Scaling applications
The capability of a service to adjust its resource footprint to a level appropriate for fulfilling client demands placed upon it is known as scalability. Scaling vertically affects the resources of existing machines (memory, CPU, disk space) whereas scaling horizontally involves the number of application units available.
Units are not always synonymous with machines however. Multiple units can be placed onto a single machine (co-location) and still be considered horizontal scaling if sufficient resources are present on the machine.
This page will describe how rudimentary scaling works with Juju as well as mention some less common situations.
Scaling up
In the context of Juju, scaling up means increasing the number of application
units and always involves the add-unit
command.
Closely resembling scaling up is the addition of a machine devoid of a unit.
This is accomplished via the add-machine
command:
juju add-machine
Note: A machine provisioned via the add-machine
command that does not yet house an application unit will, by default, be used for any subsequent application deployment (via the deploy
command).
Scaling up behind a load balancer
In many cases simply adding more units will not make an application scale properly. A load balancer or proxy is often required.
Below is an example of deploying a load balanced MediaWiki application by placing the HAProxy application in front of it:
juju deploy mediawiki juju deploy mysql juju deploy haproxy juju add-relation mediawiki:db mysql juju add-relation mediawiki haproxy juju expose haproxy
When a relation is made between MediaWiki and HAProxy, the latter will be configured to load balance requests to the MediaWiki application. This means that client requests should go to the HAProxy instance. To get the proxy's IP address run the following:
juju status haproxy
You can now scale up the MediaWiki application behind the proxy as you see fit. To add five more units (with each running in its own machine):
juju add-unit -n 5 mediawiki
Scaling up using a charm with built-in scaling
Some charms have scaling built-in where scaling up really is as simple as adding more units.
An example of this is the WordPress charm:
juju deploy mysql juju deploy wordpress juju add-relation mysql wordpress juju expose wordpress
To scale up by adding an extra unit one can simply do:
juju add-unit wordpress
This will cause a new unit (and machine) to be spawned and configured to work alongside the existing one.
By default, add-unit
will add a single unit. To request multiple units the
-n
option is needed. For example, to further scale up our current application
by adding 100 units of MySQL one would run:
juju add-unit -n 100 mysql
Scaling up through co-location
Like the deploy
command, it is possible to co-locate multiple applications on
a single machine. This is done via the --to
option.
For example, to add a unit of MySQL to the machine with an ID of '23':
juju add-unit mysql --to 23
To add a unit of the same application to existing LXC container '3' residing on host machine '24':
juju add-unit mysql --to 24/lxc/3
Note: Not all applications will happily co-exist (usually due to conflicting configuration files). It is therefore generally safer to place units on dedicated machines or containers.
Here we add a unit of MySQL to a new LXC container on host machine 25:
juju add-unit mysql --to lxc:25
Scaling up by specifying new constraints
It is possible to scale out an application by adding a unit with different hardware requirements (constraints) than those set with the initial deployment. The default behaviour is for new units to use the same, if any, constraints.
This is done by indirectly creating a machine with a constraint and then adding the unit to it. For example, to add a unit with 16 GiB of memory to the MySQL application if the initial deployment was only, say, 4 GiB:
juju add-machine --constraints mem=16G juju machines juju add-unit mysql --to 3
Above, it is presumed that juju machines
informed us that the new machine was
assigned an ID of '3'.
Read the Using constraints page for details on constraints.
Scaling down
In the context of Juju, scaling down means decreasing the number of application
units and always involves the remove-unit
command.
Closely resembling scaling down is the direct removal of a machine. This is
therefore also covered here and is accomplished via the remove-machine
command.
To scale down the MediaWiki application by removing a specific unit:
juju remove-unit mediawiki/1
Note that if this is the only unit running on the underlying machine, the machine will also be removed.
A machine cannot be manually removed if any of the following is true:
- it houses a unit
- it is being used as the only controller
- it is hosting Juju-managed containers (KVM guests or LXD containers)
For example, to remove a machine with ID of '6':
juju remove-machine 6
For more information on removing applications and machines, see the Removing Juju objects page.