Docker

Docker is a set of platform as a service (PaaS) products that use OS-level virtualization to deliver software in packages called _containers_. The service has both free and premium tiers. The software that hosts the containers is called Docker Engine.

• Project Homepage: [Home - Docker](https://www.docker.com/)
• Documentation: [Docker Documentation | Docker Documentation](https://docs.docker.com/)
• Docker hub: [Docker hub](https://hub.docker.com)

One click installation script:

• curl -fsSL https://get.docker.com -o get-docker.sh
• sudo sh get-docker.sh

Note for Debian testing just use sudo apt install docker, as testing packages the latest docker and the docker repo does not seem to have testing.

Run docker as non root user:

• sudo groupadd docker
• sudo usermod -aG docker $USER

Install Docker Engine : [Docker Engine](https://docs.docker.com/engine/install/)

Both the install methods actually use the standard package manager to install docker.

1. dpkg -l|grep docker to check the docker packages actually installed
2. sudo apt remove docker-ce
3. Check the the var/lib/docker directory, sudo du -d 1 -h var/lib/docker
4. rm -rf /var/lib/docker to remove the docker directory. I try not to store any important files in docker, so this should be low risk for me.

• Uninstall docker version installed via script

Docker applies iptables on the host machine, see Docker and iptables.

A Docker host that is acting or running as a router, e.g. using Traefik, the following may be needed to allow operation.

For some more details, tldr;

[Unit]
Description=IPTables Docker Rules
After=docker.service

[Service]
Type=oneshot
ExecStart=/usr/sbin/iptables -I DOCKER-USER -i br0 -o br0 -j ACCEPT

[Install]
WantedBy=multi-user.target

Docker has to perform some interesting network filtering both on the container host, as noted above and within containers as outlined here. It looks like this is required to allow container DNS functionality on Docker containers using bridge networking.

The containers DNS (/etc/resolv.conf) is assigned to a proxy on 127.0.0.11:53. (Note that DNS uses UDP not TCP datagrams.)

Further to this The container netfilter use NAT chains to operate on 127.0.0.11. See the following nftables info:

nft list tables

table ip nat

nft list table ip nat

# Warning: table ip nat is managed by iptables-nft, do not touch!
table ip nat {
	chain DOCKER_OUTPUT {
		meta l4proto tcp ip daddr 127.0.0.11 xt match "tcp" counter packets 0 bytes 0 xt target "DNAT"
		meta l4proto udp ip daddr 127.0.0.11 xt match "udp" counter packets 329 bytes 20249 xt target "DNAT"
	}

	chain OUTPUT {
		type nat hook output priority dstnat; policy accept;
		ip daddr 127.0.0.11 counter packets 329 bytes 20249 jump DOCKER_OUTPUT
	}

	chain DOCKER_POSTROUTING {
		meta l4proto tcp ip saddr 127.0.0.11 xt match "tcp" counter packets 0 bytes 0 xt target "SNAT"
		meta l4proto udp ip saddr 127.0.0.11 xt match "udp" counter packets 0 bytes 0 xt target "SNAT"
	}

	chain POSTROUTING {
		type nat hook postrouting priority srcnat; policy accept;
		ip daddr 127.0.0.11 counter packets 329 bytes 20249 jump DOCKER_POSTROUTING
	}
}
/ # 

On VPN setup, at least openvpn, the /etc/resolv.conf is overwritten anyway so the Docker netfilter chains become irrelevant.

On the other hand where Docker bridge network DNS container name resolution is desirable then these netfilter chains must basically remain unadulterated. So in these cases where I need to used netfilter within the container, the simple solution is to simply add extra chains without using the nft flush ruleset command first. These means that if subsequent rule changes are made the container will need to be recreated. This is effectively only for simple basic filter (input, output & forward) chains only, more complex netfilter nat requirements would need further consideration.

Portainer is a handy web view into the Docker system. It can be loaded on the host as a Docker container as per this docker documentation Install Portainer with Docker on Linux! A lot of the official Portainer web points to the business version. They do allow 5 free business installations with registration. The portainer-ee image is the business image on Docker hub, portainer-ce is the community addition. When downgrading from the business edition to the community edition the database on the volume storage needs to be downgraded or portainer-ce will not start.

The community addition command is:

docker volume create portainer_data
docker run -d -p 9443:9443 --name=portainer --restart=always \
    -v /var/run/docker.sock:/var/run/docker.sock \
    -v portainer_data:/data \
    portainer/portainer-ce:latest 

Notes:

• The port 8000 (-p 8000:8000) is not normally required on community edition. Further to this Kea dhcp control agent uses port 8000 as default.
• Some browsers give strange message if http access is requested and then require exception for dodgy https certificate.

Portainer agent allows a remote docker machine to be seen else were via the network. Default port seems to be 9001.

• First stop the agent container: docker stop portainer_agent
• Then remove the agent container: docker rm portainer_agent
• Then pull the latest portainer/agent: docker pull portainer/agent, default is latest if version is not specified.

docker run -d   -p 9001:9001   --name portainer_agent   --restart=always \
-v /var/run/docker.sock:/var/run/docker.sock \
-v /var/lib/docker/volumes:/var/lib/docker/volumes \
portainer/agent

Run Containers

Container Management:

Not that in many simple Alpine containers bash is not the default sh, ash or sh is more likely.

Container Bulk Management

Inspect / Troubleshoot Containers:

Run Commands:

Images:

Volumes:

The docker ps command output is long and often difficult to read on the terminal for this reason.

• docker ps --format 'table {{ .ID }}\t{{.Image}}\t{{ .Names }}'

CONTAINER ID   IMAGE                                           NAMES

• docker ps -s --format 'table {{ .ID }}\t{{.Status}}\t{{.Image}}\t{{ .Names }}'

CONTAINER ID   STATUS                  IMAGE                                           NAMES

• docker ps -s --format 'table {{ .ID }}\t{{.Status}}\t{{.Image}}\t{{ .Names }}\t{{.Size}}'

CONTAINER ID   STATUS                  IMAGE                                           NAMES               SIZE

• How to change the docker ps output format
• Docker Docs
  • docker container ls
  • Container size on disk

Backup docker data from inside container volumes and package it in a tarball archive.
docker run --rm --volumes-from CONTAINER -v $(pwd):/backup busybox tar cvfz /backup/backup.tar CONTAINERPATH

Restore the volume with a tarball archive.
docker run --rm --volumes-from CONTAINER -v $(pwd):/backup busybox sh -c "cd CONTAINERPATH && tar xvf /backup/backup.tar --strip 1"

• Docker Volumes: How to Create & Get Started
• NFS Docker Volumes: How to Create and Use

Docker volume types:

• Host Volumes: -v /path/on/host:/path/in/container
• Anonymous Volumes: -v /path/in/container
• Named Volumes: -v name:/path/in/container

Docker controlled volumes are stored at /var/lib/docker/volumes

Type of networks on Docker:

1. Default Bridge (The default separate isolated bridge network.)
   • This is the default network used if nothing else is defined
2. User defined bridge (DNS name resolution on Docker containers, which default bridge does not.)
   • docker network create network_named
3. Host (Appears on the host machine as if installed there, no separate network.)
4. MACVLAN
   1. MACVLAN (without subVLAN) this create a new ip address on the host machine

      • docker network create -d macvlan \
        --subnet 192.168.1.0/24 \
        --gateway 192.168.1.1 \
        -o parent=br0 \
        network_named

      • No host DHCP access so need to specify ip address when creating container (docker cli --ip 192.168.1.14). If not specified docker DHCP will assign and could cause clash with host.
      • May be problem with multiple MACs on common switch port. Need to set promiscuous mode on network, e.g. sudo ip link set br0 promisc on.
   2. MACVLAN with subVLAN - never had to use this yet myself, so do not know much about it, something about trunking on a common link.
5. IPVLAN
   1. IPVLAN on host subnet, this create a new ip address on the host machine, but not with new MAC number, it uses the host MAC number

      • docker network create -d ipvlan \
        --subnet 192.168.1.0/24 \
        --gateway 192.168.1.1 \
        -o parent=br0 \
        network_named

      • No host DHCP access so need to specify ip address when creating container (docker cli --ip 192.168.1.14). If not specified docker DHCP will assign and could cause clash with host.
      • May be problem with shared MAC with multiple IP address, but less likely than MACVLAN.
   2. IPVLAN on separate subnet using the host machine as gateway, but not with new MAC number, it uses the host MAC number

      • docker network create -d ipvlan \
        --subnet 192.168.1.0/24 \
        -o parent=br0 -o ipvlan_mode=l3 \
        --subnet 192.168.10.0/24 \
        network_named

      • No host DHCP access so need to specify ip address when creating container (docker cli --ip 192.168.1.14). If not specified docker DHCP will assign and could cause clash with host.
      • May be problem with shared MAC with multiple IP address, but less likely than MACVLAN.
6. Overlay network, an even more obscure network arrangement I know nothing about.
7. None network - no assigned network, container has no external network connectivity

A lot of containers are setup to be small and hence do not include many, if any of the tools required to diagnose problems. A small docker image netshoot includes the most common networking tools and when attached to the same docker network can be used to diagnose the network and containers networks thereon.

• docker run --rm --name netshoot --network proxy -it nicolaka/netshoot /bin/bash

• netshoot: a Docker + Kubernetes network trouble-shooting swiss-army container docker run --name netshoot --rm -it nicolaka/netshoot /bin/bash
• Docker - How to cleanup (unused) resources