defconfigs: add AWS P5.4xlarge GPU instance support

Add support for AWS P5.4xlarge GPU instance configuration and fix
existing P5.48xlarge defconfig. P5.4xlarge provides single NVIDIA H100
80GB GPU at ~2.29/hour, more cost-effective than P5.48xlarge for
single-GPU workloads. Update ansible instance type mapping to include
p5.4xlarge. Create aws-t3-micro defconfig for minimal cost testing.

Generated-by: Claude AI
Signed-off-by: Developer <[email protected]>

Author: mcgrof

defconfigs/aws-gpu-p5-48xlarge

@@ -24,8 +24,8 @@ CONFIG_TERRAFORM_AWS_REGION_US_EAST_1=y
 CONFIG_TERRAFORM_AWS_AZ_US_EAST_1A=y
 
 # P5 instance family with 8x H100 GPUs
-CONFIG_TERRAFORM_AWS_INSTANCE_TYPE_P5=y
-CONFIG_TERRAFORM_AWS_INSTANCE_P5_48XLARGE_CHOICE=y
+CONFIG_TERRAFORM_AWS_INSTANCE_FAMILY_P5=y
+CONFIG_TERRAFORM_AWS_INSTANCE_SIZE_P5_48XLARGE=y
 
 # Use Deep Learning PyTorch Ubuntu AMI
 CONFIG_TERRAFORM_AWS_USE_GPU_AMI=y
@@ -45,4 +45,4 @@ CONFIG_WORKFLOWS_DEDICATED_WORKFLOW=y
 CONFIG_KDEVOPS_WORKFLOW_ENABLE_GITR=y
 CONFIG_GITR=y
 CONFIG_GITR_INIT=y
-CONFIG_GITR_CUSTOM_COMMAND="nvidia-smi && python -c 'import torch; print(f\"H100s available: {torch.cuda.device_count()}\")'"
+CONFIG_GITR_CUSTOM_COMMAND="nvidia-smi && python -c 'import torch; print(f\"H100s available: {torch.cuda.device_count()\")'"

defconfigs/aws-gpu-p5-4xlarge

@@ -0,0 +1,43 @@
+# AWS P5.4xlarge GPU instance - Single NVIDIA H100 GPU
+#
+# This defconfig sets up an AWS P5.4xlarge instance with:
+# - 1x NVIDIA H100 80GB GPU
+# - 24 vCPUs
+# - 192 GiB memory
+# - 3.84 TB local NVMe SSD storage
+# - 400 Gbps network with EFA v2 support
+# - GPUDirect RDMA support
+#
+# Cost: ~$12.29/hour on-demand
+#
+# This is a more cost-effective option than P5.48xlarge for
+# single-GPU workloads and development/testing purposes.
+
+# Basic configuration
+CONFIG_KDEVOPS_HOSTS_PREFIX="kdevops"
+CONFIG_KDEVOPS_HOSTS_NUM_KEEP=1
+
+# Use Terraform for AWS
+CONFIG_TERRAFORM=y
+CONFIG_TERRAFORM_AWS=y
+CONFIG_TERRAFORM_AWS_DATA_ENABLE=y
+
+# Skip extra storage configuration
+CONFIG_TERRAFORM_EXTRA_VARS_AUTO=y
+
+# Use us-east-1 for best availability
+CONFIG_TERRAFORM_AWS_REGION_US_EAST_1=y
+CONFIG_TERRAFORM_AWS_AZ_US_EAST_1A=y
+
+# P5 instance with single H100 GPU
+CONFIG_TERRAFORM_AWS_INSTANCE_FAMILY_P5=y
+CONFIG_TERRAFORM_AWS_INSTANCE_SIZE_P5_4XLARGE=y
+
+# Use default VPC to avoid VPC limit issues
+# CONFIG_TERRAFORM_AWS_CREATE_VPC is not set
+CONFIG_TERRAFORM_AWS_ASSIGN_PUBLIC_IP=y
+
+# Storage configuration - suitable for ML workloads
+CONFIG_TERRAFORM_AWS_EBS_VOLUMES_PER_INSTANCE=2
+CONFIG_TERRAFORM_AWS_EBS_VOLUME_SIZE=500
+CONFIG_TERRAFORM_AWS_EBS_VOLUME_TYPE_GP3=y

defconfigs/aws-t3-micro

@@ -0,0 +1,34 @@
+# AWS t3.micro instance - minimal cost testing configuration
+#
+# This defconfig sets up a minimal AWS instance for testing
+# kdevops functionality without incurring significant costs.
+# t3.micro is eligible for AWS free tier.
+
+# Basic configuration
+CONFIG_KDEVOPS_HOSTS_PREFIX="kdevops"
+CONFIG_KDEVOPS_HOSTS_NUM_KEEP=1
+
+# Use Terraform for AWS
+CONFIG_TERRAFORM=y
+CONFIG_TERRAFORM_AWS=y
+CONFIG_TERRAFORM_AWS_DATA_ENABLE=y
+
+# Skip extra storage configuration
+CONFIG_TERRAFORM_EXTRA_VARS_AUTO=y
+
+# Use us-east-1 for best availability
+CONFIG_TERRAFORM_AWS_REGION_US_EAST_1=y
+CONFIG_TERRAFORM_AWS_AZ_US_EAST_1A=y
+
+# T3 micro instance - cheapest option, free tier eligible
+CONFIG_TERRAFORM_AWS_INSTANCE_TYPE_T3=y
+CONFIG_TERRAFORM_AWS_INSTANCE_T3_MICRO_CHOICE=y
+
+# Use default VPC to avoid VPC limit issues
+# CONFIG_TERRAFORM_AWS_CREATE_VPC is not set
+CONFIG_TERRAFORM_AWS_ASSIGN_PUBLIC_IP=y
+
+# Minimal storage - single small volume
+CONFIG_TERRAFORM_AWS_EBS_VOLUMES_PER_INSTANCE=1
+CONFIG_TERRAFORM_AWS_EBS_VOLUME_SIZE=8
+CONFIG_TERRAFORM_AWS_EBS_VOLUME_TYPE_GP3=y

playbooks/roles/gen_tfvars/tasks/main.yml

@@ -51,6 +51,7 @@
                                        'g5.24xlarge' if terraform_aws_instance_g5_24xlarge | default(false) | bool else
                                        'g5.48xlarge' if terraform_aws_instance_g5_48xlarge | default(false) | bool else
                                        'p4d.24xlarge' if terraform_aws_instance_p4d_24xlarge | default(false) | bool else
+                                       'p5.4xlarge' if terraform_aws_instance_p5_4xlarge | default(false) | bool else
                                        'p5.48xlarge' if terraform_aws_instance_p5_48xlarge | default(false) | bool else
                                        terraform_aws_instance_type | default('t2.micro') }}"
   when: kdevops_terraform_provider == "aws"