We will be creating a state directory and informing pulumi we want to use that local directory to manage the state, no need to use a cloud backend managed by pulumi or other providers (for the sake of testing).
We then start by initializing the resources. The provider which we loaded in the plugins section is also a resource that has properties (the main one now is just the mnemonic of TFChain).
Then, we create a scheduler `threefold:provider:Scheduler`, that does the planning for us. Instead of being too specific about node IDs, we just give it some generic information. For example, "I want to work against these data centers (farms)". As long as the necessary criteria are provided, the scheduler can be more specific in the planning and select the appropriate resources available on the TFGrid.
Now, that we created the scheduler, we can go ahead and create the network resource `grid:internal:Network`. Please note that the network depends on the scheduler's existence. If we remove it, the scheduler and the network will be created in parallel, that's why we have the `dependsOn` section. We then proceed to specify the network resource properties, e.g. the name, the description, which nodes to deploy our network on, the IP range of the network. In our case, we only choose one node.
To access information related to our deployment, we set the section **outputs**. This will display results that we can use, or reuse, while we develop our infrastructure further.
Now, we will check an [example](https://github.com/threefoldtech/pulumi-provider-grid/blob/development/examples/yaml/virtual_machine) on how to create a virtual machine.
We have a scheduler, and a network just like before. But now, we also have a deployment `threefold:provider:Deployment` object that can have one or more disks and virtual machines.
The deployment can be linked to a network using `network_name` and can have virtual machines in the `vms` section, and disks in the `disks` section. The disk can be linked and mounted in the VM if `disk_name` is used in the `mounts` section of the VM.
We also specify a couple of essential properties, like how many virtual cores, how much memory, what FList to use, and the environment variables in the `env_vars` section.
That's it! You can now execute `make run` to bring the infrastructure up.
We now see how to deploy a [Kubernetes cluster using Pulumi](https://github.com/threefoldtech/pulumi-provider-grid/blob/development/examples/yaml/kubernetes/Pulumi.yaml).
Now, we define the Kubernetes resource `grid:internal:Kubernetes` that has master and workers slice. You define almost everything like a normal VM except for the FLiist. Also note that the token is the `cluster token`. This will ensure that the workers and the master communicate properly.
The ThreeFold Pulumi repository also covers examples on [how to work with TFGrid gateways](https://github.com/threefoldtech/pulumi-provider-grid/blob/development/examples/yaml/gateway_name/Pulumi.yaml).
The basic idea is that you have a virtual machine workload on a specific IP, e.g. public IPv4, IPv6, or Planetary Network, and you want to access it using domains.
There are two versions to achieve this, a simple and a fully controlled version.
- Simple domain version:
- subdomain.gent01.dev.grid.tf
- This is a generous service from ThreeFold to reserve a subdomain on a set of defined gateway domains like **gent01.dev.grid.tf**.
- Fully controlled domain version:
- e.g. `mydomain.com` where you manage the domain with the name provider.
### Example of a Simple Domain Prefix
We present here the file for a simple domain prefix.
Here's an [example](https://github.com/threefoldtech/pulumi-provider-grid/blob/development/examples/yaml/gateway_fqdn/Pulumi.yaml) of a more complicated, but fully controlled domain.
