QGIS embbeded WMS/WFS/WCS asynchronous scalable server.
This is a asynchronous HTTP Qgis server written in python on top of the tornado framework and the 0MQ messaging framework for distributing requests workers.
It is based on the new Qgis 3 server API for efficiently passing requests/responses using 0MQ messaging framework to workers.
The server may be run as a self-contained single service or as a proxy server with an arbitrary number of workers running remotely or locally. Independant workers connect automatically to the front-end proxy with no need of special configuration on the proxy side. Thus, this is ideal for auto-scaling configuration for use with container orchestrator as Rancher, Swarm or Kubernetes.
The server is aimed at solving some real situations encountered in production environment: zero conf scalability, handle long-running request situation, auto restart...
- Multiples workers
- Fair queuing request dispatching
- Timeout for long running/stalled requests
- Full support of qgis server plugins
- Auto-restart trigger for workers
- Support streamed/chunked responses
- SSL support
- OS: Unix/Posix variants (Linux or OSX) (Windows not officialy supported)
- Python >= 3.5
- QGIS > 3.0 installed
- Some python knowledge about python virtualenv and package installation.
- libzmq >= 4.0.1 and pyzmq >= 17
Install in development mode
pip install -e .
Version X.Y.Z From python package archive
pip install py-qgis-server-X.Y.Z.tar.gz
Running the server
The server does not run as a daemon by itself, there is several way to run a command as a daemon.
- Use Supervisor http://supervisord.org/. Will gives you full control over logs and server status notifications.
- Use the
- Use systemd
Running the server
usage: qgisserver [-h] [-d] [-c [PATH]] [--version] [-p PORT] [-b IP] [-w NUM] [-j NUM] [-u SETUID] [--rootdir PATH] [--proxy] [--timeout SECONDS] qgis/HTTP/0MQ scalable server optional arguments: -h, --help show this help message and exit -d, --debug debug mode -c [PATH], --config [PATH] Configuration file --version Return version number and exit -p PORT, --port PORT http port -b IP, --bind IP Interface to bind to -w NUM, --workers NUM Num workers -j NUM, --jobs NUM Num server instances -u SETUID, --setuid SETUID uid to switch to --rootdir PATH Path to qgis projects --proxy Run only as proxy --timeout SECONDS Set client timeout in seconds
By default the command will run server instances with workers and use unix sockets to communicate. This can be used to run the server as a single command.
Running proxy and workers separately
--proxy option the server will act as a proxy server to talk to independant qgis workers.
Qgis workers can be run using the command:
qgisserver-worker --host=localhost --rootdir=path/to/projects
Requests to OWS services
The OWS requests use the following format:
Using with lizmap
In order to use the server with lizmap, you must set the following configuration
[services] wmsServerURL="http://my.domain:<port>/ows/" ... ; Use relative path relativeWMSPath=true
The configuration can be done either as configuration .ini file in or as environment variables.
The precedences of the gonfiguration parameters is the following (from lowest to highest)
- Defaults values
- Environment variables
- Config file
- Command line options
Here is a sample config file with the default values and the corresponding env variables
[server] port=8080 # QGSRV_SERVER_HTTP_PORT interfaces=0.0.0.0 # QGSRV_SERVER_INTERFACES workers=2 # QGSRV_SERVER_WORKERS timeout=20 # QGSRV_SERVER_TIMEOUT pluginpath= # QGSRV_SERVER_PLUGINPATH [logging] level=DEBUG # QGSRV_LOGGING_LEVEL [cache] size=10 # QGSRV_CACHE_SIZE rootdir= # QGSRV_CACHE_ROOTDIR [zmq] identity=OWS-SERVER # QGSRV_ZMQ_IDENTITY bindaddr=tcp://*:18080 # QGSRV_ZMQ_INADDR maxqueue=1000 # QGSRV_ZMQ_MAXQUEUE timeout=15000 # QGSRV_ZMQ_TIMEOUT
Have a look to
config.py for all configuration variables.
By default, the server log on stdout/stderr and you have to configure redirection and log rotation on your infrastructure environment