This will install and setup WordPress optimized to run in the cloud. This install, in particular, will
place Ngnix and php-fpm configured to scale horizontally with Nginx's reverse proxy
WordPress is a powerful blogging platform written in PHP. This charm aims to deploy WordPress in a fashion that will allow anyone to scale and grow out
a single installation.
This charm is available in the Juju Charm Store, to deploy you'll need at a minimum: a cloud environment, a working Juju installation,
and a successful bootstrap. Please refer to the Juju Getting Started documentation before continuing.
Once bootstrapped, deploy the MySQL charm then this WordPress charm:
juju deploy mysql
juju deploy wordpress
Add a relation between the two of them
juju add-relation wordpress mysql
Expose the WordPress installation
juju expose wordpress
This WordPress charm comes with several tuning levels designed to encompass the different styles in which this charm will be used.
A use case for each tuning style is outlined below:
The Bare configuration option is meant for those who wish to run the stock WordPress setup with no caching, no manipulation of data,
and no additional scale out features enabled. This is ideal if you intend to install additional plugins to deal with coordinating
WordPress units or simply wish to test drive WordPress as it is out of the box. This will still create a load-balancer when an additional
unit is created, though everything else will be turned off (WordPress caching, APC OpCode caching, and NFS file sharing).
To run this WordPress charm under a bare tuning level execute the following:
juju set wordpress tuning=bare
When running in Single mode, this charm will make every attempt to provide a solid base for your WordPress install. By running in single
the following will be enabled: Nginx microcache, APC OpCode caching, WordPress caching module, and the ability to sync files via NFS.
While Single mode is designed to allow for scaling out, it's meant to only scale out for temporary relief; say in the event of a large
traffic in-flux. It's recommended for long running scaled out versions that optimized is used. The removal of the file share speeds up
the site and servers ensuring that the most efficient set up is provided.
To run this WordPress charm under a single tuning level execute the following:
juju set wordpress tuning=single
If you need to run WordPress on more than one instance constantly, or require scaling out and in on a regular basis, then Optimized is the
recommended configuration. When you run WordPress under an Optimized tuning level, the ability to install, edit, and upgrade themes and plugins
is disabled. By doing this the charm can drop the need for an NFS mount which is inefficient and serve everything from it's local disk.
Everything else provided in Single level is available. In order to install or modify plugins with this setup you'll need to edit and commit
them to a forked version of the charm in the files/wordpress/ directory.
To run this WordPress charm under an optimized tuning level execute the following:
juju set wordpress tuning=optimized
In order to allow for custom WordPress content within the Juju charm a separate configuration option exists for pointing to any Git or Bzr
repository. An example of a valid formed wp-content repository can be found on the Juju Tools Github page.
To set the wp-content directive to a git repository, use one of the following formats making sure to replace items like host, path, and repo with their
juju set wordpress wp-content=git@host:path/repo.git
juju set wordpress wp-content=http://host/path/repo.git
juju set wordpress wp-content=git://host/path/repo.git
If you wish to use a bzr repository, then apply one of the following schemes replacing items like host, username, path, and repo with their respective values:
For LaunchPad hosted repostiories:
juju set wordpress wp-content=lp:~username/path/repo
For other Bzr repositories:
juju set wordpress wp-content=bzr://host/path/repo
juju set wordpress wp-content=bzr+ssh://host/path/repo
Setting the wp-content option to an empty string ("") will result in no further updates being pulled from that repository; however, the last pull will remain
on the system and will not be removed.
This option will create a directory _debug at the root of each unit (http://unit-address/_debug). In this directory are two scripts: info.php (/_debug/info.php)
and apc.php (/_debug/apc.php). info.php is a simple phpinfo script that will outline exactly how the environment is configured. apc.php is the APC admin portal which
provides APC caching details in addition to several administrative functions like clearing the APC cache. This should never be set to "yes" in production as it exposes
detailed information about the environments and may provide a way for an intruder to DDoS the machine.
juju set wordpress debug=yes
to disable debugging:
juju set wordpress debug=no
The debugging is disabled by default.
By default the WordPress charm will install nginx and php5-fpm to serve pages. In the event you do not wish to use nginx - for whatever reason - you can switch to Apache2.
This will provide a near identical workflow as if you were using nginx with one key difference: memcached. In nginx, the cached pages are served from memcached prior to
hitting the php contents, this isn't possible with apache2. As such memcached support still works, since it falls back to the WordPress caching engine, but it's not as robust.
Otherwise, Apache2 will still perform balancing and everything else mentioned above. You can switch between engines at will with the following:
juju set wordpress engine=apache2
Then back to nginx:
juju set wordpress engine=nginx
Any other value will result in the default (nginx) being used.
At this time WordPress + Memcached don't work on HP Cloud's standard.xsmall. To get around this deploy the WordPress charm with the
charm to at least a standard.small, to do this:
juju deploy --constraints "instance-type=standard.small" wordpress
This only is a problem when attempting to relate memcached to WordPress, otherwise an xsmall is okay though it's really not the best
sized platform for running a stable WordPress install.
If you're in Single mode and you want to/need to scale out, but you've been upgrading, modifying, and installing plugins + themes like
a normal WordPress user on a normal install; you can still scale out but you'll need to deploy a shared-fs charm first. At the time of
this writing only the NFS charm will work, but as more shared-fs charms come out (gluster, ceph, etc) that provide a shared-fs/mount
interface those should all work as well. In this example we'll use NFS:
juju deploy nfs
juju add-relation nfs wordpress:nfs
By doing so, everything in the wp-contents directory is moved to this NFS mount and then shared to all future WordPress units. It's strongly
recommended that you first deploy the nfs mount, then scale WordPress out. Failure to do so may result in data loss. Once nfs is deployed,
running, and related you can scale out the WordPress unit using the following command:
juju add-unit wordpress
In the event you want more than one unit at a time (and do not wish to run the add-unit command multiple times) you can supply a -n number
of units to add, so to add three more units:
juju add-unit -n3 wordpress
Why not? We could ALL use more caching. Deploy a memcached server and relate it to your WordPress service to add memcache caching. This will
automagically install WP-FFPC (regardless of your tuning settings) and configure it to cache
rendered pages to the memcache server. In addition to this layer of caching, Nginx will pull directly from memcached bypassing PHP altogether.
You could theoretically then turn off php5-fpm on all of your servers and just have Nginx serve static content via memcached (though, you
wouldn't be able to access the admin panel or any uncached pages - it's just a potential scenario).
juju deploy memcached
juju add-relation memcached wordpress
This setup will also synchronize the flushing of cache across all WordPress nodes, making it ideal to avoid stale caches.
A small note, when using the Apache2 engine and memcache, all request will still be sent to WordPress via Apache where typical caching
procedures will take place and wp-ffpc will render the memcached page.
There is a "hack" that will allow you to deploy multiple full services to the same machine as the bootstrap node, this has nothing to do with
the charm, but it's something that comes up more than once. Use this, of course, at your own risk. At any time the Juju developers may smart
up and decide to remove this configuration option from the environments.yaml file. Prior to your first deployment you'll need to add the
following line to your Juju Environments file:
This will say "Everything that you deploy, will go on the bootstrap node". Make sure you plan to have a big enough bootstrap node to house
both your database and WordPress install. After you've bootstrap'd the environment, deploy the MySQL and WordPress charms like you normally
would. Instead of seeing three nodes you'll only see one, but both of your services will have been deployed. FROM THIS POINT you should
either remove or comment out the placement line in the environments file. This will prevent issues from occurring when you try to deploy
additional services or try to scale out existing services.