Deployment: The Art of Uploading your Code

This wouldn’t be much of a tutorial if we didn’t at least help show you how to share your project with the world! There are a lot of neat deployment tools out there and I’m sorry, we’re not going to show you any of them. At least not in this screencast. Instead, we’ll go through the exact steps you’ll need for deployment. If you want to automate them, awesome!

To keep things simple, I’m going to “deploy” to a different directory right on my local machine. So, just pretend this is our server and I’ve already ssh’ed into it.

We already have MySQL, PHP and Apache up and running.

Step 1) Upload the Files

First, we’ve gotta get the files up to the server! The easiest way is just to clone your git repository right on the server. To do this, you’ll need to push your code somewhere accessible, like GitHub. The finished code for this tutorial already lives on GitHub, under a branch called episode4-finish.

Let’s clone this repository:

git clone

Move into the directory. If your code lives anywhere other than the master branch, you’ll need to switch to that branch:

git checkout -b episode4-finish origin/episode4-finish

GitHub might ask you to authenticate yourself or give you some public key error. If that happens, you’ll need to register the public key of your server as a deploy key for your repository. This is what gives your server permission to access the code.

GitHub has great articles on deploy keys and generating a public key.

Step 2) Configuring the Web Server

Code, check! Next, let’s configure the web server. I’m using Apache, but Symfony has a cookbook article about using Nginx. Find your Apache configuration and add a new VirtualHost that points to the web/ directory of our project. In our case, /var/www/

<VirtualHost *:80>
    DocumentRoot /var/www/

    <Directory /var/www/>
        Options Indexes FollowSymlinks
        AllowOverride All

        # Use these 2 lines for Apache 2.3 and below
        Order allow,deny
        allow from all

        # Use this line for Apache 2.4 and above
        Require all granted

    ErrorLog /var/log/apache2/events_error.log
    CustomLog /var/log/apache2/events_access.log combined

The VirtualHost is pretty simple and needs ServerName, DocumentRoot and Directory keys.

Restart your webserver. For many servers, this is done by calling service restart apache:

sudo service restart apache2

Project: First-Time Setup

Code, check! VirtualHost, check!

Since this is the first time we’ve deployed, we need to do some one-time setup.

First, download Composer and use it to install our vendor files:

curl -sS | php
php composer.phar install

At the end, it’ll ask you for values to fill into your parameters.yml file. You’ll need to have a database user and password ready.

Speaking of, let’s create the database and insert the schema. I’ll even run the fixtures to give our site some starting data:

php app/console doctrine:database:create
php app/console doctrine:schema:create
php app/console doctrine:fixtures:load

In this pretend scenario, I’ve already pointed the DNS for to my server. So let’s try it:

It’s alive! And with a big error, which might just show up as the white screen of death on your server. Symfony can’t write to the cache directory. We need to do a one-time chmod on it and the logs dir:

sudo chmod -R 777 app/cache/ app/logs/

Let’s try again. Ok, we have a site, and we can even login as Wayne. But it’s missing all the styles. Ah, right, dump the assetic assets:

php app/console assetic:dump --env=prod

Crap! Scroll up. This failed when trying to run uglifycss. I don’t have Uglifycss installed on this machine yet. To get ugly Just run npm install to fix this.

php app/console assetic:dump --env=prod

Now, the dump works, AND the site looks great!

Things to do on each Deploy

On your next deploy, things will be even easier. Here’s a simple guide:

  1. Update your Code. With our method, that’s as simple as running a git pull:
git pull origin
  1. Just in case we added any new libraries to Composer, run the install command:
php composer.phar install
  1. Update your database schema. The easy, but maybe dangerous way is with the schema update console command:
php app/console doctrine:schema:update --force

Why dangerous? Let’s say you rename a property from name to firstName. Instead of renaming the column, this task may just drop name and add firstName. That would mean that you’d lose all that data!

There’s a library called Doctrine Migrations that helps do this safely.

  1. Clear your production cache:
php app/console cache:clear --env=prod
  1. Dump your Assetic assets:
php app/console assetic:dump --env=prod

That’s it! As your site grows, you may have more and more things you need to setup. But for now, it’s simple.

Performance Setup you Need

One more thing. There are a few really easy wins to maximize Symfony’s performance.

First, when you deploy, dump Composer’s optimized autoloader:

php composer.phar dump-autoload --optimize

This helps Composer’s autoloader find classes faster, sometimes much faster. And hey, there’s no downside at all to this!


If you add the –optimize-autoloader flag, Composer will generate a class map, which will give your whole application a performance boost. Using the APC ClassLoader may give you an even bigger boost.

Next, make sure you have a byte code cache installed on your server. For PHP 5.4 and earlier, this was called APC. For 5.5 and later, it’s called OPcache. In the background, these cache the compiled PHP files, making your site much faster. Again, there’s no downside here - make sure you have one of these on your server.

And on that note, PHP typically gets faster from version to version. So staying on the latest version is good for more than just security and features. Thanks PHPeeps!

Ok, that’s it! Now google around for some deployment tools to automate this!

Leave a comment!

  • 2015-06-15 Theravadan

    Thank you very much. it all makes sense.

  • 2015-06-12 weaverryan


    If you use migrations ( then you'll be able to review the SQL in your new migration file to make sure it works. For example, if you renamed a column and the generated migration did a "drop and add" instead of a rename, you could modify that SQL until you're happy with it.

    Then, when you deploy, you *never* run doctrine:schema:update - you only ever run doctrine:migrations:migrate. If you need to rename a table, then you will have already made sure that your new migration file performs this rename. If you've done everything correct, then after your migrations run, you production database will be perfectly in-sync. In other words, if you *did* run doctrine:schema:update --dump-sql after migrations, it should come back with no changes to be made. Does that make sense? Getting the workflow correct is really important :).

    About multiple bundles and only one migration_versions table, that is by design and a good thing! Sure, you may have multiple bundles, but you only have one database, and so it makes sense to only have one set of migration files. Now, imagine that I install a third-party bundle that gives me 2 new tables. This bundle will *not* ship with its own migration file to add these 2 tables. And, that's ok! As soon as you've installed the bundle, you can run doctrine:migrations:diff, and this will cause a new migration file to be generated with the CREATE TABLE for those 2 new tables.

    Does it all make sense? Let me know!

  • 2015-06-10 Theravadan

    Hi Guys,
    you mention about doctrine migrations being a safer way for renaming the table columns, but how would you make sure that update schema does not break your database?
    In other words how do you tell update schema not to drop this table because there is a fix in doctrine migrations for example?
    Also another question is how do you deal with having multiple bundles having their own migrations, since there is only one migration_versions table?

    Thanks a million!