Cluster/Infra Creation Epic

[andrebriggs]

Spikes

• Node pools w/availability zones

Terraform work

• Create a new Terraform environment template with the following:
  • Single AKS
  • Key Vault
  • Azure Monitor
  • Node pools w/availability zones
  • ACR

CLI Work

• spk infra validate (and also spk init)
  • Validate prerequisites (e.g.terraform, az cli, helm, git)
  • Validate that user is logged in to Azure
  • Validate environment variables (e.g. ARM_SUBSCRIPTION_ID, ARM_CLIENT_ID, ARM_CLIENT_SECRET, and ARM_TENANT_ID) are set and not null
  • Create soft/hard checks for validation
  • Create a state file for initialization (in ~/.spk/spk-config.yaml)
    • Write versions of executables to spk-config.yaml
• spk infra scaffold --name <env name> --source <bedrock git url> --template <env template>
  • Validate that spk infra validate has run (check for existence of ~./spk/spk-config.yaml infra block)
  • Clone official/custom Bedrock repo provided by the --source argument (potentially to a tmp directory)
    • Support for cloning private repos
  • Create a folder for new environment
  • Create a definition.yaml file with parameters like environment name, template name, and variables from a variables.tf
  • Supports remote backend state as part of definition.yaml (if it exists)
• spk infra generate --name <env name>
  • Clone the source repo to fetch Terraform template (potentially to a tmp directory)
    • Support for cloning private repos
  • Create a "generated" directory
  • Copy Terraform template to the "generated" directory
  • Modify the Terraform template to use information provided by the definition.yaml file
  • Iterate through cluster directories (e.g. west, east), overwriting specific values in the cluster terraform.tfvars with values that are unique to the cluster definition.yaml
  • Backend State Support

Considerations:

• What happens if I move from machine A to machine B, where does the spk state live?
• What happens if the remote bedrock environment changes?
• What happens if I want to update a particular variable/etc?

[andrebriggs]

Create a state file for initialization (in ~/.spk)

Have we checked with the other groups about their need for cached data?

Clone official/custom Bedrock repo using -p argument

Are we requiring that a custom repo follows the same Bedrock directory structure?

Generates an infrastructure "HLD" (e.g. infra.json/yaml)

What exactly will be in here?

CC @yradsmikham @NathanielRose

[NathanielRose]

@andrebriggs
Have we checked with the other groups about their need for cached data?

• No we have not, will definitely surface if their is a need for other teams as well. Our thinking was targeting spk infra init then migrate to a spk init that calls other command initializations such as spk project init + spk deployment init

Are we requiring that a custom repo follows the same Bedrock directory structure?

• We are at the moment. This can later be migrated to sparse cloning a repo based on a bedrock environment path to be passed to our spk infra create

What exactly will be in here?

• Our Brainstorm YAML:

---
Bedrock_Repo: True
Deployment_Name: Azure-ECK-Simple
Terraform_Version: 0.12.3
Bedrock_Source: (URL/LocalPath)
Data_Resources: [resource_group_name]
TFvars: 
{
resource_group_name      = "resource-group-name"
cluster_name             = "cluster-name"
agent_vm_count           = "3"
dns_prefix               = "dns-prefix"
service_principal_id     = "client-id"
service_principal_secret = "client-secret"
ssh_public_key           = "public-key"
gitops_ssh_url           = "[email protected]:timfpark/fabrikate-cloud-native-manifests.git"
gitops_ssh_key           = "<path to private gitops repo key>"
vnet_name                = "<vnet name>"
}

CC @yradsmikham @NathanielRose

[yradsmikham]

@andrebriggs

After reaching out to other teams, it appears that there could be a need to have some sort of "cached data". Service Introspection may need store keys, and Service Creation potentially may need to store spk logs or PATs. Regardless, there is a need (currently) for infra.

[NathanielRose]

@andrebriggs @yradsmikham Can this be closed?

[yradsmikham]

@andrebriggs @yradsmikham Can this be closed?

Yes, it looks complete.