Upgrading from Ubuntu 14.04 -> 16.04.1 under Amazon EC2

The Upgrade Process

I finally got around to upgrading my web server from Ubuntu 14.04 to Ubuntu 16.04.1. I decided to jot down my experiences in case anyone else runs into these issues.

Preparation

The first thing I did prior to doing the upgrade was to run updraft plus to backup the Blog’s database and files. After that finished I logged into the Amazon EC2 console and created an image of my AMI. This way in the worst case scenario that I trashed my image I could just boot the backup image and be where I was before I began.

The next thing I did was check to see if I had screen installed. My thinking on these sorts of upgrades is you can easily lose your network connection and it is nice to be able to get right back to where you were without worrying about it if something goes wrong.

Beginning the upgrade

Ubuntu has a handy command:

 sudo do-release-upgrade

I invoked the command. It let me know that it would start ssh on an additional port just in case I got disconnected during the upgrade. I said yes to begin and it began downloading all the packages it needed. Amazon seems to have a ton of bandwidth so this step is super fast. Wouldn’t you know it but in the middle of this my network connection dropped. Luckily I was able to ssh back in and reattach to my screen session without issue. There were a couple of questions I had to answer about whether to keep a config file that I had modified or else replace it with the package maintainers. I made the decision for those questions based on looking at the diff to see if the differences in the files were important or not. At the end it asks to reboot to finalize the upgrade so I said yes.

The Aftermath

The machine booted up quickly and as soon as I was able to ssh in I logged into the website to see if everything was still working. I was promptly greated with a 502 Bad Gateway error. I started poking around and I could see that my nginx config hadn’t been changed and the service seemed to be running. At that point I figured it must be an issue with php-fpm. It turns out that it didn’t even have the package installed for php-fpm and the php packages that I did have were upgraded from 5.0 to 7.0. After googling around I was able to find out that the unix domain socket name and location have changed in the new version of php. I edited my default.conf for nginx and pointed it to the new socket location:

fastcgi_pass unix:/run/php/php7.0-fpm.sock;

That being done I needed to make some changes on the php-fpm side. The config file has moved where previous it was located under /etc/php5/fpm/pool.d/www.conf the new location is /etc/php/7.0/fpm/pool.d/www.conf

In this file I needed to update the following fields:

listen = /run/php/php7.0-fpm.sock
listen.owner = nginx
listen.group = nginx

In this case I didn’t change the listen value I just used that value above to update my default.conf under nginx. After setting those values and restarting php7.0-fpm and nginx services my site was back up. Hopefully this will be helpful to anyone else who hits the same issue after they upgrade.