conlink replaces the standard Docker Compose networking for all or portions of your project, providing fine-grained control over layer 2 and layer 3 networking with a declarative configuration syntax.
services:
# 1. Add conlink to your Docker Compose file, and point it to your network
# config file, which can be the Docker Compose file itself!
network:
image: lonocloud/conlink:latest
pid: host
network_mode: none
cap_add: [SYS_ADMIN, NET_ADMIN, SYS_NICE, NET_BROADCAST, IPC_LOCK]
security_opt: [ 'apparmor:unconfined' ]
volumes:
- /var/run/docker.sock:/var/run/docker.sock
- /var/lib/docker:/var/lib/docker
- ./:/test
command: /app/build/conlink.js --compose-file /test/docker-compose.yaml
node:
image: alpine
# 2. Disable standard Docker Compose networking where desired
network_mode: none
command: sleep Infinity
# 3. Create links with complete control over interface naming,
# addressing, routing, and much more! Any necessary "switches"
# (bridges) are created automatically.
x-network:
links:
- {bridge: s1, ip: 10.0.1.1/24}
Check out the runnable examples for more ideas on what is possible. This guide walks through how to run each example.
The reference documentation contains the full list of configuration options. Be sure to also read usage notes, which highlight some unique aspects of using conlink-provided networking.
Conlink also includes tools that make docker compose a much more powerful development and testing environment (refer to Compose Tools for details):
- mdc: modular management of multiple compose configurations
- wait: wait for network and file conditions before continuing
- copy: recursively copy files with variable templating
There are a number of limitations of docker-compose networking that conlink addresses:
- Operates at layer 3 of the network stack (i.e. IP).
- Has a fixed model of container interface naming: first interface is
eth0
, second iseth1
, etc. - For containers with multiple interfaces, the mapping between docker compose networks and container interface is not controllable (docker/compose#4645 (comment), docker/compose#8561 (comment))
- If a container uses the scale property, then IPs cannot be assigned and user assigned MAC addresses will be the same for every instance of that service.
- Docker bridge networking interferes with switch and bridge protocol traffic (BPDU, STP, LLDP, etc). Conlink supports the "patch" link mode that allows this type of traffic to pass correctly.
Conlink has the following features:
- Declarative network configuration (links, bridges, patches, etc)
- Event driven (container restarts and scale changes)
- Low-level control of network interfaces/links: MTU, routes, port forwarding, netem properties, etc
- Automatic IP and MAC address incrementing for scaled containers
- Central network container for easy monitoring and debug
- Composable configuration from multiple sources/locations
General:
- docker
- docker-compose version 1.25.4 or later.
Other:
- For Open vSwtich (OVS) bridging, the
openvswitch
kernel module must loaded on the host system (where docker engine is running). - For patch connections (
bridge: patch
), the kernel must support tc qdisc mirred filtering via theact_mirred
kernel module. - For podman usage (e.g. second part of
test3
), podman is required. - For remote connections/links (e.g.
test5
), thegeneve
(and/orvxlan
) kernel module must be loaded on the host system (where docker engine is running) - For CloudFormation deployment (e.g.
test6
), the AWS CLI is required.
This software is copyright Viasat and subject to the terms of the Mozilla Public License version 2.0 (MPL.20). A copy of the license is located in the LICENSE file at the top of the repository or available at https://mozilla.org/MPL/2.0/.