A while ago I made a passing comment about replacing StartSSL using Let’s Encrypt and while it’s true you can use it in the way I described, it’s not as simple as the command would imply and there are a number of pitfalls – not least that if you side-step most of them by going a truly manual route, your efforts only last for 3 months – let’s encrypt certificates expire after only 3 months.

What this all boils down to is that in order to make proper use of this excellent free service you have to automate the process of obtaining and renewing certificates. Not just that though, because their python tooling (certbot) will do this in a general fashion for you, but to truly implement a hands off one size fits all solution. I articulated this as the following requirements.

• Tool must check for impending expiry and renew in time to avoid nag e-mails
• Support should be available for adding new certificate requests to the “pool”, not just to renew existing ones
• A fixed, permanent web server must not be required
• Servers running on non-standard ports must be supproted
• Servers not ordinarily accessible to the public must be supported
• Firewall must not be required to always be open to Let’s Encrypt servers
• Tool must support servers who’s control IP (SSH etc.) differs from that to which an SSL certificate is to be assigned (A record of SSL domain)

A evening of hacking around on a Raspberry Pi to get certbot running and some hasty coding in bash revealed a solution which essentially, does the following, in order, for each domain requiring a certificate.

1. Checks to see if the domain has a cert and if it’s expiring
2. Creates a remote directory from which files will be served as part of the renewal process
3. Inserts a temporary firewall rule to allow Lets’ Encrypt servers to validate the served files
4. Starts a temporary lightweight webserver to serve said files and runs it on a port that Let’s Encrypt supports (80)
5. Mounts this served directory locally on the certbot machine so that certbot can load files into it
6. Runs certbot to create/renew the certificate
7. Unmounts the temporary server directory
8. Stops the temporary webserver
9. Removes the temporary firewall rule
10. Cleans up temporary files
11. Load the certificate files onto the remote server over SSH
12. Bounces any services that use them so that they are picked up (e.g. Apache)

Before I share the script, a few words of warning and a few pointers

• Everything needs to run as root
• I mounted a NAS share of my certificate files so that they were stored and backed up centrally. Certbot expects to find everything in /etc/letsencrypt so that will need to be your mount point if you use the NAS option
• The script assumes that the server on which it is running has SSH keys setup on all hosts it needs to connect to such that passwordless root is possible.

If you don’t know how to do any of these things then you probably shouldn’t try using the script. It’s not that you won’t be capable or anything, rather that there are a lot of moving parts and basic Linux knowledge that would allow you to achieve these 3 points will go a long way in helping you to debug any issues with the script.

Finally then, the script; configured for two dummy domains, running on the same box, one serving from apache, the other a custom process.

#!/bin/bash
# Configuration section
email=your.email@address.com
certPath=/etc/letsencrypt/live/
remotePort=80 # Lets encrypt only supports 80 - ouch - we'll need to stop servers while doing this
domains=('one.domain.com' 'two.domain.com')
hosts=('1.2.3.5' '1.2.3.6')
sshHosts=('1.2.3.4' '1.2.3.4')
sslPathOnServer=('/etc/apache2/ssl/' '/etc/specialprocess/ssl/')
keyFileNameOnServer=('www.key' 'pub.key')
certFileNameOnServer=('www.crt' 'cert.crt')
chainFileNameOnServer=('www-chain.pem' 'chain.pem')
serverStopCommand=('/etc/init.d/apache2 stop' '/etc/init.d/specialprocess stop')
serverStartCommand=('/etc/init.d/apache2 start' '/etc/init.d/specialprocess start')

# Work is done here, no further edits!
i=0
for domain in ${domains[@]}; do
        # Ascertain if the certificate will expire in 3 weeks or less - stops lets encrypt nagging e-mails
        new=0
        if [ -d "$certPath$domain" ]
        then
                openssl x509 -in $certPath$domain/cert.pem -checkend $(( 86400 * 21 )) -noout
                result=$?
        else
                new=1
                result=1
        fi
        if [[ $result == 1 ]]
        then
                # Report the status
                if [[ $new == 1 ]]
                then
                        echo "Certificate for $domain [New]"
                        echo "Running certificate creation process..."
                else
                        echo "Ceritificate for $domain [Expiring]"
                        echo "Running renewal process..."
                fi
                # Start renewal process; create remote directory, iptables rule, start remote server
                echo -n "       Setting up iptables rules on remote server, stopping server process using certs & starting temporary micro server..."
                rule="INPUT 1 -p tcp -m tcp --dport $remotePort -d ${hosts[i]} -j ACCEPT"
                ssh root@${sshHosts[i]} << ENDSSH > /dev/null 2>&1
                eval ${serverStopCommand[i]}
                mkdir /tmp/$domain
                iptables -I $rule
                cd /tmp/$domain
                python -m SimpleHTTPServer $remotePort > /dev/null 2>&1 &
ENDSSH
                echo " Done"

                # Mount remote directory locally over SSHFS
                echo -n "       Creating mount to directory served by temporary micro server..."
                if [ -d /tmp/certbot ]
                then
                        rm -rf /tmp/certbot
                fi
                mkdir /tmp/certbot
                sshfs ${sshHosts[i]}:/tmp/$domain /tmp/certbot
                echo " Done"

                # Call letsencrypt to renew/create
                echo -n "       Call letsencrypt to renew the certificate in our store..."
                letsencrypt certonly --webroot -n -w /tmp/certbot -m $email -d $domain > /dev/null 2>&1
                echo " Done"

                # Unmount remote SSHFS directory
                echo -n "       Unmounting micro server directory and cleaning mount point..."
                umount /tmp/certbot
                rm -rf /tmp/certbot
                echo " Done"

                # SCP certificate/chain/key files to remote host
                echo -n "       Copying new certificate files to the server..."
                scp $certPath$domain/cert.pem ${sshHosts[i]}:${sslPathOnServer[i]}${certFileNameOnServer[i]} > /dev/null 2>&1
                scp $certPath$domain/privkey.pem ${sshHosts[i]}:${sslPathOnServer[i]}${keyFileNameOnServer[i]} > /dev/null 2>&1
                scp $certPath$domain/chain.pem ${sshHosts[i]}:${sslPathOnServer[i]}${chainFileNameOnServer[i]} > /dev/null 2>&1
                echo " Done"

                # Cleanup remote host; stop temp server, reload default iptables, cleanup temp directory, restart server process
                echo -n "       Stop micro server, clean up iptables rule & start server process using certs... "
                ssh ${sshHosts[i]} << ENDSSH > /dev/null 2>&1
                kill \`ps -ef | grep SimpleHTTPServer | grep -v grep | awk '{ print \$2; }'\`
                iptables -D INPUT 1
                rm -rf /tmp/$domain
                eval ${serverStartCommand[i]}
ENDSSH
                echo " Done"

                # Confirm completion
                if [[ $new == 1 ]]
                then
                        echo "Certificate for $domain [Created]"
                        echo "Creation process complete"
                else
                        echo "Renewal process complete"
                        echo "Certificate for $domain [Renewed]"
                fi
        else
                # Certificate needs no renewal, indicate as such
                echo "Certificate for $domain [OK]"
        fi
        i=$i+1
done

You’ll need to modify the variables/arrays at the start to get it working, some are obvious, but I’ll explain them anyway.

• email : This is used to create, recover and access your account. Once created, the private key for accessing your account will be stored in the let’s encrypt home directory (usully /etc/letsencrypt)
• certPath : The path to the /live directory in your let’s encrypt home directory, usually /etc/letsencrypt/live
• remotePort : Leave this at 80; it’s just here to help setup iptables rules – it cannot be changed in let’s encrypt config
• domains : An array of domains for which you want certificates
• hosts : An array of IP addresses corresponding to the aforementioned certificates – keep the ordering the same!
• sshHosts : An array of IP addresses used for administering each of the aforementioned domains; for example you may have several web servers/domains all being served from the same physical host – this would then be an array of identical IPs, corresponding in array size with the number of domains
• sslPathOnServer : What is the file system path to each domain’s certificates on the host server
• keyFileNameOnServer : Within that directory, what is the key file name
• certFileNameOnServer : Within that directory, what is the certificate file name
• chainFileNameOnServer : Within that directory, what is the chain file name
• serverStopCommand : For each domain, what command stops the server process the domain is used for
• serverStartCommand : Likewise, what is the command to start the server process again

That’s it! Best of luck if you choose to go this way and do let me know if the script is useful to you.

I recently did a apt-get dist-upgrade on my Raspberry Pi running Raspbian 8 and when I rebooted the device appeared to drop clean off my network. I was away from home at the time so I decided to fix it when I was next on the premises – it wasn’t a hugely important device to have running 24/7.

When I looked at the device in person I realised that it was in fact connected to the network, but with a different IP address to that it had held before. Having MAC address based IP allocation, this seemed rather strange. Looking at the MAC address in ifconfig revealed the problem – it had changed! Thinking this to just be a fluke, I rebooted. Now, not only was the MAC still different to that which it should have been, it had changed a third time!

Investigating what had happened during the upgrade revealed that both the bootloader and the kernel had both been upgraded, but for whatever reason, the new kernel wasn’t being loaded. Modifying the /boot/config.txt as follows along with a reboot fixed this discrepancy and lo and behold the MAC returned to it’s former value and the device picked up the IP address it was supposed to have over DHCP

From

[pi1]
kernel=vmlinuz-4.4.0-1-rpi
initramfs initrd.img-4.4.0-1-rpi followkernel
# to enable DeviceTree, comment out the next line
device_tree=

To

[pi1]
kernel=vmlinuz-4.9.0-2-rpi
initramfs initrd.img-4.9.0-2-rpi followkernel
# to enable DeviceTree, comment out the next line
device_tree=

It’s worth pointing out that you may have a similar problem with a similar cause but slightly different kernel versions to me. If this is the case, or to check if it is, verify the old/new versions in /boot as follows

ls -lha /boot | grep initrd.*rpi

Which will reveal your old/new versions

-rwxr-xr-x 1 root root 5.7M Jul 3 18:26 initrd.img-4.4.0-1-rpi
-rwxr-xr-x 1 root root 13M Jul 3 18:46 initrd.img-4.9.0-2-rpi

You should then modify /boot/config.txt from the old version to the new version as per my above example.