infrastructure/servers/legion/ollama.tf

# Ollama — local LLM inference with GPU
resource "kubernetes_namespace" "ollama" {
  metadata {
    name = "ollama"
    labels = {
      managed-by = "terraform"
      tier       = "application"
    }
  }
}

resource "kubernetes_persistent_volume_claim" "ollama_models" {
  metadata {
    name      = "ollama-models"
    namespace = kubernetes_namespace.ollama.metadata[0].name
  }

  spec {
    access_modes = ["ReadWriteOnce"]
    resources {
      requests = {
        storage = "50Gi"
      }
    }
  }
}

resource "kubernetes_deployment" "ollama" {
  metadata {
    name      = "ollama"
    namespace = kubernetes_namespace.ollama.metadata[0].name
    labels = {
      app = "ollama"
    }
  }

  spec {
    replicas = 1

    strategy {
      type = "Recreate"
    }

    selector {
      match_labels = {
        app = "ollama"
      }
    }

    template {
      metadata {
        labels = {
          app = "ollama"
        }
      }

      spec {
        container {
          name  = "ollama"
          image = "ollama/ollama:latest"

          port {
            container_port = 11434
          }

          env {
            name  = "NVIDIA_VISIBLE_DEVICES"
            value = "all"
          }

          volume_mount {
            name       = "models"
            mount_path = "/root/.ollama"
          }

          resources {
            requests = {
              memory = "2Gi"
              cpu    = "500m"
            }
            limits = {
              memory = "8Gi"
              cpu    = "4"
            }
          }
        }

        volume {
          name = "models"
          persistent_volume_claim {
            claim_name = kubernetes_persistent_volume_claim.ollama_models.metadata[0].name
          }
        }
      }
    }
  }
}

resource "kubernetes_service" "ollama" {
  metadata {
    name      = "ollama"
    namespace = kubernetes_namespace.ollama.metadata[0].name
  }

  spec {
    selector = {
      app = "ollama"
    }

    port {
      port        = 11434
      target_port = 11434
      node_port   = 31434
    }

    type = "NodePort"
  }
}