GCP.GKEOnPrem reference

BareMetalCluster

Source: src/GCP/GKEOnPrem/BareMetalCluster.ts

An Anthos on bare metal user cluster registered with the GKE On-Prem API. Creating a cluster requires a connected admin cluster (Hub membership) and on-prem hardware.

Clusters have no GCP labels field, so Alchemy stamps ownership into annotations and a [alchemy …] description prefix for list / nuke. Cluster id, location, and admin membership are identity — changing them replaces the cluster. Other fields update in place.

BareMetalCluster: Creating a Bare Metal Cluster

const cluster = yield* GCP.GKEOnPrem.BareMetalCluster("Workload", {
  adminClusterMembership:
    "projects/my-project/locations/global/memberships/admin",
  bareMetalVersion: "1.28.0-gke.1",
  controlPlane: {
    controlPlaneNodePoolConfig: {
      nodePoolConfig: {
        nodeConfigs: [{ nodeIp: "10.200.0.2" }],
      },
    },
  },
  storage: {
    lvpShareConfig: {
      lvpConfig: { path: "/mnt/localpv-share", storageClass: "local-shared" },
    },
    lvpNodeMountsConfig: {
      path: "/mnt/localpv-disk",
      storageClass: "local-disks",
    },
  },
  networkConfig: {
    islandModeCidr: {
      serviceAddressCidrBlocks: ["10.96.0.0/12"],
      podAddressCidrBlocks: ["192.168.0.0/16"],
    },
  },
  loadBalancer: {
    vipConfig: { controlPlaneVip: "10.200.0.8", ingressVip: "10.200.0.9" },
    metalLbConfig: {
      addressPools: [{ pool: "pool-1", addresses: ["10.200.0.10-10.200.0.20"] }],
    },
  },
  description: "app user cluster",
});

BareMetalCluster: Updating a Bare Metal Cluster

Re-declare the same logical id with changed props; the engine keeps the physical resource and updates it in place.

const cluster = yield* GCP.GKEOnPrem.BareMetalCluster("Workload", {
  adminClusterMembership:
    "projects/my-project/locations/global/memberships/admin",
  bareMetalVersion: "1.29.0-gke.1",
  controlPlane: {
    controlPlaneNodePoolConfig: {
      nodePoolConfig: {
        nodeConfigs: [{ nodeIp: "10.200.0.2" }],
      },
    },
  },
  storage: {
    lvpShareConfig: {
      lvpConfig: { path: "/mnt/localpv-share", storageClass: "local-shared" },
    },
    lvpNodeMountsConfig: {
      path: "/mnt/localpv-disk",
      storageClass: "local-disks",
    },
  },
  networkConfig: {
    islandModeCidr: {
      serviceAddressCidrBlocks: ["10.96.0.0/12"],
      podAddressCidrBlocks: ["192.168.0.0/16"],
    },
  },
  loadBalancer: {
    vipConfig: { controlPlaneVip: "10.200.0.8", ingressVip: "10.200.0.9" },
    metalLbConfig: {
      addressPools: [{ pool: "pool-1", addresses: ["10.200.0.10-10.200.0.20"] }],
    },
  },
  description: "app user cluster v2",
});

BareMetalClustersBareMetalNodePool

Source: src/GCP/GKEOnPrem/BareMetalClustersBareMetalNodePool.ts

A worker node pool on an Anthos on bare metal user cluster.

Node pools have no GCP labels field, so Alchemy stamps ownership into annotations and a [alchemy …] displayName prefix for list / nuke. Pool id, location, and parent cluster are identity — changing them replaces the pool. Config, display name, and upgrade policy update in place.

BareMetalClustersBareMetalNodePool: Creating a Bare Metal Node Pool

const pool = yield* GCP.GKEOnPrem.BareMetalClustersBareMetalNodePool(
  "Workers",
  {
    bareMetalCluster: cluster.name,
    nodePoolConfig: {
      nodeConfigs: [
        { nodeIp: "10.200.0.11" },
        { nodeIp: "10.200.0.12" },
      ],
    },
    displayName: "app workers",
  },
);

BareMetalClustersBareMetalNodePool: Updating a Bare Metal Node Pool

Re-declare the same logical id with changed props; the engine keeps the physical resource and updates it in place.

const pool = yield* GCP.GKEOnPrem.BareMetalClustersBareMetalNodePool(
  "Workers",
  {
    bareMetalCluster: cluster.name,
    nodePoolConfig: {
      nodeConfigs: [
        { nodeIp: "10.200.0.11" },
        { nodeIp: "10.200.0.12" },
        { nodeIp: "10.200.0.13" },
      ],
    },
    displayName: "app workers v2",
  },
);

VmwareCluster

Source: src/GCP/GKEOnPrem/VmwareCluster.ts

An Anthos on VMware user cluster registered with the GKE On-Prem API. Creating a cluster requires a connected admin cluster (Hub membership) and vSphere infrastructure.

Clusters have no GCP labels field, so Alchemy stamps ownership into annotations and a [alchemy …] description prefix for list / nuke. Cluster id, location, and admin membership are identity — changing them replaces the cluster. Other fields update in place.

VmwareCluster: Creating a VMware Cluster

const cluster = yield* GCP.GKEOnPrem.VmwareCluster("Workload", {
  adminClusterMembership:
    "projects/my-project/locations/global/memberships/admin",
  onPremVersion: "1.28.0-gke.1",
  controlPlaneNode: { cpus: "4", memory: "8192", replicas: "1" },
  networkConfig: {
    serviceAddressCidrBlocks: ["10.96.0.0/12"],
    podAddressCidrBlocks: ["192.168.0.0/16"],
    dhcpIpConfig: { enabled: true },
  },
  loadBalancer: {
    vipConfig: { controlPlaneVip: "10.200.0.8", ingressVip: "10.200.0.9" },
    metalLbConfig: {
      addressPools: [{ pool: "pool-1", addresses: ["10.200.0.10-10.200.0.20"] }],
    },
  },
  description: "app user cluster",
});

VmwareCluster: Updating a VMware Cluster

Re-declare the same logical id with changed props; the engine keeps the physical resource and updates it in place.

const cluster = yield* GCP.GKEOnPrem.VmwareCluster("Workload", {
  adminClusterMembership:
    "projects/my-project/locations/global/memberships/admin",
  onPremVersion: "1.29.0-gke.1",
  controlPlaneNode: { cpus: "4", memory: "8192", replicas: "1" },
  networkConfig: {
    serviceAddressCidrBlocks: ["10.96.0.0/12"],
    podAddressCidrBlocks: ["192.168.0.0/16"],
    dhcpIpConfig: { enabled: true },
  },
  loadBalancer: {
    vipConfig: { controlPlaneVip: "10.200.0.8", ingressVip: "10.200.0.9" },
    metalLbConfig: {
      addressPools: [{ pool: "pool-1", addresses: ["10.200.0.10-10.200.0.20"] }],
    },
  },
  description: "app user cluster v2",
});

VmwareClustersVmwareNodePool

Source: src/GCP/GKEOnPrem/VmwareClustersVmwareNodePool.ts

A worker node pool on an Anthos on VMware user cluster.

Node pools have no GCP labels field, so Alchemy stamps ownership into annotations and a [alchemy …] displayName prefix for list / nuke. Pool id, location, and parent cluster are identity — changing them replaces the pool. Config, display name, version, and autoscaling update in place.

VmwareClustersVmwareNodePool: Creating a VMware Node Pool

const pool = yield* GCP.GKEOnPrem.VmwareClustersVmwareNodePool("Workers", {
  vmwareCluster: cluster.name,
  config: {
    imageType: "ubuntu_containerd",
    replicas: "3",
    cpus: "4",
    memoryMb: "8192",
  },
  displayName: "app workers",
});

VmwareClustersVmwareNodePool: Updating a VMware Node Pool

Re-declare the same logical id with changed props; the engine keeps the physical resource and updates it in place.

const pool = yield* GCP.GKEOnPrem.VmwareClustersVmwareNodePool("Workers", {
  vmwareCluster: cluster.name,
  config: {
    imageType: "ubuntu_containerd",
    replicas: "5",
    cpus: "4",
    memoryMb: "8192",
  },
  displayName: "app workers v2",
});