Opposite the manual deployment process is the automatic deployment process whereby our platform takes care of deploying game servers and utilities (side cars) for you. This is done according to the Deployment Configuration you have setup for a Fleet.
For automatic deployment to be possible, you must first create the following elements:
- a Deployment Environment
- a Fleet with:
Fleet operational status
- Automatic deployment
- Automatic scaling
These two modes are explained in the following chapters:
In this mode, the amount of instances defined by DeploymentProfile.minimumCapacity will be (statically) deployed, per region. So if you want to deploy a certain, static amount of application instances, you can use this option.
You can adjust the number of deployed instances at any time by changing the DeploymentProfile.minimumCapacity value. The system will then create or destroy the amount of instances necessary to match the new value. Note that minimumCapacity can be overridden per region as well.
For more information, refer to the Minimum capacity deployments chapter.
In this mode, at least the amount of instances defined by DeploymentProfile.minimumCapacity will be deployed, per region, as described in the previous paragraph. When additional capacity is required due to game servers becoming allocated or occupied, more game servers will automatically be deployed, first on your bare metal servers, or when these have been saturated, on cloud VMs as defined in the DeploymentContainerLocations of each DeploymentRegion.
The same is valid for downscaling. If there are too many (empty) instances, the platform will remove any unused instances (on a per region basis).
For a full explanation on how automatic scaling works, please refer to the Automatic Scaling chapter.
Minimum capacity deployments
When you enable automatic deployment or automatic scaling on a Fleet, our scaling mechanism will start by deploying the number of Application Instances defined by the DeploymentProfile.minimumCapacity value. The number of deployed instances will never drop below this value, even if there are no players at all. The minimumCapacity value as such can be used in preparation for a new release, to ensure you have enough game servers to begin with upon release of your game. Coincidentally, minimumCapacity could also be used by you to manually scale game servers, by changing this value whenever you need more or less game servers.
If the DeploymentProfile.minimumCapacity value exceeds your bare metal capacity, the scaler will continue to deploy in the cloud, provided cloud locations have been defined for the relevant regions.
The DeploymentProfile.minimumCapacity value defines minimumCapacity for each individual region. You can override this value per region in case you would like more instances deployed in a region where you expect more players. But keep in mind, minimumCapacity is mainly meant for the initial deployment of instances. Our scaling mechanism will of course start scaling instances if more players are needed [in a certain region].
The bufferValue only applies to the automatic scaling mode of a Fleet.
Besides the minimumCapacity setting, there is the bufferValue setting which determines the size of the pool of ApplicationInstances that must be available at all times. The purpose of this buffer is to catch sudden increases in the number of players and as such, to attempt (to ensure) there are always sufficient free game servers for your clients to join. In other words, the bufferValue is there to compensate for the time it takes to deploy game servers. The longer that takes (related to the size of your build files), the larger the buffer should be. The larger the buffer, the less risc there is of running out of available game servers. It is therefore important not to set this value too low.