Note: This is an experimental feature which has issues. We welcome contributions from the community to improve it.

Discovery

Weave Discovery retrieves and watches the list of peers in a Weave network from a discovery backend. Discovery uses this information for telling a Weave Router about new peers it must connect to, as well as peers that are not available anymore and must be forgotten. Discovery can also advertise the Weave Router in the same backend so other peers can join it.

Installation

Just download the discovery script from the repository, or use this URL:

curl -L -O https://raw.githubusercontent.com/weaveworks/discovery/master/discovery
chmod a+x discovery

This will download the discovery script to the current directory.

Usage

Once you have the script, joining a new discovery URL is as simple as running:

$ ./discovery join <URL>

where the endpoint URL can be something like

• etcd://<etcd_ip>/<path>
• consul://<consul_addr>/<path>
• zk://<zookeeper_addr1>,<zookeeper_addr2>/<path>
• token://<swarm_cluster_id>
• file://<path>

Example:

$ ./discovery join etcd://192.168.9.1/mycluster

(you can get more information about the supported backends as well as the URL formats here)

The discovery endpoint must provide a valid list of peers (as IP:port of Weave routers) that will be part of the network. This information can be filled up manually, but Discovery can automatically advertise its Router with the --advertise-router flag:

$ ./discovery join --advertise-router etcd://192.168.9.1/mynetwork

This will advertise the local router using then IP of the default interface. But sometimes it can be useful to advertise a particular IP:port instead of the automatically obtained values (for example, when using a non standard port). In this case, the --advertise argument will do the job:

$ ./discovery join --advertise=193.144.60.100:16783 etcd://192.168.9.1/mynetwork

Discovery could also advertise the guessed external IP with the --advertise-external flag. However this will not perform any NAT mapping, so make sure the router is accessible at this IP and port by the rest of the peers in the backend.

By default, Discovery will configure the Weave Router in the local machine. However, you can point Discovery to a different Router with the --weave flag. For example, we could point Discovery to the Weave HTTP API found at 192.168.9.3:6784 with:

$ ./discovery join --weave=192.168.9.3:6784 etcd://192.168.9.1/mynetwork

Using Discovery with Swarm

Discovery uses the same backends used by Docker Swarm, so the same endpoint used for keeping the list of Swarm nodes can be also used for keeping the list of peers in the Weave network.

In this case, for each node in the cluster, you should:

• launch the Weave Router, Proxy and Discovery
• launch the Swarm agent, advertising the Proxy IP:port (eg, 10.246.2.2:12375)

The Swarm Agent will register the Proxy IP:port in the discovery backend, so you don't need to --advertise anything from Discovery: Swarm will do it for you.

However, Swarm will advertise the Proxy IP and port. While the IP is correct, the port is not the port where a Weave Router is listening at. You must fix this by replacing the port discovered by the port where the Router can be found:

$ ./discovery join --discovered-port=6783 etcd://192.168.9.1/mynetwork

Note that this presents a limitation when using Discovery with Swarm: all the Routers must be located at the same port (and you must specify that port on the --discovered-port arguments).

Example

A Vagrant file has been included for testing the functionaly of Discovery with Swarm. The Vagrant will start three VMs, each one with:

• a Weave router and proxy
• a Swarm agent that uses some token and points to the Weave proxy
• a Discovery instance that uses the same token and points to the Weave router


You can launch the Vagrant VMs with:

	$ vagrant up --provision

Machines will be provisioned and a new Swarm token ID will be automatically generated. You will see the ID in the console messages. After a few seconds, you will be able to check the list of peers that have successfully registered in this token with:

	$ swarm list token://<TOKEN>
	10.246.2.3:12375
	10.246.2.4:12375
	10.246.2.2:12375

You can start the cluster manager in the host machine (or in any other machine with connectivity to the Vagrants, like another VM):

	$ docker run -d -p 22375:2375 swarm manage token://<TOKEN>

This Swarm manager will then provide a Docker API at port 22375. Point your Docker client to this port for getting the list of the machines in the cluster with:

	$ docker -H 127.0.0.1:22375 info
	
	Containers: 17
    Images: 26
    Storage Driver: 
    Role: primary
    Strategy: spread
    Filters: affinity, health, constraint, port, dependency
    Nodes: 3
     minion-0: 10.246.2.2:12375
      └ Containers: 6
      └ Reserved CPUs: 0 / 2
      └ Reserved Memory: 0 B / 2.052 GiB
      └ Labels: executiondriver=native-0.2, kernelversion=3.19.0-21-generic, operatingsystem=Ubuntu 15.04, storagedriver=overlay
     minion-1: 10.246.2.3:12375
      └ Containers: 6
      └ Reserved CPUs: 0 / 2
      └ Reserved Memory: 0 B / 2.052 GiB
      └ Labels: executiondriver=native-0.2, kernelversion=3.19.0-21-generic, operatingsystem=Ubuntu 15.04, storagedriver=overlay
     minion-2: 10.246.2.4:12375
      └ Containers: 5
      └ Reserved CPUs: 0 / 2
      └ Reserved Memory: 0 B / 2.052 GiB
      └ Labels: executiondriver=native-0.2, kernelversion=3.19.0-21-generic, operatingsystem=Ubuntu 15.04, storagedriver=overlay
    Execution Driver: 
    Kernel Version: 
    Operating System: 
    CPUs: 6
    Total Memory: 6.155 GiB
    Name: 
    ID: 

Now you can run a Ubuntu shell in some machine in the cluster, with:

	$ docker -H 127.0.0.1:22375 run -ti  ubuntu
	root@7f1007c28c1c:/# 

and this container will be automatically attached to the Weave network.