Using MoinMoin with FastCgi
Contents
FastCGI is a method which enables apache to communicate with long-running scripts. This has the advantage that the script is only started and initialized one time, and that data could be cached in memory from request to request, enhancing the performance of the CGI application.
Deploying on Apache
To deploy MoinMoin using FastCGI you need an apache with mod_fastcgi. Please refer to the documentation of mod_fastcgi how to build it.
Follow the basic apache installation for your operating system as described in other parts of the MoinMoin installation documentation. This is HelpOnInstalling/ApacheOnLinux in most cases. In places where the documentation refers to moin.cgi you use moin.fcg instead. Be sure that .fcg is handled by the FastCGI module (AddHandler fastcgi-script .fcg in your apache config).
Normally apache will start CGI scripts with its own user and group, or with the user and group of the VirtualHost if you are using the suexec wrapper. To enable this with FastCGI you need to use FastCgiWrapper On in your apache config (see FastCGI Documentation).
Be sure to restart your apache after you changed py files (i.e. the config) for a running FastCGI server, or you won't see any changes!
Running as an external application and/or on Windows
MoinMoin can be run as an external application that answers FastCGI request via a local TCP/IP socket. This works on Windows as well. All you need to do (after having installed mod_fastcgi and a working MoinMoin instance) is this:
- Select a port number for the internal communication. It should be larger than 1023. For this example, we chose 8888 (just for fun).
Add the following lines to your httpd.conf:
Alias /moin.fcg "/your/path/to/moin.fcg" FastCgiExternalServer "/your/path/to/moin.fcg" -host localhost:8888
Edit moin.fcg. Replace
fcg = thfcgi.THFCGI(handle_request)
with
fcg = thfcgi.THFCGI(handle_request, 0, 8888)
Start the file moin.fcg manually like a Python script:
python moin.fcg
- Start Apache.
Fallback to CGI if FastCGI is not available
Install and test MoinMoin according to HelpOnInstalling/ApacheOnLinux. Then make and test the changes to run mod_fastcgi. If you are satisfied, you can add the following block to your apache config:
<IfModule !mod_fastcgi.c> ScriptAlias /mywiki "/your/path/to/moin.cgi" </IfModule> <IfModule mod_fastcgi.c> AddHandler fastcgi-script .fcg ScriptAlias /mywiki "/your/path/to/moin.fcg" </IfModule>
Now Apache will use mod_fastcgi if available and otherwise use the slow cgi script.
Deploying on lighttpd
The best option is to run moin as external application. In this case, you don't need to restart the web server when you want to restart moin.
Configuration
fastcgi.server = ( "/mywiki" => ( "localhost" => ( "host" => "127.0.0.1", "port" => 8888, "check-local" => "disable", "broken-scriptfilename" => "enable", ) ) )
You will see a lot of "2005-12-02 16:30:11: (mod_fastcgi.c.2373) FastCGI-stderr:" errors in
lighttpd's error log, the reason is not clear.
Starting MoinMoin instance
To start a MoinMoin instance, either follow the changes in "Running as an external application and/or on Windows" above, or run moin.fcg using spawn-fcgi (installed with lighttpd).
cd /www/org.mywiki/bin sudo -u www /usr/local/bin/spawn-fcgi -f ./moin.fcg -P moin.pid
To stop the instance later:
sudo kill `cat moin.pid`
Running spawn-fcgi as root and using -u www -g www fail because of Python ImportError, the reason is not clear. sudo -u www seems to have to same effect.
Multiple moin processes
If you start multiple moin instances on different ports, lighttpd balance the load automatically between them, using all the cpus.
Use this configuration:
fastcgi.server = ( "/mywiki" => ( "localhost" => ( "host" => "127.0.0.1", "port" => 1080, "check-local" => "disable", "broken-scriptfilename" => "enable", ), ( "host" => "127.0.0.1", "port" => 1081, "check-local" => "disable", "broken-scriptfilename" => "enable", ), ) )
MoinMoin Startup script
Here is a startup script for Mac OS X, using DarwinPortsStartup system. It is probably useful for other unix like platforms.
#!/bin/sh # Start and stop multiple moin fast cgi instances runnings on PORTS NAME="moin" DIR="/www/org.mywiki/bin" FCGIAPP="./moin.fcg" PREFIX="/usr/local" # List of ports to start moin instances on, separated with whitesapce # Keep in sync with fastcgi.server in lighttpd.conf PORTS="1080 1081" start_on_port () { # Start moin instance on port, leaving pid file port=$1 cd "${DIR}" && sudo -u www "${PREFIX}/bin/spawn-fcgi" \ -f "${FCGIAPP}" \ -p $port \ -P "${NAME}-${port}.pid" \ > /dev/null } kill_on_port () { # Try to kill process using pid in pid file, then remove the pid file pidFile="${DIR}/${NAME}-$1.pid" kill `cat "$pidFile"` && rm -f "$pidFile" > /dev/null } start () { for port in $PORTS; do start_on_port $port; done } stop () { for port in $PORTS; do kill_on_port $port; done } case "$1" in start) # XXX starting twice will break pid files (bug in spawn-fcgi) start && echo -n " $NAME" ;; stop) stop && echo -n " $NAME" ;; restart) stop start && echo -n " $NAME" ;; *) echo "Usage: `basename $0` {start|stop|restart}" >&2 ;; esac exit 0
Adding MoinMoin startup script on Mac OS X
With this script, moin instances will be started automatically on startup.
Install drarwinports
Install DarwinPortsStartup package:
sudo port install DarwinPortsStartup
Copy moin.sh into /opt/local/etc/rc.d:
sudo cp moin.sh /opt/local/etc/rc.d