Deploy Fano application with Apache load balancer module

Apache provides reverse proxy load balancer to distribute load to one or more application instances. When one application is unable to handle request, load balancer distributes load to other application instance thus improving performance, availability and scalability.

Requirement

Setting up mod_proxy_balancer

In Apache 2.4, mod_proxy_balancer is installed but not enabled. Make sure that mod_proxy_balancer is enabled. For example, in Debian-based, run

$ sudo a2enmod proxy_balancer

Also you need to enable one or more scheduler algorithm module. For example,

$ sudo a2enmod lbmethod_byrequests

Deploy Fano Application with load balancer with Fano CLI

Apart from task for scaffolding web application, Fano CLI also provides --deploy-lb-scgi=[domain name], --deploy-lb-fcgi=[domain name], --deploy-lb-uwsgi=[domain name] or --deploy-lb-http=[domain name] to help setup load balancer during development for SCGI, FastCGI, uwsgi or http web application, respectively.

After you create SCGI, FastCGI, uwsgi or http project with --project-scgi, --project-fcgi, --project-uwsgi or --project-mhd, for example

$ fanocli --project-scgi=myapp

From inside Fano SCGI web application project directory, run

$ sudo fanocli --deploy-lb-scgi=myapp.fano

Command above, will create virtual host for Apache web server that utilize mod_proxy_balancer module, enabled virtual host configuration, reload Apache web server configuration and add entry to myapp.fano domain in /etc/hosts.

Replace with --deploy-lb-fcgi, --deploy-lb-uwsgi or --deploy-lb-http for setting up FastCGI, uwsgi or http web application respectively.

Skip adding domain name entry in /etc/hosts

By default --deploy-* parameter will cause domain name entry is added in /etc/hosts file. You may want to setup domain name with DNS server manually or you do not want to mess up with /etc/hosts file. You can avoid it by adding --skip-etc-hosts parameter.

$ sudo fanocli --deploy-lb-scgi=myapp.fano --skip-etc-hosts

Generate virtual host config to standard output

If you want to generate virtual host configuration without actually modifying web server configuration, you can use --stdout command line option. This option will generate virtual host configuration and print it to standard output. It is useful if you want to deploy configuration manually.

Because it will not change any web server configuration, you do not need to run it with root privilege. So following code is suffice.

$ fanocli --deploy-lb-fcgi=myapp.me --stdout

Deploy Fano Application with load balancer manually

Skip this section if you deploy using Fano CLI.

If you prefer setting up virtual host manually, create new file in /etc/httpd/conf.d or /etc/apache/sites-available or /usr/local/etc/apache24 directory for Fedora-based, Debian-based Linux or FreeBSD respectively.

Add, for example, following code,

<VirtualHost *:80>
    ServerName myapp.fano
    DocumentRoot /path/to/my/app/doc/root

    ErrorLog /var/nginx/log/myapp.fano-error.log
    CustomLog /var/nginx/log/myapp.fano-access.log combined

    <Directory /path/to/my/app/doc/root>
        Options -MultiViews -FollowSymlinks +SymlinksIfOwnerMatch +ExecCGI
        AllowOverride FileInfo Indexes
        Require all granted
    </Directory>

    <Proxy balancer://myapp.fano>
        BalancerMember scgi://127.0.0.1:20477
        BalancerMember scgi://127.0.0.1:20478
        ProxySet lbmethod=by_requests
    </Proxy>

    ProxyRequests Off
    ProxyPassMatch "/css|js|images|img|plugins|bower_components(.*)" !
    ProxyPassMatch ^/(.*)$ "balancer://myapp.fano"
</VirtualHost>

Replace BalancerMember url according to protocol, host and port of the application. For example, replace with fcgi://127.0.0.1:20477, uwsgi://127.0.0.1:20477 or http://127.0.0.1:20477 for FastCGI, uwsgi or http protocol.

Running multiple applications with load balancer

Build the application and then run two applications at once with consecutive listening ports.

$ ./build.sh
$ ./bin/app.cgi --port=20477 & ./bin/app.cgi --port=20478 &

& is required to make sure that both applications are running in parallel.

To stop all applications,

$ pkill app.cgi

Access application from browser

Open http://myapp.fano you should see main controller is invoked. There is no visual indication compare to application running without load balancer.

Indication that your application is running with mod_proxy_balancer is availability of new environment variables, for examples BALANCER_NAME, BALANCER_WORKER_NAME and BALANCER_WORKER_ROUTE.

Try to access route that does not exist, so that environment variables are printed on the browser. It will show that load balancer distributes the load by observing value of BALANCER_WORKER_NAME variable.

Set balancer member

By default, if parameter --members not set, it is assumed that you will use two application instances, running on 127.0.0.1:20477 and 127.0.0.1:20478 respectively.

--members parameter allows set multiple balancer members separated by coma.

$ sudo fanocli --deploy-lb-scgi=myapp.fano --members=127.0.0.1:20000,localhost:20001

Run applications at once with consecutive listening ports.

$ ./bin/app.cgi --host=127.0.0.1 --port=20000 & ./bin/app.cgi --host=localhost --port=20001 &

Developer is responsible to make sure that no other application is using those listening ports.

Change load balancing scheduler algorithm

By default, if parameter --lbmethod is not set, byrequests is assumed, which will use mod_lbmethod_byrequests module to distribute requests. Please refer to mod_proxy_balancer documentation for available algorithms.

For example, to use mod_lbmethod_bybusyness module,

$ sudo fanocli --deploy-lb-scgi=myapp.fano --lbmethod=bybusyness

Explore more

Documentation