Please enable JavaScript.

Coggle requires JavaScript to display documents.

CI Brainstorm (CI (Circle CI (Travis like, Output is splitted in steps, …

- - - - De facto standard for CI
      - Admin can delegate configuration to super user in each projects
      - Extensive set of plugins available
      - Can "pause" build with UI (button, text fields,...)
    - - Single Master
      - Java
  - - - Cristal clear integration with GitHub
      - De facto standard for OpenSource projects
      - Self Hosted: https://enterprise.travis-ci.com/
    - - Single output
  - - - Really good documentation
      - autoscale runner
        
        docker machine
        
        kubernetes
      - go language / single binary / linux, mac, win
      - add dynamicaly additional runners => no config files to update
      - handle deployment in different environment auto & manual
      - all in one
      - branch protection
        
        do not merge on build failure
        
        automerge on build success
    - - Build system do not support "matrix combinations" feature
      - all in one
      - not fully open sourced (enterprise version w/ more features)
        
        detailed stats
        
        kerberos
        
        squash and merge
        
        push rules restrictions
        
        merge request approval
    - - IT Juliette would take care of it
      - Preference of Juliette (over GitHub)
      - Need local mirror in TLS
- - - - groups management
      - compatible w/ jenkins
      - import project from github
      - git LFS support
      - group/subgroup management
      - todo list updated based on mention
      - move ticket from one project to another
- - - - Reconnection timeout
      - Can automatically restore from a given message ID
- - - - Cross host request load balancing (1 instances per host). Recommended ("Service Mesh" model).
        Use daemonset on kubernetes
      - Sidecar model: per pod. More expansive
  - - - Inject tracing context to trace a request in log (calls visualization a la Chrome's Developer Panel Performance)
      - Grafana dashboard
      - Kibana ingest
    - - Zk, etc are fully consistent but service stability is an "eventually consistent problem. Can be easily handled by Envoy (because is linked to every service and can health check them)
      - Configuration needed for health checks (can delegate this to kubernetes?) :!?:
        Eventhough, duplicate active health check is not a performance issue
      - Don't seem to be linkable to etcd/zk/... :explode:
    - - Exponential retry by default
      - outlier detection (penalty box for service returning HTTP 500)
      - No sign of retry budget ?
    - - external service (also open source)
      - Shadowing (test new feature in prod on a shadow service)
    - - Istio brings Envoy to Kubernetes :explode:
    - - Zone aware (stay in same DC)
      - Zone stats, Canary specify stats
      - Blue/Green deployment, Canary deployment
- - - - Per project workflow
      - Scrum Workflow available
    - - Complex to setup
  - - - Built in GitHub
      - Per project
    - - less configurable than Jira
- - - - Authenticated
      - Encrypted
      - Auditable
      - Logged
        
        each request should be logged somewhere
- - - - Cons
        
        Need a patch for each change
      - Pros
        
        Versioned along with the code
    - - Used by Traefik
      - Everythink in Markdown