docker-vxlan-plugin is a vxlan plugin for docker designed for on premise deployments where users have full control over their network. It is purposefully built so that users can manage routing as part of their larger infrastructure, and avoid NAT and multi-homed containers. This plugin alone handles communication between containers on a single vxlan, even (directly) between hosts. It requires an external routing system to route between vxlans. Please see our work on docker-drouter for a distributed routing system that enables short-cut routing between vxlans.
The out of box networking options for docker are built around a use case common to deployment on managed virtual servers where each host has a single public IP address. They are not well suited to bare metal datacenter deployments where routing can be controlled and multiple layers of nat is undesirable. This plugin does not nat your containers and assumes that you're running it in an environment where you know how to distribute routes to your vxlan network. You must set the gateway to the IP address of another router that already exists on your vxlan network.
Follow the Tutorial for how to get up and running quickly with docker-machine.
How it works
When a container joins a network created with the vxlan driver if doesn't already exist a vxlan interface is created. MacVlan interfaces are created for each container and attached to the vxlan interface.
Running the plugin
Run from docker
docker run -v /run/docker/plugins/:/run/docker/plugins -v /var/run/docker.sock:/var/run/docker.sock --privileged --net=host TrilliumIT/docker-vxlan-plugin
The plugin must be run in privileged mode with host networking to be able to add network links to the system.
Run outside docker
Download the latest release binary and execute it.
scope of the plugin. Can be either
global. Default is
-scope=global is specified the network options will be published to the docker cluster key-value store and containers can be brought up on the network on any host in the cluster after the network has been created. The global scope will also allow the default global IPAM driver to be used which will coordinate IP address allocation between all hosts in a docker cluster. Note that the gateway address specified during network creation will not be assigned to the host, but it will still be passed to containers as their default route. This applies in both
The device to use as the VTep endpoint. If this is specified as a daemon option it takes presidence over a VtepDev specified as a
network create option.
Network create options
The following options can be passed to
docker network create as
-o option=value. Please consult the man page for ip link and see the vxlan section for more details on some of these options.
Name of the vxlan interface
MTU of the vxlan interface
MAC Address of the vxlan interface
Transaction Queue Length of the vxlan interface
specifies the VXLAN Network Identifer (or VXLAN Segment Identifier) to use.
specifies the physical device to use for tunnel endpoint communication.
specifies the source IP address to use in outgoing packets.
specifies the multicast IP address to join.
specifies the TTL value to use in outgoing packets.
specifies the TOS value to use in outgoing packets.
specifies if unknown source link layer addresses and IP addresses are entered into the VXLAN device forwarding database.
specifies ARP proxy is turned on.
specifies if route short circuit is turned on.
specifies if netlink LLADDR miss notifications are generated.
specifies if netlink IP ADDR miss notifications are generated.
Do not age FDB entries.
enables the Group Policy extension (VXLAN-GBP).
specifies the lifetime in seconds of FDB entries learnt by the kernel.
specifies the maximum number of FDB entries.
specifies the minimum UDP source port
specifies the maximum UDP source port