konduktor-nightly 0.1.0.dev20250806105405__tar.gz → 0.1.0.dev20250807224131__tar.gz

{konduktor_nightly-0.1.0.dev20250806105405 → konduktor_nightly-0.1.0.dev20250807224131}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: konduktor-nightly
-Version: 0.1.0.dev20250806105405
+Version: 0.1.0.dev20250807224131
 Summary: GPU Cluster Health Management
 Author: Andrew Aikawa
 Author-email: asai@berkeley.edu

{konduktor_nightly-0.1.0.dev20250806105405 → konduktor_nightly-0.1.0.dev20250807224131}/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 = '75e0a60af7c427a1419fda58e1ccc6dff519b4ba'
+_KONDUKTOR_COMMIT_SHA = '45add2e516f1b7bb1f16ed063e5af87b8e5609cf'
 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.dev20250806105405'
+__version__ = '1.0.0.dev0.1.0.dev20250807224131'
 __root_dir__ = os.path.dirname(os.path.abspath(__file__))

{konduktor_nightly-0.1.0.dev20250806105405 → konduktor_nightly-0.1.0.dev20250807224131}/konduktor/cli.py

@@ -653,6 +653,14 @@ def status(all_users: bool):
         'Default is 1000.'
     ),
 )
+@click.option(
+    '--node-rank',
+    '--node_rank',
+    '-N',
+    default=0,
+    type=int,
+    help='The node rank to tail logs from.',
+)
 @click.argument('job_id', type=str, nargs=1)
 # TODO(zhwu): support logs by job name
 def logs(
@@ -660,6 +668,7 @@ def logs(
     job_id: str,
     follow: bool,
     num_lines: int,
+    node_rank: int,
 ):
     # NOTE(dev): Keep the docstring consistent between the Python API and CLI.
     """Retrieve/tail the log of a job."""
@@ -689,7 +698,12 @@ def logs(
         'Logs are tailed from 1 hour ago, ' 'to see more logs, check Grafana.',
         fg='yellow',
     )
-    log_utils.tail_logs(job_id, follow=follow, num_logs=num_lines)
+    log_utils.tail_logs(
+        job_id,
+        worker_id=node_rank,
+        follow=follow,
+        num_logs=num_lines,
+    )
 
 
 @cli.command(cls=_DocumentedCodeCommand)
@@ -1483,7 +1497,11 @@ def serve_status(all_users: bool):
 
 
 def main():
-    return cli()
+    try:
+        return cli(standalone_mode=False)
+    except click.exceptions.Abort:
+        click.secho('Detaching...', fg='yellow', bold=True)
+        return
 
 
 if __name__ == '__main__':

{konduktor_nightly-0.1.0.dev20250806105405 → konduktor_nightly-0.1.0.dev20250807224131}/konduktor/utils/log_utils.py

@@ -30,7 +30,7 @@ import requests
 import websockets
 
 from konduktor import config, logging
-from konduktor.utils import subprocess_utils
+from konduktor.utils import kubernetes_utils, subprocess_utils
 
 logger = logging.get_logger(__name__)
 
@@ -337,8 +337,13 @@ def tail_loki_logs_ws(
 
 
 def tail_vicky_logs(
-    job_name: str, worker_id: int = 0, num_logs: int = 1000, follow: bool = True
+    job_name: str,
+    worker_id: int = 0,
+    num_logs: int = 1000,
+    follow: bool = True,
 ):
+    context = kubernetes_utils.get_current_kube_config_context_name()
+    namespace = kubernetes_utils.get_kube_config_context_namespace(context)
     query: Dict[str, Any] = {}
     if num_logs > 5000:
         # TODO(asaiacai): we should not have a limit on the number of logs, but rather
@@ -365,8 +370,9 @@ def tail_vicky_logs(
         logger.debug(f'Vicky URL: {vicky_url}')
 
         query['query'] = (
-            'k8s.namespace.name: "default" AND '
-            f'batch.kubernetes.io/job-name: "{job_name}-workers-{worker_id}"'
+            f'k8s.namespace.name: "{namespace}" AND '
+            f'batch.kubernetes.io/job-name: "{job_name}-workers-0" AND '
+            f'batch.kubernetes.io/job-completion-index: "{worker_id}"'
         )
         query['start_offset'] = '1h'
 
@@ -406,7 +412,10 @@ def tail_vicky_logs(
 
 
 def tail_logs(
-    job_name: str, worker_id: int = 0, num_logs: int = 1000, follow: bool = True
+    job_name: str,
+    worker_id: int = 0,
+    num_logs: int = 1000,
+    follow: bool = True,
 ):
     logs_backend = config.get_nested(('logs', 'backend'), None)
     if logs_backend == LogBackend.VICTORIA:

{konduktor_nightly-0.1.0.dev20250806105405 → konduktor_nightly-0.1.0.dev20250807224131}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "konduktor-nightly"
-version = "0.1.0.dev20250806105405"
+version = "0.1.0.dev20250807224131"
 description = "GPU Cluster Health Management"
 packages = [
     {include = "konduktor"}