skypilot-nightly 1.0.0.dev20240902__tar.gz → 1.0.0.dev20240903__tar.gz

{skypilot_nightly-1.0.0.dev20240902/skypilot_nightly.egg-info → skypilot_nightly-1.0.0.dev20240903}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: skypilot-nightly
-Version: 1.0.0.dev20240902
+Version: 1.0.0.dev20240903
 Summary: SkyPilot: An intercloud broker for the clouds
 Author: SkyPilot Team
 License: Apache 2.0

{skypilot_nightly-1.0.0.dev20240902 → skypilot_nightly-1.0.0.dev20240903}/sky/__init__.py

@@ -5,7 +5,7 @@ from typing import Optional
 import urllib.request
 
 # Replaced with the current commit when building the wheels.
-_SKYPILOT_COMMIT_SHA = '2e204d001b037cd4ff84abbaa4f2c81d3442ba22'
+_SKYPILOT_COMMIT_SHA = '0203971a36dbcf0a6d3615e6ed25f3f6d11b53f6'
 
 
 def _get_git_commit():
@@ -35,7 +35,7 @@ def _get_git_commit():
 
 
 __commit__ = _get_git_commit()
-__version__ = '1.0.0.dev20240902'
+__version__ = '1.0.0.dev20240903'
 __root_dir__ = os.path.dirname(os.path.abspath(__file__))
 
 

{skypilot_nightly-1.0.0.dev20240902 → skypilot_nightly-1.0.0.dev20240903}/sky/backends/cloud_vm_ray_backend.py

@@ -3112,7 +3112,8 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
             setup_script = log_lib.make_task_bash_script(setup,
                                                          env_vars=setup_envs)
             encoded_script = shlex.quote(setup_script)
-            if detach_setup or _is_command_length_over_limit(encoded_script):
+
+            def _dump_setup_script(setup_script: str) -> None:
                 with tempfile.NamedTemporaryFile('w', prefix='sky_setup_') as f:
                     f.write(setup_script)
                     f.flush()
@@ -3121,6 +3122,9 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
                                  target=remote_setup_file_name,
                                  up=True,
                                  stream_logs=False)
+
+            if detach_setup or _is_command_length_over_limit(encoded_script):
+                _dump_setup_script(setup_script)
                 create_script_code = 'true'
             else:
                 create_script_code = (f'{{ echo {encoded_script} > '
@@ -3128,20 +3132,42 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
 
             if detach_setup:
                 return
+
             setup_log_path = os.path.join(self.log_dir,
                                           f'setup-{runner.node_id}.log')
-            returncode = runner.run(
-                f'{create_script_code} && {setup_cmd}',
-                log_path=setup_log_path,
-                process_stream=False,
-                # We do not source bashrc for setup, since bashrc is sourced
-                # in the script already.
-                # Skip an empty line and two lines due to the /bin/bash -i and
-                # source ~/.bashrc in the setup_cmd.
-                #   bash: cannot set terminal process group (7398): Inappropriate ioctl for device # pylint: disable=line-too-long
-                #   bash: no job control in this shell
-                skip_lines=3,
-            )
+
+            def _run_setup(setup_cmd: str) -> int:
+                returncode = runner.run(
+                    setup_cmd,
+                    log_path=setup_log_path,
+                    process_stream=False,
+                    # We do not source bashrc for setup, since bashrc is sourced
+                    # in the script already.
+                    # Skip an empty line and two lines due to the /bin/bash -i
+                    # and source ~/.bashrc in the setup_cmd.
+                    #   bash: cannot set terminal process group (7398): Inappropriate ioctl for device # pylint: disable=line-too-long
+                    #   bash: no job control in this shell
+                    skip_lines=3)
+                return returncode
+
+            returncode = _run_setup(f'{create_script_code} && {setup_cmd}',)
+            if returncode == 255:
+                is_message_too_long = False
+                with open(setup_log_path, 'r', encoding='utf-8') as f:
+                    if 'too long' in f.read():
+                        is_message_too_long = True
+
+                if is_message_too_long:
+                    # If the setup script is too long, we retry it with dumping
+                    # the script to a file and running it with SSH. We use a
+                    # general length limit check before but it could be
+                    # inaccurate on some systems.
+                    logger.debug(
+                        'Failed to run setup command inline due to '
+                        'command length limit. Dumping setup script to '
+                        'file and running it with SSH.')
+                    _dump_setup_script(setup_script)
+                    returncode = _run_setup(setup_cmd)
 
             def error_message() -> str:
                 # Use the function to avoid tailing the file in success case
@@ -3223,7 +3249,8 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
 
         code = job_lib.JobLibCodeGen.queue_job(job_id, job_submit_cmd)
         job_submit_cmd = ' && '.join([mkdir_code, create_script_code, code])
-        if _is_command_length_over_limit(job_submit_cmd):
+
+        def _dump_code_to_file(codegen: str) -> None:
             runners = handle.get_command_runners()
             head_runner = runners[0]
             with tempfile.NamedTemporaryFile('w', prefix='sky_app_') as fp:
@@ -3238,6 +3265,9 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
                                   target=script_path,
                                   up=True,
                                   stream_logs=False)
+
+        if _is_command_length_over_limit(job_submit_cmd):
+            _dump_code_to_file(codegen)
             job_submit_cmd = f'{mkdir_code} && {code}'
 
         if managed_job_dag is not None:
@@ -3263,6 +3293,16 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
                                                       job_submit_cmd,
                                                       stream_logs=False,
                                                       require_outputs=True)
+        if returncode == 255 and 'too long' in stdout + stderr:
+            # If the setup script is too long, we retry it with dumping
+            # the script to a file and running it with SSH. We use a general
+            # length limit check before but it could be inaccurate on some
+            # systems.
+            _dump_code_to_file(codegen)
+            returncode, stdout, stderr = self.run_on_head(handle,
+                                                          job_submit_cmd,
+                                                          stream_logs=False,
+                                                          require_outputs=True)
 
         # Happens when someone calls `sky exec` but remote is outdated
         # necessitating calling `sky launch`.

{skypilot_nightly-1.0.0.dev20240902 → skypilot_nightly-1.0.0.dev20240903}/sky/cli.py

@@ -29,6 +29,7 @@ import functools
 import multiprocessing
 import os
 import shlex
+import shutil
 import signal
 import subprocess
 import sys
@@ -368,7 +369,9 @@ def _install_shell_completion(ctx: click.Context, param: click.Parameter,
                 echo "{bashrc_diff}" >> ~/.bashrc'
 
         cmd = (f'(grep -q "SkyPilot" ~/.bashrc) || '
-               f'[[ ${{BASH_VERSINFO[0]}} -ge 4 ]] && ({install_cmd})')
+               f'([[ ${{BASH_VERSINFO[0]}} -ge 4 ]] && ({install_cmd}) || '
+               f'(echo "Bash must be version 4 or above." && exit 1))')
+
         reload_cmd = _RELOAD_BASH_CMD
 
     elif value == 'fish':
@@ -390,7 +393,10 @@ def _install_shell_completion(ctx: click.Context, param: click.Parameter,
         ctx.exit()
 
     try:
-        subprocess.run(cmd, shell=True, check=True, executable='/bin/bash')
+        subprocess.run(cmd,
+                       shell=True,
+                       check=True,
+                       executable=shutil.which('bash'))
         click.secho(f'Shell completion installed for {value}', fg='green')
         click.echo(
             'Completion will take effect once you restart the terminal: ' +

{skypilot_nightly-1.0.0.dev20240902 → skypilot_nightly-1.0.0.dev20240903}/sky/exceptions.py

@@ -100,9 +100,13 @@ class CommandError(Exception):
         self.command = command
         self.error_msg = error_msg
         self.detailed_reason = detailed_reason
+
         if not command:
             message = error_msg
         else:
+            if len(command) > 100:
+                # Chunck the command to avoid overflow.
+                command = command[:100] + '...'
             message = (f'Command {command} failed with return code '
                        f'{returncode}.\n{error_msg}')
         super().__init__(message)

{skypilot_nightly-1.0.0.dev20240902 → skypilot_nightly-1.0.0.dev20240903/skypilot_nightly.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: skypilot-nightly
-Version: 1.0.0.dev20240902
+Version: 1.0.0.dev20240903
 Summary: SkyPilot: An intercloud broker for the clouds
 Author: SkyPilot Team
 License: Apache 2.0

{skypilot_nightly-1.0.0.dev20240902 → skypilot_nightly-1.0.0.dev20240903}/tests/test_smoke.py

@@ -34,6 +34,7 @@ import shutil
 import subprocess
 import sys
 import tempfile
+import textwrap
 import time
 from typing import Dict, List, NamedTuple, Optional, Tuple
 import urllib.parse
@@ -3436,6 +3437,43 @@ def test_gcp_zero_quota_failover():
     run_one_test(test)
 
 
+def test_long_setup_run_script(generic_cloud: str):
+    name = _get_cluster_name()
+    with tempfile.NamedTemporaryFile('w', prefix='sky_app_',
+                                     suffix='.yaml') as f:
+        f.write(
+            textwrap.dedent(""" \
+            setup: |
+              echo "start long setup"
+            """))
+        for i in range(1024 * 120):
+            f.write(f'  echo {i}\n')
+        f.write('  echo "end long setup"\n')
+        f.write(
+            textwrap.dedent(""" \
+            run: |
+              echo "run"
+        """))
+        for i in range(1024 * 120):
+            f.write(f'  echo {i}\n')
+        f.write('  echo "end run"\n')
+        f.flush()
+
+        test = Test(
+            'long-setup-run-script',
+            [
+                f'sky launch -y -c {name} --cloud {generic_cloud} --detach-setup --detach-run --cpus 2+ {f.name}',
+                f'sky exec --detach-run {name} "echo hello"',
+                f'sky exec --detach-run {name} {f.name}',
+                f'sky logs {name} --status 1',
+                f'sky logs {name} --status 2',
+                f'sky logs {name} --status 3',
+            ],
+            f'sky down -y {name}',
+        )
+        run_one_test(test)
+
+
 # ---------- Testing skyserve ----------