![DOCKER - DOCKERFILE
DOCKERFILE
FROM mhart/alpine-node:6.3.1
# Copy app to a directory called /app
COPY . /app
# Set the working dir to the newly created /app directory
WORKDIR /app
# Run npm install
RUN npm install --production
EXPOSE 1337
CMD ["node", "app.js"]
tracking api - docker ๏ฌle](https://image.slidesharecdn.com/d7417382-fe33-4280-9c70-d02394fb10c0-160921170136/85/ecs-presentation-10-320.jpg)
ecs-presentation
- 1. EC2 CONTAINER SERVICE AWS + DOCKER +
- 2. DOCKER
- 3. DOCKER DOCKER ISโฆ โธ Application virtualisation โธ A container with everything your app needs โธ code; runtime; tools; environment; OS โธ Building, running and management of these containers
- 4. DOCKER WHY DOCKER โธ Empowers creativity โธ isolation from platform, machine, language etc.. free developers from constraints โธ No more โworks on my machine :troll:โ โธ Sharing is trivial (image repositories, docker hub etcโฆ) โธ Quick to scale โธ Gives us a platform to pull out services today
- 5. DOCKER HOW IT WORKS โธ Each container shares the underlying host kernel (itโs why they boot so quickly) โธ Containers are to VMโs what threads are to processes โธ A container isnโt really a VM, but you can treat it like one for the most part. HARDWARE HOST KERNEL DOCKER ENGINE GUEST OS GUEST OS BINS/LIBS BINS/LIBS APP A APP B
- 7. DOCKER - RUNNING A CONTAINER > DOCKER RUN HELLO-WORLD
- 8. DOCKER - RUNNING A CONTAINER > DOCKER RUN CENTOS
- 9. DOCKERFILE
- 10. DOCKER - DOCKERFILE DOCKERFILE FROM mhart/alpine-node:6.3.1 # Copy app to a directory called /app COPY . /app # Set the working dir to the newly created /app directory WORKDIR /app # Run npm install RUN npm install --production EXPOSE 1337 CMD ["node", "app.js"] tracking api - docker ๏ฌle
- 11. DOCKER REGISTRY
- 12. DOCKER - REGISTRY EC2 CONTAINER REGISTRY (ECR) โธ A repository for docker images โธ AWSโs version of Docker Hub
- 14. ECS
- 15. ECS ECS ISโฆ โธ Container orchestration, scheduling and networking โธ Collections of de๏ฌned tasks running within a cluster โธ AWSโs answer to Kubernetes, Mesos, Fleet and Swarm
- 16. ECS WHY ECS โธ Cost - We only pay for the resources we use, much cheaper than beanstalk โธ Flexibility - Horizontal scaling โธ Good integration with AWS (duh) โธ Ef๏ฌciency - Less idle and responsive to spikes (also see the point about cost) โธ Quickness and ease of release (both new and existing containers)
- 17. ECS WHY ECS โธ Reliability - If a task dies, ECS will kill it and launch a new on within seconds. โธ Smart updates - deterministic update policies โธ Opens doorsโฆ
- 18. WHAT WE REALLY CARE ABOUT
- 19. ECS - WHAT WE REALLY CARE ABOUT 3 PARTS โธ Task De๏ฌnitions โธ Services โธ Load Balancers - application vs classic
- 20. ECS - WHAT WE REALLY CARE ABOUT TASK DEFINITIONS Task de๏ฌnitions specify the container information for your application stack, such as how many containers are part of your task, what system resources they need, how they are linked together, which host ports they will use. TASK = MEMORY CPU ENVIRONMENT VARIABLES COMMANDS VOLUMES AND STORAGE DOCKER IMAGE(S)
- 21. ECS - WHAT WE REALLY CARE ABOUT TASK DEFINITIONS โฃ One or more containers
- 22. ECS - WHAT WE REALLY CARE ABOUT TASK DEFINITIONS โฃ Immutable - when a task is updated, it is classed as a revision
- 23. ECS - WHAT WE REALLY CARE ABOUT SERVICES A service runs and maintains a โdesired countโ of a task de๏ฌnition simultaneously in an ECS cluster.
- 24. ECS - WHAT WE REALLY CARE ABOUT SERVICES โธ If any task should fail or stop for any reason, the service scheduler launches another instance of your task de๏ฌnition to replace it and maintain the desired count of tasks in the service. โธ Optionally runs the service behind a load balancer (important for apiโs) โธ Uses a speci๏ฌed minimum and maximum health percentage to determine update strategy.
- 25. ECS - WHAT WE REALLY CARE ABOUT SERVICES If the minimum healthy percent is below 100%, the service scheduler can ignore the desiredCount temporarily during a deployment. For example, if your service has a desiredCount of four tasks, a minimum healthy percent of 50% allows the scheduler to stop two existing tasks before starting two new tasks. Deployment Configuration - minimum health
- 26. ECS - WHAT WE REALLY CARE ABOUT SERVICES If a service has a desiredCount of four tasks, a maximumPercent value of 200% starts four new tasks before stopping the four older tasks (provided that the cluster resources required to do this are available). Deployment Configuration - maximum health
- 27. UNDER THE HOOD
- 28. ECS - UNDER THE HOOD AWS RESOURCES โธ EC2 โธ Auto Scaling Groups โธ Launch Con๏ฌguration โธ ELB โธ IAM
- 29. ECS - UNDER THE HOOD COMPUTE CLUSTER = ECS AGENT Amazon Linux x64 ECS AGENT Amazon Linux x64 AUTO SCALING GROUP Desired Count = 2
- 30. ECS - UNDER THE HOOD LOAD BALANCERS - CLASSIC CLUSTER :80/:443 :80/:443 host port - 8000 host port - 9000 TASK INSTANCE TASK TASK INSTANCE TASK ATTACHED TO HOST PORT ON INSTANCE
- 31. ECS - UNDER THE HOOD LOAD BALANCERS - APPLICATION CLUSTER :80/:443 :80/:443 TASK INSTANCE TASK TASK INSTANCE TASK MAPPED AT THE APPLICATION LEVEL Dynamic port Dynamic port
- 32. DEV PIPELINE
- 33. DEV PIPELINE CREATE 1. Code 2. Docker๏ฌle 3. Git
- 34. DEV PIPELINE TEAMCITY 1. Linux Agent (w/ docker & aws cli) 2. Pulls code from Github 3. Builds docker image and tags it with version 4. Pushes image to the docker repository (AWS ECR in our case)
- 35. DEV PIPELINE OCTOPUS DEPLOY 1. Create release 2. Uses semver to determine which image to use:โจ e.g. Release 1.0.2 will target image at {repositoryurl}/api:v1.0.2โจ Importantly, release 1.0.2.1 will also target {repositoryurl}/api:v1.0.2 3. Create or use a ELB if we need one 4. Create the task de๏ฌnition 5. Create the service
- 36. BIT DEMOโฆ