Let's Encrypt DC/OS!
This repository is now deprecated, and this project has moved to https://github.com/dcos-labs/letsencrypt-dcos
This is a sample Marathon app for encrypting your Marathon-lb HAProxy endpoints using Let's Encrypt. With this, you can automatically generate and renew valid SSL certs with Marathon-lb.
Getting started
Clone (or manually copy) this repo, and modify the letsencrypt-dcos.json file to include:
- The list of hostnames (must be FQDNs) for which you want to generate SSL certs (in
HAPROXY_0_VHOST) - An admin email address for your certificate (in
LETSENCRYPT_EMAIL) - The Marathon API endpoint (in
MARATHON_URL) - The Marathon-lb app ID (in
MARATHON_LB_ID) - Ensure you have at least 2 or more public agents in your DC/OS cluster, and that marathon-lb is scaled out to more than 1 public agent. Deploying this app requires this since it entails restarting marathon-lb.
Now launch the letsencrypt-dcos Marathon app:
$ dcos marathon app add letsencrypt-dcos.json
There are 2 test apps included, based on openresty, which you can use to test everything. Have a look in the test/ directory within the repo.
How does it work?
The app includes 2 scripts: run.sh and post_cert.py. The first script (run.sh) will generate the initial SSL cert and POST the cert to Marathon for Marathon-lb. It will then attempt to renew & update the cert every 24 hours. The post_cert.py script will compare the current cert in Marathon to the current live cert, and update it as necessary. post_cert.py is called after the initial cert is generated, and again every 24 hours after a renewal attempt.
A persistent volume called data is mounted inside the container at /etc/letsencrypt which contains the certificates and other generated state.
Limitations
- You may only have up to 100 domains per cert.
- Let's Encrypt currently has rate limits, such as issuing a maximum of 5 certs per set of domains per week.
- Currently, when the cert is updated, it requires a full redeploy of Marathon-lb. This means there may be a few seconds of downtime as the deployment occurs. This can be mitigated by placing another LB (such as an ELB or F5) in front of HAProxy.