来自 google devops 经验落地实践-sre.pptx

SRE是什么鬼GoogleSRE07-14•YouTubeStreaming•Videotranscoding,streaming,storage(>1PB/month)•work(>10Knodes,peaking10Tbpsegress).GoogleSRE07-14•GoogleCloudPlatform•Machinelifecyclemanagement(>Xclustersglobally,>Ymachines)•Borg,Omega(>Xmillionjobsscheduledeveryweek)SITERELIABILITYENGINEERING说白了就是DevOps一回事Site•生产线管理员•Ensureuser-visibleuptimeandservicequality•Authorityoverproductionenvironment.•跟网站一起成长•Steeplearningcurve,plexity•Continuousretraining,sitesalwaysbeingimproved•基础架构设施•Specializationsforsharedinfrastructure•ponentshavegoodreliabilityReliability•itjustworks•ServiceLevelObjective(SLO)•Monitoring/Deployment•CapacityPlanning•以一敌百•Teammanagesmonitoringanddevelopsautomation•Impliesuseofscriptinganddataanalysistools•Mostfailuresneedautomatedrecoveriesinplace•救火队员和纵火犯合体•Elevatedriskduringconvenientworkinghours•Learnofagemortalityriskduringprecedingworkday•InfantmortalityideallyalsoavoidsmealsEngineering•码农•Notadministration•报警系统重度(中毒)用户•urs•Routinelyusemultiplelayers,levelsandviewpoints•Designthemanualandautomaticescalationpaths•对未来负责•Responsibleforenablinggrowthandscaling•Planforrequirements,identifyinefficiencies•Filebugsand,whereappropriate,fixthemtooWhoareSRE•跑偏了的程序员•50-50mixofsoftwarebackgroundsystemsengineeringbackground.•重度强迫症和处女座•“ept doingthingsoverandoverbyhand.“–BenTreynor•脸皮厚•DEV/OPSEternalconflict•DEV•Theincentiveofthedevelopmentteamistoget featureslaunchedandtogetuserstoadoptthe product.•OPS•Theincentivesofateamwithoperationaldutiesisto ensurethatthethingdoesn'tblowupontheirwatch.