Update A3U template design

GoogleCloudPlatform · Dec 22, 2024 · 6642287 · 6642287
1 parent 66804e7
commit 6642287
Show file tree

Hide file tree

Showing 2 changed files with 19 additions and 18 deletions.
diff --git a/examples/gke-a3-ultragpu/gke-a3-ultragpu.yaml b/examples/gke-a3-ultragpu/gke-a3-ultragpu.yaml
@@ -53,21 +53,22 @@ deployment_groups:
 - group: primary
   modules:
   - id: gke-a3-ultra-net-0
-    source: github.com/GoogleCloudPlatform/cluster-toolkit.git//modules/network/vpc?ref=e0c690b
+    source: modules/network/vpc
     settings:
-      network_name: gke-a3-ultra-net-0
+      network_name: $(vars.deployment_name)-net-0
       subnetworks:
-      - subnet_name: gke-a3-ultra-sub-0
+      - subnet_name: $(vars.deployment_name)-sub-0
         subnet_region: $(vars.region)
         subnet_ip: 192.168.0.0/18
-      secondary_ranges:
-        gke-a3-ultra-sub-0:
+      secondary_ranges_list:
+      - subnetwork_name: $(vars.deployment_name)-sub-0
+        ranges:
         - range_name: pods
           ip_cidr_range: 10.4.0.0/14
         - range_name: services
           ip_cidr_range: 10.0.32.0/20
       firewall_rules:
-      - name: gke-a3-ultra-internal-0
+      - name: $(vars.deployment_name)-internal-0
         ranges: [192.168.0.0/16]
         allow:
         - protocol: tcp
@@ -77,16 +78,16 @@ deployment_groups:
         - protocol: icmp
 
   - id: gke-a3-ultra-net-1
-    source: github.com/GoogleCloudPlatform/cluster-toolkit.git//modules/network/vpc?ref=e0c690b
+    source: modules/network/vpc
     settings:
-      network_name: gke-a3-ultra-net-1
+      network_name: $(vars.deployment_name)-net-1
       mtu: $(vars.mtu_size)
       subnetworks:
-      - subnet_name: gke-a3-ultra-sub-1
+      - subnet_name: $(vars.deployment_name)-sub-1
         subnet_region: $(vars.region)
         subnet_ip: 192.168.64.0/18
       firewall_rules:
-      - name: gke-a3-ultra-internal-1
+      - name: $(vars.deployment_name)-internal-1
         ranges: [192.168.0.0/16]
         allow:
         - protocol: tcp
@@ -96,20 +97,20 @@ deployment_groups:
         - protocol: icmp
 
   - id: gke-a3-ultra-rdma-net
-    source: github.com/GoogleCloudPlatform/cluster-toolkit.git//community/modules/network/rdma-vpc?ref=98c49fe
+    source: modules/network/gpu-rdma-vpc
     settings:
-      network_name: gke-a3-ultra-rdma-net
+      network_name: $(vars.deployment_name)-rdma-net
       mtu: $(vars.mtu_size)
       network_profile: https://www.googleapis.com/compute/beta/projects/$(vars.project_id)/global/networkProfiles/$(vars.zone)-vpc-roce
       network_routing_mode: REGIONAL
       subnetworks_template:
-        name_prefix: gke-a3-ultra-rdma-sub
+        name_prefix: $(vars.deployment_name)-rdma-sub
         count: 8
         ip_range: 192.168.128.0/18
         region: $(vars.region)
 
   - id: a3-ultragpu-cluster
-    source: github.com/GoogleCloudPlatform/cluster-toolkit.git//modules/scheduler/gke-cluster?ref=e0c690b
+    source: modules/scheduler/gke-cluster
     use: [gke-a3-ultra-net-0]
     settings:
       release_channel: RAPID
@@ -146,7 +147,7 @@ deployment_groups:
     outputs: [instructions]
 
   - id: a3-ultragpu-pool
-    source: github.com/GoogleCloudPlatform/cluster-toolkit.git//modules/compute/gke-node-pool?ref=e0c690b
+    source: modules/compute/gke-node-pool
     use: [a3-ultragpu-cluster]
     settings:
       machine_type: a3-ultragpu-8g
@@ -183,11 +184,11 @@ deployment_groups:
     outputs: [instructions]
 
   - id: topology-aware-scheduler-install
-    source: github.com/GoogleCloudPlatform/cluster-toolkit.git//community/modules/compute/gke-topology-scheduler?ref=e0c690b
+    source: community/modules/compute/gke-topology-scheduler
     use: [a3-ultragpu-cluster]
 
   - id: workload-manager-install
-    source: github.com/GoogleCloudPlatform/cluster-toolkit.git//modules/management/kubectl-apply?ref=8c26d4a
+    source: modules/management/kubectl-apply
     use: [a3-ultragpu-cluster]
     settings:
       kueue:

diff --git a/tools/cloud-build/daily-tests/tests/gke-a3-ultragpu.yml b/tools/cloud-build/daily-tests/tests/gke-a3-ultragpu.yml
@@ -20,7 +20,7 @@ test_name: gke-a3ultra
 deployment_name: gke-a3ultra-{{ build }}
 workspace: /workspace
 blueprint_yaml: "{{ workspace }}/examples/gke-a3-ultragpu/gke-a3-ultragpu.yaml"
-network: gke-a3-ultra-net-0
+network: "{{ deployment_name }}-net-0"
 region: europe-west1
 zone: europe-west1-b
 remote_node: "{{ deployment_name }}-remote-node-0"