High Availability Cluster with Pound

It is very easy to scale web2py using a cluster:

• Install web2py on all the worker machines in your cluster (192.168.0.1:80, 192.168.0.2:80, etc. etc.). We recommend using this install procedure
• Setup one machine as a load balancer.

Here is how to setup and start the load balancer:

wget http://www.apsis.ch/pound/Pound-2.4.5.tgz
tar zxvf Pound-2.4.5.tgz
cd Pound-2.4.5
./configure
make
sudo make install
echo '
          ListenHTTP
              Address 0.0.0.0
              Port    80
          End
          Session
              Type    IP
              TTL     3600
          End
          Service
              BackEnd
                  Address 192.168.0.1
                  Port    80
              End
              BackEnd
                  Address 192.168.0.2
                  Port    80
              End
              # etc etc
          End
  ' > pound.cfg
/usr/local/sbin/pound -f pound.cfg

And that is all. You will be able to access the web2py cluster via port 80 of the load balancer.

Caveats

Pound provides sticky sessions (the same client IP is mapped to the same worker node) so you do not have to worry about sessions.

You are running different web2py instances. Make sure they connect to the same database.

You may have to worry about uploaded files. So make sure you app stores them in the database and/or make your app uploads folder a shared folder. It is a good to idea to make it a backed-up NFS mounted folder.

Each instance will store error tickets locally. You can write cross mount the error folder, you can write your own script to consolidate error tickets on one machine or you can run this provided script to move error tickets into the database:

  web2py/script/tickets2db.py