Consul + CoreOS - hassles
While CoreOS already has etcd built in and it is very similar to Consul, it is missing some of the service discovery features offered by Consul. Namely the service catalog and service health checking. These things are nice to have.
This also allows you to run Registrator on CoreOS clusters and still get the Consul-backed extra features (such as health-check support).
- Clone this repo:
git clone https://github.com/cap10morgan/consul-coreos.git.
- Check if the file
/etc/environmentexists on your CoreOS machine(s).
- On AWS and probably some other cloud environments it gets created automatically. But it will not exist if you're running CoreOS on bare metal, for example.
- If it doesn't exist, create it like this:
echo 'COREOS_PRIVATE_IPV4=your.ip.address' > /etc/environment
- Submit and run the included service file on your CoreOS cluster:
fleetctl submit consul@.service fleetctl start consul@1 fleetctl start consul@2 ... as many more as you want
Fleet will run one Consul container per CoreOS node, so you probably want to start a consul@N up to the number of CoreOS nodes in your cluster. Although be aware that currently this runs all of them as Consul servers, which is not recommended past a ceratin cluster size (some of them should start being run as clients past a certain number of nodes). Pull requests welcome!
It uses etcd to coordinate the bootstrapping of a Consul cluster on CoreOS. It runs Consul inside Docker containers (as nature intended) using progrium's excellent Consul Docker image.
- Switch to adding client nodes (as opposed to more server nodes) when the cluster is above a certain size. For now this has been tested on a 3-node cluster and should work on clusters close to that size.