For a pet project I needed Pligg which is kinda a social network site in the style of the old-digg (with karma and this kind of things).
Installing a LAMP system is kinda boring stuff, so, for that and for the sake of learning I decided to go with Docker & Fig. Sadly or luckily, every time that I dockerize something I find a lot of unexpected problems that slow me down, but… you always learn something!
This post is to explain a little bit the process and the problems that I found and how I solved them. It’s not a how to, it’s more a explanation of what I did and perhaps you can provide a better solution on the comments.
The site is “simple” so the only required stuff will be a MySQL server and an
Apache2 server. You could run 2 Docker instances manually, or… use the magic
provided by Fig. This is the
fig.yml file that explains my service:
web: build: . links: - db volumes: - /var/log:/var/www/logs ports: - "80:80" environment: - MYSQL_PASSWORD - MY_BASE_URL db: image: mysql volumes: - /var/lib/mysql:/var/lib/mysql environment: - MYSQL_DATABASE=dbpligg - MYSQL_USER=pligguser - MYSQL_ROOT_PASSWORD - MYSQL_PASSWORD ports: - "3306:3306"
Basically I am saying that I will have 2 servers: “web” & “db” and specifying their volumes, environment variables and exported ports.
First thing that we see here:
have any value, this is because Fig is going to got those values from the
current Docker host. I need them for specified some settings file that
originally were written in a normal file (difficult to change while deploying).
MY_BASE_URL is a small/ugly trick. It seems that Pligg requires to know
the host were it’s running to serve static assets as CSS or JS.
If you take a look to my repo
will see the slightly modified versions of two configuration files for Pligg
that are making use of this environment variables.
For running Pligg we need a minimal DB structure I’ve found 2 different ways of creating this data in my data container, but none of them are optimal for me, mainly because they require an extra step:
I’ve created minimal SQL dump with some default values and on the
provide a quick way of ingest this data using the same container:
$ docker exec dockerpligg_db_1 \ mysql -u pligguser -p$MYSQL_PASSWORD dbpligg < "`cat pligg.sql`"
Another way is going through all the installation process accessing to http://example.com/install
I think that perhaps creating a script that checks is the DB if empty and if it is it uses the SQL to dump the DB back would be an option, but seems kinda dangerous to automate that process in a live environment (anybody said delete data by a mistaken dump?).
@eloycoto has recommended me to use inheritance of
containers. But I am not happy with that solution either: I would need two
Dockerfiles and possibly two
fig.yml files as well or add some weird magic
to replace one container with the other after the installation.
After the first time that you run the installation you need to manually remove
the install path (this seems quite common in PHP apps?). I am doing that
docker exec to that container, but I would prefer to manually
remove it with the
Dockerfile. Why I don’t do so? Because if I remove that
folder I am forcing all the users of my configuration to use the dump SQL
method explained above and I don’t give them any change.
I suspect that it’s normal to have that kind of problems trying to “migrate” an application that was never used before in a Docker container.
To be honest with Pligg, the only changes that I required were minimal, but I
don’t know if that “install” part could be just removed with my own settings
file, I tried that and it was asking me to repeat values that were already set
I am not happy with the dump/install solution that I’ve found, but it works™!
I am sure that if you were using Docker or Fig before you will have plenty of complaints about my article, let me know leaving a comment or just tweet me something.