Skip to content

Latest commit

 

History

History
191 lines (122 loc) · 6.99 KB

learning_docker.md

File metadata and controls

191 lines (122 loc) · 6.99 KB
Raw
path title
/learnings/docker
Learnings: Docker

Table Of Contents

<<Learning_Docker>>

Parts

  • daemon <-- REST API to instruct containerd to do things
  • runc <-- OCI container runtime spec
  • containerd <-- manages lifecycle operations (start, stop, pause, etc etc)

Docker Storage <<Learning_Docker_Storage>>

Docker Storage / Overlay Mount Types <<Learning_Docker_Storage_Types>>

Name Desc
Volumes what you know and love
Bind mounts
tmpfs RAM Disks for Docker!!!!! (... or virtual RAM disks, depending on the wills of Learning_Ops_Unix_Virtual_Memory_kswapd ). Only works on Linux / Unix containers, not Windows. Can not be shared across containers.

Writing Data in your container for state <<Learning_Docker_Storage_In_Container_State>>

Normally, the top-most layer of a Docker image is read+write. You can treat this as a normal file system.

Remember:

  • Data written here will go away when container restarted - only store re-creatable data or state here
  • Could docker commit to save these changes into a new, read-only later.

The major difference between a container and an image is the top writable layer. All writes to the container that add new or modify existing data are stored in this writable layer. When the container is deleted, the writable layer is also deleted.

Where everything is stored on the Docker host <<Learning_Docker_Storage_Where_It_All_Is>>

docker info <-- Storage driver

/var/lib/docker <-- ls

$ ls -la /var/lib/docker
drwx--x--x 12 root root  4096 Dec 13 10:57 .
drwxr-xr-x 22 root root  4096 Mar 19 03:40 ..
drwx------  4 root root  4096 Mar  3 00:31 containers
drwx------  3 root root  4096 Nov 14 14:33 image
drwx------  2 root root 16384 Nov 14 14:32 lost+found
drwxr-x---  3 root root  4096 Nov 14 14:33 network
drwx------ 21 root root  4096 Mar  3 00:58 overlay2
drwx------  4 root root  4096 Nov 14 14:33 plugins
drwx------  2 root root  4096 Nov 14 14:33 swarm
drwx------  2 root root  4096 Mar  2 21:30 tmp
drwx------  2 root root  4096 Nov 14 14:33 trust
drwx------  2 root root  4096 Nov 14 14:33 volumes

This Docker host uses overlay2 as a storage driver, so layers will be in /var/lib/docker/overlay2.

See also:

  • Q: "So... what volume does /var/lib/docker live on?" A: See Learning_Ops_Unix_Mount_Points
Using a folder not /var/lib/docker

you can use docker daemon -g / the graph parameter in daemon.json.

See also:

Docker Networking <<Learning_Docker_Networking>>

And Using Docker Networks to join two Docker Compose Clusters <<Learning_Docker_Networking_Compose>>

Docker Compose boots a separate network for each Docker compose cluster you are running. Thus if you docker-compose up with two separate docker-compose.yml, these two clusters will not be able to talk to each other.

Thus you want your infrastructure in one Docker Compose, and your app in another Docker compose... but the apps need to talk to the services (and each Docker Compose is in a seperate network that don't talk to each other).

In Docker Compose file

First, define the network at all (this is a top level key):

networks:
  MY_NETWORK_NAME:
    driver: bridge

NOTE: this will be prefixed by the (same name the docker images are prefixed with). If you don't want this, create the networks externally, like so:

networks:
  MY_NETWORK_NAME:
    external:
        name: MY_NETWORK_NAME

then in the services section define what network this is on

services:
  my_thing:
    networks:
      - network_one
      - network_two

<<Learning_Docker_Compose>>

Using Init flag: <<Learning_Ops_Docker_Init_And_Compose>>

NOTE: This is only supported in Docker Compose format 2.2, not Docker format 3.x!!

version: '2.2'
services:
  web:
    image: 'co/thingy'
    init: true

<<Learning_Docker_Build>>

<<Learning_Docker_Build_Parameters>>

<<Learning_Docker_Build_Parameters_Example_Base_Image_Example>>

Dockerfile snippet

ARG ROOT_IMAGE

FROM $ROOT_IMAGE

Building this on the CLI

$ docker build -t my_image --build-arg 'ROOT_IMAGE=scratch'

<<Learning_Docker_Windows>>

Windows 10 (or Win Server 2016)? Use the new Docker (HyperV)

Windows version earlier than Windows 10? Use Docker Machine

<<Learning_Docker_Windows_Docker_Machine>>

See Docker Toolbox (Windows)

Q: Can my machine handle (Docker Machine's) virtualization well?

Under Windows 8, check Task Manager -> Performance -> Virtualization Windows 7? Download a tool from Microsoft to check

^^^^^^ if there's no hardware supported virtualization that also will likely mean you can't run VirtualBox with PAE (see Learning_Unix_Memory_PAE). Linux Kernels > Ubuntu 12or so require pae to boot.

<<Learning_Docker_Windows_Docker_Machine_Remote_Hosts>>

With Windows / Docker machine you can point it to non local Docker hosts in different cloud providers:

Preprovisioned VMs:

    $ docker-machine create --driver none --url=tcp://50.134.234.20:2376 custombox
    $ docker-machine ls

Book Recommendations