stensonb / etcd

A service for autodiscovery and configuration of applications running in containers

Autopilot pattern etcd

Autopilot Pattern implementation of etcd

This repo is a demonstration of etcd designed for self-operation according to the Autopilot pattern.

An etcd cluster needs an external source of data for all the nodes to find each other initially. This can be a bootstrap service (another etcd cluster) or an SRV record. Triton CNS does not yet support SRV records, so until it does we're standing up a temporary single-node cluster to bootstrap the cluster. After the cluster is scaled-up we can remove the bootstrap node.

Getting started

1. Get a Joyent account and add your SSH key.
2. Install the Docker Toolbox (including docker and docker-compose) on your laptop or other environment.
3. Install the the Joyent Triton CLI (triton replaces our old sdc-* CLI tools) and set up your Triton profile.

At this point you're ready to start the cluster. A script ./start.sh has been provided. It uses Docker Compose and creates the initial discovery token (see the etcd docs on cluster discovery for details) before scaling up the cluster. You can pass an environment variable SCALE to the ./start.sh script to set the cluster size to something other than the default 3 nodes.

$ ./start.sh
Using discovery node for bootstrapping 3-node cluster.
e_bootstrap_1 is up-to-date
{"action":"set","node":{"key":"/discovery/DD6E6035-9795-49E7-8762-49DDEE8455A9/_config/size","value":"3","modifiedIndex":4,"createdIndex":4}}
e_bootstrap_1 is up-to-date
Creating e_etcd_1
Creating and starting e_etcd_2 ... done
Creating and starting e_etcd_3 ... done

$ docker ps CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES dd08958fb3e6 autopilotpattern/etcd "/usr/local/bin/conta" 45 seconds ago Up 36 seconds 0.0.0.0:2379-2380->2379-2380/tcp, 0.0.0.0:4001->4001/tcp e_etcd_2 8fa74cc415de autopilotpattern/etcd "/usr/local/bin/conta" 57 seconds ago Up 42 seconds 0.0.0.0:2379-2380->2379-2380/tcp, 0.0.0.0:4001->4001/tcp e_etcd_3 59b1028eabec autopilotpattern/etcd "/usr/local/bin/conta" About a minute ago Up About a minute 0.0.0.0:2379-2380->2379-2380/tcp, 0.0.0.0:4001->4001/tcp e_etcd_1 adf54d657250 autopilotpattern/etcd "/usr/local/bin/etcd " 3 minutes ago Up 3 minutes 0.0.0.0:2379-2380->2379-2380/tcp, 0.0.0.0:4001->4001/tcp e_bootstrap_1

And we can now check the cluster health:

$ docker exec -it e_etcd_2 /usr/local/bin/etcdctl cluster-health member 2c2faabfc9a8dee9 is healthy: got healthy result from http://192.168.129.168:2379 member 5d4b828adc525786 is healthy: got healthy result from http://192.168.129.169:2379 member f556ea8cb6e4c0f0 is healthy: got healthy result from http://192.168.129.170:2379 cluster is healthy