Fix POP/SMTP not working in Sophos UTM

It all started with an innocent enough e-mail:

Data Disk is filling up - please check. Current usage: 100%

I couldn’t find any clear information about what to do about this on Sophos’ forums. My data disk was full. What to do?

I can tell you what not to do – delete random files. I thought my solution would be to log into the UTM’s console and run a du -hsx /* to see where the space was. I found a large folder inside /var/storage – /var/storage/cores/httpd.16438. I removed it, because why not?

It turns out that did some weird things to my UTM. After removing that folder I kept getting spammed with these e-mails, once every hour:

Pop3 proxy not running - restarted

It took me a while to realize, but this also meant all e-mails relayed to the UTM were not being delivered. The entire POP/SMTP subsystem of the Sophos UTM was hosed. I could not find anything on the Sophos forums. After scratching my head I decided to have a deeper look at the logs. From the command line I issued

ls -ltr /var/log

and began reading the most recent logs.

pop3.log let me know what the problem was:

pop3proxy[9270]: Can't connect to database, retrying in 10 seconds: could not connect to server: Connection refused

I could not find any useful fixes for this error. I kept digging.

selfmon.log wasn’t much help other than to confirm that pop3 was having some serious issues. It was an endless abyss of repeated error messages:

selfmonng[3818]: W NOTIFYEVENT Name=pop3proxy_running Level=INFO Id=117 suppressed
selfmonng[3818]: W actionCmd(+): '/var/mdw/scripts/pop3 restart'

system.log put me on the right track:

ulogd[5107]: pg1: connect: could not connect to server: No such file or directory


Finally, we’re getting somewhere! After some searching I learned that pg1 is the postgresql database Sophos uses. I found a way to rebuild from this forum post.

One simple command did the trick:

/etc/init.d/postgresql92 rebuild

This rebuilt the postgresql database that I apparently corrupted when I removed files with reckless abandon. My e-mails work again!

Install Guacamole 0.9.8 in CentOS 7

Lately I’ve embarked in installing the latest version of Guacamole, 0.9.8, in a fresh installation of CentOS 7. Kudos go to the excellent guide I found from here.  Derek’s guide is for 0.9.7 but it also works for 0.9.8. I ran into a few hangups but after I figured them out it worked beautifully.

First, fetch the needed binaries:

rpm -Uvh   # EPEL Repo
yum -y install wget   # wget
wget && mv home\:felfert.repo /etc/yum.repos.d/   # Felfert Repo
yum -y install tomcat libvncserver freerdp libvorbis libguac libguac-client-vnc libguac-client-rdp libguac-client-ssh
yum -y install cairo-devel pango-devel libvorbis-devel openssl-devel gcc pulseaudio-libs-devel libvncserver-devel terminus-fonts \
freerdp-devel uuid-devel libssh2-devel libtelnet libtelnet-devel tomcat-webapps tomcat-admin-webapps java-1.7.0-openjdk.x86_64

Next, install guac server (the latest as of this writing is 0.9.8)

mkdir ~/guacamole && cd ~/
tar -xzf guacamole-server-0.9.8.tar.gz && cd guacamole-server-0.9.8
./configure --with-init-dir=/etc/init.d
make install

I received an error while running ./configure :

checking for jpeg_start_compress in -ljpeg... no
configure: error: "libjpeg is required for writing jpeg messages"

It means I didn’t have libjpeg dev libraries installed. Easily fixed:

yum install libjpeg-turbo-devel

Next, install the guacamole war files

mkdir -p /var/lib/guacamole && cd /var/lib/guacamole/
 wget -O guacamole.war
 ln -s /var/lib/guacamole/guacamole.war /var/lib/tomcat/webapps/
 rm -rf /usr/lib64/freerdp/
 ln -s /usr/local/lib/freerdp/ /usr/lib64/freerdp/

Next comes configuring the database

#Install database and connector
yum -y install mariadb mariadb-server
 mkdir -p ~/guacamole/sqlauth && cd ~/guacamole/sqlauth
 tar -zxf guacamole-auth-jdbc-0.9.8.tar.gz
 tar -zxf mysql-connector-java-5.1.32.tar.gz
 mkdir -p /usr/share/tomcat/.guacamole/{extensions,lib}
 mv guacamole-auth-jdbc-0.9.8/mysql/guacamole-auth-jdbc-mysql-0.9.8.jar /usr/share/tomcat/.guacamole/extensions/
 mv mysql-connector-java-5.1.32/mysql-connector-java-5.1.32-bin.jar /usr/share/tomcat/.guacamole/lib/
 systemctl restart mariadb.service

#Configure database
mysqladmin -u root password MySQLRootPass
mysql -u root -p   # Enter above password
create database guacdb;
create user 'guacuser'@'localhost' identified by 'guacDBpass';
grant select,insert,update,delete on guacdb.* to 'guacuser'@'localhost';
flush privileges;
cd ~/guacamole/sqlauth/guacamole-auth-jdbc-0.9.8/mysql/schema/
cat ./*.sql | mysql -u root -p guacdb   # Enter SQL root password set above

Now we need to configure guacamole to use our new database.

mkdir -p /etc/guacamole/ && vi /etc/guacamole/
# MySQL properties
mysql-hostname: localhost
mysql-port: 3306
mysql-database: guacdb
mysql-username: guacuser
mysql-password: guacDBpass

# Additional settings
mysql-disallow-duplicate-connections: false

Link the file you just made to the tomcat configuration directory

ln -s /etc/guacamole/ /usr/share/tomcat/.guacamole/

Cleanup temporary files and enable necessary services on boot

cd ~ && rm -rf guacamole*
systemctl enable tomcat.service && systemctl enable mariadb.service && chkconfig guacd on
systemctl reboot

Lastly, open the firewall up for port 8080 (thanks stack overflow)

firewall-cmd --permanent --add-port=8080/tcp
firewall-cmd --reload

Navigate to guacamole in your browser: http://<IP address>/guacamole:8080. You should see the guacamole login screen.

Additional hiccup

This new version of guacamole has a different user interface. It took me longer than I’d like to admit to realize how to get out of a guacamole session once it’s started. Sessions are now full screen with no obvious way to exit.

The way to exit the full screen guacamole session is to press the magic key combination of ctrl, alt, and shift. It will reveal a menu from the side. This is all clearly defined in the user documentation, but my lack of willingness to read it caused me to waste much time. Lesson learned!

Get free SSL certificates from startssl

SSL certificates can be a pain, especially if you have to pay for them. It turs out you can get free SSL certificates from, though, so at least your wallet doesn’t have to suffer!

In order to create an account with them, head over to Their account creation process is a little strange. Follow their instructions for generating a certificate for authentication (they don’t use passwords.)

Note: if you are getting frustrated because you follow their certificate login process only to have your browser tell you there is no cert it’s likely due to some caching of the certificate error page in your browser. Clear cache and cookies (or open a browser in incognito / privacy mode) and try again to log in.

First, validate your domain using their validation wizard. Once your domain is validated, head over to the Certificates wizard to generate a certificate.

I don’t trust any website that generates private SSL keys for you, so I recommend you create your own with the openssl command (steps copied from my sophos SSL certificate tutorial) and skip the creation step on their website.

  1. Generate a Certificate Signing Request (CSR) by creating a key and using it to generate the CSR
  2. openssl genrsa -aes256 -out <keyname>.key 2048
    openssl req -new -key keyname.key -out csrname.csr

2. Copy the content of the csr file into the CSR form box and click Next

3. If you’re lucky, you’ll be provided the key files immediately. Sometimes it takes a few hours for them to approve the certificate creation first.

4. Once the certificate is created, head over to Toolbox / Retrieve certificate. After selecting the appropriate certificate, copy everything in the box and paste it into a crt file.

5. Obtain Startcom’s intermediate and root CA files by going to Toolbox / Startcom CA Certificates. Download the “Server Certificate Bundle with CRLs” file.

6. Combine the generated certificate and Startcom certificate bundle into a single file:

cat ca-bundle.pem generated_crt_file.crt  > combined.crt

Sometimes you will need to wait 6-12 hours after getting key before installing it. This allows for OCSP to propagate as explained here. If you get certificate errors after installing, this may be the cause.

7. Profit.



Install Virtualbox on a Chromebook

I am really enjoying my Chromebook Pixel 2015. Recently I needed to spin up a few VMs on this box. I tried to install virtualbox but it turns out that the kernel for the chromebook does not include virtualbox headers. Fortunately it’s fairly easy to add them, thanks to divx118’s scripts.

First, enable the necessary chromebook kernel flags. Run these commands in a crosh shell (not in a chroot)

cd ~/Downloads
sudo sh ~/Downloads/change-kernel-flags

Note: You will need to repeat the above steps after each chromeos update.

Next, open up a chroot shell and do the following:

cd ~
sudo sh

Lastly, reboot the chromebook. Once you’re back up, enter your chrooted environment and install Virtualbox from their Oracle’s download page. Don’t use the virtualbox repositories – they don’t work.

Install the downloaded deb file with dpkg

dpkg -i virtualbox-5.0_5.0.10-104061~Ubuntu~trusty_amd64.deb

You might get this error:

dpkg: dependency problems prevent configuration of virtualbox-5.0:
 virtualbox-5.0 depends on libqt4-opengl (>= 4:4.7.2); however:
  Package libqt4-opengl is not installed.

The fix to that error is to run the following command to install missing dependencies:

apt-get -f install

If you get weird errors about VT-X not being enabled in the BIOS, try running the script and rebooting again.