foxhunt/infra/modules/kapsule/main.tf

resource "scaleway_vpc_private_network" "foxhunt" {
  name   = "${var.cluster_name}-pn"
  region = var.region
}

resource "scaleway_k8s_cluster" "foxhunt" {
  name                        = var.cluster_name
  version                     = var.k8s_version
  cni                         = "cilium"
  region                      = var.region
  delete_additional_resources = true
  private_network_id          = scaleway_vpc_private_network.foxhunt.id

  auto_upgrade {
    enable                        = true
    maintenance_window_start_hour = 4
    maintenance_window_day        = "sunday"
  }

  autoscaler_config {
    disable_scale_down            = false
    scale_down_delay_after_add    = "10m"
    scale_down_unneeded_time      = "10m"
    estimator                     = "binpacking"
    ignore_daemonsets_utilization = true
  }
}

# Foxhunt app services pool (api-gateway, trading-service, etc.)
# Renamed from "services" — was 2× DEV1-L shared with databases,
# now 1× DEV1-L dedicated to foxhunt microservices only.
resource "scaleway_k8s_pool" "foxhunt" {
  cluster_id  = scaleway_k8s_cluster.foxhunt.id
  name        = "foxhunt"
  node_type   = var.foxhunt_type
  size        = 1
  min_size    = 1
  max_size    = var.foxhunt_max_size
  autoscaling        = true
  autohealing        = true
  public_ip_disabled = var.public_ip_disabled
  region             = var.region

  lifecycle {
    ignore_changes = [size]
  }
}

# Platform infrastructure pool (postgres, redis, minio, questdb, monitoring)
resource "scaleway_k8s_pool" "platform" {
  count       = var.enable_platform_pool ? 1 : 0
  cluster_id  = scaleway_k8s_cluster.foxhunt.id
  name        = "platform"
  node_type   = var.platform_type
  size        = 1
  min_size    = 1
  max_size    = 1
  autoscaling        = false
  autohealing        = true
  public_ip_disabled = var.public_ip_disabled
  region             = var.region
}

# GitLab CE node pool (dedicated GP1-XS for GitLab + Runner manager)
resource "scaleway_k8s_pool" "gitlab" {
  count       = var.enable_gitlab_pool ? 1 : 0
  cluster_id  = scaleway_k8s_cluster.foxhunt.id
  name        = "gitlab"
  node_type   = var.gitlab_type
  size        = 1
  min_size    = 1
  max_size    = 1
  autoscaling        = false
  autohealing        = true
  public_ip_disabled = var.public_ip_disabled
  region             = var.region
}

# L40S training pool (48GB VRAM, CUDA CC 89 — hyperopt + supervised training)
# Nodes auto-labeled: k8s.scaleway.com/pool-name=ci-training-l40s
resource "scaleway_k8s_pool" "ci_training_l40s" {
  count       = var.enable_ci_training_l40s_pool ? 1 : 0
  cluster_id  = scaleway_k8s_cluster.foxhunt.id
  name        = "ci-training-l40s"
  node_type   = var.ci_training_l40s_type
  size        = 1
  min_size    = 0
  max_size    = var.ci_training_l40s_max_size
  autoscaling        = true
  autohealing        = true
  public_ip_disabled = var.public_ip_disabled
  region             = var.region

  lifecycle {
    ignore_changes = [size]
  }
}

# H100 training pool (80GB VRAM, CUDA CC 90 — RL hyperopt + large model training)
# Nodes auto-labeled: k8s.scaleway.com/pool-name=ci-training-h100
resource "scaleway_k8s_pool" "ci_training_h100" {
  count       = var.enable_ci_training_h100_pool ? 1 : 0
  cluster_id  = scaleway_k8s_cluster.foxhunt.id
  name        = "ci-training-h100"
  node_type   = var.ci_training_h100_type
  size        = 1
  min_size    = 0
  max_size    = var.ci_training_h100_max_size
  autoscaling        = true
  autohealing        = true
  public_ip_disabled = var.public_ip_disabled
  region             = var.region

  lifecycle {
    ignore_changes = [size]
  }
}

# H100x2 training pool (2× 80GB VRAM — multi-GPU parallel hyperopt)
# Nodes auto-labeled: k8s.scaleway.com/pool-name=ci-training-h100x2
resource "scaleway_k8s_pool" "ci_training_h100x2" {
  count       = var.enable_ci_training_h100x2_pool ? 1 : 0
  cluster_id  = scaleway_k8s_cluster.foxhunt.id
  name        = "ci-training-h100x2"
  node_type   = var.ci_training_h100x2_type
  size        = 1
  min_size    = 0
  max_size    = var.ci_training_h100x2_max_size
  autoscaling        = true
  autohealing        = true
  public_ip_disabled = var.public_ip_disabled
  region             = var.region

  lifecycle {
    ignore_changes = [size]
  }
}

# H100-SXM training pool (2× 80GB SXM VRAM — NVLink multi-GPU)
# Nodes auto-labeled: k8s.scaleway.com/pool-name=ci-training-h100-sxm
resource "scaleway_k8s_pool" "ci_training_h100_sxm" {
  count       = var.enable_ci_training_h100_sxm_pool ? 1 : 0
  cluster_id  = scaleway_k8s_cluster.foxhunt.id
  name        = "ci-training-h100-sxm"
  node_type   = var.ci_training_h100_sxm_type
  size        = 1
  min_size    = 0
  max_size    = var.ci_training_h100_sxm_max_size
  autoscaling        = true
  autohealing        = true
  public_ip_disabled = var.public_ip_disabled
  region             = var.region

  lifecycle {
    ignore_changes = [size]
  }
}

# H100-SXM-8 training pool (8× 80GB SXM VRAM, NVLink — full multi-GPU saturation)
# Nodes auto-labeled: k8s.scaleway.com/pool-name=ci-training-h100-sxm8
resource "scaleway_k8s_pool" "ci_training_h100_sxm8" {
  count       = var.enable_ci_training_h100_sxm8_pool ? 1 : 0
  cluster_id  = scaleway_k8s_cluster.foxhunt.id
  name        = "ci-training-h100-sxm8"
  node_type   = var.ci_training_h100_sxm8_type
  size        = 1
  min_size    = 0
  max_size    = var.ci_training_h100_sxm8_max_size
  autoscaling        = true
  autohealing        = true
  public_ip_disabled = var.public_ip_disabled
  region             = var.region

  lifecycle {
    ignore_changes = [size]
  }
}

# CPU compile pool (POP2 — high CPU, no GPU, for cargo check/build)
# Nodes auto-labeled: k8s.scaleway.com/pool-name=ci-compile-cpu
resource "scaleway_k8s_pool" "ci_compile_cpu" {
  count       = var.enable_ci_compile_cpu_pool ? 1 : 0
  cluster_id  = scaleway_k8s_cluster.foxhunt.id
  name        = "ci-compile-cpu"
  node_type   = var.ci_compile_cpu_type
  size        = 1
  min_size    = 0
  max_size    = var.ci_compile_cpu_max_size
  autoscaling        = true
  autohealing        = true
  public_ip_disabled = var.public_ip_disabled
  region             = var.region

  lifecycle {
    ignore_changes = [size]
  }
}