quartz/content/notes/11-continuous-integration-2.md
2022-04-09 14:51:02 +12:00

2.5 KiB

title sr-due sr-interval sr-ease tags
11-continuous-integration-2 2022-04-07 3 250
cosc202
lecture
  1. Apprecitae that GitLab is a complex software
  2. Understand where CI jobs scripts get run
  3. explain why repository servers can host websites
  4. Understand how gitblab dternmimines awhen a CI script failed
  5. Describe a way in which CI scripts scan handle secrets
  6. Outline uses of local git hook scripts

CI runs pipelines defined in .gitlab-ci.yaml asynchronously

CI usually tests and builds your projects

Runs on a repo server. Usually persistent, internet accessible

1 GitLab overall architecture

: not in exam

  • many services used

2 GitLab runners

run CI scripts

  • gitlab.com is a cloud computing service
  • allows elf hosting which is what CS does
    • altitude is a GitLab instance at CS
      • servers to host runners that run CI scripts
      • servers that host websites, e.g., cspages.otago.nz
  • GitLab can invoke runners that you host
    • e.g., to use a particular GPU, or other hardware you have
    • GitLab runner itself is a small program written in Go

2.1 Runner architecture

  • runs jobs
    • on isolated infrastructure
    • ... to maintain security
  • that is set up on demand
    • ... handle load variation
  • suits cloud computing

RHS shows GitLab.com's CI hosting: uses google cloud

: not in exam

3 How CI changed website hosting

  • need to share artifacts produced by CI jobs
    • using the web to share artifacts is ideal
    • so now most repo servers also host websites
    • these are static websites: all content is fixed
  • CI can run static website generators (SSGs)
    • git repo contains source code of website
    • CI pipelines transforms source code into HTML files
    • HTML files then hosted as a website by repo sever

e.g., https://cosc202.cspages.otago.ac.nz

4 Debugging CI scripts

  • First ensure config files YAML is valid
    • built in GitLab editor checks YAML as you type
  • commands run from shell that fail return an exit code
    • most Unix shells store exit code of previous commands in $
    • So if variable $? (return code of previous command) is non-zero, the previous command failed
  • Git lab considers CI job as failed if any command fails
    • your shell scripting can choose to hide this exit code
      • e.g., if command supposed to fail; then true; else true; fi
    • Complex scripting? Best to put script in a file and run it from CI

5 Secrets used by CI scripts