / ...
    ref. https://mesosphere.com/why-mesos/
    

    View Slide

15. OVERVIEW
    Marathon
    ▸ A container orchestration platform for Mesos
    and DC/OS
    ▸ Long Running Services
    ▸ The mesos task is executed only once
    ▸ Marathon keeps task state
    ▸ JSON format
    ref. https://github.com/mesosphere/marathon-ui
    {
    "parse": {
    "id": "/iqon/crawler/parse",
    "container": {
    "type": "DOCKER",
    "docker": {
    "image": "xxxxxx"
    }
    }
    }
    }
    

    View Slide

16. OVERVIEW
    Crawler Infrastructure
    AWS EC2 Spot Fleet
    with AutoScaling
    

    View Slide

17. OVERVIEW
    Crawler Infrastructure
    
    Today's theme: production deployment
    AWS EC2 Spot Fleet
    with AutoScaling
    

    View Slide

18. TOC
    ▸ Introduction
    ▸ VASILY Crawler Overview
    ▸ Crawler Application
    ▸ Crawler Infrastructure
    ▸ Container Production Deployment
    ▸ deploy application
    ▸ update container image
    

    View Slide

19. Deploy Application
    Container Production Deployment
    

    View Slide

20. DEPLOY APPLICATION
    deployment strategy: Is container image containg source code?
    > >
    +
    a container image
    NOT containing source code
    a container image
    containing source code
    Icons made by Roundicons from www.flaticon.com is licensed by Creative Commons BY 3.0
    

    View Slide

21. DEPLOY APPLICATION
    deployment strategy: Is source code contained container image ?
    ▸ A container image containing source code is very
    simple and comfortable.
    ▸ but ...
    ▸ A large number of tags are registered in the container registry
    ▸ Can not separate application and environment
    ▸ Container image caching does not work at deployment time
    

    View Slide

22. > >
    DEPLOY APPLICATION
    deployment strategy
    Icons made by Roundicons from www.flaticon.com is licensed by Creative Commons BY 3.0
    +
    adoption
    

    View Slide

23. DEPLOY APPLICATION
    deployment strategy
    >
    A nice mechanism
    to work when
    deploying
    Marathon fetch
    =
    Icons made by Roundicons from www.flaticon.com is licensed by Creative Commons BY 3.0
    

    View Slide

24. DEPLOY APPLICATION
    Marathon fetch
    ▸ The list of URIs to fetch before the
    task starts.
    ▸ fetch (HTTP, HTTPS, FTP, FTPS, ...)
    ▸ [option] extract
    ▸ [option] executable (chmod +x)
    ▸ [option] cache
    ▸ Marathon fetch is build on the
    mechanism of Apache Mesos fetcher.
    {
    "parse": {
    "id": "/iqon/crawler/parse",
    "container": {
    "type": "DOCKER",
    "docker": {
    "image": "xxxxxx"
    },
    "volumes": [
    {
    "containerPath": "/var/app",
    "hostPath": "./iqon_crawler"
    }
    ]
    },
    "fetch": [
    {
    "uri": "https:// xxxxx /iqon_crawler.tar.gz",
    "executable": false,
    "extract": true,
    "cache": false
    }
    ]
    }
    }
    

    View Slide

25. DEPLOY APPLICATION
    Marathon fetch
    Mesos Task Sandbox
    4. docker run --volume "/ path / to / sandbox / root / iqon_crawler : /var/app" ...
    1. fetch: https:// xxxxx /iqon_crawler.tar.gz
    2. iqon_crawler.tar.gz
    3. tar xzf iqon_cralwer.tar.gz
    ‘
    

    View Slide

26. DEPLOY APPLICATION
    Deploy Application
    AWS EC2 Spot Fleet
    with AutoScaling
    Today's theme: production deployment
    

    View Slide

27. 4. pull

    container image
    DEPLOY APPLICATION
    Deploy Application
    AWS
    EC2
    Spot
    Fleet
    1. copy

    source code
    3. fetch

    source code
    5. mount

    & run
    Amazon S3 quay.io (container registory service)
    2. http request
    

    View Slide

28. Update Container Image
    Container Production Deployment
    

    View Slide

29. UPDATE CONTAINER IMAGE
    repository design: Is the Dockerfile placed in the same repository as the application?
    $ tree -L 1 iqon_crawler
    iqon_crawler/
    !"" Dockerfile
    !"" Gemfile
    !"" Gemfile.lock
    !"" README.md
    !"" lib/
    ...
    $ tree -L 1 iqon_crawler
    iqon_crawler/
    !"" Gemfile
    !"" Gemfile.lock
    !"" README.md
    !"" lib/
    ...
    $ tree -L 2 iqon_infra/
    iqon_infra/
    !"" iqon_crawler/
    # $"" Dockerfile
    !"" README.md
    ...
    the application repository
    NOT containing Dockerfile
    the application repository
    containing Dockerfile
    

    View Slide

30. UPDATE CONTAINER IMAGE
    repository design
    ▸ application repository
    ▸ iqon_crawler
    EC2 INSTANCE
    ▸ infrastructure repository
    ▸ iqon_infra
    ▸ Chef / Packer / Vagrant / DNS configure / ...
    APPLICATION APPLICATION APPLICATION
    

    View Slide

31. UPDATE CONTAINER IMAGE
    repository design
    ▸ application repository
    ▸ iqon_crawler
    EC2 INSTANCE
    ▸ infrastructure repository
    ▸ iqon_infra
    ▸ Chef / Packer / Vagrant / DNS configure / ...
    ʁ
    Dockerfile
    ʁ
    APPLICATION APPLICATION APPLICATION
    DOCKER CONTAINER DOCKER CONTAINER DOCKER CONTAINER
    

    View Slide

32. ▸ Easy unified management
    ▸ Easy updating of application circumstances
    ▸ Accelerate deployment (described later)
    UPDATE CONTAINER IMAGE
    Merit of containing Dockerfile Merit of not containing Dockerfile
    Icons made by Freepik from www.flaticon.com is licensed by CC 3.0 BY
    update application
    update Dockerfile
    update
    application
    update
    Dockerfile
    

    View Slide

33. UPDATE CONTAINER IMAGE
    VASILY Repositories
    ▸ application repository
    ▸ iqon_crawler
    EC2 INSTANCE
    ▸ infrastructure repository
    ▸ iqon_infra
    ▸ Chef / Packer / Vagrant / DNS configure / ...
    Dockerfile
    APPLICATION APPLICATION APPLICATION
    DOCKER CONTAINER DOCKER CONTAINER DOCKER CONTAINER
    ʁ
    adoption
    

    View Slide

34. UPDATE CONTAINER IMAGE
    update container image
    branch: master
    branch: dockerbuild
    1. ...
    2. COPY Gemfile / Gemfile.lock
    3. RUN apk add ... && bundle install
    4. ...
    docker build docker push
    Bundle install is unnecessary at the time of deployment
    

    View Slide

35. Conclusion
    

    View Slide

36. CONCLUSION
    Production deployment of the Docker container with Marathon
    ▸ VASILY is operating a containerized crawler system
    ▸ The container image does not contain the source code and it is mounted at deployment time
    ▸ Dockerfile and application are located in the same repository.
    ▸ As the application is updated, the container image is appropriately updated
    

    View Slide

37. CONCLUSION
    Production deployment of the Docker container with Marathon
    ▸ If you are interested, please also read this article
    ▸ Docker / Apache Mesos / Marathon ʹΑΔ3ഒ଎͍IQONΫϩʔϥʔͷߏங

    http://bit.ly/2nD7qxG
    ▸ Reason for containerization
    ▸ Architecture
    ▸ Operational device
    ▸ ...
    

    View Slide

38. We’re Hiring!
    

    View Slide