Docker Networking

The Container Netwotk Model (CNM) is the design document for Docker Networking which is implemented via ’libnetwork’ in golang

  • Sandbox - isolated network stavk
  • Endpount - Virtual Ethernet Interfece
  • Network - Virtual Switch (bridge)

Note - Libnetwork -> implemenrs the control and management plane functions

Network Specific Drivers -> implemenrs the data plane while also handling the connectivity and isolation

Docker Native Network Drivers

  • Bridge - uses a software bridge whicha allows containers connected to the same bridge network to communicate

  • Host - Connects the docker host and containers to use hosts networking directly

  • Overlay - used for multi-host networking which allows a single network to span multiple hosts such that containers on diffenet hosts can communicate at layer 2

  • Macvlan - Useful for lagacy and monitor apps which expect to directly conneted to physical network as it assigns completely the networkinh stack on container . No IP will be configuared with this driver

Docker Networking ( Hands-on )

  $ docker network --help
  
  1. check what networks are already present in the docker by default
  $ dokcer run -itd --name netcon ubuntu /bin/bash 
  

check for the network placement of this container in the above networks. We see that by default all containers got attached to bridge network using network inspect command this could be done

  $ docker network inspect bridge 
  

In the Container option abobe the information about container name , IPv4 etc present

Run another container net2con from busybox image and check its network placement same as above

  docker run -itd --name net2con busybox /bin/bash 
  

lets us now run nginx container as net3con-web with exposed port

  docker run -itd --name net3con-web -p 5000:80 nginx 
  

to check for the port correctly setup and exposed , one can use port command followed by container name .

  
docker port net3con-web
  

to visualize only the relavant information which is the Containers block from inspect command output use -f option to format the JSON and parse it via jq ulaitity

  docker network inspect bridege -f "{{json.Containers}}" | jq
  

one can also oberserve the associated bridge.id to docker0 default bridge with all the containers attached via vethxxxxxx virtual ethernet interfces

  $brctl show 
  

concusion

Cgroups and Namespaces provides essential isolation and limits the container resources. This is highly valuable for the Docker host system in case of DoS Attacks wherein if resources are set minimally, less harm will be there. For resource isolation, feel free to look over to the Linux ‘cgcreate’, ‘cgdelete’, ‘cgset’ and related parameters like ‘cpuset’, ‘cpuacct’, etc. Running containers and walking around the namespaces, cgroups, container networking

Last updated 03 Jun 2024, 13:43 +0530 . history