xpk 0.12.0__tar.gz → 0.13.0__tar.gz

{xpk-0.12.0/src/xpk.egg-info → xpk-0.13.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: xpk
-Version: 0.12.0
+Version: 0.13.0
 Summary: xpk helps Cloud developers to orchestrate training jobs on accelerators on GKE.
 Author-email: XPK team <xpk-code-reviewers@google.com>
 License: Apache-2.0

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/commands/batch.py

@@ -31,6 +31,7 @@ from ..core.kjob import (
 )
 from ..core.kueue import LOCAL_QUEUE_NAME
 from ..utils.console import xpk_exit, xpk_print
+from ..utils.execution_context import is_dry_run
 from .kind import set_local_cluster_command
 from .kjob_common import add_gpu_networking_annotations_to_command, add_TAS_annotations_to_command
 
@@ -51,18 +52,16 @@ def batch(args: Namespace) -> None:
     if set_cluster_command_code != 0:
       xpk_exit(set_cluster_command_code)
 
-  err_code = prepare_kjob(args)
-  if err_code > 0:
-    xpk_exit(err_code)
-  setup_k8s_service_accounts()
+  if not is_dry_run():
+    err_code = prepare_kjob(args)
+    if err_code > 0:
+      xpk_exit(err_code)
+    setup_k8s_service_accounts()
 
   submit_job(args)
 
 
 def submit_job(args: Namespace) -> None:
-
-  setup_k8s_service_accounts()
-
   cmd = (
       'kubectl kjob create slurm'
       f' --profile {AppProfileDefaults.NAME.value}'
@@ -73,7 +72,8 @@ def submit_job(args: Namespace) -> None:
   cmd = add_gpu_networking_annotations_to_command(args, cmd)
   cmd = add_TAS_annotations_to_command(args, cmd)
 
-  for annotation in get_storage_annotations(args):
+  annotations = [] if is_dry_run() else get_storage_annotations(args)
+  for annotation in annotations:
     cmd += f' --pod-template-annotation {annotation}'
 
   if args.ignore_unknown_flags:

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/commands/cluster.py

@@ -76,6 +76,7 @@ from ..core.vertex import create_vertex_tensorboard
 from ..core.workload import get_workload_list
 from ..utils.console import get_user_input, xpk_exit, xpk_print
 from ..utils.file import write_tmp_file
+from ..utils.execution_context import is_dry_run
 from . import cluster_gcluster
 from .common import set_cluster_command
 import shutil
@@ -128,9 +129,10 @@ def cluster_adapt(args) -> None:
 
   get_cluster_credentials(args)
 
-  k8s_client = setup_k8s_env(args)
+  if not is_dry_run():
+    k8s_client = setup_k8s_env(args)
+    install_storage_crd(k8s_client)
 
-  install_storage_crd(k8s_client)
   install_storage_csis(args)
 
   # create Vertex Tensorboard for new and existing clusters if create-vertex-tensorboard is set
@@ -251,9 +253,10 @@ def cluster_create(args) -> None:
   if update_coredns_command_code != 0:
     xpk_exit(update_cluster_command_code)
 
-  k8s_client = setup_k8s_env(args)
+  if not is_dry_run():
+    k8s_client = setup_k8s_env(args)
+    install_storage_crd(k8s_client)
 
-  install_storage_crd(k8s_client)
   install_storage_csis(args)
 
   # create Vertex Tensorboard for new and existing clusters if create-vertex-tensorboard is set
@@ -409,10 +412,8 @@ def cluster_cacheimage(args) -> None:
       nodeSelectorKey=node_selector_key,
   )
   tmp = write_tmp_file(yml_string)
-  command_apply = f'kubectl apply -f {str(tmp.file.name)}'
-  command_delete = (
-      f'kubectl delete -f {str(tmp.file.name)} --ignore-not-found=true'
-  )
+  command_apply = f'kubectl apply -f {str(tmp)}'
+  command_delete = f'kubectl delete -f {str(tmp)} --ignore-not-found=true'
 
   return_code = run_command_with_updates(
       command_delete, 'Deleting Cached Image', args

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/commands/common.py

@@ -18,6 +18,7 @@ from ..core.commands import run_command_with_updates_retry
 from ..core.capacity import H100_MEGA_DEVICE_TYPE, CapacityType
 from ..core.gcloud_context import zone_to_region
 from ..utils.console import xpk_print, xpk_exit
+from ..utils.execution_context import is_dry_run
 from ..core.system_characteristics import (
     SystemCharacteristics,
 )
@@ -63,6 +64,9 @@ def is_TAS_possible(
     True if possible and False otherwise.
   """
 
+  if is_dry_run():
+    return True
+
   if system_characteristics is None:
     xpk_print('system_characteristics data was not found in configmaps.')
     xpk_exit(1)

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/commands/inspector.py

@@ -346,7 +346,7 @@ def inspector(args) -> None:
       )
 
   # Summarize inspector:
-  xpk_print(f'Find xpk inspector output file: {inspector_file.name}')
+  xpk_print(f'Find xpk inspector output file: {inspector_file}')
 
   if final_return_code != 0:
     xpk_print(

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/commands/job.py

@@ -28,6 +28,28 @@ from ..utils.console import xpk_exit, xpk_print
 from .kind import set_local_cluster_command
 
 
+JOBS_DRY_RUN_YAML = """
+items:
+- apiVersion: slurm.k8s.io/v1alpha1
+  kind: SlurmJob
+  metadata:
+    annotations:
+      kjobctl.x-k8s.io/script: echo hello
+    creationTimestamp: '2024-04-29T12:00:00Z'
+    labels:
+      kjobctl.x-k8s.io/app-profile: default
+    name: golden-job
+    namespace: default
+  spec:
+    script: echo hello
+"""
+
+PODS_DRY_RUN_RESULT = """
+foo-pod 2/2 Running 0 2d
+bar-pod 1/1 Evicted 0 1d
+"""
+
+
 def job_info(args):
   """Run commands obtaining information about a job given by name.
 
@@ -52,7 +74,10 @@ def job_info(args):
       f' metadata.name=={job_name}'
   )
   job_code, job_text = run_command_for_value(
-      job_command, 'Getting job info', args
+      job_command,
+      'Getting job info',
+      args,
+      dry_run_return_val=JOBS_DRY_RUN_YAML,
   )
   if job_code != 0:
     xpk_print(f'Job info request returned ERROR {job_code}')
@@ -60,7 +85,10 @@ def job_info(args):
 
   pods_command = f'kubectl get pods -l=job-name={job_name} --no-headers'
   pods_code, pods_text = run_command_for_value(
-      pods_command, 'Getting pods list', args
+      pods_command,
+      'Getting pods list',
+      args,
+      dry_run_return_val=PODS_DRY_RUN_RESULT,
   )
   if pods_code != 0:
     xpk_print(f'Pods list request returned ERROR {pods_code}')

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/commands/storage.py

@@ -58,6 +58,7 @@ from ..core.storage import (
 )
 from ..utils.console import get_user_input, xpk_exit, xpk_print
 from ..utils.kubectl import apply_kubectl_manifest
+from ..utils.execution_context import is_dry_run
 
 
 def storage_create(args: Namespace) -> None:
@@ -243,8 +244,10 @@ def enable_csi_drivers_if_necessary(args: Namespace) -> None:
 
 
 def storage_list(args: Namespace) -> None:
-  k8s_api_client = setup_k8s_env(args)
-  storages = list_storages(k8s_api_client)
+  storages = []
+  if not is_dry_run():
+    k8s_api_client = setup_k8s_env(args)
+    storages = list_storages(k8s_api_client)
   print_storages_for_cluster(storages)
 
 

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/commands/workload.py

@@ -97,6 +97,7 @@ from ..core.workload_decorators import (
 )
 from ..utils.console import get_user_input, xpk_exit, xpk_print
 from ..utils.file import write_tmp_file
+from ..utils.execution_context import is_dry_run
 from . import cluster_gcluster
 from .common import is_TAS_possible
 
@@ -306,8 +307,10 @@ def workload_create(args) -> None:
   Returns:
     0 if successful and 1 otherwise.
   """
-  k8s_api_client = setup_k8s_env(args)
-  setup_k8s_service_accounts()
+  k8s_api_client = None
+  if not is_dry_run():
+    k8s_api_client = setup_k8s_env(args)
+    setup_k8s_service_accounts()
 
   workload_exists = check_if_workload_exists(args)
 
@@ -383,8 +386,10 @@ def workload_create(args) -> None:
   all_storages = []
   # Currently storage customization is not supported for Pathways workloads. b/408468941
   if not args.use_pathways:
-    storages: list[Storage] = get_storages_to_mount(
-        k8s_api_client, args.storage
+    storages: list[Storage] = (
+        []
+        if k8s_api_client is None
+        else get_storages_to_mount(k8s_api_client, args.storage)
     )
     gcs_fuse_storages = list(
         filter(lambda storage: storage.type == GCS_FUSE_TYPE, storages)
@@ -569,14 +574,14 @@ def workload_create(args) -> None:
         pod_failure_policy=pod_failure_policy,
     )
   tmp = write_tmp_file(yml_string)
-  command = f'kubectl apply -f {str(tmp.file.name)}'
+  command = f'kubectl apply -f {str(tmp)}'
   return_code = run_command_with_updates(command, 'Creating Workload', args)
 
   if return_code != 0:
     xpk_print(f'Create Workload request returned ERROR {return_code}')
     xpk_exit(return_code)
 
-  if not args.use_pathways:
+  if not args.use_pathways and not is_dry_run():
     add_bucket_iam_members(args, storages)
 
   # Get GKE outlier dashboard for TPU
@@ -725,7 +730,11 @@ def workload_delete(args) -> None:
       )
     else:
       return_code = run_commands(
-          commands, 'Delete Workload', task_names, batch=100
+          commands,
+          'Delete Workload',
+          task_names,
+          batch=100,
+          dry_run=args.dry_run,
       )
 
     if return_code != 0:
@@ -743,8 +752,6 @@ def workload_list(args) -> None:
   Returns:
     0 if successful and 1 otherwise.
   """
-  xpk_print(args)
-
   xpk_print('Starting workload list', flush=True)
   add_zone_and_project(args)
   get_cluster_credentials(args)

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/core/cluster.py

@@ -442,7 +442,11 @@ def setup_k8s_env(args) -> k8s_client.ApiClient:
   if not getattr(args, 'kind_cluster', False):
     add_zone_and_project(args)
     get_cluster_credentials(args)
-    args.project_number = project_id_to_project_number(args.project)
+    args.project_number = (
+        project_id_to_project_number(args.project)
+        if not args.dry_run
+        else abs(hash(args.project) % (10**12))  # 12 digit hash
+    )
 
   config.load_kube_config()
   return k8s_client.ApiClient()

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/core/cluster_private.py

@@ -19,6 +19,7 @@ from ..utils.network import (
     add_current_machine_to_networks,
     is_current_machine_in_any_network,
 )
+from ..utils.execution_context import is_dry_run
 from ..utils.objects import is_text_true
 from .commands import run_command_for_value, run_command_with_updates
 from .gcloud_context import zone_to_region
@@ -37,7 +38,7 @@ def authorize_private_cluster_access_if_necessary(args) -> int:
     if not args.private and args.authorized_networks is None:
       xpk_print('Cluster is public and no need to authorize networks.')
       return 0
-    else:
+    elif not is_dry_run():
       xpk_print(
           'Cannot convert an existing public cluster to private. The arguments'
           ' --private and --authorized-networks are not acceptable for public'
@@ -164,6 +165,7 @@ def get_cluster_authorized_networks(args) -> list[str]:
       command,
       'Fetching the list of authorized network from cluster describe.',
       args,
+      dry_run_return_val='127.0.0.1/32',
   )
 
   if return_code != 0:

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/core/commands.py

@@ -78,14 +78,13 @@ def run_command_batch(commands, jobname, per_command_name, output_logs):
     The max return code and a list of all the return codes.
   """
 
+  files = [open(f, 'w', encoding='utf-8') for f in output_logs]
   children = []
   start_time = datetime.datetime.now()
-  for i, command in enumerate(commands):
+  for command, file in zip(commands, files):
     children.append(
         # subprocess managed by list pylint: disable=consider-using-with
-        subprocess.Popen(
-            command, stdout=output_logs[i], stderr=output_logs[i], shell=True
-        )
+        subprocess.Popen(command, stdout=file, stderr=file, shell=True)
     )
 
   while True:
@@ -99,7 +98,7 @@ def run_command_batch(commands, jobname, per_command_name, output_logs):
       slow_worker_text = per_command_name[slow_worker_index]
       slow_str = (
           f', task {slow_worker_text} still working, logfile'
-          f' {output_logs[slow_worker_index].name}'
+          f' {output_logs[slow_worker_index]}'
       )
     else:
       slow_str = ''
@@ -116,7 +115,7 @@ def run_command_batch(commands, jobname, per_command_name, output_logs):
       )
       xpk_print(
           f'Failure is {per_command_name[failing_index]}'
-          f' and logfile {output_logs[failing_index].name}'
+          f' and logfile {output_logs[failing_index]}'
       )
       for child in children:
         child.terminate()
@@ -126,6 +125,10 @@ def run_command_batch(commands, jobname, per_command_name, output_logs):
       break
 
     time.sleep(1)
+
+  for file in files:
+    file.close()
+
   return max_returncode, returncodes
 
 
@@ -351,6 +354,6 @@ def run_command_with_full_controls(
 
 def run_kubectl_apply(yml_string: str, task: str, args: Namespace) -> int:
   tmp = write_tmp_file(yml_string)
-  command = f'kubectl apply -f {str(tmp.file.name)}'
+  command = f'kubectl apply -f {str(tmp)}'
   err_code = run_command_with_updates(command, task, args)
   return err_code

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/core/config.py

@@ -22,7 +22,7 @@ from ..utils import file
 from ..utils.console import xpk_print
 
 # This is the version for XPK PyPI package
-__version__ = 'v0.12.0'
+__version__ = 'v0.13.0'
 XPK_CURRENT_VERSION = __version__
 XPK_CONFIG_FILE = os.path.expanduser('~/.config/xpk/config.yaml')
 

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/core/docker_image.py

@@ -21,6 +21,7 @@ import string
 
 from ..utils.console import xpk_exit, xpk_print
 from ..utils.file import write_tmp_file
+from ..utils.execution_context import is_dry_run
 from .commands import run_command_with_updates
 
 DEFAULT_DOCKER_IMAGE = 'python:3.10'
@@ -75,7 +76,9 @@ def build_docker_image_from_base_image(args, verbose=True) -> tuple[int, str]:
   """
 
   # Pick a name for the docker image.
-  docker_image_prefix = os.getenv('USER', 'unknown')
+  docker_image_prefix = (
+      'dry-run' if is_dry_run() else os.getenv('USER', 'unknown')
+  )
   docker_name = f'{docker_image_prefix}-runner'
 
   script_dir_dockerfile = """FROM {base_docker_image}
@@ -94,7 +97,7 @@ def build_docker_image_from_base_image(args, verbose=True) -> tuple[int, str]:
   )
   tmp = write_tmp_file(docker_file)
   docker_build_command = (
-      f'docker buildx build --platform={PLATFORM} -f {str(tmp.file.name)} -t'
+      f'docker buildx build --platform={PLATFORM} -f {str(tmp)} -t'
       f' {docker_name} {args.script_dir}'
   )
   xpk_print(f'Building {args.script_dir} into docker image.')
@@ -114,10 +117,16 @@ def build_docker_image_from_base_image(args, verbose=True) -> tuple[int, str]:
 
   # Pick a randomly generated `tag_length` character docker tag.
   tag_length = 4
-  tag_random_prefix = ''.join(
-      random.choices(string.ascii_lowercase, k=tag_length)
+  tag_random_prefix = (
+      'prefix'
+      if is_dry_run()
+      else ''.join(random.choices(string.ascii_lowercase, k=tag_length))
+  )
+  tag_datetime = (
+      'current'
+      if is_dry_run()
+      else datetime.datetime.now().strftime('%Y-%m-%d-%H-%M-%S')
   )
-  tag_datetime = datetime.datetime.now().strftime('%Y-%m-%d-%H-%M-%S')
   tag_name = f'{tag_random_prefix}-{tag_datetime}'
   cloud_docker_image = f'gcr.io/{args.project}/{docker_name}:{tag_name}'
   xpk_print(f'Adding Docker Image: {cloud_docker_image} to {args.project}')

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/core/docker_resources.py

@@ -20,6 +20,7 @@ from .capacity import H100_DEVICE_TYPE, H100_MEGA_DEVICE_TYPE, H200_DEVICE_TYPE
 from .cluster import setup_k8s_env
 from .storage import GCS_FUSE_TYPE, GCP_FILESTORE_TYPE, PARALLELSTORE_TYPE, GCE_PD_TYPE, LUSTRE_TYPE, Storage, get_storages_to_mount
 from .system_characteristics import AcceleratorType, SystemCharacteristics
+from ..utils.execution_context import is_dry_run
 
 
 def get_main_container_resources(
@@ -272,8 +273,10 @@ def get_volumes(args, system: SystemCharacteristics) -> str:
               - name: shared-data
               """
 
-  storages: list[Storage] = get_storages_to_mount(
-      setup_k8s_env(args), args.storage
+  storages: list[Storage] = (
+      []
+      if is_dry_run()
+      else get_storages_to_mount(setup_k8s_env(args), args.storage)
   )
   for storage in storages:
     if storage.type in {
@@ -325,8 +328,10 @@ def get_volume_mounts(args, system: SystemCharacteristics) -> str:
   elif system.accelerator_type == AcceleratorType['GPU']:
     volume_mount_yaml = ''
 
-  storages: list[Storage] = get_storages_to_mount(
-      setup_k8s_env(args), args.storage
+  storages: list[Storage] = (
+      []
+      if is_dry_run()
+      else get_storages_to_mount(setup_k8s_env(args), args.storage)
   )
   for storage in storages:
     if storage.type in {

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/core/jobset.py

@@ -134,7 +134,7 @@ def update_jobset_resources_if_necessary(args):
       memory_limit_size=new_memory_limit,
   )
   tmp = write_tmp_file(yml_string)
-  command = f'kubectl apply -f {str(tmp.file.name)}'
+  command = f'kubectl apply -f {str(tmp)}'
 
   task = 'Updating jobset Controller Manager resources'
   return_code = run_command_with_updates_retry(command, task, args)

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/core/kjob.py

@@ -23,6 +23,7 @@ from kubernetes.client import ApiClient
 from kubernetes.client.rest import ApiException
 
 from ..utils import templates
+from ..utils.execution_context import is_dry_run
 from ..utils.console import xpk_exit, xpk_print
 from .capacity import H100_DEVICE_TYPE, H100_MEGA_DEVICE_TYPE, H200_DEVICE_TYPE
 from .cluster import DEFAULT_NAMESPACE, XPK_SA, setup_k8s_env
@@ -368,8 +369,10 @@ def create_pod_template_instance(args: Namespace, service_account: str) -> int:
 def prepare_kjob(args: Namespace) -> int:
   system = get_cluster_system_characteristics(args)
 
-  k8s_api_client = setup_k8s_env(args)
-  storages = get_auto_mount_storages(k8s_api_client)
+  storages = []
+  if not is_dry_run():
+    k8s_api_client = setup_k8s_env(args)
+    storages = get_auto_mount_storages(k8s_api_client)
 
   service_account = ""
   if len(storages) > 0:

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/core/kueue.py

@@ -436,6 +436,8 @@ def install_kueue_crs(
       cluster_hardware_name=cluster_hardware_name,
       resource_type=resource_type,
       total_chips=total_chips,
+      cpu_limit=args.cpu_limit,
+      memory_limit=args.memory_limit,
   )
   topology_label = ''
   if system.device_type in [
@@ -474,7 +476,7 @@ def install_kueue_crs(
     yml_string = topology_yaml + yml_string
 
   tmp = write_tmp_file(yml_string)
-  command = f'kubectl apply -f {str(tmp.file.name)}'
+  command = f'kubectl apply -f {str(tmp)}'
 
   task = 'Applying Kueue Custom Resources'
   return_code = run_command_with_updates_retry(command, task, args)
@@ -484,7 +486,7 @@ def install_kueue_crs(
 
 
 def get_kueue_covered_resources_config(
-    cluster_hardware_name, resource_type, total_chips
+    cluster_hardware_name, resource_type, total_chips, cpu_limit, memory_limit
 ) -> str:
   """Gets Kueue covered resources configuration.
 
@@ -497,17 +499,31 @@ def get_kueue_covered_resources_config(
     A string of Kueue covered resources configuration.
   """
   config_format = """
-  - coveredResources: ["{resource_type}"]
+  - coveredResources: {resource_types}
     flavors:
     - name: {cluster_hardware_name}
       resources:
       - name: "{resource_type}"
-        nominalQuota: {total_chips}
-  """
+        nominalQuota: {total_chips}"""
+  resource_types = [resource_type]
+  if cpu_limit:
+    config_format = config_format + """
+      - name: "cpu"
+        nominalQuota: {cpu_limit}"""
+    resource_types.append('cpu')
+  if memory_limit:
+    config_format = config_format + """
+      - name: "memory"
+        nominalQuota: {memory_limit}"""
+    resource_types.append('memory')
+
   config_string = config_format.format(
       cluster_hardware_name=cluster_hardware_name,
+      resource_types=resource_types,
       resource_type=resource_type,
       total_chips=total_chips,
+      cpu_limit=cpu_limit,
+      memory_limit=memory_limit,
   )
   return config_string
 
@@ -536,7 +552,7 @@ def update_kueue_resources_if_necessary(args):
       memory_limit_size=new_memory_limit, KUEUE_VERSION=KUEUE_VERSION
   )
   tmp = write_tmp_file(yml_string)
-  command = f'kubectl apply -f {str(tmp.file.name)}'
+  command = f'kubectl apply -f {str(tmp)}'
 
   task = 'Updating Kueue Controller Manager resources'
   return_code = run_command_with_updates_retry(command, task, args)

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/core/nap.py

@@ -250,7 +250,7 @@ def create_autoprovisioning_config(
       zones=f'- {args.zone}',
   )
   autoprovisioning_config = AutoprovisioningConfig(
-      config_filename=write_tmp_file(yml_string).name,
+      config_filename=write_tmp_file(yml_string),
       minimum_chips=minimum,
       maximum_chips=maximum,
   )

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/core/network.py

@@ -221,7 +221,7 @@ def create_cluster_network_config(args) -> int:
   """
   yml_string = CLUSTER_NETWORK_YAML.format(cluster_name=args.cluster)
   tmp = write_tmp_file(yml_string)
-  command = f'kubectl apply -f {str(tmp.file.name)}'
+  command = f'kubectl apply -f {str(tmp)}'
 
   return_code = run_command_with_updates(
       command, 'GKE Cluster Create Network Config', args

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/core/nodepool.py

@@ -265,7 +265,9 @@ def run_gke_node_pool_create_command(
       )
       configmap_yml = {}
       configmap_yml[resources_configmap_name] = resources_yml
-      return_code = create_or_update_cluster_configmap(configmap_yml)
+      return_code = create_or_update_cluster_configmap(
+          configmap_yml, args.dry_run
+      )
       if return_code != 0:
         return 1
 
@@ -461,7 +463,7 @@ def get_nodepool_zone(args, nodepool_name) -> tuple[int, str | None]:
       f' --region={zone_to_region(args.zone)} --format="value(locations)"'
   )
   return_code, nodepool_zone = run_command_for_value(
-      command, 'Get Node Pool Zone', args
+      command, 'Get Node Pool Zone', args, dry_run_return_val=args.zone
   )
   if return_code != 0:
     xpk_print(f'Get Node Pool Zone returned ERROR {return_code}')
@@ -570,7 +572,10 @@ def upgrade_gke_nodepools_version(args, default_rapid_gke_version) -> int:
   for i, command in enumerate(commands):
     xpk_print(f'To complete {task_names[i]} we are executing {command}')
   max_return_code = run_commands(
-      commands, 'Update GKE node pools to default RAPID GKE version', task_names
+      commands,
+      'Update GKE node pools to default RAPID GKE version',
+      task_names,
+      dry_run=args.dry_run,
   )
   if max_return_code != 0:
     xpk_print(

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/core/pathways.py

@@ -19,6 +19,7 @@ from ..core.docker_container import get_user_workload_container
 from ..core.gcloud_context import zone_to_region
 from ..core.nodepool import get_all_nodepools_programmatic
 from ..utils.console import xpk_exit, xpk_print
+from ..utils.execution_context import is_dry_run
 from .system_characteristics import AcceleratorType, SystemCharacteristics
 
 
@@ -79,7 +80,10 @@ def ensure_pathways_workload_prerequisites(args, system) -> bool:
   # Ensure the cluster and CPU nodepools were created with create-pathways
   all_node_pools = get_all_nodepools_programmatic(args)
   desired_pw_cpu_node_pools = {'cpu-np'}
-  if not desired_pw_cpu_node_pools.issubset(set(all_node_pools[0])):
+  if (
+      not desired_pw_cpu_node_pools.issubset(set(all_node_pools[0]))
+      and not is_dry_run()
+  ):
     xpk_print(
         'Cluster needs to be created with `xpk create-pathways` to run'
         ' Pathways workloads.'
@@ -322,7 +326,7 @@ def try_to_delete_pathwaysjob_first(args, workloads) -> bool:
     return_code = run_command_with_updates(commands[0], 'Delete Workload', args)
   else:
     return_code = run_commands(
-        commands, 'Delete Workload', task_names, batch=100
+        commands, 'Delete Workload', task_names, batch=100, dry_run=args.dry_run
     )
 
   if return_code != 0:

{xpk-0.12.0 → xpk-0.13.0}/src/xpk/core/ray.py

@@ -132,7 +132,7 @@ def install_ray_cluster(args, system) -> int:
   )
 
   tmp = write_tmp_file(yml_string)
-  command = f'kubectl apply -f {str(tmp.file.name)}'
+  command = f'kubectl apply -f {str(tmp)}'
   task = 'Applying RayCluster'
   retry_attempts = 1
   return_code = run_command_with_updates_retry(