Let’s Encrypt when your server is behind a firewall and you can’t use DNS Challenge

Sometimes you want to give a server that is  behind a firewall a valid SSL certificate. Let’s Encrypt provides a nice solution for this called DNS Challenge. The problem with DNS Challenge is it may require some manual configuring to on your server to set it up. That is all fine and good but sometimes the server won’t allow you to do that configure. A classic example is the hass.io (Home Assistant) appliance or FreePBX

I couldn’t find a coherent set of instructions for setting up Let’s Encrypt DNS Challenges with hassio, especially as my DNS provider requires custom scripts in order to achieve this.

However, I came up with a work around. First let’s think about how Let’s Encrypt (LE) usually works. Suppose you want to get a certificate for my.example.com (MEC), but my.example.com is NATed behind example.com’s (EC) IP address. When you run the LE client on MEC, it requests the LE servers to do a challenge request to MEC. MEC will have the same public IP as EC. the client on MEC will some information on the MEC server at the <webroot>/.well-known/acme-challenge. The LE servers will try and query it. If they find what they are looking for, they issue you the certificate.

So I got to thinking, as this request takes place on port 80, it must contain the domain name header, and so, if you run an web server on port 80 of EC and A entry for  MEC the same public IP address as EC, EC could simply proxy that request to the MEC server.

I looked up how to do this and set it up on my server and low and behold, I was able to get a valid certificate.

In my case I did it with nginx , so the config looked like this:

On my server on port 80 (example.com) of the firewall in /etc/nginx/sites-available/my.example.com:

server {
       server_name my.example.com;
       location / {
		proxy_set_header Host $http_host;

Then symlink it to /etc/nginx/sites-enabled/my.example.com and test the nginx config:

$ sudo ln -s /etc/nginx/sites-available/my.example.com  /etc/nginx/sites-enabled/my.example.com
$ sudo nginx -t
[sudo] password for jason: 
nginx: the configuration file /etc/nginx/nginx.conf syntax is ok
nginx: configuration file /etc/nginx/nginx.conf test is successful

Restart nginx:

sudo systemctl restart nginx.service

Now start your Let’s Encrypt certificate request on your server and you should be good to go.

MSSQL ODBC Client on Debian 9 Stretch

Many years ago fREW Schmidt wrote a very handy article on installing MSSQL ODBC drivers in debian. Since then Microsoft have improved their support of debian somewhat and things have changed making it easier to install, however they still (as at 2017-11-06) have not released proper drivers for Debian Stretch.

Add the microsoft repo

echo  'deb [arch=amd64] https://packages.microsoft.com/debian/8/prod jessie main' > /etc/apt/sources.list.d/mssql-release.list

Add the key

curl https://packages.microsoft.com/keys/microsoft.asc | sudo apt-key add -

then update and install

sudo apt update
sudo apt install msodbcsql mssql-tools

If you try and connect to sql now you get an obscure error like this:

/opt/mssql-tools/bin/sqlcmd -S -U xxx -P yyy
Sqlcmd: Error: Microsoft ODBC Driver 13 for SQL Server : Can't open lib '/opt/microsoft/msodbcsql/lib64/libmsodbcsql-13.1.so.9.1' : file not found.

But the file exists:

$ file /opt/mssql-tools/bin/sqlcmd
/opt/mssql-tools/bin/sqlcmd: ELF 64-bit LSB executable, x86-64, version 1 (GNU/Linux), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 2.6.32, BuildID[sha1]=21a353af470e7849544daab892ec9b1bfc36dc87, not stripped

This somewhat misleading error is actually due to the libmsodbcsql lib being linked against a very specific version of  libssl. Check the output of ldd:

$ ldd /opt/microsoft/msodbcsql/lib64/libmsodbcsql-13.1.so.9.1 | grep 'not found'
libcrypto.so.1.0.0 => not found
libssl.so.1.0.0 => not found

But it turns out this can easily be solved by manually installing the libssl package from Debian Jessie:

wget "http://security.debian.org/debian-security/pool/updates/main/o/openssl/libssl1.0.0_1.0.1t-1+deb8u7_amd64.deb"
sudo apt install ./libssl1.0.0_1.0.1t-1+deb8u7_amd64.deb

then you should be back in business.

convert mp3 to m4b on the linux command line

I quite often want to convert mp3 files to m4b, Apple’s proprietary nastiness. There is pacpl of course but that has issues these days, particularly with copying the id3 tags to the new file.

FFmpeg can do it when you build from source (I just found a nice script that will build it from source for you). So I came up with this way to do it from the command line. It uses an awesome feature of xargs which will run it in parallel, so one for each core.

find -type f -name \*.mp3 \
| xargs -n 1 -P $(getconf _NPROCESSORS_ONLN || echo 1) -i \
  bash -c \
 'i="{}"; ffmpeg -y -i "$i" -map_metadata 0 \
        -c:a libfdk_aac -b:a 128k\
        -map_metadata:s:a 0:s:a -f ipod "${i%.*}.m4b"'

What that does is:

  1. find all the mp3 files and pipe their names into xargs
  2. then get the number of cores available and pass that to the -P argument
  3. and run ffmpeg and map all the metadata to the new file
  4. and strip off the mp3 file extension and add m4b to it

There! Nothing to it!

Update 2017-03-23: Added double quotes around the bash variable $i to cope with spaces. You should always do this (and so should I)

Using fail2ban to mitigate WordPress xmlrpc.php DDoS attacks

The other day my WordPress network went down. Upon investigation it turned out it was receiving a massive amount of http posts to the xmlrpc.php file. Apparently there is a WordPress DDoS that uses this mechanism. It brings apache and mysql to their knees as they can’t process the posts fast enough. If you search google for WordPress xmlrpc.php DDoS you can find lot more info about this.

An temporary fix is to block all access to that file from your apache configs with something like:

<Files xmlrpc.php>
  Order allow,deny
  Deny from all

That brought the load back to normal so I could at least access the WordPress backend.

After googling around for a solution it appeared that fail2ban could help. Luckily there is a plugin for that. WP fail2ban has two parts. The first is a plugin that enables logging of xmlrpc events and authentication events to /var/log/auth. It is important to keep these events separate from the normal http access logs as the access log file can get very large and fail2ban can raise the load significantly just processing it.

You also need to add a few configuration options to wp-config.php

# prevent user enumeration
# block some obviously invalid users

See the readme for more details about what they do

The second part is enabling filters and gaols in fail2ban. Luckily this is also provided by the WP fail2ban plugin. Copy the wordpress.conf file from the wp-fail2ban directory to the fail2ban config directory:

~# cp /var/www/wp-content/plugins/wp-fail2ban/wordpress.conf \

Then edit /etc/jail.local and insert:

enabled = true
filter = wordpress
logpath = /var/log/auth.log

# set the ban time to 1 hour - probably could be even higher for good measure
bantime = 3600

# needed for debian wheezy otherwise fail2ban doesn't start and reports
#   errors with the config
port = http,https

Now restart fail2ban:

~# /etc/init.d/fail2ban restart
[ ok ] Restarting authentication failure monitor: fail2ban.

Remove the block on the xmlrpc.php file from your apache config and restart apache. Then you should see in your fail2ban logs something like:

2014-08-09 23:18:30,405 fail2ban.actions: WARNING [wordpress] Ban
2014-08-09 23:20:49,090 fail2ban.actions: WARNING [wordpress] Ban
2014-08-09 23:20:50,108 fail2ban.actions: WARNING [wordpress] Ban
2014-08-09 23:21:04,162 fail2ban.actions: WARNING [wordpress] Ban
2014-08-09 23:21:28,206 fail2ban.actions: WARNING [wordpress] Ban
2014-08-09 23:21:36,234 fail2ban.actions: WARNING [wordpress] Ban
2014-08-09 23:21:36,294 fail2ban.actions: WARNING [wordpress] Ban
2014-08-09 23:21:44,346 fail2ban.actions: WARNING [wordpress] Ban
2014-08-09 23:21:46,400 fail2ban.actions: WARNING [wordpress] Ban
2014-08-09 23:21:52,423 fail2ban.actions: WARNING [wordpress] Ban
2014-08-09 23:22:12,488 fail2ban.actions: WARNING [wordpress] Ban
2014-08-09 23:22:12,509 fail2ban.actions: WARNING [wordpress] Ban
2014-08-09 23:22:42,564 fail2ban.actions: WARNING [wordpress] Ban
2014-08-09 23:22:53,590 fail2ban.actions: WARNING [wordpress] Ban
2014-08-09 23:22:53,607 fail2ban.actions: WARNING [wordpress] Ban

I found however that I was being hit from over 1800 unique IP addresses and despite fail2ban successfully banning them, it was taking too long to ban enough that the load would return to normal so I re-blocked the xmlrpc.php file for 24 hours. After that, I enabled it and it seemed as though the DDoS had gone away. So far so good.

Howto quickly find your Beaglebone Black’s IP address

beaglebone black board with thought bubble containing an IP address

Whenever I connect my Beaglebone Black (BBB) to a network, I have to work out it’s IP address so I can ssh into it. This can be tricky. Some of your options are:

  1. connect to the serial terminal, or connect over the usb network interface which gives the BBB the address, log in and issue the command ip addr.
  2. use nmap to try and search out its IP address on your subnet but this is time consuming and not very accurate I have found.
  3. use avahi-browse -rat (thanks Madox for that tip.)

Last night I came up with a Better Way™. Rather than trying to determine the BBB’s address, why not use a fully qualified domain name and a dynamic dns service? I could then just type ssh myfqdn.com or whatever and log in. Think how that would simplify one’s life!

To implement this, set up a dynamic DNS somewhere with a FQDN for your BBB. If you happen to have your own domain name already you can use a sub-domain from that. I think its fairly common for DNS hosts to offer an API to update your IP address. I happen to use Rimu Hosting and they have their own simple web api.

Then you just need to write a little script to update the IP address every time the DHCP client receives a new IP address, and drop it into /etc/dhcp/dhclient-exit-hooks.d/

Here is my script. This will only work with Rimu Hosting as they have their own privately developed API, and you’d need to insert your own KEY into the script.

# update ip address with rimu hosting. See https://rimuhosting.com/dns/dyndns.jsp
if [[ ! -z ${new_ip_address} ]]
   echo $(date +"%F %T") ${new_ip_address} &gt;&gt; /root/ddns.txt
   curl "https://rimuhosting.com/dns/dyndns.jsp?action=SET&amp;name=clock.emacstragic.net&amp;value=${new_ip_address}&amp;type=A&amp;api_key=XXX"
    echo "got no ip"


I discovered this didn’t work at home. Turns out that dnsmasq in OpenWRT is set to ignore this kind of dns request, due to potential security risks. There is a solution to that. Add a list rebind_domain line to your /etc/config/dhcp line on the router.

config dnsmasq
	list    rebind_domain 'clock.emacstragic.net'

Thanks to Michal Čihař for the solution to that.

Backup your Beaglebone Black

image of a beaglebone black with a backup icon over the top of it

Using a method I previously wrote about, it’s quite easy to backup your Beaglebone Black over the network.

ssh root@bbb 'tar cf - / 2>/dev/null' | pv -cN tar \
   | bzip2 | pv -cN bzip2 > bbb-backup-tar.bz2

It runs the bzip compression locally because presumably it will be faster than on the Beaglebone Black. I didn’t actually test that hypothesis though.

pv gives nice little indicators:

    bzip2: 1.81MB 0:00:12 [ 995kB/s] [      <=>                               ]
      tar: 36.2MB 0:00:12 [3.84MB/s] [                             <=>        ]

Printing from Windows to a samba shared CUPS-PDF printer sometimes fails

cups log with a red error cross on it

I had this problem where prints to our CUPS-PDF printer sometimes failed to be processed on the server. The job would disappear as though it has been printed but nothing else would happen. Printing from the same application to a Windows based PDF printer, and then printing the resulting PDF via Adobe Acrobat to the CUPS PDF printer would work fine. Printing the same PDF via Sumatra PDF to CUPS-PDF would also fail.

Further investigation revealed that the resulting print job files would differ. The jobs that fail looked like they contained a lot of binary data but the ones that succeeded looked like normal PDF files.

Then I discovered this entry in the Windows Event Viewer:

The document XXXX, owned by jason, failed to print on printer \\server\PDF. Try to print the document again, or restart the print spooler.
Data type: NT EMF 1.008. Size of the spool file in bytes: 2555904. Number of bytes printed: 0. Total number of pages in the document: 1. Number of pages printed: 0. Client computer: \\CLIENT. Win32 error code returned by the print processor: 0. The operation completed successfully.

Googleing that error took me to this RPi forum which had a solution buried down near the bottom. Thanks to Chemirocha for that tip. This bug has been plaguing me for a few years on and off!

automysqlbackup ERRORS REPORTED: MySQL Backup error Log Warning: Skipping the data of table mysql.event. Specify the –events option explicitly.

I was receiving email error messages from cron like this from my autbackupmysql on a regular basis:
ERRORS REPORTED: MySQL Backup error Log for somehost.com.dxa - 2014-05-01_06h26m
-- Warning: Skipping the data of table mysql.event. Specify the --events option explicitly.

It turns out that mysqldump now warns you if the events table is not being dumped. So to get rid of the warning either ensure the table gets dumped when you do a backup or tell mysql explicitly not to dump it. I chose the former approach as it is a backup after all.
Simply add the following line to /etc/mysql/my.cnf


This tells the mysqldump program to explicity include the events table, and removes the warning. You can see a discussion about this option here.

If you are using debian, you will need to add that section to the /etc/mysql/debian.cnf file also as automysqlbackup uses that file for its configuration instead. See debian bug report for more details.

increase wordpress Maximum upload file size beyond 8MB.

There is at least three places you need to set the maximum upload file size in WordPress.

First check your php.ini

upload_max_filesize = 64M;
post_max_size = 64M;

Restart Apache after making this change.

Then if you are using WordPress multi-site, you need change a setting in the backend. Visit My Sites -> Network Admin -> Dashboard -> Settings -> Network Settings. Then scroll down to the Upload Settings area and change that.

Thanks to dorr13 for that tip.