Building  Applications With  Docker 

Cloud  Computing  Lingo 

No content

Project  Infrastructures 

Traditional  single-­‐tenant  application 

Traditional  multi-­‐tenant  application 

Docker multi-­‐tenant  application 

Containers 

Problem:  hay  from  field  to  bakery ??? 

No content

No content

No content

Problem:  code  from  DEV  to  PRD manual  setup (-­‐) nearly  impossible  to  replicate  environments scripted  setup (+) +/-­‐ easy  to  replicate  environments (-­‐) knowlegderequired  about  what  is  inside  to   create  the  scripts containers (+) very  easy  to  replicate  (just  drop  it  into  place) (+) no  knowlegderequired  about  what  is  inside 

Docker 

No content

No content

Build: #  docker build  -­‐t  trescst/phprodemo . Ship:   #  docker push  trescst/phprodemo Run:   #  docker run  -­‐d  trescst/phprodemo 

Demo 

Twelve  Factor  App 

1 Codebase 

2  Dependencies • Explicitly declare  and  isolate  dependencies • Composer  json +  lock  file • Bower  /  NPM  /  Bundle 

3  Config • Store  config in  the  environment  variables • Never  group  configuration  in  “environments” • Each  env var is  independently  managed  for  each  deploy • Easy  upscaling 

4  Backing  services 

5  Build,  release,  run 

6  Process • Execute  the  app  as  one  or  more  stateless  processes • Share  nothing • Data  must  be  stored  in  statefull backing  services 

7.  Port  binding • Include  an  HTTP  server • Export  services  via  port  binding • Add  a  routing  layer  for  mapping  the  port  to  a  subdomain • The  services  can  communicate  over  containers  by  port 

8  Concurrency • Scale  out  via  the  process  model • Nothing  is  shared • Reliable  and  easy 

9  Disposability • All  processes  can  be  stopped  /  started  at  any  time. • Minimum  startup  time • Gracefull shutdown  in  SIGTERM • Maximum  robustness • Fast  deploys 

10  Dev/prod  parity • Keep  development,  staging,  and  production  as  similar  as  possible Traditional  app Twelve-­‐factor  app Time  between  deploys Weeks Hours Code  authors  vs  code  deployes Different  people Same  people Dev  vs  production  environments Divergent As  similar  as  possible 

11  Logs • Treat  logs  as  event  streams • Always  send  logs  to  the  output  stream • Never  concern  about  storage  of  the  output • Storage  is  handled  by  the  executing  environment. 

monolog: handlers: docker: type:  stream path:  "php://stdout" level:  debug formatter:  monolog.formatter.logstash 

CloudWatch Lambda ElastiCache 

12  Admin  processes • Run  management  tasks  as  one-­‐off  processes • CLI  scripts • Migrations 

Demo 

Any  questions?!