The purpose of this image is to dynamically add/remove containers to/from nginx as they become available in tutum. It provides NO default configuration for nginx, but gives you the ability to write your own using simple templating.
For this to have any meaning you must handle redeploying your containers in a way that they are never all down at the same time. Check out my script for this issue.
First of, create the service you will want to load balance in tutum. No need to start it, just make sure it exists.
Then create your load balancer from this image, either give it full access, or set TUTUM_USER and TUTUM_APIKEY to your username and apikey.
Set LB_SERVICE to the uuid of the service you want to loadbalance.
You can find this if you navigate to your service in the tutum web interface and look at the url. It will be something like
https://dashboard.tutum.co/container/service/show/<this is the uuid you are looking for>/
It is not required to link the containers.
You will have to provide your own configuration for nginx and mount these on /opt/conf/ in the load balancer.
But I would recommend to create your own image based on this one, where you ship the configs with it.
The files are interpretted as mako templates, with the only variable being containers, a list of the running containers in the service you specified.
The whole container object is accessible, as described in the tutum api docs.
Every .conf file in /opt/conf/ will be used.
This is a simple example on how to create an upstream from your containers:
upstream backend {
% for container in containers:
server ${container.private_ip}:8080;
% endfor
}
Optionally, you may set GRACE_PERIOD to a number (a float, really) of seconds to wait before reloading nginx, to give your service some time to start.