Let’s Encrypt makes TLS much easier for pretty much everyone, but can still be annoying to use. It took me a while to smooth over the cracks in my Let’s Encrypt configuration across my (large) fleet of different TLS-enabled services. I wanted to take a quick moment to share setup with you.
2020-01-02 update: acme-client is unmaintained and caught the BSD disease anyway. I use uacme and my current procedure is documented on my new server checklist. It might not be exactly applicable to your circumstances, YMMV.
The main components are:
- acme-client
- nginx
- cron
nginx and cron need no introduction, but acme-client deserves a closer look. The acme client blessed by Let’s Encrypt is certbot, but BOY is it complicated. It’s a big ol’ pile of Python and I’ve found it fragile, complicated, and annoying. The goal of maintaining your nginx and apache configs for you is well intentioned but ultimately useless for advanced users. The complexity of certbot is through the roof, and complicated software breaks.
I bounced between alternatives for a while but when I found acme-client, it totally clicked. This one is written in C with minimal dependencies (LibreSSL and libcurl, no brainers IMO). I bring a statically linked acme-client binary with me to new servers and setup time approaches zero as a result.
I use nginx to answer challenges (and for some services, to use the final
certificates for HTTPS - did you know you can use Let’s Encrypt for more
protocols than just HTTPS?). I quickly mkdir -p /var/www/acme/.well-known/acme-challenge, make sure nginx can read it, and add
the following rules to nginx to handle challenges:
server {
listen 80;
listen [::]:80;
server_name example.org;
location ^~ /.well-known/acme-challenge {
alias /var/www/acme;
}
}
If I’m not using the certificates for HTTPS, this is all I need. But assuming I have some kind of website going, the full configuration usually looks more like this:
server {
listen 80;
listen [::]:80;
server_name example.org;
location / {
return 302 https://$server_name$request_uri;
}
location ^~ /.well-known/acme-challenge {
alias /var/www/acme;
}
}
server {
listen 443 ssl;
listen [::]:443 ssl;
server_name example.org;
ssl_certificate /etc/ssl/acme/$server_name/fullchain.pem;
ssl_certificate_key /etc/ssl/acme/$server_name/privkey.pem;
location ^~ /.well-known/acme-challenge {
alias /var/www/acme;
}
# ...application specific rules...
}
This covers the nginx side of things. To actually do certificate negotiation, I have a simple script I carry around:
exec >>/var/log/acme 2>&1
date
acme() {
site=$1
shift
acme-client -vNn \
-c /etc/ssl/acme/$site/ \
-k /etc/ssl/acme/$site/privkey.pem \
$site $*
}
acme example.org subd1.example.org subd2.example.org
nginx -s reload
The first two lines set up a log file in /var/log/acme I can use to debug any
issues that arise. Then I have a little helper function that wires up
acme-client the way I like it, and I can call it for each domain I need certs
for on this server. The last line changes if I’m doing something other than
HTTPS with the certs (for example, postfix reload).
One gotcha is that acme-client will bail out if the directories don’t exist when
you run it, so a quick mkdir -p /etc/ssl/acme/example.org when adding new
sites is necessary
The final step is a simple cron entry that runs the script daily:
0 0 * * * /usr/local/bin/acme-update-certs
It’s that easy. It took me a while to get a Let’s Encrypt setup that was simple and satisfactory, but I believe I’ve settled on this one. I hope you find it useful!