CloudML comes along with a shell component, which allows you to load, modify and enact CloudML models. The CloudML shell can be run in two modes: a batch mode and an interactive mode. The batch mode lets you run a single command, whereas the interactive mode lets you input commands in a step-wise fashion.
Here on the right, a CloudML shell is available in interactive mode.
Try to display information about the available commands.
CloudML> help
The CloudML shell also provides a set of commands, which enable interaction with the application running in the Cloud.
Before actually enacting a deployment, a model has to be loaded in the deployment engine.
Try to load de deployment model from a file.
Let's consider a file called SensAppAdmin.json.
You can explore the content of this file here
load deployment from SensAppAdmin.json
Once a deployment models has been loaded in the deployment engine, the deployment can effectively be deployed.
Try to Deploy the application.
In this example we are simulating a deployment.
Real deployment may take several minutes.
Deploy
Once a deployment has been loaded, informations about the various components that compose your model can be obtained from the shell. Both the list of component instances and types are available.
Try to retrieve the list of component instances.
Details on a specific instance or type
can be obtained using the view command
list instances
It is possible to scale out a VM instance using the scale out command. This command automatically update the model, thus generating the necessary dependencies. Then if needed it creates an image of the VM to be scaled, start a new VM using this image and finally restart all necessary software components installed hosted on this VM.
Try to scale the VM called sensapp-sl1.
Names are not uniques whilst IDs are!
IDs are given at deployment-time
scale out RegionOne/243472e4-3bfd-499c-8f5f-241946ca58eb
At any time you can save in JSON or XMI format the deployment model currently stored in the memory of the deployment engine.
Try to save the scaled model as a JSON file.
The serialization format is automatically
deduced based on the filename's extension.
store deployment to file.json