Fix tests on development branch (#263)

* apply pathways fixes Signed-off-by: Piotr Pawłowski <ppawl@google.com> --------- Signed-off-by: Piotr Pawłowski <ppawl@google.com>
AI-Hypercomputer · Jan 10, 2025 · 10816ca · 10816ca
1 parent ea2bbab
commit 10816ca
Show file tree

Hide file tree

Showing 3 changed files with 5 additions and 14 deletions.
diff --git a/.github/workflows/build_tests.yaml b/.github/workflows/build_tests.yaml
@@ -43,6 +43,7 @@ env:
   ZONE: us-central2-a
   REGION: us-central2
   STORAGE_NAME: test-storage
+  CLUSTER_ARGUMENTS: "--network=${{secrets.NETWORK_NAME}} --maintenance-window=23:50"
 
 jobs:
   run-unit-tests:
@@ -150,7 +151,7 @@ jobs:
       run: |
         python3 xpk.py cluster create-pathways --cluster $TPU_CLUSTER_NAME --tpu-type=v4-8  --num-slices=2 \
         --zone=us-central2-b --default-pool-cpu-machine-type=n1-standard-16 --default-pool-cpu-num-nodes=16 \
-        --reservation='${{ secrets.GCP_TPU_V4_RESERVATION }}' --enable-workload-identity --enable-gcsfuse-csi-driver
+        --reservation='${{ secrets.GCP_TPU_V4_RESERVATION }}' --enable-workload-identity --enable-gcsfuse-csi-driver --custom-cluster-arguments="${CLUSTER_ARGUMENTS}"\
     - name: Authenticate Docker
       run: gcloud auth configure-docker --quiet
     - name: Create auto-mount Storage instance

diff --git a/src/xpk/commands/cluster.py b/src/xpk/commands/cluster.py
@@ -33,7 +33,6 @@
     set_jobset_on_cluster,
     set_up_cluster_network_for_gpu,
     setup_k8s_env,
-    update_cluster_with_clouddns_if_necessary,
     update_cluster_with_gcsfuse_driver_if_necessary,
     update_cluster_with_workload_identity_if_necessary,
     zone_to_region,
@@ -132,12 +131,6 @@ def cluster_create(args) -> None:
       xpk_exit(update_cluster_command_code)
 
   # Update Pathways clusters with CloudDNS if not enabled already.
-  if args.enable_pathways:
-    update_cluster_command_code = update_cluster_with_clouddns_if_necessary(
-        args
-    )
-    if update_cluster_command_code != 0:
-      xpk_exit(update_cluster_command_code)
 
   get_cluster_credentials(args)
 

diff --git a/src/xpk/core/core.py b/src/xpk/core/core.py
@@ -1150,17 +1150,14 @@ def is_cluster_using_clouddns(args) -> bool:
   command = (
       f'gcloud container clusters describe {args.cluster}'
       f' --project={args.project} --region={zone_to_region(args.zone)}'
-      ' 2> /dev/null | grep "clusterDns: CLOUD_DNS" | wc -l'
+      ' 2> /dev/null | grep "clusterDns: CLOUD_DNS"'
   )
-  return_code, cloud_dns_matches = run_command_for_value(
+  return_code, _ = run_command_for_value(
       command,
       'Check if Cloud DNS is enabled in cluster describe.',
       args,
   )
-  if return_code != 0:
-    xpk_exit(return_code)
-  cloud_dns_matches = int(cloud_dns_matches)
-  if cloud_dns_matches > 0:
+  if return_code == 0:
     xpk_print('Cloud DNS is enabled on the cluster, no update needed.')
     return True
   return False