Docker image for QGIS server

Usage

The Docker container needs to have access to all files of the QGIS project to be published. Either you create another image to add the files or you inject them using a volume. For example, if your QGIS project is stored in ./qgis/project.qgz:

docker run --detach --publish=8380:80 --volume=${PWD}/qgis:/etc/qgisserver camptocamp/qgis-server

With the previous command, you'll get to your server with this URL: http://localhost:8380/?MAP=/etc/qgisserver/project.qgz&SERVICE=WMS&REQUEST=GetCapabilities

Apache Tunings

You can use the following variables (-e option in docker run):

• QGIS_CATCH_SEGV: Set to 1 if you want stacktraces in the logs in case of segmentation faults by running tail -f /var/log/qgis.log in the container
• FCGID_MAX_REQUESTS_PER_PROCESS: The number of requests a QGIS server will serve before being restarted by apache
• FCGID_MIN_PROCESSES: The minimum number of fcgi processes to keep (defaults to 1)
• FCGID_MAX_PROCESSES: The maximum number of fcgi processes to keep (defaults to 5)
• FCGID_IO_TIMEOUT: This is the maximum period of time the module will wait while trying to read from or write to a FastCGI application (default is 40)
• FCGID_BUSY_TIMEOUT: The maximum time limit for request handling (defaults to 300)
• FCGID_IDLE_TIMEOUT: Application processes which have not handled a request for this period of time will be terminated (defaults to 300)
• FILTER_ENV: Filter the environment variables with e.g.: | grep -vi _SERVICE_ | grep -vi _TCP | grep -vi _UDP | grep -vi _PORT to remove the default Kubernetes environment variables (default in an empty string)
• GET_ENV: alternative to FILTER_ENV, a command that return the environment variables (defaults to env)

See also QGIS server documentation

Fonts present in the /etc/qgisserver/fonts directory will be installed and thus usable by QGIS.

Lighttpd

You can also use lighttpd as the web server.

The main benefit of that is to have only one running process per container, that's useful especially on Kubernetes.

For that you need tow containers, one for the MapServer and spawn-fcgi, and one for lighttpd.

The environment variable needed by mapserver should be on the spawn-fcgi container.

The MapServer logs will be available on the 'lighttpd' container.

Used environment variables:

• LIGHTTPD_CONF: The lighttpd configuration file (defaults to /etc/lighttpd/lighttpd.conf)
• LIGHTTPD_PORT: The port lighttpd will listen on (defaults to 8080)
• LIGHTTPD_FASTCGI_HOST: The host of the FastCGI server (spawn-fcgi, defaults to spawn-fcgi)
• LIGHTTPD_FASTCGI_PORT: The port of the FastCGI server (spawn-fcgi, defaults to 3000)
• LIGHTTPD_FASTCGI_SOCKET: The socket of the FastCGI server (defaults to '')
• LIGHTTPD_ACCESSLOG_FORMAT: The format of the access log (defaults to "%h %V %u %t \"%r\" %>s %b")

Get a stack trace in case of segfault

To get a good stack trace you should use the -debug image.

Run the querty that makes the server crash, then:

Open a bash as root on the container with something like: docker compose exec --user=root qgisserver bash, then:

CORE_FILENAME=$(ls -tr1 /tmp/|grep core|tail -n 1)
gdb /usr/local/bin/qgis_mapserv.fcgi /tmp/${CORE_FILENAME}

The command bt will give you the stack trace.

Running the client

If you want to edit a project file, you can run the client from a Linux machine with the following command:

docker run --rm -ti --env=DISPLAY=unix${DISPLAY} --volume=/tmp/.X11-unix:/tmp/.X11-unix --volume=${HOME}:${HOME} camptocamp/qgis-server:master-desktop

Changelog

QGIS 3.22

We removed the default values for the following environment variables to better fit with the QGIS documentation:

• QGIS_SERVER_LOG_LEVEL, was 0
• QGIS_PROJECT_FILE, was /etc/qgisserver/project.qgs
• MAX_CACHE_LAYERS, was ""
• QGIS_AUTH_DB_DIR_PATH, was /etc/qgisserver/
• PGSERVICEFILE, was /etc/qgisserver/pg_service.conf

Contributing

Install the pre-commit hooks:

pip install pre-commit
pre-commit install --allow-missing-config