xpk 0.10.1__py3-none-any.whl → 0.12.0__py3-none-any.whl

xpk/commands/cluster.py

@@ -78,6 +78,8 @@ from ..utils.console import get_user_input, xpk_exit, xpk_print
 from ..utils.file import write_tmp_file
 from . import cluster_gcluster
 from .common import set_cluster_command
+import shutil
+import os
 
 
 def cluster_adapt(args) -> None:
@@ -90,7 +92,7 @@ def cluster_adapt(args) -> None:
 
   system, return_code = get_system_characteristics(args)
 
-  if return_code > 0:
+  if return_code > 0 or system is None:
     xpk_print('Fetching system characteristics failed!')
     xpk_exit(return_code)
 
@@ -139,8 +141,6 @@ def cluster_adapt(args) -> None:
     if not tensorboard_config:
       xpk_exit(1)
 
-  # Provision node pools dynamically based on incoming workloads:
-  # Currently autoprovisioning is not supported with Pathways.
   autoprovisioning_config = None
   if args.enable_autoprovisioning:
     xpk_print('Enabling Autoprovisioning')
@@ -199,7 +199,7 @@ def cluster_create(args) -> None:
   """
   system, return_code = get_system_characteristics(args)
 
-  if return_code > 0:
+  if return_code > 0 or system is None:
     xpk_print('Fetching system characteristics failed!')
     xpk_exit(return_code)
 
@@ -215,13 +215,13 @@ def cluster_create(args) -> None:
     xpk_exit(0)
 
   return_code, gke_server_config = get_gke_server_config(args)
-  if return_code != 0:
+  if return_code != 0 or gke_server_config is None:
     xpk_exit(return_code)
 
   return_code, gke_control_plane_version = get_gke_control_plane_version(
       args, gke_server_config
   )
-  if return_code != 0:
+  if return_code != 0 or gke_control_plane_version is None:
     xpk_exit(return_code)
 
   create_cluster_command_code = create_cluster_if_necessary(
@@ -247,6 +247,10 @@ def cluster_create(args) -> None:
 
   get_cluster_credentials(args)
 
+  update_coredns_command_code = update_coredns_if_necessary(args)
+  if update_coredns_command_code != 0:
+    xpk_exit(update_cluster_command_code)
+
   k8s_client = setup_k8s_env(args)
 
   install_storage_crd(k8s_client)
@@ -288,7 +292,7 @@ def cluster_create(args) -> None:
   # Provision node pools dynamically based on incoming workloads:
   # Currently autoprovisioning is not supported with Pathways.
   autoprovisioning_config = None
-  if not args.enable_pathways and args.enable_autoprovisioning:
+  if args.enable_autoprovisioning:
     xpk_print('Enabling Autoprovisioning')
     autoprovisioning_config, return_code = enable_autoprovisioning_on_cluster(
         args, system
@@ -392,7 +396,7 @@ def cluster_cacheimage(args) -> None:
   get_cluster_credentials(args)
   system, return_code = get_system_characteristics(args)
 
-  if return_code > 0:
+  if return_code > 0 or system is None:
     xpk_print('Fetching system characteristics failed!')
     xpk_exit(return_code)
 
@@ -702,6 +706,263 @@ def cluster_create_ray_cluster(args) -> None:
   cluster_create(args)
 
 
+def install_jq(args):
+  """Installs 'jq' utility."""
+  if shutil.which('jq'):
+    xpk_print("Task: 'Install jq' skipped, jq already installed.")
+    return
+  command_jq_install = 'sudo apt install jq -y'
+  xpk_print("Task: 'Install jq' in progress.")
+  return_code = run_command_with_updates(command_jq_install, 'Install jq', args)
+  if return_code != 0:
+    xpk_print(f'Install jq error {return_code}')
+    xpk_exit(return_code)
+
+
+def clone_coredns_deployment_repo(args, coredns_repo_full_path: str):
+  """Clones the CoreDNS deployment repository if it doesn't exist."""
+  if os.path.exists(coredns_repo_full_path):
+    xpk_print(
+        f"Directory '{coredns_repo_full_path}' already exists, skip git clone."
+    )
+    return
+  command_git_clone = (
+      'git clone https://github.com/coredns/deployment.git'
+      f' {coredns_repo_full_path}'
+  )
+  xpk_print(
+      "Task: 'Clone deployment' in progress, Target"
+      f' directory:{coredns_repo_full_path}.'
+  )
+  return_code = run_command_with_updates(
+      command_git_clone, 'Clone deployment', args
+  )
+  if return_code != 0:
+    xpk_print(f'Clone deployment error {return_code}')
+    xpk_exit(return_code)
+
+
+def deploy_coredns_manifests(args, coredns_k8s_path: str):
+  """Deploys CoreDNS manifests to the cluster."""
+  if not os.path.isdir(coredns_k8s_path):
+    xpk_print(
+        f"Error：CoreDNS Kubernetes path '{coredns_k8s_path}' does not exist."
+        ' Has git clone been successful?'
+    )
+    xpk_exit(1)
+  original_cwd = os.getcwd()
+  try:
+    os.chdir(coredns_k8s_path)
+    xpk_print(f'Current working directory changed to: {os.getcwd()}')
+
+    command_deploy_coredns = './deploy.sh | kubectl apply -f -'
+    xpk_print(
+        f"Task: 'Deploy CoreDNS' in progress, Located at '{coredns_k8s_path}'"
+    )
+    return_code = run_command_with_updates(
+        command_deploy_coredns, 'Deploy CoreDNS', args
+    )
+    if return_code != 0:
+      xpk_print(f'Deploy CoreDNS error {return_code}')
+
+  finally:
+    xpk_print(f'Restoring working directory to: {original_cwd}')
+    os.chdir(original_cwd)
+  if return_code != 0:
+    xpk_exit(return_code)
+
+
+def scale_down_deployment(
+    args, deployment_name: str, namespace: str = 'kube-system'
+):
+  """Scales down a specified Kubernetes deployment to 0 replicas."""
+  command = (
+      f'kubectl scale deployment {deployment_name} --replicas=0'
+      f' --namespace={namespace}'
+  )
+  xpk_print(f"Task: 'Scaling down {deployment_name}' in progress")
+  return_code = run_command_with_updates(
+      command, f'Scale down {deployment_name}', args
+  )
+  if return_code != 0:
+    xpk_print(f'Scale down {deployment_name} error {return_code}')
+    xpk_exit(return_code)
+  xpk_print(f'\n{deployment_name} has been scaled down.')
+
+
+def scale_up_coredns(args, replicas: int = 15, namespace: str = 'kube-system'):
+  """Scales up the CoreDNS deployment to a specified number of replicas."""
+  command_coredns_scale = (
+      f'kubectl scale deployment coredns --replicas={replicas} -n {namespace}'
+  )
+  xpk_print(f"Task: 'Scale CoreDNS' in progress (to {replicas} replicas)")
+  return_code = run_command_with_updates(
+      command_coredns_scale, 'Scale CoreDNS', args
+  )
+  if return_code != 0:
+    xpk_print(f'Scale CoreDNS error {return_code}')
+    xpk_exit(return_code)
+
+
+def check_deployment_exists(args, deployment_name: str, namespace: str) -> bool:
+  """Check for the existence of a specific Deployment in a given namespace."""
+  # TODO: rewrite this to be more obvious, check if it is correct
+  command = (
+      f'kubectl get deployment {deployment_name} -n'
+      f' {namespace} --ignore-not-found'
+  )
+  result = run_command_with_updates(
+      command, 'Waiting for kubeDNS to be checked.', args
+  )
+  return result != 0
+
+
+def verify_coredns_readiness(
+    args, timeout: int = 240, namespace: str = 'kube-system'
+):
+  """Verifies CoreDNS readiness using kubectl wait commands."""
+  xpk_print('Now verifying CoreDNS readiness...')
+  kube_dns_exists = check_deployment_exists(args, 'kube-dns', namespace)
+  if kube_dns_exists:
+    # Wait for kube-dns to be fully scaled down
+    command_kube_dns_wait_scaled_down = (
+        'kubectl wait deployment/kube-dns'
+        " --for=jsonpath='{.status.replicas}'=0"
+        f' --namespace={namespace} --timeout={timeout}s'
+    )
+    xpk_print('Verifying if kube-dns has scaled down...')
+    return_code_kube_dns = run_command_with_updates(
+        command_kube_dns_wait_scaled_down, 'Wait for kube-dns scale down', args
+    )
+    if return_code_kube_dns != 0:
+      xpk_print('kube-dns did not scale down successfully within the timeout.')
+      xpk_exit(1)  # Exit if kube-dns cannot scale down
+    else:
+      xpk_print('kube-dns has successfully scaled down.')
+  else:
+    xpk_print('kube-dns deployment not found.')
+  # Wait for CoreDNS to be fully scaled up and available
+  command_coredns_wait_available = (
+      'kubectl wait deployment/coredns --for=condition=Available=true'
+      f' --namespace={namespace} --timeout={timeout}s'
+  )
+  xpk_print('Verifying if CoreDNS is available...')
+  return_code_coredns = run_command_with_updates(
+      command_coredns_wait_available, 'Wait for coredns available', args
+  )
+  if return_code_coredns != 0:
+    xpk_print(
+        'CoreDNS verification failed, it might not have fully started within'
+        ' the timeout.'
+    )
+    xpk_exit(1)  # Exit if coredns cannot become available
+
+  xpk_print('CoreDNS has successfully started and passed verification.')
+
+
+def cleanup_coredns_repo(coredns_repo_full_path: str):
+  """Deletes the cloned CoreDNS deployment directory."""
+  xpk_print(
+      "Task: 'Deleting CoreDNS deployment directory' in progress:"
+      f' {coredns_repo_full_path}'
+  )
+  try:
+    shutil.rmtree(coredns_repo_full_path)
+    xpk_print(f'Successfully deleted directory: {coredns_repo_full_path}')
+  except OSError as e:
+    xpk_print(f'Error deleting directory {coredns_repo_full_path}: {e}')
+
+
+def update_coredns(args) -> int:
+  """Updates and deploys CoreDNS within a cluster.
+
+  Args:
+    args: user provided arguments for running the command.
+
+  Returns:
+    0 if successful and 1 otherwise.
+  """
+  coredns_repo_dir = os.path.expanduser('/tmp/')
+  coredns_repo_dir_name = 'deployment'
+  coredns_repo_full_path = os.path.join(coredns_repo_dir, coredns_repo_dir_name)
+  coredns_k8s_path = os.path.join(coredns_repo_full_path, 'kubernetes')
+  # 1. Install jq
+  install_jq(args)
+
+  # 2. Clone CoreDNS deployment repository
+  clone_coredns_deployment_repo(args, coredns_repo_full_path)
+
+  # 3. Deploy CoreDNS to the cluster
+  deploy_coredns_manifests(args, coredns_k8s_path)
+
+  # 4. Scale down kube-dns-autoscaler
+  scale_down_deployment(args, 'kube-dns-autoscaler')
+
+  # 5. Scale down kube-dns
+  scale_down_deployment(args, 'kube-dns')
+
+  # 6. Scale up coredns and verify readiness
+  scale_up_coredns(args, replicas=15)
+  verify_coredns_readiness(args, timeout=120)
+
+  xpk_print('The CoreDNS setup process has been completed.')
+
+  # 7. Cleanup
+  cleanup_coredns_repo(coredns_repo_full_path)
+
+  return 0
+
+
+def coredns_deployment_exists(args, namespace: str = 'kube-system') -> bool:
+  """Checks if the CoreDNS deployment exists in the given namespace.
+
+  Args:
+    namespace: The Kubernetes namespace to check for the CoreDNS deployment.
+
+  Returns:
+    True if the 'coredns' deployment exists, False otherwise.
+  """
+  command = f'kubectl get deployment coredns -n {namespace}'
+  xpk_print(
+      "Task: 'Checking CoreDNS deployment existence' in progress for"
+      f' namespace: {namespace}'
+  )
+  return_code = run_command_with_updates(
+      command, f'Check CoreDNS deployment in {namespace}', args
+  )
+  if return_code == 0:
+    verify_coredns_readiness(args)
+    xpk_print(f"CoreDNS deployment 'coredns' found in namespace '{namespace}'.")
+    return True
+  else:
+    xpk_print(
+        f"CoreDNS deployment 'coredns' NOT found in namespace '{namespace}' or"
+        ' an error occurred.'
+    )
+    return False
+
+
+def update_coredns_if_necessary(args) -> int:
+  """Updates and deploys CoreDNS within the cluster if it's not already present.
+
+  This function checks for the existence of the CoreDNS deployment.
+  If it's not found, it proceeds to deploy and configure CoreDNS.
+
+  Args:
+    args: User-provided arguments for running the command.
+
+  Returns:
+    0 if successful (CoreDNS was already present or successfully deployed),
+    and 1 otherwise.
+  """
+  if coredns_deployment_exists(args, namespace='kube-system'):
+    xpk_print('Skipping CoreDNS deployment since it already exists.')
+    return 0
+  else:
+    xpk_print('CoreDNS deployment not found. Proceeding with CoreDNS setup.')
+    return update_coredns(args)
+
+
 def create_cluster_if_necessary(
     args, gke_control_plane_version: str, system: SystemCharacteristics
 ) -> int:
@@ -842,6 +1103,7 @@ def run_gke_cluster_create_command(
       f' {args.custom_cluster_arguments}'
       f' {rapid_release_cmd}'
       ' --enable-dns-access'
+      ' --autoscaling-profile=optimize-utilization'
   )
 
   enable_ip_alias = False

xpk/commands/cluster_gcluster.py

@@ -310,4 +310,5 @@ def generate_blueprint(
           system_node_pool_machine_type=args.default_pool_cpu_machine_type,
           system_node_pool_min_node_count=args.default_pool_cpu_num_nodes,
       )
-  return None
+  xpk_print('Device type is not supported.')
+  xpk_exit(1)

xpk/commands/common.py

@@ -50,8 +50,8 @@ def set_cluster_command(args) -> int:
 
 
 def is_TAS_possible(
-    system_characteristics: SystemCharacteristics,
-    capacity_type: CapacityType,
+    system_characteristics: SystemCharacteristics | None,
+    capacity_type: CapacityType | None,
     flex: bool,
 ) -> bool:
   """Check cluster's machine_type and capacity type to determine if Kueue TAS is possible
@@ -71,7 +71,7 @@ def is_TAS_possible(
     xpk_print('capacity_type data was not found in configmaps.')
     xpk_exit(1)
 
-  if flex:
+  if not flex:
     return False
 
   if (

xpk/commands/info.py

@@ -51,19 +51,19 @@ def info(args: Namespace) -> None:
   cqs = run_kueuectl_list_clusterqueue(args)
   quotas = get_nominal_quotas(cqs)
 
-  if lq:
+  if lq and lqs is not None:
     print_formatted_lqs(lqs, quotas)
 
   if cq:
     print_formatted_cqs(cqs, quotas)
 
 
-def get_nominal_quotas(cqs: list[dict]) -> dict[str, dict[str, str]]:
+def get_nominal_quotas(cqs: str) -> dict[str, dict[str, str]]:
   """Get quotas from clusterqueues.
   This function retrieves how much of resource in each flavor is assigned to cluster queue.
   It parses flavors of passed cluster queues.
   Args:
-    - cqs - list of cluster queues.
+    - cqs - string containing a list of cluster queues in JSON format.
   Returns:
     - dictionary of cluster queues resources quotas in format:
     {cq_name:{"flavorName:resourceName":quota}}
@@ -75,7 +75,7 @@ def get_nominal_quotas(cqs: list[dict]) -> dict[str, dict[str, str]]:
     xpk_print(cqs)
     xpk_exit(1)
 
-  quotas = {}
+  quotas: dict[str, dict] = {}
   for cq in cq_list:
     spec = cq['spec']
     cq_name = cq['metadata']['name']
@@ -89,7 +89,7 @@ def get_nominal_quotas(cqs: list[dict]) -> dict[str, dict[str, str]]:
   return quotas
 
 
-def print_formatted_cqs(cqs: list[dict], nominalQuotas) -> None:
+def print_formatted_cqs(cqs: str, nominalQuotas) -> None:
   try:
     cq_list = json.loads(cqs)['items']
   except ValueError:
@@ -105,7 +105,7 @@ def print_formatted_cqs(cqs: list[dict], nominalQuotas) -> None:
   )
 
 
-def print_formatted_lqs(lqs: list[dict], nominalQuotas) -> None:
+def print_formatted_lqs(lqs: str, nominalQuotas) -> None:
   try:
     lq_list = json.loads(lqs)['items']
   except ValueError:
@@ -143,18 +143,18 @@ def parse_queue_lists(
 
 
 def get_flavors_resources_reservations(
-    cq_name: str, flavors_res: list[dict]
+    cq_name: str, flavors_res: dict
 ) -> dict[str, dict[str, str]]:
   """Get usage of flavors resources.
   This function parser flavorsReservation section of clusterQueue of LocalQueue.
   Args:
     - cq_name - name of ClusterQueue to which flavors belong.
-    - flavors_res - list of reservations made by flavors
+    - flavors_res - dict of reservations made by flavors
   Returns:
     Dict containing usage of each resource in flavor for each flavor in cluster or local queue.
     Dict format: {cq_name: {{flavor:resource}:reservation}}
   """
-  reservations = {}
+  reservations: dict[str, dict] = {}
   reservations[cq_name] = {}
   for flavor_name, flavor_resources_reservation_list in flavors_res.items():
     for resource in flavor_resources_reservation_list:
@@ -167,15 +167,15 @@ def get_flavors_resources_reservations(
 
 def get_flavors_usage(
     q_entry: dict, res_field: str, flavor_resource_quotas: dict
-) -> list[dict]:
+) -> dict[str, str]:
   """Parse q_entry to retrieve list of each resource usage in flavour.
   Args:
     q_entry - single entry into either LocalQueue or ClusterQueue structured as json
     flavor_resource_quotas - nominalQuota of flavors resource usage for each clusterqueue
   Returns:
-    list of dicts where each list entry is in format (key, entry) where:
+    Dict where for each (key, value):
     - key is flavorName:resourceName
-    - entry is flavorResourceReservation/flavorResourceQuota
+    - value is string formatted as 'flavorResourceReservation/flavorResourceQuota'
   """
   status = q_entry['status']
   flavors_res = status[res_field]

xpk/commands/job.py

@@ -18,6 +18,7 @@ import re
 import sys
 
 from ruamel.yaml import YAML
+from typing import cast
 
 from ..core.commands import run_command_for_value, run_command_with_updates
 from ..core.cluster import get_cluster_credentials
@@ -84,7 +85,7 @@ def job_info(args):
 
 
 def get_profile(job_yaml: dict) -> str:
-  containers = (
+  containers: list[dict] = (
       job_yaml.get('spec', {})
       .get('template', {})
       .get('spec', {})
@@ -96,13 +97,13 @@ def get_profile(job_yaml: dict) -> str:
 
 
 def get_mounts(job_yaml: dict) -> list[dict]:
-  containers = (
+  containers: list[dict] = (
       job_yaml.get('spec', {})
       .get('template', {})
       .get('spec', {})
       .get('containers', [])
   )
-  mounts = next(iter(containers), {}).get('volumeMounts', [])
+  mounts: list[dict] = next(iter(containers), {}).get('volumeMounts', [])
   return mounts
 
 
@@ -112,23 +113,24 @@ def get_kjob_env_vars(job_desc_text: str) -> list[tuple[str, str]]:
   return search_res
 
 
-def get_pods(pods_text: str) -> list[str]:
+def get_pods(pods_text: str) -> list[dict[str, str]]:
   pods_lines = pods_text.strip().split('\n')
-  pods_lines = [line.split() for line in pods_lines]
+  pods_lines_tokenized = [line.split() for line in pods_lines]
   return [
       {
-          'Name': line[0],
-          'Status': line[2],
+          'Name': tokens[0],
+          'Status': tokens[2],
       }
-      for line in pods_lines
+      for tokens in pods_lines_tokenized
   ]
 
 
 def get_script_name(job_yaml: dict) -> str | None:
-  return (
+  return cast(
+      str | None,
       job_yaml.get('metadata', {})
       .get('annotations', {})
-      .get('kjobctl.x-k8s.io/script', '')
+      .get('kjobctl.x-k8s.io/script', ''),
   )
 
 

xpk/commands/kjob_common.py

@@ -33,6 +33,7 @@ from ..core.resources import get_cluster_capacity_type, get_cluster_system_chara
 def add_gpu_networking_annotations_to_command(args, cmd: str) -> str:
   gpu_type = get_gpu_type_from_cluster(args)
 
+  annotations: tuple
   if gpu_type == H100_MEGA_DEVICE_TYPE:
     annotations = get_a3mega_pod_template_annotations(args)
   elif gpu_type == H200_DEVICE_TYPE:
@@ -40,7 +41,7 @@ def add_gpu_networking_annotations_to_command(args, cmd: str) -> str:
   elif gpu_type == B200_DEVICE_TYPE:
     annotations = get_a4_pod_template_annotations(args)
   else:
-    annotations = []
+    annotations = tuple()
 
   flags = [
       f" --pod-template-annotation {annotation} " for annotation in annotations

xpk/commands/storage.py

@@ -141,7 +141,7 @@ def storage_delete(args: Namespace) -> None:
 
 def storage_attach(args: Namespace) -> None:
   add_zone_and_project(args)
-  manifest = [{}]
+  manifest: list[dict] = [{}]
   if args.type == GCP_FILESTORE_TYPE:
     if args.instance is None:
       args.instance = args.name

xpk/commands/workload.py

@@ -84,6 +84,7 @@ from ..core.system_characteristics import (
 from ..core.vertex import create_vertex_experiment
 from ..core.workload import (
     check_if_workload_exists,
+    get_jobsets_list_gcp_link,
     get_workload_list,
     wait_for_job_completion,
     zone_to_region,
@@ -226,7 +227,8 @@ spec:
             metadata:
               labels:
                 xpk.google.com/workload: {args.workload}
-              annotations: {annotations}
+              annotations:
+                {annotations}
             spec:
               priorityClassName: {args.priority}
               restartPolicy: Never
@@ -319,7 +321,7 @@ def workload_create(args) -> None:
   xpk_print('Starting workload create', flush=True)
   system, return_code = get_system_characteristics(args)
 
-  if return_code > 0:
+  if return_code > 0 or system is None:
     xpk_print('Fetching system characteristics failed!')
     xpk_exit(return_code)
 
@@ -345,7 +347,7 @@ def workload_create(args) -> None:
   ):
     xpk_print(
         'Warning: Cluster has been created using XPK version:'
-        f' {cluster_config_map["xpk_version"]} but the XPK version you are'
+        f' {cluster_xpk_version} but the XPK version you are'
         f' using to schedule workload is: {XPK_CURRENT_VERSION}. Some features'
         ' might not be available for this cluster. We recommend to'
         ' upgrade/downgrade your XPK version or cluster by running `xpk'
@@ -354,7 +356,7 @@ def workload_create(args) -> None:
 
   debugging_dashboard_id = None
 
-  tensorboard_config = {}
+  tensorboard_config: dict | None = {}
   if VERTEX_TENSORBOARD_FEATURE_FLAG and args.use_vertex_tensorboard:
     tensorboard_config = create_vertex_experiment(args)
     # exit if failed to create Experiment in Vertex AI
@@ -450,8 +452,8 @@ def workload_create(args) -> None:
       - action: FailJobSet
         onJobFailureReasons:
         - PodFailurePolicy"""
-    restart_on_exit_codes = get_restart_exit_codes(args)
-    restart_on_exit_codes = ','.join(map(str, restart_on_exit_codes))
+    restart_on_exit_codes_list = get_restart_exit_codes(args)
+    restart_on_exit_codes = ','.join(map(str, restart_on_exit_codes_list))
     pod_failure_policy = f"""
           podFailurePolicy:
             rules:
@@ -760,4 +762,8 @@ def workload_list(args) -> None:
     xpk_print(f'List Job request returned ERROR {return_code}')
     xpk_exit(return_code)
   xpk_print(f'Workload List Output:\n{return_value}')
+
+  workload_list_gcp_link = get_jobsets_list_gcp_link(project=args.project)
+  xpk_print(f'See your workloads in Cloud Console: {workload_list_gcp_link}')
+
   xpk_exit(0)

xpk/core/blueprint/blueprint_generator.py

@@ -34,7 +34,7 @@ from ..system_characteristics import get_system_characteristics_by_device_type
 from .blueprint_definitions import Blueprint, DeploymentGroup, DeploymentModule
 from ..kueue import KUEUE_VERSION
 
-yaml = yaml.YAML()
+yaml_parser = yaml.YAML()
 
 a3high_device_type = H100_DEVICE_TYPE
 a3mega_device_type = H100_MEGA_DEVICE_TYPE
@@ -52,7 +52,7 @@ blueprint_dependencies_dir = {
 }
 
 cluster_toolkit_url = "github.com/GoogleCloudPlatform/cluster-toolkit"
-cluster_toolkit_version = "v1.57.1"
+cluster_toolkit_version = "v1.62.2"
 
 
 class BlueprintGeneratorOutput:
@@ -1019,7 +1019,7 @@ class BlueprintGenerator:
   ) -> str:
     blueprint_path = self._get_blueprint_path(blueprint_name, prefix)
     with open(blueprint_path, "w+", encoding="utf-8") as blueprint_file:
-      yaml.dump(xpk_blueprint, blueprint_file)
+      yaml_parser.dump(xpk_blueprint, blueprint_file)
     return blueprint_path
 
   def _get_blueprint_path(self, blueprint_name, prefix: str = ""):
@@ -1033,7 +1033,7 @@ class BlueprintGenerator:
     ensure_directory_exists(storage_path_with_prefix)
     return storage_path_with_prefix
 
-  def blueprint_exists(self, blueprint_name, prefix: str = ""):
+  def blueprint_exists(self, blueprint_name, prefix: str = "") -> bool:
     blueprint_path = self._get_blueprint_path(blueprint_name, prefix)
     return os.path.exists(blueprint_path)
 
@@ -1061,6 +1061,6 @@ class BlueprintGenerator:
     }
 
 
-yaml.register_class(Blueprint)
-yaml.register_class(DeploymentGroup)
-yaml.register_class(DeploymentModule)
+yaml_parser.register_class(Blueprint)
+yaml_parser.register_class(DeploymentGroup)
+yaml_parser.register_class(DeploymentModule)