DevOps conf 2015 (9/1)

*因為沒有提供投影片，所以抄的零零落落

開場

• 搶先一步的IT競爭力
• 南非Standard Bank加快發佈軟體2000倍
• 反之，若是競爭對手的速度快2000倍
• 實例Standard Bank
• 除了Dev與Ops Team外，更應用在流程面
• 新服務申請新的Server，層層的人工簽核
• 由於改變流程，因此花時間讓高層瞭解
• 熟悉用詞
• 導入衝擊
• 建立文化
• 將Dev, Ops於同一辦公室工作
• 宣言服裝(ChopChop)

Conquer CI Server - 樂天 直井和久

• 因無人專職維運CI Server，同仁也不太會用，所以效能低落
• 專職維運CI Server
• Project Metrics
• 指標太多，訂出要衡量什麼，才能取捨
• Show progress easily
• observer the transition
• Reduce exchange between Dev & Ops team
• 許多環境設定waste time on communication
• 透過CHEF, Docker
• Developer可自行測試
• Operation直接放設定檔
• 調整CI Alert mail
• 寫看的懂，留useful的資訊
• you don't need to measure everything. only useful information
• 逐步設定其他的metric
• how we nurture member
• share knowledge face-to-face weekly.
• Conclusion
• Find/measure/review Project Metrics to consolidate ... 忘了
• Use useful information for communication
• 相關軟體
• Cobertura
• Jacoco

Enterprise DevOps - Michael Ducy

• Cultural and Professional movement
• Book: The Phoenix Project
• Increasing the flow of work through the IT system
• DevOps are CALMS
• Culture (learning, improvement, failure and betterment)
• Automation (work flow)
• Lean 
• Measurement (看的到，才知道要改善什麼)
• Sharing 
• "... software is eating the world" Marc Anderson
• Lean
• 每次上版要花多少時間在process
• it's a waste
• 30 Day Challenges (DevOps Dojo)
• Time boxed Project
• Give origination a chance to experiment
• Eliminate old process to prove new 
• Eliminate waste
• Culture 無法被管理
• 是有最底層的membr形成
• Managers are no longer a manager
• "the team i support"
• The Coaching kata
• 
  

Yahoo 持續交付 - 應百怡

• CD如何保持Speed & Quality
• Speed
• Eliminate human involvement
• Small batch
• Easy to identity issue
• Check & revise
• Quality
• Build the  right thing
• Small batch
• Build the thing right
• Lots test
• No human allowed
• Yahoo CEO
• NO CD, NO LAUNCH, NO FOOLING
• 5 Stage

雲端OS Team落實CI - 雙子星執行長 符儒嘉

• Biz -> Dev -> Test -> Ops -> Market
• Dev過程有Agile
• Test有CD
• Ops有 infrastructure as code
• Branching Methodology
• feature 
• release integration
• release candidate

Whoscall - 葉秉哲

• get a little better every single day
• 要monitoring什麼
• 從Risk management出發(risk有機會的意思)
• CPU Usage, 不busy的可調整
• Application 
• Metric
• availible
• response time
• error
• Service
• pingdom => slack
• Crashlytics => slack
• 分析crash report
• DB
• Metric
• R/W lock
• Size
• Load
• Speed
• Service
• MongoDB Cloud Manager
• Poor Performance
• Metric
• instance metrics
• application metrics
• resource
• Malfunction
• log分析
• fluentd->elesticsearch
• 較複雜的用BigQuery
• New Realic
• 看stack trace
• 各段執行時間
• Java Metrics
• 程式自己定metric，不用再刻前後執行執行時間
• sentry
• slack
• hook在app裡
• 
  
• Service
• 統一的fluentD 格式 
• fluentD
• ruby實作，部份用c，效能較好
• logstash
• java實作，需jvm，效能較差 (講師：java魔人不要罵我)
• 普羅米修期
• 多維度，時間序列