Page Contents

To Read





Both virtualisation and containerisation decouple workloads from underlying hardware.
	Containers are OS LEVEL virtualisation.

Containers are lighter-weight. They add an additional layer of abstraction ontop of the host OS.
The still allow sandboxing applications and resource allocation and control etc.

Run apps on one machine in a way where they won't "meddle" with other apps on the same machine.
Eg. N projects, each with its own set of dependencies! Docker images help "sandbox" each project so
you can keep them all on the same machine without getting cross-interference.

More resource efficient than full VMs. Docker shares common ground between projects. Runs on Linux.
On windows its just a VM!

Portability across machines and environments - e.g. dependency version mismatch.

Docker container is not a VM on Linux. On Windows it can be, but depends on your windows version 
and method of Docker-install - Docker Toolbox v.s. Docker for Windows. The former uses VirtualBox
(type 2 hypervisor) and can only be accessed from the Docker Quickstart Terminal. The latter uses
the Windows hypervisor (type 1 hypervisor) and can be accessed from any terminal you like. There is
currently the restriction with the latter that you cannot run both the type1 hypervisor and
VirtualBox at the same time :( Toolbox access the daemon at and have to use the
Docker terminal. On the latter you can use local host. 

VM - isolate system
Docket - isolate application

Docker and WSL

First Way I Tried

!!! NEXT SECTION                                                                                !!!

Taken from here:
Replicated verbatim:

	Install Docker

	# Update the apt package list.
	sudo apt-get update -y

	# Install Docker's package dependencies.
	sudo apt-get install -y \
	    apt-transport-https \
	    ca-certificates \
	    curl \

	# Download and add Docker's official public PGP key.
	curl -fsSL | sudo apt-key add -

	# Verify the fingerprint.
	sudo apt-key fingerprint 0EBFCD88

	# Add the `stable` channel's Docker upstream repository.
	# If you want to live on the edge, you can change "stable" below to "test" or
	# "nightly". I highly recommend sticking with stable!
	sudo add-apt-repository \
	   "deb [arch=amd64] \
	   $(lsb_release -cs) \

	# Update the apt package list (for the new apt repo).
	sudo apt-get update -y

	# Install the latest version of Docker CE.
	sudo apt-get install -y docker-ce

	# Allow your user to access the Docker CLI without needing root access.
	sudo usermod -aG docker $USER

	Connect to a remote Docker daemon with this 1 liner:
	echo "export DOCKER_HOST=tcp://localhost:2375" >> ~/.bashrc && source ~/.bashrc
	# ALSO make sure in your Docker-for-windows settings that you have enabled the
	# "Expose daemon on tcp://localhost:2375 without TLS" option.

	Ensure Volume Mounts Work
	When using WSL, Docker for Windows expects you to supply your volume paths in a format that matches this: /c/Users/nick/dev/myapp.
	But, WSL doesn’t work like that. Instead, it uses the /mnt/c/Users/nick/dev/myapp format.

		Running Windows 10 18.03+ or Newer?
		sudo nano /etc/wsl.conf

		# Now make it look like this and save the file when you're done:
		root = /
		options = "metadata"

		# Now reboot your windows device.

		Running Windows 10 17.09?
		udo mkdir /c
		sudo mount --bind /mnt/c /c

A Better, Newer Way
Looks like Docker Desktop support for WSL 2 backend got a lot better :)

Docker's Architecture

Docker daemon, provides REST API, and the docker CLI client, which uses the REST API.

Docker container is a running instance of a docker image

Image - file system and parmeters - like a package containing everything you need to run your application
      - can download, build and run.
      - images are stored inside a registry. it's similar to a code repository, but holds Docker
      - lightwight executable packages that contain everything required in them, software, app
        code, dependencies etc etc.
      - Considered a read only artifact.
           !! NOTE this means that anything created in the Docker container that is not mapped to !!
           !! an external resource on the host is lost when the container is shutdown             !!
      - Created using Docker files - these detail how the Docker image should be built.
            - docker build

Container == running instance of a Docker image - like instance of a class
          Is immutable so any changes made whilst running them are lost on exit
          The container only has access to the resources provided in the Docker image unless
            external resources are mapped into the container.

Docker has 3 main components
	1. Docker daemon
		Install on host - "brain" of docker - performs/actons commands from the CLI and setup/run
		Docker containers.
	2. Docker CLI
		Way to interact with Docker - issue commands.
	3. Image registry
		Makes it possible for docker daemon to download and run an image

##-- alpine is only about 3MB so much better to download than ubuntu which could be seveal hundred MB.
##-- althernative is "slim"
docker run -it --rm --name alpine alpine sh 
           ^^^ ^^^^ ^^^^^^
           ^^^ ^^^^ Custom unique name or handle to refer to instance by
           ^^^ Completely remove container when stopped
           be interactive and terminal (handle ctrl-c and colours etc)

docker stop alpine # to shut down the container

docker hub as auatomated build hooks - so if i push a change to my GitHub the docker registery can rebuild my image for me automatgiclaly.

The Docker Build Process


1. Run docker container, modify then do "docker commit"
2. Docker file - blueprint or recipe for what you want your docker image to be - the best way (apparently)

Docker images are layers, which are kinda like self-contained files and the docker image is the result of stacking up these layers.

First non-comment in docker file must be the "FROM" instruction (instruction names in caps) - import a base image to build on.
 	# FROM tag:version
 	FROM python:2.7-alpine
 	FROM ruby:2.4-alpine ## etc etc

FROM - The base image this new Docker image will be based on
     - It will be the very first instruction in any Docker file.
     - For example
           FROM python:2.7-alpine
                ^^^^^^ ^^^^^^^^^^
                Image  Verion of image.
                name   (If left out, Docker assumes you want latest)

RUN - To run a command(s)
    - It will execute whatever commands you define in a shell
	# RUN -any script you can run in your base OS without Docker-
	RUN mkdir /app
	WORKDIR /app # change working directory for subsequent RUN commands
	COPY requirements.txt # You can use relative paths but cannot copy in files above the dockerfile.
	LABEL value="something" # Attach arbitrary metadata to an image (must use quotes for value)
	CMD - default command to  instance run.
		note RUN - build of image
		CMD - run when instance of image starts up]
      By default this is run as an argument to '/bin/sh -c'.
      The CMD instruction is being passed in as an argument to and ENTRYPOINT script (see later)

EXPOSE - Expose a port on which the container can access the outside world on
         e.g. EXPOSE 8080 # Expose this port for the webapp.

ENTRYPOINT - Lets you define what script or command to run when the container starts up
             E.g. ENTRYPOINT [ "python" ]

CMD - Execute commands at runtime, difference to ENTRYPOINT is that you can provide default values
      which can be overriden at runtime. Entrypoint can be overridden.
      E.g. CMD [ "" ]

FROM python:2.7-alpine

RUN mkdir /app

COPY requirements.txt requirements.txt # goes to /app/requirements.txt
RUN pip install -r requirements.txt
COPY . . #again /app - copies everything from local machine relative to dockerfile into our docker image.
LABEL maintainer="James James <>" \
CMD flask run --host= --port=5000

Union File Systems and Copy on Write
VERY IMPORTANT HOW YOU ORDER THINGS! Each command/instruction will be a cached-layer so make sure
that more static large layers come first so that if something changes Docker won't have to redo a
lot of stuff (because cached layers below the layer that change stay the same but layers above it
must be redone).

Chain commands together where possible to avoid creating unecessary additional layers.

Layers can be cached across multiple Docker images for memory usage efficiency.

To see what layers the image has use the `Docker inspect IMG_ID` command. 

To achieve this layer Docker uses a UNION FILE SYSTEM. Many lays of filesystems appears to the
image as one single FS.
	If higher layer mods file in lower layer, copy is made in higher layer and modified then used.
	If it didn't do this every Docker image sharing the lower layer would see the change and this
	is not what is wanted! Called COPY-ON-WRITE strategy.

	Sharing layers helps reduce image size.

Docker adds one writable layer on top of your image so you can modify the FS, store things etc.
BUT TAKE CARE - this is not persistent by default and any data stored/chaged etc is lost when the
container is shutdown.

Cleanup / File System Size Easy Wins
1. Purge any apps you installed to setup but won't need for the image to run:
       apt-get purge -y 
2. Do a general house clean:
       apt-get autoremove -y
       apt-get clean -y
       rm -rf /var/lib/apt/lists/*
       rm -rf /var/cache/apt/archives
       rm -fr /var/cache/debconf/*
 3. Get rid of man pages
       rm -rf /usr/share/doc/
       rm -rf /usr/share/man/
       rm -rf /usr/share/locale/

Docker CLI Command Cheat Sheet

# Build Docker Images from CLI
$ docker image build -t web1:1.0 .
                     ^^     ^^^
                     ^^     the tag version
                     tag the image so that we dont have to use its random hash

$ docker image inspect web1

$docker image ls [-a]
$docker image rm <tag:version>|<hash (image ID)>

# Run/stop Docker Instances
$ docker container ls  [-a] # List running docker images (akak containers)
$ docker container run web1
$ docker container stop web1 or container-ID

# Ports
$ docker container run -p bind-port-on-host:bind-port-in-container

# Env var
$ docker container run -e SOME_VAR=some-value

# Logs and stats
$ docker container logs [-f] container-ID-or-name
$ docker container stats

# Auto restart
use docker run ... --restart on-failure ...

# Run as daemon and get logs
use '-d' flag
Then docker logs CONTAINER-ID

# Windows


Avoid the Kill-build-reload cycle every time you change your code. I.e. you change it, kill the
docker instance, rebuild the image, and re-run it... sigh!

Mount source code from main operating system straight into the running container. To do this use the
'-v' option.

   docker container run ... -v abs-path-on-host:path-to-mount-in-target

To specify a path on C:\ in windows do -v //c/path/to/wherever

Dynamically Connect To Container And Run Commands

Here you are connecting to an _already running_ container

$ docker container exec [--user] "$(id -u):$(id -g)"] -it the-name (bash|sh|cmd)
                        When running on linux host so that files created in docker are not owned
                        by root, but by you.

or to run a command once without an interactive shell
$ docker container exec "echo james" alpine

Link Docker Containers Over A Network

$docker network ls

Docker will have installed a virtual network adaptor. On windows I see it as:

   Ethernet adapter vEthernet (DockerNAT):

      Connection-specific DNS Suffix  . :
      IPv4 Address. . . . . . . . . . . :
      Subnet Mask . . . . . . . . . . . :
      Default Gateway . . . . . . . . . :

On Linux I see it as:


See the docker networks using
$ docker network ls
NETWORK ID          NAME                DRIVER              SCOPE
5c6e2a9ca9b2        bridge              bridge              local << This is the virtual network adapter created oh your host as seen above (DockerNAT)
0aa3a20ed51a        host                host                local
bb9b8f0ffcd6        none                null                local

The bridge network is used by default. Inspect it using:
$ docker network inspect bridge

NOTE: Conatiners are, by default, added to the bridge network unless you specify otherwise

To get DNS "out of the box" you need to configure your own network. 
   docker metwork create --driver bridge mynetworkmame

With new network supply "--net mynetworkname" to the "docker conatiner run..." command line.

Once this is done each container can be accessed using its name, rather than the IP address. Docker has
a little DNS server running for us that maps the container name to the IP address Docker assigned it :)

Data Volumes a.k.a. Named Volumes

Allow you to persist data - but note containers really should be STATELESS and PORTABLE.

Create using "docker volume create my_volume_name"
Use "docker volume inspect my_volume_name" to see things like its mountpoint in the container.

In stead of doing -v /local/path:/path/on/conatiner do my_volume_name:/path/on/container
                                                        Name if volume created in prev step


A command in the Docker file that auto-runs a script after the container has started


From some dynamic config you can create an environment variable tha the script can read to
change its execution directly

   Use -e ENV_VAR_NAME=value on the command line to the "docker container run" command.

Last line of your script must read
   exec "$@"
This is because, in your docker file, the CMD instruction is being passed in as an 
argument to your ENTRYPOINT script, so you must take care to run it, otherwise whatever was
in the CMD line in the docker file will not get run!

Managing Docker

docker system info
docker system df [-v] #< See disk usage
docker system prune [-a] [-f] (Caution when using -a!!)
docker image ls #< See images: images with <none> and dangling and can be killed
docker container stop $(docker container ls -a -q) #< Stop all running containers