Last week, I needed to create a docker image doing a simple thing:

• downloading 1 XML file
• converting it to JSON
• serve both files

But I’ve a hard part: every 15 minutes, I need to look if a new version of the XML file is available and in case, running the entire process.

Let’s go to see how I did it

Requirements

• alpine linux
• gawk
• nginx unit
• python
• wget
• xml2json

Docker image creation

I use Python container based on alpine linux as the base, adding an environment variable to avoid PIP cache.

To have multiple processes inside the contianer, I use s6-overlay and I think to rewrite the global path because by default it’s not including the path where Python binaries (pip, python) are installed.

Process with dependencies and type of run

S6-overlay permits to define dependencies between processus and it’s exactly what I need because:

• docker-entrypoint needs to run before running the real nginx daemon but without persistance, it’s a oneshot at boot
• cron daemon needs to run in background and has 0 dependency
• nginx unit needs to run in background when docker-entrypoint has ended the configuration

cron

To setup cron, we have few steps inside the Dockerfile:

• copy the shell script to execute cron as the daemon
• set this script has “longrun”
• add it to be run at boot

nginx unit

Steps:

• copy the shell script to execute nginx unit as a daemon
• set unit log file to output on /dev/stdout
• set nginx access log to output on /dev/stdout
• set this script has “longrun”
• add docker-entrypoint has a dependency
• add it to be run at boot

docker-entrypoint

Steps:

• create the directory where we’ll had config files
• set this script has “oneshot”
• set the command to execute the docker-entrypoint (I don’t copy the file like for other service)

Entrypoint

Finally, we’ll use s6 has the entrypoint for this container

Link

The repository: xmltv_tnt_json_unit