Deployment to a production server

As already described in the Installation section, no specific deployment procedures are for PyWPS when using flask-based server. But this formula is not intended to be used in a production environment. For production, Apache httpd or nginx servers are more advised. PyWPS is runs as a WSGI application on those servers. PyWPS relies on the Werkzeug library for this purpose.

Deploying an individual PyWPS instance

PyWPS should be installed in your computer (as per the Installation section). As a following step, you can now create several instances of your WPS server.

It is advisable for each PyWPS instance to have its own directory, where the WSGI file along with available processes should reside. Therefore create a new directory for the PyWPS instance:

$ sudo mkdir /path/to/pywps/

# create a directory for your processes too
$ sudo mkdir /path/to/pywps/processes

Note

In this configuration example it is assumed that there is only one instance of PyWPS on the server.

Each instance is represented by a single WSGI script (written in Python), which:

  1. Loads the configuration files
  2. Serves processes
  3. Takes care about maximum number of concurrent processes and similar

Creating a PyWPS WSGI instance

An example WSGI script is distributed along with PyWPS-Demo service, as described in the Installation section. The script is actually straightforward - in fact, it’s a just wrapper around the PyWPS server with a list of processes and configuration files passed as arguments. Here is an example of a PyWPS WSGI script:

$ $EDITOR /path/to/pywps/pywps.wsgi

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
#!/usr/bin/env python3

from pywps.app.Service import Service

# processes need to be installed in PYTHON_PATH
from processes.sleep import Sleep
from processes.ultimate_question import UltimateQuestion
from processes.centroids import Centroids
from processes.sayhello import SayHello
from processes.feature_count import FeatureCount
from processes.buffer import Buffer
from processes.area import Area

processes = [
    FeatureCount(),
    SayHello(),
    Centroids(),
    UltimateQuestion(),
    Sleep(),
    Buffer(),
    Area()
]

# Service accepts two parameters:
# 1 - list of process instances
# 2 - list of configuration files
application = Service(
    processes,
    ['/path/to/pywps/pywps.cfg']
)

Note

The WSGI script is assuming that there are already some processes at hand that can be directly included. Also it assumes, that the configuration file already exists - which is not the case yet.

The Configuration is described in next chapter (Configuration), as well as process creation and deployment (Processes).

Deployment on Apache2 httpd server

First, the WSGI module must be installed and enabled:

$ sudo apt-get install libapache2-mod-wsgi
$ sudo a2enmod wsgi

You then can edit your site configuration file (/etc/apache2/sites-enabled/yoursite.conf) and add the following:

# PyWPS-4
WSGIDaemonProcess pywps home=/path/to/pywps user=www-data group=www-data processes=2 threads=5
WSGIScriptAlias /pywps /path/to/pywps/pywps.wsgi process-group=pywps

<Directory /path/to/pywps/>
    WSGIScriptReloading On
    WSGIProcessGroup pywps
    WSGIApplicationGroup %{GLOBAL}
    Require all granted
</Directory>

Note

WSGIScriptAlias points to the pywps.wsgi script created before - it will be available under the url http://localhost/pywps

Note

Please make sure that the logs, workdir, and outputpath directories are writeable to the Apache user. The outputpath directory need also be accessible from the URL mentioned in outputurl configuration.

And of course restart the server:

$ sudo service apache2 restart

Deployment on nginx

Note

We are currently missing documentation about nginx. Please help documenting the deployment of PyWPS to nginx.

You should be able to deploy PyWPS on nginx as a standard WSGI application. The best documentation is probably to be found at Readthedocs.

Testing the deployment of a PyWPS instance

Note

For the purpose of this documentation, it is assumed that you’ve installed PyWPS using the localhost server domain name.

As stated, before, PyWPS should be available at http://localhost/pywps, we now can visit the url (or use wget):

# the --content-error parameter makes sure, error response is displayed
$ wget --content-error -O - "http://localhost/pywps"

The result should be an XML-encoded error message.

<?xml version="1.0" encoding="UTF-8"?>
<!-- PyWPS 4.0.0-alpha2 -->
<ows:ExceptionReport xmlns:ows="http://www.opengis.net/ows/1.1" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.opengis.net/ows/1.1 http://schemas.opengis.net/ows/1.1.0/owsExceptionReport.xsd" version="1.0.0">
    <ows:Exception exceptionCode="MissingParameterValue" locator="service" >
        <ows:ExceptionText>service</ows:ExceptionText>
    </ows:Exception>
</ows:ExceptionReport>

The server responded with the pywps.exceptions.MissingParameterValue exception, telling us that the parameter service was not set. This is compliant with the OGC WPS standard, since each request mast have at least the service and request parameters. We can say for now, that this PyWPS instance is properly deployed on the server, since it returns proper exception report.

We now have to configure the instance by editing the pywps.cfg file and adding some processes.