konduktor-nightly 0.1.0.dev20250821104804__tar.gz → 0.1.0.dev20250823104438__tar.gz

{konduktor_nightly-0.1.0.dev20250821104804 → konduktor_nightly-0.1.0.dev20250823104438}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: konduktor-nightly
-Version: 0.1.0.dev20250821104804
+Version: 0.1.0.dev20250823104438
 Summary: GPU Cluster Health Management
 Author: Andrew Aikawa
 Author-email: asai@berkeley.edu

{konduktor_nightly-0.1.0.dev20250821104804 → konduktor_nightly-0.1.0.dev20250823104438}/konduktor/__init__.py

@@ -11,7 +11,7 @@ from konduktor.task import Task
 __all__ = ['launch', 'Resources', 'Task', 'Serving']
 
 # Replaced with the current commit when building the wheels.
-_KONDUKTOR_COMMIT_SHA = 'eee38d922bf4c7cb8a2e6e730092dde0ae372500'
+_KONDUKTOR_COMMIT_SHA = 'face26aca22b99192e740dca4875261fcffa2a55'
 os.makedirs(os.path.expanduser('~/.konduktor'), exist_ok=True)
 
 
@@ -45,5 +45,5 @@ def _get_git_commit():
 
 
 __commit__ = _get_git_commit()
-__version__ = '1.0.0.dev0.1.0.dev20250821104804'
+__version__ = '1.0.0.dev0.1.0.dev20250823104438'
 __root_dir__ = os.path.dirname(os.path.abspath(__file__))

{konduktor_nightly-0.1.0.dev20250821104804 → konduktor_nightly-0.1.0.dev20250823104438}/konduktor/backends/deployment_utils.py

@@ -47,8 +47,6 @@ _DEPLOYMENT_METADATA_LABELS = {
     'model_name_label': AIBRIX_NAME_LABEL,
 }
 
-_RUN_DURATION_ANNOTATION_KEY = 'kueue.x-k8s.io/maxRunDurationSeconds'
-
 
 def render_specs(
     task: 'konduktor.Task',
@@ -781,7 +779,12 @@ def show_status_table(namespace: str, all_users: bool):
             )
             if port_obj and port_obj.port:
                 port_str = str(port_obj.port)
-        endpoint_str = f'{ip_str}:{port_str}' if port_str else ip_str
+
+        # For vLLM deployments, don't append port since external routing is on port 80
+        if AIBRIX_NAME_LABEL in labels:
+            endpoint_str = ip_str
+        else:
+            endpoint_str = f'{ip_str}:{port_str}' if port_str else ip_str
 
         # Replicas
         ready_replicas = (

{konduktor_nightly-0.1.0.dev20250821104804 → konduktor_nightly-0.1.0.dev20250823104438}/konduktor/backends/pod_utils.py

@@ -329,15 +329,6 @@ def inject_deployment_pod_metadata(
     if task.resources and task.resources.labels:
         pod_spec['metadata']['labels'].update(task.resources.labels)
 
-    # Add max run duration annotation
-    assert task.resources is not None and task.resources.labels is not None
-    maxRunDurationSeconds = task.resources.labels.get('maxRunDurationSeconds', None)
-    if not maxRunDurationSeconds:
-        raise ValueError('maxRunDurationSeconds is required')
-    pod_spec['metadata']['annotations'][_RUN_DURATION_ANNOTATION_KEY] = str(
-        maxRunDurationSeconds
-    )
-
     # Set restart policy for deployments
     pod_spec.setdefault('spec', {})
     pod_spec['spec']['restartPolicy'] = 'Always'

{konduktor_nightly-0.1.0.dev20250821104804 → konduktor_nightly-0.1.0.dev20250823104438}/konduktor/cli.py

@@ -104,6 +104,7 @@ def _make_task_with_overrides(
     instance_type: Optional[str] = None,
     num_nodes: Optional[int] = None,
     max_restarts: Optional[int] = None,
+    completions: Optional[int] = None,
     image_id: Optional[str] = None,
     disk_size: Optional[int] = None,
     env: Optional[List[Tuple[str, str]]] = None,
@@ -166,7 +167,9 @@ def _make_task_with_overrides(
     if max_restarts is not None:
         assert task.resources is not None
         task.resources.job_config['max_restarts'] = max_restarts
-
+    if completions is not None:
+        assert task.resources is not None
+        task.resources.job_config['completions'] = completions
     if num_nodes is not None:
         task.num_nodes = num_nodes
     if name is not None:
@@ -215,6 +218,16 @@ _TASK_OPTIONS = [
             'supplied.'
         ),
     ),
+    click.option(
+        '--completions',
+        required=False,
+        type=int,
+        help=(
+            'Number of successful completions required. Overrides YAML.'
+            'Overrides the "completions" config in the YAML if both are '
+            'supplied.'
+        ),
+    ),
     click.option(
         '--cpus',
         default=None,
@@ -798,6 +811,7 @@ def launch(
     memory: Optional[str],
     num_nodes: Optional[int],
     max_restarts: Optional[int],
+    completions: Optional[int],
     image_id: Optional[str],
     env_file: Optional[Dict[str, str]],
     env: List[Tuple[str, str]],
@@ -822,6 +836,7 @@ def launch(
         memory=memory,
         num_nodes=num_nodes,
         max_restarts=max_restarts,
+        completions=completions,
         image_id=image_id,
         env=env,
         disk_size=disk_size,
@@ -1686,6 +1701,7 @@ def serve_launch(
     memory: Optional[str],
     num_nodes: Optional[int],
     max_restarts: Optional[int],
+    completions: Optional[int],
     image_id: Optional[str],
     env_file: Optional[Dict[str, str]],
     env: List[Tuple[str, str]],
@@ -1714,6 +1730,7 @@ def serve_launch(
         memory=memory,
         num_nodes=num_nodes,
         max_restarts=max_restarts,
+        completions=completions,
         image_id=image_id,
         env=env,
         disk_size=disk_size,

{konduktor_nightly-0.1.0.dev20250821104804 → konduktor_nightly-0.1.0.dev20250823104438}/konduktor/resource.py

@@ -387,7 +387,13 @@ class Resources:
 
     def get_completions(self) -> Optional[int]:
         value = self.job_config.get('completions')
-        return int(value) if value is not None else None
+        if value is not None:
+            value = int(value)
+            if value <= 0:
+                with ux_utils.print_exception_no_traceback():
+                    raise ValueError('completions must be a positive integer')
+            return value
+        return None
 
     def get_max_restarts(self) -> Optional[int]:
         value = self.job_config.get('max_restarts')

{konduktor_nightly-0.1.0.dev20250821104804 → konduktor_nightly-0.1.0.dev20250823104438}/konduktor/templates/jobset.yaml.j2

@@ -19,7 +19,7 @@ jobset:
     annotations: {}
   spec:
     ttlSecondsAfterFinished: 31536000 # 1 year (365 days)
-    {% if max_restarts %}
+    {% if max_restarts is not none %}
     failurePolicy:
       maxRestarts: {{ max_restarts }}
     {% endif %}
@@ -29,7 +29,11 @@ jobset:
         spec:
           ttlSecondsAfterFinished: 600 # 5 minutes
           parallelism: {{ num_nodes }}
+          {% if completions %}
+          completions: {{ completions }}
+          {% else %}
           completions: {{ num_nodes }}
+          {% endif %}
           backoffLimit: 0
           template: {}
           podFailurePolicy:

{konduktor_nightly-0.1.0.dev20250821104804 → konduktor_nightly-0.1.0.dev20250823104438}/konduktor/utils/schemas.py

@@ -467,7 +467,8 @@ def get_job_schema():
         'additionalProperties': False,
         'properties': {
             'completions': {
-                'type': 'number',
+                'type': 'integer',
+                'minimum': 1,
             },
             'max_restarts': {
                 'type': 'integer',

{konduktor_nightly-0.1.0.dev20250821104804 → konduktor_nightly-0.1.0.dev20250823104438}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "konduktor-nightly"
-version = "0.1.0.dev20250821104804"
+version = "0.1.0.dev20250823104438"
 description = "GPU Cluster Health Management"
 packages = [
     {include = "konduktor"}