Let's Encrypt integration with Kong
This repository provides with a cron-based certbot agent that will attempt to acquire Let's Encrypt certificates you control for a list of subdomains you provide, and provision Kong with them.
Ideal for integrating a Kong deployment in Kubernetes with Let's Encrypt.
There's an example kubernetes cronjob you can use as a guide to deploy wherever you need.
- Kong Certbot Agent 3.x: compatible with Kong 0.14 and 1.x. Kubernetes examples compatible with Kubernetes >= 1.8
- Kong Certbot Agent 2.x: compatible with Kong 0.14 and 1.x. Kubernetes examples compatible with Kubernetes <= 1.8
- Kong Certbot Agent 1.x: compatible with Kong <= 0.13. Kubernetes examples compatible with Kubernetes <= 1.8
- The container takes 3 environment variables to operate:
- KONG_ENDPOINT: this will be the http endpoint your kong admin is at, without its path. ie
http://kong:8001
- EMAIL: this is the email address linked to your let's encrypt certificates.
- DOMAINS: this is a comma-separated list of domains we'll be asking certificates for.
- KONG_ENDPOINT: this will be the http endpoint your kong admin is at, without its path. ie
- Deploy container in your environment.
- It will automagically run the updater script every 24th of the month.
- Profit!
In order for the challenge to work correctly, you need to open up a service and a route in Kong pointing to the container at a very specific URL path. It MUST respond on every domain you're requesting certs for.
When it comes the time to run certbot, it will open an HTTP server, put some stuff on a specific path, then ping Let's Encrypt, which will attempt to read that from the domain requested. If successful, a certificate is generated.
This is a service definition example in Kong admin:
{
"host": "kong-certbot-agent",
"created_at": 1543512083,
"connect_timeout": 60000,
"id": "service-id-foo",
"protocol": "http",
"name": "KongCertbot",
"read_timeout": 60000,
"port": 80,
"updated_at": 1543513810,
"retries": 5,
"write_timeout": 60000
}
This assumes that http://kong-certbot-agent
is correctly pointing to the agent's container.
Then, associate this route to it:
{
"created_at": 1543512115,
"strip_path": false,
"hosts": [
"your.list",
"of.domains",
"for.the",
"same.certificate"
],
"preserve_host": false,
"regex_priority": 0,
"updated_at": 1543513584,
"paths": [
"/.well-known/acme-challenge"
],
"service": {
"id": "service-id-foo"
},
"methods": [
"GET"
],
"protocols": [
"http"
],
"id": "route-id-foo"
}
Here's a kubernetes cronjob example.
Your k8s service SHOULD always time out since there's nothing listening on HTTP except for when certbot itself is running and requesting certs from LE.
You can, alternatively, run the actual command yourself. This will allow you to use your own scheduling around it, as it's done on the kubernetes cronjob example.
# Get a certificate for three subdomains, and submit to kong
docker run -it --rm \
-e KONG_ENDPOINT=http://kong-admin:8001 \
-e EMAIL=foo@bar.com \
-e DOMAINS=bar.com,foo.bar.com,www.bar.com \
phpdockerio/kong-certbot-agent
# Get a TEST certificate for three subdomains, and submit to kong
docker run -it --rm \
-e KONG_ENDPOINT=http://kong-admin:8001 \
-e EMAIL=foo@bar.com \
-e DOMAINS=bar.com,foo.bar.com,www.bar.com \
-e TEST_CERT=true \
phpdockerio/kong-certbot-agent
You can give the agent a pretty big list of domains to acquire certificates for (100), but bear in mind it will be one certificate shared among all of them. You might want to set up different cronjobs for different sets of certificates, grouped in a manner that makes sense to you. Also, if one of the domains you're getting a certificate from fails the HTTP challenge, cert acquisition for the whole group fails.
Unfortunately, certbot does not support http challenges on wildcard certs, needing to resort to other types (like DNS). Due to the way certbot agent works, this will never be supported by the agent.
Yes. Certbot has a limit of 50 certificate requests per domain per week - it is very easy to go over this limit during your initial set up while you manage to get all your stuff lined up together nicely:
- Use test certs initially, allowances are more generous. You can modify the command to
command: [ "/workdir/certbot-agent", "certs:update", "$(KONG_ENDPOINT)", "$(EMAIL)", "$(DOMAINS)", "--test-cert" ]
until you have everything right. - Ensure your scheduling does not retry a failed command. It's very unlikely it will succeed a second time with the same parameters and you'll go over the limit quicker than fast, especially in Kubernetes which by default will retry until your cluster goes down. The example kubernetes cronjob specifically stops this from happening
By default, certbot has a limit of 50 certificate requests per domain per week as mentioned earlier, so bear this in mind. Also, certs are good for 3 months. Let's Encrypt themselves recommend once every 60 days. The example kubernetes cronjob is setup like so.
You can certainly do it more often, but there's no point in spamming Let's Encrypt with extra requests - remember this is a shared resource, free as in freedom and beer, and someone surely pays for it. Be considerate.