A simple docker container that runs QGIS Server
This image uses the QGIS Desktop docker image as its base.
Note You should revise the security etc. of this implementation before using in a production environment.
Please see the canonical documentation for QGIS Server if you need more general info on how QGIS Server works.
To use the image, either pull the latest trusted build from https://registry.hub.docker.com/u/kartoza/qgis-server/ by doing this:
docker pull kartoza/qgis-server:LTR
Note that the LTR build will always track the most recent QGIS Long Term Release build.
We use versioned images and highly recommend that you track a specific version for your orchestrated services since things may break between versions.
In this repository you will find a subdirectory for each QGIS version supported. Each directory contains a self contained docker project and we will maintain all the versioned builds from these containers.
You can build the image yourself like this:
git clone git://github.com/kartoza/docker-qgis-server
cd docker-qgis-server/2.18
docker build -t kartoza/qgis-server .
To run a container do:
docker run --name "qgis-server" -p 8080:80 -d -t kartoza/qgis-server
http://localhost:8080 should show QGIS Server, with an error because it
couldn't find a QGIS project to parse.
NOTE: Again we would like to recommend you run a tagged version rather than just the latest since your configuration may break if we change something.
If you have a single QGIS project on your host to publish, you can use this command:
docker run --name "qgis-server" -v /path/to/your/project_folder/:/project -p 8080:80 -d -t kartoza/qgis-server:LTR
# Use the default test project provided by this repository in the project folder:
docker run --name "qgis-server" -v $(PWD)/project/:/project -p 8080:80 -d -t kartoza/qgis-server:LTR
The container will try to load the project located in /project called project.qgs.
You can override this setting using environment variable, see below about apache configuration.
http://localhost:8080 should show QGIS Server with your correct project.
http://localhost:8080/?SERVICE=WMS&VERSION=1.3.0&REQUEST=GetCapabilities should give the GetCapabilities.
However, if you have many QGIS projects, you need to give the path to the project in the URL:
We mount our project folder in /gis in this example to show that the name is not important.
Be careful: you need to set the environment variable QGIS_PROJECT_FILE to none,
as your project won't be found under /project/project.qgs.
docker run --name "qgis-server" -e QGIS_PROJECT_FILE='' -v /path/to/your/project_folder:/gis -p 8080:80 -d -t kartoza/qgis-server:LTR
# Use the default test project provided by this repository in the project folder:
docker run --name "qgis-server" -e QGIS_PROJECT_FILE='' -v $(PWD)/project/:/gis -p 8080:80 -d -t kartoza/qgis-server:LTR
http://localhost:8080/?SERVICE=WMS&VERSION=1.3.0&REQUEST=GetCapabilities&MAP=/gis/project.qgs should give the GetCapabilities.
Here is a contrived example showing how you can run QGIS Server
from in a docker container using docker-compose. Example docker-compose follows:
db:
image: kartoza/postgis:9.6-2.4
environment:
- USERNAME=docker
- PASS=docker
qgisserver:
image: kartoza/qgis-server:LTR
hostname: qgis-server
volumes:
# Wherever you want to mount your data from
- ./project:/project
links:
- db:db
ports:
- "8080:80"
To run the example do:
docker-compose up
You should see QGIS server start up. For more detailed approaches to using and building the QGIS Server container, see below.
http://localhost:8080 should show QGIS Server.
Note: The database in the above example is stateless (it will be deleted when
running docker-compose rm). If you want to connect to the PG database from docker
use the following info:
- host: db
- database: gis
- user: docker
- password: docker
Please see the provided docker-compose example which includes examples for both QGIS Server 2/LTR and QGIS Server 3.
Apache will make of the following environment variables. You can tweak these by replacing these options in your docker-compose.yml or docker run command.
APACHE_CONFDIR /etc/apache2
APACHE_ENVVARS $APACHE_CONFDIR/envvars
APACHE_RUN_USER www-data
APACHE_RUN_GROUP www-data
APACHE_RUN_DIR /var/run/apache2
APACHE_PID_FILE $APACHE_RUN_DIR/apache2.pid
APACHE_LOCK_DIR /var/lock/apache2
APACHE_LOG_DIR /var/log/apache2
LANG C
The following variables (with defaults shown) are QGIS specific options you can tweak by replacing these options in your docker-compose.yml or docker run command.
QGIS_DEBUG 5
QGIS_LOG_FILE /proc/self/fd/1
QGIS_SERVER_LOG_FILE /proc/self/fd/1
QGIS_SERVER_LOG_LEVEL 5
PGSERVICEFILE /project/pg_service.conf
QGIS_PROJECT_FILE /project/project.qgs
QGIS_PLUGINPATH /opt/qgis-server/plugins
Note: please consult the QGIS Server documentation for details on the options you can pass to QGIS Server.
Probably you will want to mount the /project folder with local volume that contains some QGIS projects. As you can see above, if no project file is specified, QGIS will try to serve up /project/project.qgs by default so if you are looking for an easy to share WMS/WFS url, simply call your project file project.qgs and mount it in the /project directory.
./build.sh; docker kill server; docker rm server;
docker run --name="qgis-server" \
-d -p 8080:80 \
kartoza/qgis-server:LTR
docker logs qgis-server
Replace <path_to_local_qgis_project_folder> with an absolute path on your
filesystem. That folder should contain the .qgs project files you want to
publish and all the data should be relative to the project files and within the
mounted volume. See https://github.com/kartoza/maps.kartoza.com for an example
of a project layout that we use to power http://maps.kartoza.com
An example project folder is provided here for convenience (and we use it for validation testing).
Accessing the services:
Simply entering the URL of the docker container with its port number will respond with a valid OGC response if you use this url in QGIS 'add WMS layer' dialog.
http://localhost:8080 should show QGIS Server.
The container uses Apache as web frontend by default (out of the box). We provided default configuration for running QGIS Server as a standalone container immediately. For production usage, or different use case, you might want to extend or change the conf file.
You can do so by providing the conf file into the container,
either by extending the base image and committing a conf file in the image, or just volume mount the conf file via docker-compose.
Either way, specify your new conf file location in an environment variable called QGIS_CONF_FILE.
The entrypoint script then will locate your conf file and activates it by putting the link to /etc/apache2/conf-enabled folder.
If you don't need this kind of behaviour when extending the image, simply dump all of your config file into /etc/apache2/conf-enabled directory in the container.
Then turn off the entrypoint script. Regular Docker common sense applies.
As our special use case, we also provide built in conf file in /etc/apache2/conf-available/qgis-single-worker.
This is intended to make Apache work in single thread worker only, because the scaling is handled by container management such as Rancher/Kubernetes.
This conf file might not be optimized for general use case (it will be slow if deployed as one container only).
QGIS Server supports server-side plugins written in python. The following plugins are shipped by default with this image:
- On-the-fly project creation: This new service parameter will create a QGIS project using one or many layers existing on the file system. See the project README for details.
Tim Sutton (tim@kartoza.com) - May 2014
Acknowledgement:
During the Girona QGIS hackfest in 2016, Patrick Valsecchi did an almost complete re-write of this image recipe which I have heavily based this and the recipe in docker-qgis-desktop on. Thanks Patrick!