This course from Stephen teach how to create and deploy any web apps into a Web services.
Docker is a platform or ecosystem around creating and running containers
Docker wants to make it really easy and straight forward for run software or install in any computer as in webserver as well without worrying about all bunch of setup or dependencies.
Single file with all the dependencies and config or setup required to run a very specific program.
for example a NODEJS, NGINX, or REDIS etc
Is a instance of an Image to runs a program.
It's a program with his own isolated set of hardware resources, it has own set of memory, it has own space of networking technology, it has own spcae of hard-drive.
eg:
docker run hello-world
eg:
docker run busybox echo hello world!
eg:
docker ps
#or
docker ps --all
eg:
# create a container
docker create redis
# start a container
docker start 4b263884282
eg:
docker start -a 4b263884282
eg:
docker system prune
eg:
docker log 4b263884282
stop command use to take more time for shutdown the container.
kill command use to shutdown the container immediately.
we have two separate containers. we want to include redis-cli container into redis-server container to run together
eg:
docker exec -it 4b263884282 redis-cli
When you running docker on your computer or machine every single container you are running is running inside a virtual machine running Linux.
The IT flag is two separate flag
-it
# or
-i -t
-i, --interactive Keep STDIN open even if not attached
-t, --tty Allocate a pseudo-TTY | make sure all the text nicely format | auto-complete
You will not want to execute without having execute same command.
"sh" is a command processor or a shell its allow to type command in and will be execute inside the container.
eg:
docker exec -it 4b263884282 sh
The containers do not automatically share their files system
why use new command?
docker build .
the build command it's will be use to take docker file and generating it
the convention to tagging an Image
eg:
docker build -t localhost/redis:latest .
# and run
docker run localhost/redis
In common, we use image to create container. We can manually create a container runs command inside container and generate an image. In straight word we can do manually the same thing Dockerfile does.
eg:
# IMPORTANT YOU DON'T WANT USE THIS WAY IN GENERAL
# BETTER USE Dockerfile APPROACH
docker commit -c 'CMD["redis-server"]' 4b263884282
To solve the issue "npm not available on a base image"
#Dockerfile
FROM node: alpine
alpine is a term in docker role for a small incompact images. Many popular repository were going to offer alpine version of their images.
None of the files inside your root directory are available inside the container by default. Completely segmented out unless you specifically allowed inside your Dockerfile.
To solve 'no such file or directory'
eg:
#Dockerfile
# Install some dependencies
COPY ./ ./
#Default command
.....
We do not setup port-porting inside Dockerfile, a port-porting stuff is strictly a run time constrain, in other words its something we only change when we run a container or start a container.
eg:
docker run -p 8080:8080 localhost/simpleweb
eg:
# Dockerfile
WORKDIR /usr/app
to check the working directory is no longer in image root directory we can check by
eg:
#open 1st cli
docker run -p 5001:5001 <initial-name>/<initial-docker-container>
#open 2nd cli to check
docker exec -it <id-container> sh
how to avoid having completely reinstall all dependencies just because we made a change in source code file?
eg:
#Dockerfile
COPY ./package.json ./ #just copying current specify directory to WORKDIR directory
RUN npm install # just run once
COPY ./ ./ # copy over everything else except package.json
#docker-compose.yml
services:
redis-server: # 1st image
image: 'redis'
node-app: # 2nd image
...
...
...
## app.js
const client = redis.createClient({
// add docker images
host: 'redis-server',
// default redis-server
port: 6379 # specify port for redis-server
});
#docker-compose.yml
version: '3'
services:
redis-server:
image: 'redis'
node-app:
restart: always # Restart Policies
build: .
ports:
- "5001:5001"
eg:
#in cli with folder related docker-compose.yml
docker-compose ps
eg:
#create Dockerfile.dev in root folder directory
#run in cli
docker build -f Dockerfile.dev -t "<images-name>/<container-name>:latest" . # -f is stand for looking a specify Dockerfile
to solve this problem, just delete node_modules on root folder.
.
├── node_modules # Delete this for avoid duplicate files in images
├── public
└── src
-v $(pwd):/app
-v # volume list | Bind mount a volume
$(pwd) # pwd stand of "present working directory"
:/app # when we use a ":" we want to map out a folder inside the container to the folder outside container
-v /app/node_modules # just a placeholder for the folder that's inside the container
# open 1st cli
docker-compose up # to build an image
# open 2nd cli
docker ps # to copy the run CONTAINER-ID
docker exec -it <container-id> npm run test
# docker-compose.yml
# add this code
version "3"
services:
....
....
....
# for test purpose
test:
build:
context: .
dockerfile: Dockerfile.dev
volumes:
- /app/node_modules
- .:/app
# override command to run test
command: ["npm", "run", "test"]
# run cli
docker-compose up --build
but we have an issue here,
so we make two different images to solve this issue.
Travis CI is a hosted, distributed continuous integration service used to build and test projects hosted at Github. Travis CI automatically detects when a commit has been made and push to a Github repository that is using Travis CI, and each time this happen, it will try to build project and run test.
#.travis.yml
script :
- docker build -e CI=true -t <images-name>/<container-name>:<env> npm run test -- --coverage
# added a options
-- --coverage # to make automatically run command exit
-e CI=true # it tells docker the env list
The benefit of using Elastic Beanstalk is monitors the amount of the traffic that come into our virtual machines and automatically scale everything up.
for bucket_name:
eg:
# add this into .travis.yml
deploy:
provider: elasticbeanstalk
region: "us-west-2" # the region you choose
app: "docker" # a name that you setup in aws
env: "Docker-env" # a name that revere as the environment
bucket_name: "elasticbeanstalk-us-west-2-<your-app-id>"
bucket_path: "docker" # the name same as app names
on:
branch: master
Set an API_KEYS to give access to our aws account over Travis-CI.
At travis CI add your access_key_id and secret_access_id,
# add this into .travis.yml
deploy:
provider: elasticbeanstalk
region: "us-west-2" # the region you choose
app: "docker" # a name that you setup in aws
env: "Docker-env" # a name that revere as the environment
bucket_name: "elasticbeanstalk-us-west-2-<your-app-id>"
bucket_path: "docker" # the name same as app names
on:
branch: master
# a new line setting
access_key_id: $AWS_ACCESS_KEY
secret_access_key:
secure: "$AWS_SECRET_KEY" # make sure you put double quote
# at Dockerfile | production-env
EXPOSE 80 # is a communication for each developers
if you hit an error when deploy to aws, maybe you forget to config EXPOSE at Dockerfile
At aws-elascticbeanstalk is little bit different , elasticbeanstalk when it's start up docker container is gonna look at Dockerfile and gonna look to EXPOSE instruction, and what ever port you listed in there, is what elasticbeanstalk is going to map directly automatically.
We build the images multiple times, we build out our images over travis-CI when we run our task, and we also build image a second time after we push all over code through travis-CI over to Elastic Beanstalk. This is not the best approach because we essentially taking web-server or web-apps and we using it to build the images, chance are we really want the web-server to be just concern running the web-server and not to take any extra process of building images. So we concern to not allow the images to build in active running web-server.
Nginx # to do some routing
React Server # get some front-end mockup | assets
Express Server # to set backend API
Worker # a separate backend NodeJS process
Redis # a memory data-store for housing temporary value
Postgres # a database which is very similar with mysql
The purpose is to make dev Dockerfiles for each one is if we make a change client, server, worker we ensure to not rebuild the entire images to get changes into a fact that makes really slow development workflow.
eg:
# create docker-compose.yml file on complex root dir and write on it:
version: '3'
services:
postgres:
image: 'postgres:latest'
container_name: postgres-complex
or see:
variableName=value
when you use this command its mean you run 2 step process,
-
first step process you build an images, that's kind of preparation part create a new images.
-
second step process when some point on the future we run a container, we actually take an images and create instance of container out of it.
variableName
So if you have env-var setup on your machine like some secret API-KEY that maybe you want to use this syntax
or see the files:
On production environment nginx server has to listen on port 3000
.travis.yml for multi-images
eg:
# in .travis.yml
# ATTENTION ON DETAIL
# change <your-docker-username> as your docker profile
after_success:
- docker build -t <your-docker-username>/multi-client ./client
- docker build -t <your-docker-username>/multi-nginx ./nginx
- docker build -t <your-docker-username>/multi-server ./server
- docker build -t <your-docker-username>/multi-worker ./worker
# Log in to docker CLI
- echo $"DOCKER_PASSWORD" | docker login -u "$DOCKER_ID" --password-stdin
# Take those images and push them to docker hub
- docker push <your-docker-username>/multi-client
- docker push <your-docker-username>/multi-nginx
- docker push <your-docker-username>/multi-server
- docker push <your-docker-username>/multi-workerWe have a couple different folder in each of them has a separate a Dockerfile, so any time we want to run multiple separate containers on AWS EB we need to create a special file.
The new file going to be a JSON file, that to tell EB exactly where to pull
images from, what resources to allocate for each one, how to setup
a port-mapping, and some associated information.
docker-compose.yaml have direction how to build an images and Dockerrun.aws.json the image has been build just specify images to use.
AWS EB # AWS ELASTIC-BEANSTALK
Is not immediately clear when you start reading AWS documentation to how to customize Dockerrun.aws.json. So lets look the AWS documentation,
#Dockerrun.aws.json
{
"AWSEBDockerrunVersion":2,
"containerDefinitions": [
{
"name": "client",
"image": "<your-docker-id>/multi-client",
"hostname": "client", # it's name same at your docker-compose.yml services
"essential": false # if we remakrs with true and this container ever crashes all the other container in this group will be close-down
},
{
"name": "server",
"image": "<your-docker-id>/mulit-server",
"hostname": "api", # renaming hostnmae from server, cause Nginx try to redirect traffic upstream
"essential": false
},
{
"name": "worker",
"image": "<your-docker-id>/multi-worker",
"hostname": "worker",
"essential": false
},
{
"name": "nginx",
"image": "<your-docker-id>/multi-nginx",
"hostname": "nginx", # just optional require because no other services need to directly access nginx
"essential": true, # if this container crash with any reason all container must be automatically shutdown at the same time
"portMappings": [
{
"hostPort": 80, # open up a port on the host or on the machine
"containerPort": 80 # default nginx mapped port from host
}
],
"links": ["client", "server"] # forming links from nginx over to the client container and server container
}
]
}
!! ATTENTION when you create a project without a free-plane, make sure you DELETE any project or AWS will charge any instances.
VPC # Virtual Private cloud
Security Group (firewall Rules): is a rules describing what different services or sources of internet traffic can connect to different services running inside your VPC's
Q: Now we understand what VPC's is and Security Group is, how are we going to form a connection between EB Instance with RDS (Postgress) and EC (Redis)?
A: we gonna create a new security group, and new security group is going to say essentially as a rule let any traffic access this instances if it belong to the security group and we gonna attached it to all three of this different services. So all the services is gonna belong to this one common security group. And Security group essentially says if another AWS instance belong to a new security group then let the traffic flow through and let different services talk to each other.
rds-launch-wizard # a new security group automated created when you make postgress instances
ElastiCache (redis)
AMAZON RDS (Postgres)
EB Instances
ATTENTION
-
When you put your Environment properties in EB the values didn't hidden, so when you entry properties, potentially other people come to this page can see database password.
-
ElastiCache Redis at Primary Endpoint, we do not copy the port.
eg:
# add this in .travis.yml
deploy:
provider: elasticbeanstalk
region: us-west-1
app: multi-docker
env: MultiDocker-env
bucket_name: <your-bucket-name>
bucket_path: docker-multi
on:
branch: master
access_key_id: $AWS_ACCESS_KEY
secret_access_key:
secure: $AWS_SECRET_KEY
Is a system for running many different containers over multiple different machines
When you need to run many different containers with different images
minikube # a command line tools to setup a tiny kubernetes cluster
kubectl # is a program use to interact kubernetes cluster in general
k8s # a convention term | abbervation for kubernetes
explanation config file
apiVersion: v1 # scope or limits the type of object
kind: Pod # a specific type of object
metadata: client-pod # most of use is for logging purposes
name: client-pod
labels:
component: web # a label selector system to connect with client-node-port.yaml
when we make config file kubernetes we not quite making a container me make something different we make an object.
Q: what is an object is on kubernetes?
A: a config file we set (make) the term object is refer to a think that exist inside at kubernetes cluster, so we don't specifically says we make an object so much, reality we make specific type of object
Q: what is an object use for on kubernetes?
A: object is essentially think that going to be created inside kuberntes cluster to get application to work the way we might expect. Every object or type of object have slightly different purpose
Q: what is Object types of Pod use for?
A: a Pod use to running a container
Q: what is object types ReplicaController use for?
A: a ReplicaController use for monitoring a container
Q: what is object types Service use for?
A: a Service use for up networking
When we start to load-up the configuration file into kubectl is going to create a Pod inside Virtual-Machine (we refer VM as a Node). A Pod it self it's a grouping of containers with very common purpose.
Q: We might be wondering why me making a Pod that has a grouping a container?
A: In the kubernetes world there is no such think as just creating a container on a cluster
back with EB, docker-compose we were creating containers really old-day no shoe what so ever. In the world kubernetes we do not have the ability to just run one naked single container by it self with no associated over had. The smallest think you can deploy is a Pod.
It always to be declare or deploying containers within a Pod, as the smallest think we can deploy to run a single container.
Q: why me make a Pod?
A: we cannot deploy individual containers by them self as we could with docker-compose, DB requirement of a Pod we must run one or more containers inside of it.
In the world of Pod we start to grouping together containers that have a very discrete or very tightly couple relationship, in other words these are containers absolutely have a tight immigration and must be executed with each others.
We use this Second object (service) types any times we want to setup some amount of networking inside of kubernetes cluster.
eg:
# client-node-port.yaml
apiVersion: V1
kind: Service
metadata:
name: client-node-port
spec:
types: NodePort
ports:
- port: 3050
targetPort: 3000
nodePort: 31515 # Expose a container to the outside world | for dev environment purpose
selector: # a label-selector-system
component: webkube-proxy # a single window to the outside world
rather then referring to the Service to connect to the client-pod.yaml, we instead using in kubernetes label-selector-system. To connect between client-node-port.yaml with client-pod.yaml
A component: web is arbitrary key-value pair.
A targetPort: 3000 is identical to the containerPort: 3000 over the Pod definition
A nodePort the most IMPORTANT is to communicated between developer to access multi-client Pod