AWS Elastic Beanstalk provides a compelling platform for deployment of applications. My web site, the one you’re viewing this page on now, has historically been deployed on a Linux computer hosted somewhere. Well, ok, the software from which you’re reading this is WordPress but it’s hosted on the same Apache server as the main web site.
I recently redesigned the main site and in the process purposefully made the site more portable. This essentially means that I can create a zip file with the site, including relevant Apache configuration bits contained in an htaccess file and then deploy them onto any equivalent server, regardless of that server’s underlying Apache or filesystem configuration.
That got me thinking: Can I deploy the site onto Elastic Beanstalk in Amazon Web Services? The answer: Yes I can. The path that I followed was essentially to clone a clean copy of the repository from git and then create a zip file with the contents but not the .git-related bits. Here’s the command, executed from within the directory with the code:
zip -r braingia.zip . --exclude=\*.git\*
The next step is to then deploy this into AWS Elastic Beanstalk. That’s relatively straightforward using the wizard in AWS. In a few minutes, AWS had deployed a micro instance with the code on it. I needed to undo some hard-coded path information and also redo the bit within the templating system that relied on a local wordpress file for gathering recent blog posts. It wasn’t immediately clear what the issue was though, and the biggest challenge I encountered was debugging.
Debugging PHP in Elastic Beanstalk
My workflow would normally call for ssh’ing into the server and looking at error logs. However, I found that adding the ability to display errors was helpful. That setting is found within the Configuration area of Elastic Beanstalk:
However, that’s not a setting I would use on a production site. There are two other ways to troubleshoot Elastic Beanstalk. First, you can view and download Apache logs and other related logs from Elastic Beanstalk. For example, adding error_log() statements to the code would result in log entries in these easily-viewable logs.
The other debug option is to enable an EC2 key pair for the EC2 instance associated with the application. This is done at application creation or later through the Configuration. Therefore, I simply deployed another application with the same Application Version and chose an EC2 key pair this time.
Note that AWS replaces the instance entirely if you change this key at a later date, so if you have Single Instance-hosted version of the application, the site will be unavailable while a new instance is spun up.
Once the key pair is enabled on the server, it’s simply a matter of ssh’ing into the EC2 instance using the key and the ec2-user account, like so:
ssh -i mykey.pem email@example.com
Doing so, it’s easy to navigate around to see everything on the server, including log files.
Note that the current version of the application can be found within the /var/app/current directory on Elastic Beanstalk-deployed PHP applications. You can even edit files directly there, but I wouldn’t recommend it since it breaks the zip-based deployment paradigm and architecture.
In summary, Elastic Beanstalk deployment and debugging were much easier and much more powerful than I envisioned.