CloudVpc -- Quick Start v1.0.0

1Change into the configuration directory

cd cloudvpc/aws-asg-oci-pool
cp terraform.tfvars.example terraform.tfvars

terraform.tfvars.example lists every variable the configuration declares, with its default; the required ones are at the top. Keep the directory layout you received: the configuration refers to its modules by relative path.

2Set the values

Replace the contents of terraform.tfvars with the following, using your own values:

# --- required ------------------------------------------------------------
license_key          = "CSTW1-..."                          # your licence key
delivery_url         = "https://api.gateway.cloudspectra.ai"
package_version      = "1.0.0"                              # an exact version, never "latest"
oci_tenancy_ocid     = "ocid1.tenancy.oc1..aaaa..."
oci_compartment_ocid = "ocid1.compartment.oc1..aaaa..."

# Required by the configuration but not used by the install path:
# workers fetch the client through the Cloud Spectra downloader.
client_apt_base_url  = "https://downloads.cloudspectra.ai/gateway/apt"

# --- the relay image, and where workers download the client --------------
relay_ami_id         = "ami-0123456789abcdef0"              # arm64 Cloud Spectra Gateway AMI, your Region
public_artifact_base = "https://downloads.cloudspectra.ai"

# --- placement and naming ------------------------------------------------
name       = "cloudvpc-demo"                                # unique per deployment in these accounts
aws_region = "us-east-1"
oci_region = "us-ashburn-1"                                 # your OCI home region

# --- reach the demo application from your workstation --------------------
lb_internal = false
admin_cidrs = ["203.0.113.10/32"]                           # your public IPv4, as a /32

3Initialise and apply

terraform init
terraform apply

Review the plan and confirm. Terraform creates the AWS VPC, the relay Auto Scaling group, the load balancer and the peer map table, then the OCI network and the worker pool.

4Let the fleet pair

When terraform apply returns, the instances are still booting. Each relay starts and claims its rendezvous address; each worker downloads the client, finds a relay over IPv6, enrols and brings its tunnel up. Watch the relays claim their addresses:

eval "$(terraform output -raw watch_claims)"

The table lists each network interface in the VPC with its IPv6 rendezvous address and its private IPv4 address. A relay with an IPv6 address listed has claimed its slot.

5Check the load balancer targets

eval "$(terraform output -raw watch_targets)"

Every relay should be registered and healthy. The health check is the relay's own health endpoint, and the Auto Scaling group replaces a relay that keeps failing it.

6Send requests through AWS and see OCI answer

terraform output -raw app_url
eval "$(terraform output -raw verify)"

Each line reports the cloud that served the request and the AWS instance the worker presents itself as:

{"served":"oci","as":"i-0a1b2c3d4e5f60718"}
{"served":"oci","as":"i-0f9e8d7c6b5a41302"}

Two things in that output are the product:

• served is oci: the request entered an AWS load balancer and was answered by a machine in OCI.
• as is an AWS instance ID -- its relay's. The worker read it from the VPC's instance metadata service through the tunnel. To the VPC, the worker is that instance.

7Change one number

The relay Auto Scaling group is the only place the fleet size is decided:

aws autoscaling set-desired-capacity \
  --auto-scaling-group-name "$(terraform output -raw asg_name)" --desired-capacity 3

A third relay launches and claims an address; the workers read the group's size, grow the OCI pool to match, and the new worker pairs with the new relay. Run verify again to see a third instance ID. Scale back to 2 the same way: the worker whose relay left removes itself from the pool. relay_min and relay_max (2 and 6 by default) bound the capacity.

8Destroy everything this configuration created

terraform destroy

This removes the OCI pool and network, the relays, the load balancer, the peer map table and the VPC. If the destroy stops with DependencyViolation and a message that the network has some mapped public address(es), the relays are still terminating and their public addresses are still mapped, so the internet gateway cannot be detached yet. Wait a minute and run terraform destroy again.

Destroying the deployment does not cancel your Cloud Spectra subscription; manage that on your account page.