konduktor-nightly 0.1.0.dev20250810104857__tar.gz → 0.1.0.dev20250811105223__tar.gz

{konduktor_nightly-0.1.0.dev20250810104857 → konduktor_nightly-0.1.0.dev20250811105223}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: konduktor-nightly
-Version: 0.1.0.dev20250810104857
+Version: 0.1.0.dev20250811105223
 Summary: GPU Cluster Health Management
 Author: Andrew Aikawa
 Author-email: asai@berkeley.edu

{konduktor_nightly-0.1.0.dev20250810104857 → konduktor_nightly-0.1.0.dev20250811105223}/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 = '0f0b36c3a67aa7c60d6cb33240631b7c8ccaed03'
+_KONDUKTOR_COMMIT_SHA = '92fe69bd3f29e7b191de663c598dfcf10738f87a'
 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.dev20250810104857'
+__version__ = '1.0.0.dev0.1.0.dev20250811105223'
 __root_dir__ = os.path.dirname(os.path.abspath(__file__))

{konduktor_nightly-0.1.0.dev20250810104857 → konduktor_nightly-0.1.0.dev20250811105223}/konduktor/backends/jobset_utils.py

@@ -232,8 +232,12 @@ def stop_jobset(namespace: str, job_name: str) -> Optional[Dict[str, Any]]:
             'spec': {'suspend': True},
             'metadata': {
                 'annotations': {
-                    backend_constants.STOP_USERID_LABEL: common_utils.user_and_hostname_hash(),
-                    backend_constants.STOP_USERNAME_LABEL: common_utils.get_cleaned_username(),
+                    backend_constants.STOP_USERID_LABEL: (
+                        common_utils.user_and_hostname_hash()
+                    ),
+                    backend_constants.STOP_USERNAME_LABEL: (
+                        common_utils.get_cleaned_username()
+                    ),
                 }
             },
         }

{konduktor_nightly-0.1.0.dev20250810104857 → konduktor_nightly-0.1.0.dev20250811105223}/konduktor/cli.py

@@ -855,16 +855,23 @@ def launch(
             'Serving information detected. Use '
             '`konduktor serve launch` instead for serving.'
         )
+    try:
+        _launch_with_confirm(
+            task,
+            dryrun=dryrun,
+            detach_run=detach_run,
+            no_confirm=yes,
+            serving=bool(task.serving),
+        )
+    except KeyboardInterrupt:
+        click.secho(
+            f'Detaching... manage your job {task.name} with the following commands:',
+            fg='yellow',
+            bold=True,
+        )
 
-    job_name = _launch_with_confirm(
-        task,
-        dryrun=dryrun,
-        detach_run=detach_run,
-        no_confirm=yes,
-        serving=bool(task.serving),
-    )
     click.secho(
-        ux_utils.command_hint_messages(ux_utils.CommandHintType.JOB, job_name),
+        ux_utils.command_hint_messages(ux_utils.CommandHintType.JOB, task.name),
         fg='green',
         bold=True,
     )
@@ -1139,6 +1146,12 @@ def stop(
     for job in track(jobs, description='Suspending job(s)...'):
         jobset_utils.stop_jobset(namespace, job)
 
+    click.secho(
+        ux_utils.command_hint_messages(ux_utils.CommandHintType.JOB_STOP, jobs),
+        fg='green',
+        bold=True,
+    )
+
 
 @cli.command(cls=_DocumentedCodeCommand)
 @click.argument(
@@ -1836,7 +1849,7 @@ def main():
         return cli(standalone_mode=False)
     except click.exceptions.Abort:
         click.secho('Detaching...', fg='yellow', bold=True)
-        return
+        return None
 
 
 if __name__ == '__main__':

{konduktor_nightly-0.1.0.dev20250810104857 → konduktor_nightly-0.1.0.dev20250811105223}/konduktor/logging.py

@@ -75,12 +75,14 @@ def get_logger(name: str):
     fh.setFormatter(FORMATTER)
     logger.addHandler(fh)
 
-    # --- Console logging: DEBUG level only if KONDUKTOR_DEBUG=1 ---
+    # --- Console logging: INFO level by default, DEBUG if KONDUKTOR_DEBUG=1 ---
+    ch = logging.StreamHandler()
     if os.environ.get('KONDUKTOR_DEBUG') == '1':
-        ch = logging.StreamHandler()
         ch.setLevel(logging.DEBUG)
-        ch.setFormatter(FORMATTER)
-        logger.addHandler(ch)
+    else:
+        ch.setLevel(logging.INFO)
+    ch.setFormatter(FORMATTER)
+    logger.addHandler(ch)
 
     logger.propagate = False
     return logger

{konduktor_nightly-0.1.0.dev20250810104857 → konduktor_nightly-0.1.0.dev20250811105223}/konduktor/utils/ux_utils.py

@@ -6,7 +6,7 @@ import os
 import sys
 import traceback
 import typing
-from typing import Callable, Optional, Union
+from typing import Callable, List, Optional, Union
 
 import colorama
 import rich.console as rich_console
@@ -196,27 +196,41 @@ def spinner_message(
 
 class CommandHintType(enum.Enum):
     JOB = 'JOB'
+    JOB_STOP = 'JOB_STOP'
 
 
-def command_hint_messages(hint_type: CommandHintType, job_id: str) -> str:
+def command_hint_messages(
+    hint_type: CommandHintType,
+    job_id: Union[str, List[str]],
+) -> str:
     """Gets the command hint messages for the given job id."""
+    hint_str = '\n📋 Useful Commands'
     if hint_type == CommandHintType.JOB:
         job_hint_str = (
             f'\nJob ID: {job_id}'
-            f'\n{INDENT_SYMBOL}To cancel the job:\t\t'
-            f'{BOLD}konduktor down {job_id} {RESET_BOLD}'
             f'\n{INDENT_SYMBOL}To stream job logs:\t\t'
             f'{BOLD}konduktor logs {job_id} {RESET_BOLD}'
             f'\n{INDENT_SYMBOL}To list all jobs:\t\t'
             f'{BOLD}konduktor status{RESET_BOLD}'
+            f'\n{INDENT_SYMBOL}To suspend the job:\t\t'
+            f'{BOLD}konduktor stop {job_id} {RESET_BOLD}'
+            f'\n{INDENT_SYMBOL}{colorama.Fore.RED}To delete the job:\t\t'
+            f'{BOLD}konduktor down {job_id} {RESET_BOLD}{colorama.Style.RESET_ALL}'
         )
-        hint_str = '\n📋 Useful Commands'
         hint_str += f'{job_hint_str}'
-        if config.get_nested(('tailscale', 'secret_name'), None) is not None:
-            hint_str += (
-                f'\n{INDENT_SYMBOL}To tailscale ssh:\t\t'
-                f'{BOLD}ssh root@{job_id}-workers-0-0 {RESET_BOLD}'
-            )
-        return hint_str
+    elif hint_type == CommandHintType.JOB_STOP:
+        assert isinstance(job_id, list), 'job_id must be a list of strings'
+        job_ids_str = ' '.join(job_id)
+        hint_str += (
+            f'\n{INDENT_SYMBOL}To resume the following jobs:\t\t'
+            f'{BOLD}konduktor start {job_ids_str} {RESET_BOLD}'
+        )
     else:
         raise ValueError(f'Invalid hint type: {hint_type}')
+
+    if config.get_nested(('tailscale', 'secret_name'), None) is not None:
+        hint_str += (
+            f'\n{INDENT_SYMBOL}To tailscale ssh:\t\t'
+            f'{BOLD}ssh root@{job_id}-workers-0-0 {RESET_BOLD}'
+        )
+    return hint_str

{konduktor_nightly-0.1.0.dev20250810104857 → konduktor_nightly-0.1.0.dev20250811105223}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "konduktor-nightly"
-version = "0.1.0.dev20250810104857"
+version = "0.1.0.dev20250811105223"
 description = "GPU Cluster Health Management"
 packages = [
     {include = "konduktor"}