executorlib 1.10.1__tar.gz → 1.10.3__tar.gz

{executorlib-1.10.1 → executorlib-1.10.3}/PKG-INFO

@@ -1,6 +1,6 @@
-Metadata-Version: 2.4
+Metadata-Version: 2.5
 Name: executorlib
-Version: 1.10.1
+Version: 1.10.3
 Summary: Up-scale python functions for high performance computing (HPC) with executorlib.
 Project-URL: Homepage, https://github.com/pyiron/executorlib
 Project-URL: Documentation, https://executorlib.readthedocs.io
@@ -49,31 +49,31 @@ Classifier: Programming Language :: Python :: 3.14
 Classifier: Topic :: Scientific/Engineering :: Physics
 Requires-Python: <3.15,>3.9
 Requires-Dist: cloudpickle<=3.1.2,>=2.0.0
-Requires-Dist: pyzmq<=27.1.0,>=25.0.0
+Requires-Dist: pyzmq<=27.2.0,>=25.0.0
 Provides-Extra: all
 Requires-Dist: h5py<=3.16.0,>=3.6.0; extra == 'all'
 Requires-Dist: ipython<=9.9.0,>=7.33.0; extra == 'all'
 Requires-Dist: mpi4py<=4.1.2,>=3.1.4; extra == 'all'
 Requires-Dist: networkx<=3.6.1,>=2.8.8; extra == 'all'
-Requires-Dist: pydantic<=2.13.4,>=2.5.3; extra == 'all'
-Requires-Dist: pygraphviz<=1.14,>=1.10; extra == 'all'
-Requires-Dist: pysqa==0.4.3; extra == 'all'
+Requires-Dist: pydantic<=2.13.5,>=2.5.3; extra == 'all'
+Requires-Dist: pygraphviz<=2.0.1,>=1.10; extra == 'all'
+Requires-Dist: pysqa==1.0.0; extra == 'all'
 Provides-Extra: cache
 Requires-Dist: h5py<=3.16.0,>=3.6.0; extra == 'cache'
 Provides-Extra: cluster
 Requires-Dist: h5py<=3.16.0,>=3.6.0; extra == 'cluster'
-Requires-Dist: pysqa==0.4.3; extra == 'cluster'
+Requires-Dist: pysqa==1.0.0; extra == 'cluster'
 Provides-Extra: graph
 Requires-Dist: networkx<=3.6.1,>=2.8.8; extra == 'graph'
-Requires-Dist: pygraphviz<=1.14,>=1.10; extra == 'graph'
+Requires-Dist: pygraphviz<=2.0.1,>=1.10; extra == 'graph'
 Provides-Extra: graphnotebook
 Requires-Dist: ipython<=9.9.0,>=7.33.0; extra == 'graphnotebook'
 Requires-Dist: networkx<=3.6.1,>=2.8.8; extra == 'graphnotebook'
-Requires-Dist: pygraphviz<=1.14,>=1.10; extra == 'graphnotebook'
+Requires-Dist: pygraphviz<=2.0.1,>=1.10; extra == 'graphnotebook'
 Provides-Extra: mpi
 Requires-Dist: mpi4py<=4.1.2,>=3.1.4; extra == 'mpi'
 Provides-Extra: validation
-Requires-Dist: pydantic<=2.13.4,>=2.5.3; extra == 'validation'
+Requires-Dist: pydantic<=2.13.5,>=2.5.3; extra == 'validation'
 Description-Content-Type: text/markdown
 
 # executorlib

{executorlib-1.10.1 → executorlib-1.10.3}/pyproject.toml

@@ -1,9 +1,9 @@
 [build-system]
 requires = [
-    "hatchling>=1.27.0,<=1.30.1",
+    "hatchling>=1.27.0,<=1.32.0",
     "hatch-vcs==0.5.0",
     "cloudpickle>=2.0.0,<=3.1.2",
-    "pyzmq>=25.0.0,<=27.1.0",
+    "pyzmq>=25.0.0,<=27.2.0",
 ]
 build-backend = "hatchling.build"
 
@@ -31,7 +31,7 @@ classifiers = [
 ]
 dependencies = [
     "cloudpickle>=2.0.0,<=3.1.2",
-    "pyzmq>=25.0.0,<=27.1.0",
+    "pyzmq>=25.0.0,<=27.2.0",
 ]
 dynamic = ["version"]
 
@@ -43,28 +43,28 @@ Repository = "https://github.com/pyiron/executorlib"
 [project.optional-dependencies]
 cache = ["h5py>=3.6.0,<=3.16.0"]
 graph = [
-    "pygraphviz>=1.10,<=1.14",
+    "pygraphviz>=1.10,<=2.0.1",
     "networkx>=2.8.8,<=3.6.1",
 ]
 graphnotebook = [
-    "pygraphviz>=1.10,<=1.14",
+    "pygraphviz>=1.10,<=2.0.1",
     "networkx>=2.8.8,<=3.6.1",
     "ipython>=7.33.0,<=9.9.0",
 ]
 mpi = ["mpi4py>=3.1.4,<=4.1.2"]
 cluster = [
-    "pysqa==0.4.3",
+    "pysqa==1.0.0",
     "h5py>=3.6.0,<=3.16.0",
 ]
 validation = [
-    "pydantic>=2.5.3,<=2.13.4",
+    "pydantic>=2.5.3,<=2.13.5",
 ]
 all = [
     "mpi4py>=3.1.4,<=4.1.2",
-    "pysqa==0.4.3",
+    "pysqa==1.0.0",
     "h5py>=3.6.0,<=3.16.0",
-    "pydantic>=2.5.3,<=2.13.4",
-    "pygraphviz>=1.10,<=1.14",
+    "pydantic>=2.5.3,<=2.13.5",
+    "pygraphviz>=1.10,<=2.0.1",
     "networkx>=2.8.8,<=3.6.1",
     "ipython>=7.33.0,<=9.9.0",
 ]
@@ -98,12 +98,16 @@ ignore = [
     "E501",
     # ignore imports which are not at the top of a module
     "PLC0415",
+    # Too many branches
+    "PLR0912",
     # Too many arguments in function definition
     "PLR0913",
+    # Too many statements
+    "PLR0915",
+    # Too many not-keyword-only arguments,
+    "PLR0917",
     # Magic value used in comparison
     "PLR2004",
-    # Too many branches
-    "PLR0912",
 ]
 
 [tool.hatch.build]

{executorlib-1.10.1 → executorlib-1.10.3}/src/executorlib/__init__.py

@@ -44,6 +44,35 @@ def get_cache_data(cache_directory: str) -> list[dict]:
     return get_cache_data(cache_directory=cache_directory)
 
 
+def get_cache_data_queue(
+    cache_directory: str,
+    queue_type: Optional[str] = None,
+    config_directory: Optional[str] = None,
+) -> list[dict]:
+    """
+    Collect all HDF5 files in the cache directory and check their status in the queue system
+
+    Args:
+        cache_directory (str): The directory to store cache files.
+        queue_type (str, optional): The type of the queue system ["slurm", "flux"].
+        config_directory (str, optional): The directory containing the configuration for the queue system.
+
+    Returns:
+        list[dict]: List of dictionaries each representing on of the HDF5 files in the
+                    cache directory with their status in the queue system.
+    """
+    from executorlib.standalone.hdf import get_cache_data
+    from executorlib.task_scheduler.file.spawner_pysqa import (
+        get_queue_system_cache_data,
+    )
+
+    return get_queue_system_cache_data(
+        cache_dict=get_cache_data(cache_directory=cache_directory),
+        queue_type=queue_type,
+        config_directory=config_directory,
+    )
+
+
 def get_future_from_cache(
     cache_directory: str,
     cache_key: str,

{executorlib-1.10.1 → executorlib-1.10.3}/src/executorlib/_version.py

@@ -18,7 +18,7 @@ version_tuple: tuple[int | str, ...]
 commit_id: str | None
 __commit_id__: str | None
 
-__version__ = version = '1.10.1'
-__version_tuple__ = version_tuple = (1, 10, 1)
+__version__ = version = '1.10.3'
+__version_tuple__ = version_tuple = (1, 10, 3)
 
 __commit_id__ = commit_id = None

{executorlib-1.10.1 → executorlib-1.10.3}/src/executorlib/executor/single.py

@@ -393,7 +393,7 @@ class TestClusterExecutor(BaseExecutor):
         )
         if not plot_dependency_graph:
             from executorlib.task_scheduler.file.spawner_subprocess import (
-                execute_in_subprocess,
+                subprocess_execute,
             )
             from executorlib.task_scheduler.file.task_scheduler import (
                 create_file_executor,
@@ -415,7 +415,7 @@ class TestClusterExecutor(BaseExecutor):
                     block_allocation=block_allocation,
                     init_function=init_function,
                     disable_dependencies=disable_dependencies,
-                    execute_function=execute_in_subprocess,
+                    execute_function=subprocess_execute,
                     wait=wait,
                     refresh_rate=refresh_rate,
                     validator=validate_resource_dict_with_optional_keys,

executorlib-1.10.3/src/executorlib/standalone/command_pysqa.py

@@ -0,0 +1,148 @@
+import contextlib
+import os
+import subprocess
+from time import monotonic
+from typing import Optional, Union
+
+from pysqa import QueueAdapter
+
+# Minimum time between two queries of the queuing system for the status of a task whose output
+# file has not appeared yet. Detecting a dead job (timeout, OOM, node failure, scancel, ...) relies
+# on this status query, but it must not be issued on every poll of the (much faster) refresh_rate
+# loop, as that would flood the queuing system commands (e.g. squeue/sacct) with requests.
+_JOB_STATUS_CHECK_INTERVAL = 30.0
+
+
+def pysqa_job_output_validation(
+    task_key: str,
+    file_name: str,
+    queue_id: int,
+    status_check_dict: Optional[dict],
+    pysqa_config_directory: Optional[str] = None,
+    backend: Optional[str] = None,
+    job_status_check_interval: float = _JOB_STATUS_CHECK_INTERVAL,
+) -> bool:
+    """
+    Check whether the queuing system job backing a task has died without ever writing its output
+    file. Only applies to queuing system backends (pysqa) and is throttled to at most once every
+    ``job_status_check_interval`` seconds per task, to avoid flooding the queuing system with
+    status queries on every poll of the (much faster) refresh_rate loop.
+
+    A dead job is recognized in two ways, since queuing systems differ in whether they drop
+    terminated jobs from their listing: slurm's squeue removes a job as soon as it is gone
+    (status None), while flux's "flux jobs -a" keeps listing inactive jobs and instead reports
+    pysqa's terminal-failure status "error" (the same status pysqa_terminate already treats as
+    not alive).
+
+    Args:
+        task_key (str): The key of the task.
+        file_name (str): Path of the expected output HDF5 file.
+        queue_id (int, optional): The queuing system ID of the task.
+        pysqa_config_directory (str, optional): path to the pysqa config directory.
+        backend (str, optional): name of the backend used to spawn tasks ["slurm", "flux"].
+        status_check_dict (dict): Dictionary tracking when each task's job status was last queried.
+        job_status_check_interval (float): Minimum time interval between job status checks for the same task.
+
+    Returns:
+        bool: True if the job is no longer known to the queuing system, or is reported as having
+            errored out, and still has no output.
+    """
+    now = monotonic()
+    last_checked = (
+        status_check_dict.get(task_key, 0.0) if status_check_dict is not None else 0.0
+    )
+    if now - last_checked < job_status_check_interval:
+        return False
+    if status_check_dict is not None:
+        status_check_dict[task_key] = now
+    status = pysqa_get_status_of_job(
+        queue_id=queue_id,
+        config_directory=pysqa_config_directory,
+        backend=backend,
+    )
+    return (status is None or status == "error") and not os.path.exists(file_name)
+
+
+def pysqa_terminate(
+    queue_id: int,
+    config_directory: Optional[str] = None,
+    backend: Optional[str] = None,
+):
+    """
+    Delete job from queuing system
+
+    Args:
+        queue_id (int): Queuing system ID of the job to delete.
+        config_directory (str, optional): path to the config directory.
+        backend (str, optional): name of the backend used to spawn tasks ["slurm", "flux"].
+    """
+    qa = QueueAdapter(
+        directory=config_directory,
+        queue_type=backend,
+        execute_command=pysqa_execute_command,
+    )
+    status = qa.get_status_of_job(process_id=queue_id)
+    if status is not None and status not in ["finished", "error"]:
+        with contextlib.suppress(subprocess.CalledProcessError):
+            qa.delete_job(process_id=queue_id)
+
+
+def pysqa_get_status_of_job(
+    queue_id: int,
+    config_directory: Optional[str] = None,
+    backend: Optional[str] = None,
+) -> Optional[str]:
+    """
+    Query the status of a job from the queuing system.
+
+    Args:
+        queue_id (int): Queuing system ID of the job to check.
+        config_directory (str, optional): path to the config directory.
+        backend (str, optional): name of the backend used to spawn tasks ["slurm", "flux"].
+
+    Returns:
+        str: status of the job as reported by the queuing system, None if the queuing system no
+             longer knows about the job (e.g. it timed out, was killed or already finished).
+    """
+    qa = QueueAdapter(
+        directory=config_directory,
+        queue_type=backend,
+        execute_command=pysqa_execute_command,
+    )
+    return qa.get_status_of_job(process_id=queue_id)
+
+
+def pysqa_execute_command(
+    commands: str,
+    working_directory: Optional[str] = None,
+    split_output: bool = True,
+    shell: bool = False,
+    error_filename: str = "pysqa.err",
+) -> Union[str, list[str]]:
+    """
+    A wrapper around the subprocess.check_output function. Modified from pysqa to raise an exception if the subprocess
+    fails to submit the job to the queue.
+
+    Args:
+        commands (str): The command(s) to be executed on the command line
+        working_directory (str, optional): The directory where the command is executed. Defaults to None.
+        split_output (bool, optional): Boolean flag to split newlines in the output. Defaults to True.
+        shell (bool, optional): Additional switch to convert commands to a single string. Defaults to False.
+        error_filename (str, optional): In case the execution fails, the output is written to this file. Defaults to "pysqa.err".
+
+    Returns:
+        Union[str, List[str]]: Output of the shell command either as a string or as a list of strings
+    """
+    if shell and isinstance(commands, list):
+        commands = " ".join(commands)
+    out = subprocess.check_output(
+        commands,
+        cwd=working_directory,
+        stderr=subprocess.STDOUT,
+        universal_newlines=True,
+        shell=not isinstance(commands, list),
+    )
+    if out is not None and split_output:
+        return out.split("\n")
+    else:
+        return out

{executorlib-1.10.1 → executorlib-1.10.3}/src/executorlib/standalone/interactive/communication.py

@@ -78,19 +78,17 @@ class SocketInterface:
         while len(response_lst) == 0:
             response_lst = self._poller.poll(self._time_out_ms)
             if not self._spawner.poll():
-                raise ExecutorlibSocketError(
-                    "SocketInterface crashed during execution."
-                )
+                return {
+                    "error": ExecutorlibSocketError(
+                        "SocketInterface crashed during execution."
+                    ),
+                }
         data = self._socket.recv(zmq.NOBLOCK)
         if self._logger is not None:
             self._logger.warning(
                 "Received dictionary of size: " + str(sys.getsizeof(data))
             )
-        output = cloudpickle.loads(data)
-        if "result" in output:
-            return output["result"]
-        else:
-            raise output["error"]
+        return cloudpickle.loads(data)
 
     def send_and_receive_dict(self, input_dict: dict) -> dict:
         """
@@ -152,9 +150,11 @@ class SocketInterface:
         """
         result = None
         if self._spawner.poll():
-            result = self.send_and_receive_dict(
+            output = self.send_and_receive_dict(
                 input_dict={"shutdown": True, "wait": wait}
             )
+            if "result" in output:
+                result = output["result"]
             self._spawner.shutdown(wait=wait)
         self._reset_socket()
         return result

{executorlib-1.10.1 → executorlib-1.10.3}/src/executorlib/standalone/serialize.py

@@ -34,7 +34,7 @@ def serialize_funct(
     fn_kwargs: Optional[dict] = None,
     resource_dict: Optional[dict] = None,
     cache_key: Optional[str] = None,
-) -> tuple[str, dict]:
+) -> tuple[str, dict, Optional[Exception]]:
     """
     Serialize a function and its arguments and keyword arguments into an HDF5 file.
 
@@ -56,7 +56,8 @@ def serialize_funct(
                                    overwritten by setting the cache_key.
 
     Returns:
-        Tuple[str, dict]: A tuple containing the task key and the serialized data.
+        Tuple[str, dict, Optional[Exception]]: A tuple containing the task key, the serialized data and an
+                                               exception if the serialization failed.
 
     """
     if fn_args is None:
@@ -68,21 +69,25 @@ def serialize_funct(
     if cache_key is not None:
         task_key = cache_key
     else:
-        binary_all = cloudpickle.dumps(
-            {
-                "fn": fn,
-                "args": fn_args,
-                "kwargs": fn_kwargs,
-            }
-        )
-        task_key = _get_function_name(fn=fn) + _get_hash(binary=binary_all)
+        try:
+            binary_all = cloudpickle.dumps(
+                {
+                    "fn": fn,
+                    "args": fn_args,
+                    "kwargs": fn_kwargs,
+                }
+            )
+        except Exception as e:
+            return "", {}, e
+        else:
+            task_key = _get_function_name(fn=fn) + _get_hash(binary=binary_all)
     data = {
         "fn": fn,
         "args": fn_args,
         "kwargs": fn_kwargs,
         "resource_dict": resource_dict,
     }
-    return task_key, data
+    return task_key, data, None
 
 
 def _get_hash(binary: bytes) -> str:

{executorlib-1.10.1 → executorlib-1.10.3}/src/executorlib/task_scheduler/file/shared.py

@@ -8,7 +8,7 @@ from typing import Any, Callable, Optional
 from executorlib.standalone.command import get_cache_execute_command
 from executorlib.standalone.hdf import get_cache_files, get_output, get_queue_id
 from executorlib.standalone.serialize import serialize_funct
-from executorlib.task_scheduler.file.spawner_subprocess import terminate_subprocess
+from executorlib.task_scheduler.file.spawner_subprocess import subprocess_terminate
 
 
 class FutureItem:
@@ -59,6 +59,7 @@ def execute_tasks_h5(
     execute_function: Callable,
     executor_kwargs: dict,
     terminate_function: Optional[Callable] = None,
+    validate_function: Optional[Callable] = None,
     pysqa_config_directory: Optional[str] = None,
     backend: Optional[str] = None,
     disable_dependencies: bool = False,
@@ -76,6 +77,7 @@ def execute_tasks_h5(
                               - cwd (str/None): current working directory where the parallel python task is executed
         execute_function (Callable): The function to execute the tasks.
         terminate_function (Callable): The function to terminate the tasks.
+        validate_function (Callable): The function to validate the tasks.
         pysqa_config_directory (str, optional): path to the pysqa config directory (only for pysqa based backend).
         backend (str, optional): name of the backend used to spawn tasks.
         disable_dependencies (boolean): Disable resolving future objects during the submission.
@@ -92,6 +94,7 @@ def execute_tasks_h5(
     cache_dir_dict: dict = {}
     file_name_dict: dict = {}
     duplicate_dict: dict = {}
+    status_check_dict: dict = {}
     while True:
         task_dict = None
         with contextlib.suppress(queue.Empty):
@@ -104,7 +107,9 @@ def execute_tasks_h5(
                 process_dict=process_dict,
                 duplicate_dict=duplicate_dict,
                 cache_dir_dict=cache_dir_dict,
+                status_check_dict=status_check_dict,
                 terminate_function=terminate_function,
+                validate_function=validate_function,
                 pysqa_config_directory=pysqa_config_directory,
                 backend=backend,
                 refresh_rate=refresh_rate,
@@ -124,7 +129,7 @@ def execute_tasks_h5(
                     executor_kwargs=executor_kwargs,
                 )
             )
-            task_key, data_dict = serialize_funct(
+            task_key, data_dict, serialize_exception = serialize_funct(
                 fn=task_dict["fn"],
                 fn_args=task_args,
                 fn_kwargs=task_kwargs,
@@ -132,7 +137,9 @@ def execute_tasks_h5(
                 cache_key=cache_key,
             )
             data_dict["error_log_file"] = error_log_file
-            if task_key not in memory_dict:
+            if serialize_exception is not None:
+                task_dict["future"].set_exception(serialize_exception)
+            elif task_key not in memory_dict:
                 if os.path.join(
                     cache_directory, task_key + "_o.h5"
                 ) not in get_cache_files(cache_directory=cache_directory):
@@ -151,25 +158,31 @@ def execute_tasks_h5(
                                 )
                             )
                         task_dependent_lst = []
-                    process_dict[task_key] = execute_function(
-                        command=get_cache_execute_command(
+                    try:
+                        process_dict[task_key] = execute_function(
+                            command=get_cache_execute_command(
+                                file_name=file_name,
+                                cores=task_resource_dict["cores"],
+                                backend=backend,
+                                exclusive=task_resource_dict.get("exclusive", False),
+                                openmpi_oversubscribe=task_resource_dict.get(
+                                    "openmpi_oversubscribe", False
+                                ),
+                                pmi_mode=pmi_mode,
+                            ),
                             file_name=file_name,
-                            cores=task_resource_dict["cores"],
+                            data_dict=data_dict,
+                            task_dependent_lst=task_dependent_lst,
+                            resource_dict=task_resource_dict,
+                            config_directory=pysqa_config_directory,
                             backend=backend,
-                            exclusive=task_resource_dict.get("exclusive", False),
-                            openmpi_oversubscribe=task_resource_dict.get(
-                                "openmpi_oversubscribe", False
-                            ),
-                            pmi_mode=pmi_mode,
-                        ),
-                        file_name=file_name,
-                        data_dict=data_dict,
-                        task_dependent_lst=task_dependent_lst,
-                        resource_dict=task_resource_dict,
-                        config_directory=pysqa_config_directory,
-                        backend=backend,
-                        cache_directory=cache_directory,
-                    )
+                            cache_directory=cache_directory,
+                        )
+                    except Exception as e:
+                        task_dict["future"].set_exception(e)
+                        memory_dict[task_key] = task_dict["future"]
+                        future_queue.task_done()
+                        continue
                 file_name = os.path.join(cache_directory, task_key + "_o.h5")
                 file_name_dict[task_key] = file_name
                 queue_id = get_queue_id(file_name=file_name)
@@ -188,7 +201,9 @@ def execute_tasks_h5(
                 cache_dir_dict=cache_dir_dict,
                 process_dict=process_dict,
                 duplicate_dict=duplicate_dict,
+                status_check_dict=status_check_dict,
                 terminate_function=terminate_function,
+                validate_function=validate_function,
                 pysqa_config_directory=pysqa_config_directory,
                 backend=backend,
                 refresh_rate=refresh_rate,
@@ -199,15 +214,30 @@ def _check_task_output(
     task_key: str,
     future_obj: Future,
     cache_directory: str,
+    queue_id: Optional[int] = None,
+    pysqa_config_directory: Optional[str] = None,
+    backend: Optional[str] = None,
+    validate_function: Optional[Callable] = None,
+    status_check_dict: Optional[dict] = None,
     duplicate_dict: Optional[dict] = None,
 ) -> Future:
     """
     Check the output of a task and set the result of the future object if available.
 
+    If the output file is missing and the task is running on a queuing system backend, this also
+    detects jobs which died without producing output (e.g. walltime TIMEOUT, OOM, NODE_FAIL or an
+    external scancel) by periodically querying the job status via pysqa and fails the future
+    instead of leaving it pending forever.
+
     Args:
         task_key (str): The key of the task.
         future_obj (Future): The future object associated with the task.
         cache_directory (str): The directory where the HDF5 files are stored.
+        queue_id (int, optional): The queuing system ID of the task, if submitted via pysqa.
+        pysqa_config_directory (str, optional): path to the pysqa config directory.
+        backend (str, optional): name of the backend used to spawn tasks ["slurm", "flux"].
+        status_check_dict (dict): Dictionary tracking when each task's job status was last queried,
+            used to throttle calls to the queuing system.
         duplicate_dict (dict): The dictionary mapping task keys to their associated duplicate future objects.
     Returns:
         Future: The updated future object.
@@ -215,8 +245,32 @@ def _check_task_output(
     """
     file_name = os.path.join(cache_directory, task_key + "_o.h5")
     if not os.path.exists(file_name):
-        return future_obj
-    exec_flag, no_error_flag, result = get_output(file_name=file_name)
+        if (
+            backend is None
+            or queue_id is None
+            or validate_function is None
+            or not validate_function(
+                task_key=task_key,
+                file_name=file_name,
+                queue_id=queue_id,
+                pysqa_config_directory=pysqa_config_directory,
+                backend=backend,
+                status_check_dict=status_check_dict,
+            )
+        ):
+            return future_obj
+        exec_flag, no_error_flag, result = (
+            True,
+            False,
+            RuntimeError(
+                f"executorlib: queue job {queue_id} for task {task_key} terminated without "
+                "producing output (timeout, out-of-memory, node failure or cancellation)."
+            ),
+        )
+    else:
+        exec_flag, no_error_flag, result = get_output(file_name=file_name)
+    if status_check_dict is not None:
+        status_check_dict.pop(task_key, None)
     _update_future(
         future_obj=future_obj,
         exec_flag=exec_flag,
@@ -321,7 +375,9 @@ def _refresh_memory_dict(
     cache_dir_dict: dict,
     process_dict: dict,
     duplicate_dict: Optional[dict] = None,
+    status_check_dict: Optional[dict] = None,
     terminate_function: Optional[Callable] = None,
+    validate_function: Optional[Callable] = None,
     pysqa_config_directory: Optional[str] = None,
     backend: Optional[str] = None,
     refresh_rate: float = 0.01,
@@ -334,7 +390,10 @@ def _refresh_memory_dict(
         cache_dir_dict (dict): dictionary with task keys and cache directories
         process_dict (dict): dictionary with task keys and process reference.
         duplicate_dict (dict): dictionary with task keys and duplicate future objects.
+        status_check_dict (dict): dictionary with task keys and the last time their queuing system
+            job status was queried, used to throttle detection of jobs that died without output.
         terminate_function (callable): The function to terminate the tasks.
+        validate_function (callable): The function to validate the tasks.
         pysqa_config_directory (str): path to the pysqa config directory (only for pysqa based backend).
         backend (str): name of the backend used to spawn tasks.
         refresh_rate (float): The rate at which to refresh the result. Defaults to 0.01.
@@ -351,11 +410,19 @@ def _refresh_memory_dict(
         pysqa_config_directory=pysqa_config_directory,
         backend=backend,
     )
+    if status_check_dict is not None:
+        for key in cancelled_lst:
+            status_check_dict.pop(key, None)
     memory_updated_dict = {
         key: _check_task_output(
             task_key=key,
             future_obj=value,
             cache_directory=cache_dir_dict[key],
+            queue_id=process_dict.get(key),
+            pysqa_config_directory=pysqa_config_directory,
+            backend=backend,
+            validate_function=validate_function,
+            status_check_dict=status_check_dict,
             duplicate_dict=duplicate_dict,
         )
         for key, value in memory_dict.items()
@@ -381,7 +448,7 @@ def _cancel_processes(
         pysqa_config_directory (str): path to the pysqa config directory (only for pysqa based backend).
         backend (str): name of the backend used to spawn tasks.
     """
-    if terminate_function is not None and terminate_function == terminate_subprocess:
+    if terminate_function is not None and terminate_function == subprocess_terminate:
         for task in process_dict.values():
             terminate_function(task=task)
     elif terminate_function is not None and backend is not None:
@@ -443,7 +510,9 @@ def _shutdown_executor(
     process_dict: dict,
     cache_dir_dict: dict,
     duplicate_dict: Optional[dict] = None,
+    status_check_dict: Optional[dict] = None,
     terminate_function: Optional[Callable] = None,
+    validate_function: Optional[Callable] = None,
     pysqa_config_directory: Optional[str] = None,
     backend: Optional[str] = None,
     refresh_rate: float = 0.01,
@@ -466,6 +535,9 @@ def _shutdown_executor(
         process_dict (dict): Mapping of task keys to process handles or queue IDs.
         duplicate_dict (dict): Mapping of task keys to lists of duplicate Future objects.
         cache_dir_dict (dict): Mapping of task keys to the cache directory for each task.
+        status_check_dict (dict): Mapping of task keys to the last time their queuing system job
+            status was queried, used to throttle detection of jobs that died without output.
+        validate_function (Callable, optional): Function used to validate the tasks.
         terminate_function (Callable, optional): Function used to terminate running processes.
         pysqa_config_directory (str, optional): Path to the pysqa config directory.
         backend (str, optional): Name of the backend ("slurm", "flux", or None for subprocess).
@@ -478,7 +550,9 @@ def _shutdown_executor(
                 cache_dir_dict=cache_dir_dict,
                 process_dict=process_dict,
                 duplicate_dict=duplicate_dict,
+                status_check_dict=status_check_dict,
                 terminate_function=terminate_function,
+                validate_function=validate_function,
                 pysqa_config_directory=pysqa_config_directory,
                 backend=backend,
                 refresh_rate=refresh_rate,
@@ -493,7 +567,9 @@ def _shutdown_executor(
                 cache_dir_dict=cache_dir_dict,
                 process_dict=process_dict,
                 duplicate_dict=duplicate_dict,
+                status_check_dict=status_check_dict,
                 terminate_function=terminate_function,
+                validate_function=validate_function,
                 pysqa_config_directory=pysqa_config_directory,
                 backend=backend,
                 refresh_rate=refresh_rate,
@@ -512,7 +588,9 @@ def _shutdown_executor(
             cache_dir_dict=cache_dir_dict,
             process_dict=process_dict,
             duplicate_dict=duplicate_dict,
+            status_check_dict=status_check_dict,
             terminate_function=terminate_function,
+            validate_function=validate_function,
             pysqa_config_directory=pysqa_config_directory,
             backend=backend,
             refresh_rate=refresh_rate,

{executorlib-1.10.1 → executorlib-1.10.3}/src/executorlib/task_scheduler/file/spawner_pysqa.py

@@ -2,14 +2,15 @@ import os
 from subprocess import CalledProcessError
 from typing import Optional
 
+from pandas import DataFrame
 from pysqa import QueueAdapter
 
+from executorlib.standalone.command_pysqa import pysqa_execute_command, pysqa_terminate
 from executorlib.standalone.hdf import dump, get_queue_id
 from executorlib.standalone.inputcheck import check_file_exists
 from executorlib.standalone.interactive.spawner import (
     set_current_directory_in_environment,
 )
-from executorlib.standalone.scheduler import pysqa_execute_command, terminate_with_pysqa
 
 
 def execute_with_pysqa(
@@ -122,7 +123,7 @@ def terminate_tasks_in_cache(
     for f in hdf5_file_lst:
         queue_id = get_queue_id(f)
         if queue_id is not None:
-            terminate_with_pysqa(
+            pysqa_terminate(
                 queue_id=queue_id,
                 config_directory=pysqa_config_directory,
                 backend=backend,
@@ -148,9 +149,64 @@ def terminate_task_in_cache(
     file_name = os.path.join(cache_directory, cache_key + "_i.h5")
     queue_id = get_queue_id(file_name=file_name)
     if queue_id is not None:
-        terminate_with_pysqa(
+        pysqa_terminate(
             queue_id=queue_id,
             config_directory=pysqa_config_directory,
             backend=backend,
         )
     os.remove(file_name)
+
+
+def get_queue_system_cache_data(
+    cache_dict: list[dict],
+    queue_type: Optional[str] = None,
+    config_directory: Optional[str] = None,
+) -> list[dict]:
+    """
+    Get the status of the queue system.
+
+    Args:
+        cache_dict (list[dict]): List of dictionaries each representing on of the HDF5 files in the cache directory.
+        queue_type (str, optional): The type of the queue system ["slurm", "flux"].
+        config_directory (str, optional): The directory containing the configuration for the queue system.
+
+    Returns:
+        cache_dict (list[dict]): List of dictionaries each representing on of the HDF5 files in the cache directory.
+    """
+    return _merge_cache_with_queue_status(
+        cache_dict=cache_dict,
+        dataframe=QueueAdapter(
+            directory=config_directory,
+            queue_type=queue_type,
+        ).get_queue_status(),
+    )
+
+
+def _merge_cache_with_queue_status(
+    cache_dict: list[dict], dataframe: DataFrame
+) -> list[dict]:
+    """
+    Merge the cache data with the queue status.
+
+    Args:
+        cache_dict (list[dict]): List of dictionaries each representing on of the HDF5 files in the cache directory.
+        dataframe (DataFrame): DataFrame containing the status of the queue system.
+
+    Returns:
+        cache_dict (list[dict]): List of dictionaries each representing on of the HDF5 files in the cache directory.
+    """
+    cache_updated_dict = []
+    for row in cache_dict:
+        if "output" in row and row["output"]:
+            row["status"] = "finished"
+        elif "queue_id" in row and row["queue_id"] in dataframe["jobid"].values:
+            row["status"] = dataframe[dataframe["jobid"] == row["queue_id"]][
+                "status"
+            ].values[-1]
+        elif "queue_id" not in row:
+            row["status"] = "running"
+        else:
+            row["status"] = "aborted"
+        cache_updated_dict.append(row)
+
+    return cache_updated_dict

{executorlib-1.10.1 → executorlib-1.10.3}/src/executorlib/task_scheduler/file/spawner_subprocess.py

@@ -10,7 +10,7 @@ from executorlib.standalone.interactive.spawner import (
 )
 
 
-def execute_in_subprocess(
+def subprocess_execute(
     command: list,
     file_name: str,
     data_dict: dict,
@@ -65,7 +65,7 @@ def execute_in_subprocess(
     return subprocess.Popen(command, universal_newlines=True, cwd=cwd)
 
 
-def terminate_subprocess(task):
+def subprocess_terminate(task):
     """
     Terminate a subprocess and wait for it to complete.
 

{executorlib-1.10.1 → executorlib-1.10.3}/src/executorlib/task_scheduler/file/task_scheduler.py

@@ -12,17 +12,21 @@ from executorlib.standalone.inputcheck import (
 from executorlib.task_scheduler.base import TaskSchedulerBase, validate_resource_dict
 from executorlib.task_scheduler.file.shared import execute_tasks_h5
 from executorlib.task_scheduler.file.spawner_subprocess import (
-    execute_in_subprocess,
-    terminate_subprocess,
+    subprocess_execute,
+    subprocess_terminate,
 )
 
 try:
-    from executorlib.standalone.scheduler import terminate_with_pysqa
+    from executorlib.standalone.command_pysqa import (
+        pysqa_job_output_validation,
+        pysqa_terminate,
+    )
     from executorlib.task_scheduler.file.spawner_pysqa import execute_with_pysqa
 except ImportError:
     # If pysqa is not available fall back to executing tasks in a subprocess
-    execute_with_pysqa = execute_in_subprocess  # type: ignore
-    terminate_with_pysqa = None  # type: ignore
+    execute_with_pysqa = subprocess_execute  # type: ignore
+    pysqa_terminate = None  # type: ignore
+    pysqa_job_output_validation = None  # type: ignore
 
 
 class FileTaskScheduler(TaskSchedulerBase):
@@ -74,6 +78,7 @@ class FileTaskScheduler(TaskSchedulerBase):
             "future_queue": self._future_queue,
             "execute_function": execute_function,
             "terminate_function": terminate_function,
+            "validate_function": pysqa_job_output_validation,
             "pysqa_config_directory": pysqa_config_directory,
             "backend": backend,
             "disable_dependencies": disable_dependencies,
@@ -126,10 +131,10 @@ def create_file_executor(
     check_executor(executor=flux_executor)
     check_nested_flux_executor(nested_flux_executor=flux_executor_nesting)
     check_flux_log_files(flux_log_files=flux_log_files)
-    if execute_function != execute_in_subprocess:
-        terminate_function = terminate_with_pysqa  # type: ignore
+    if execute_function != subprocess_execute:
+        terminate_function = pysqa_terminate  # type: ignore
     else:
-        terminate_function = terminate_subprocess  # type: ignore
+        terminate_function = subprocess_terminate  # type: ignore
     return FileTaskScheduler(
         executor_kwargs=executor_kwargs,
         pysqa_config_directory=pysqa_config_directory,

{executorlib-1.10.1 → executorlib-1.10.3}/src/executorlib/task_scheduler/interactive/blockallocation.py

@@ -80,35 +80,37 @@ class BlockAllocationTaskScheduler(TaskSchedulerBase):
         executor_kwargs["restart_limit"] = restart_limit
         self._process_kwargs = executor_kwargs
         self._max_workers = max_workers
-        self_id = random.getrandbits(128)
-        self._self_id = self_id
+        self._self_id = random.getrandbits(128)
         _interrupt_bootup_dict[self._self_id] = False
-        alive_workers = [max_workers]
-        alive_workers_lock = Lock()
-        bootup_events = [Event() for _ in range(self._max_workers)]
-        bootup_events[0].set()
+        self._alive_workers = [max_workers]
+        self._alive_workers_lock = Lock()
+        self._bootup_events = [Event() for _ in range(self._max_workers)]
+        self._bootup_events[0].set()
         self._set_process(
             process=[
                 Thread(
                     target=_execute_multiple_tasks,
-                    kwargs=executor_kwargs
-                    | {
-                        "worker_id": worker_id,
-                        "stop_function": lambda: _interrupt_bootup_dict[self_id],
-                        "bootup_event": bootup_events[worker_id],
-                        "next_bootup_event": (
-                            bootup_events[worker_id + 1]
-                            if worker_id + 1 < self._max_workers
-                            else None
-                        ),
-                        "alive_workers": alive_workers,
-                        "alive_workers_lock": alive_workers_lock,
-                    },
+                    kwargs=self._worker_kwargs(worker_id),
                 )
                 for worker_id in range(self._max_workers)
             ],
         )
 
+    def _worker_kwargs(self, worker_id: int) -> dict:
+        self_id = self._self_id
+        return self._process_kwargs | {
+            "worker_id": worker_id,
+            "stop_function": lambda: _interrupt_bootup_dict[self_id],
+            "bootup_event": self._bootup_events[worker_id],
+            "next_bootup_event": (
+                self._bootup_events[worker_id + 1]
+                if worker_id + 1 < len(self._bootup_events)
+                else None
+            ),
+            "alive_workers": self._alive_workers,
+            "alive_workers_lock": self._alive_workers_lock,
+        }
+
     @property
     def max_workers(self) -> int:
         return self._max_workers
@@ -126,12 +128,20 @@ class BlockAllocationTaskScheduler(TaskSchedulerBase):
                         process for process in self._process if process.is_alive()
                     ]
             elif self._max_workers < max_workers:
+                old_max_workers = self._max_workers
+                self._bootup_events.extend(
+                    Event() for _ in range(max_workers - old_max_workers)
+                )
+                for idx in range(old_max_workers, max_workers):
+                    self._bootup_events[idx].set()
+                with self._alive_workers_lock:
+                    self._alive_workers[0] += max_workers - old_max_workers
                 new_process_lst = [
                     Thread(
                         target=_execute_multiple_tasks,
-                        kwargs=self._process_kwargs,
+                        kwargs=self._worker_kwargs(worker_id),
                     )
-                    for _ in range(max_workers - self._max_workers)
+                    for worker_id in range(old_max_workers, max_workers)
                 ]
                 for process_instance in new_process_lst:
                     process_instance.start()
@@ -379,10 +389,9 @@ def _set_init_function(
 ) -> Optional[Exception]:
     interface_initialization_exception = None
     if init_function is not None and interface.status:
-        try:
-            _ = interface.send_and_receive_dict(
-                input_dict={"init": True, "fn": init_function, "args": (), "kwargs": {}}
-            )
-        except Exception as init_exception:
-            interface_initialization_exception = init_exception
+        output = interface.send_and_receive_dict(
+            input_dict={"init": True, "fn": init_function, "args": (), "kwargs": {}}
+        )
+        if "error" in output:
+            interface_initialization_exception = output["error"]
     return interface_initialization_exception

{executorlib-1.10.1 → executorlib-1.10.3}/src/executorlib/task_scheduler/interactive/shared.py

@@ -102,12 +102,16 @@ def _execute_task_without_cache(
         bool: True if the task was submitted successfully, False otherwise.
     """
     try:
-        future_obj.set_result(interface.send_and_receive_dict(input_dict=task_dict))
-    except Exception as thread_exception:
-        if isinstance(thread_exception, ExecutorlibSocketError):
+        output = interface.send_and_receive_dict(input_dict=task_dict)
+    except Exception as e:
+        future_obj.set_exception(exception=e)
+    else:
+        if "result" in output:
+            future_obj.set_result(output["result"])
+        elif isinstance(output["error"], ExecutorlibSocketError):
             return False
         else:
-            future_obj.set_exception(exception=thread_exception)
+            future_obj.set_exception(exception=output["error"])
     return True
 
 
@@ -132,28 +136,34 @@ def _execute_task_with_cache(
     """
     from executorlib.standalone.hdf import dump, get_cache_files, get_output
 
-    task_key, data_dict = serialize_funct(
+    task_key, data_dict, serialize_exception = serialize_funct(
         fn=task_dict["fn"],
         fn_args=task_dict["args"],
         fn_kwargs=task_dict["kwargs"],
         resource_dict=task_dict.get("resource_dict", {}),
         cache_key=cache_key,
     )
-    file_name = os.path.abspath(os.path.join(cache_directory, task_key + "_o.h5"))
-    if file_name not in get_cache_files(cache_directory=cache_directory):
-        try:
+    if serialize_exception is not None:
+        future_obj.set_exception(exception=serialize_exception)
+    else:
+        file_name = os.path.abspath(os.path.join(cache_directory, task_key + "_o.h5"))
+        if file_name not in get_cache_files(cache_directory=cache_directory):
             time_start = time.time()
-            result = interface.send_and_receive_dict(input_dict=task_dict)
-            data_dict["output"] = result
-            data_dict["runtime"] = time.time() - time_start
-            dump(file_name=file_name, data_dict=data_dict)
-            future_obj.set_result(result)
-        except Exception as thread_exception:
-            if isinstance(thread_exception, ExecutorlibSocketError):
+            try:
+                output = interface.send_and_receive_dict(input_dict=task_dict)
+            except Exception as e:
+                future_obj.set_exception(exception=e)
+                return True
+            if "result" in output:
+                data_dict["output"] = output["result"]
+                data_dict["runtime"] = time.time() - time_start
+                dump(file_name=file_name, data_dict=data_dict)
+                future_obj.set_result(output["result"])
+            elif isinstance(output["error"], ExecutorlibSocketError):
                 return False
             else:
-                future_obj.set_exception(exception=thread_exception)
-    else:
-        _, _, result = get_output(file_name=file_name)
-        future_obj.set_result(result)
+                future_obj.set_exception(exception=output["error"])
+        else:
+            _, _, result = get_output(file_name=file_name)
+            future_obj.set_result(result)
     return True

{executorlib-1.10.1 → executorlib-1.10.3}/src/executorlib/task_scheduler/interactive/spawner_pysqa.py

@@ -5,12 +5,12 @@ from typing import Callable, Optional
 
 from pysqa import QueueAdapter
 
+from executorlib.standalone.command_pysqa import pysqa_execute_command, pysqa_terminate
 from executorlib.standalone.inputcheck import validate_number_of_cores
 from executorlib.standalone.interactive.spawner import (
     BaseSpawner,
     set_current_directory_in_environment,
 )
-from executorlib.standalone.scheduler import pysqa_execute_command, terminate_with_pysqa
 from executorlib.task_scheduler.base import validate_resource_dict
 from executorlib.task_scheduler.interactive.blockallocation import (
     BlockAllocationTaskScheduler,
@@ -162,7 +162,7 @@ class PysqaSpawner(BaseSpawner):
             wait (bool, optional): Whether to wait for the interface to shutdown. Defaults to True.
         """
         if self._process is not None:
-            terminate_with_pysqa(
+            pysqa_terminate(
                 queue_id=self._process,
                 config_directory=self._config_directory,
                 backend=self._backend,

executorlib-1.10.1/src/executorlib/standalone/scheduler.py

@@ -1,65 +0,0 @@
-import contextlib
-import subprocess
-from typing import Optional, Union
-
-from pysqa import QueueAdapter
-
-
-def terminate_with_pysqa(
-    queue_id: int,
-    config_directory: Optional[str] = None,
-    backend: Optional[str] = None,
-):
-    """
-    Delete job from queuing system
-
-    Args:
-        queue_id (int): Queuing system ID of the job to delete.
-        config_directory (str, optional): path to the config directory.
-        backend (str, optional): name of the backend used to spawn tasks ["slurm", "flux"].
-    """
-    qa = QueueAdapter(
-        directory=config_directory,
-        queue_type=backend,
-        execute_command=pysqa_execute_command,
-    )
-    status = qa.get_status_of_job(process_id=queue_id)
-    if status is not None and status not in ["finished", "error"]:
-        with contextlib.suppress(subprocess.CalledProcessError):
-            qa.delete_job(process_id=queue_id)
-
-
-def pysqa_execute_command(
-    commands: str,
-    working_directory: Optional[str] = None,
-    split_output: bool = True,
-    shell: bool = False,
-    error_filename: str = "pysqa.err",
-) -> Union[str, list[str]]:
-    """
-    A wrapper around the subprocess.check_output function. Modified from pysqa to raise an exception if the subprocess
-    fails to submit the job to the queue.
-
-    Args:
-        commands (str): The command(s) to be executed on the command line
-        working_directory (str, optional): The directory where the command is executed. Defaults to None.
-        split_output (bool, optional): Boolean flag to split newlines in the output. Defaults to True.
-        shell (bool, optional): Additional switch to convert commands to a single string. Defaults to False.
-        error_filename (str, optional): In case the execution fails, the output is written to this file. Defaults to "pysqa.err".
-
-    Returns:
-        Union[str, List[str]]: Output of the shell command either as a string or as a list of strings
-    """
-    if shell and isinstance(commands, list):
-        commands = " ".join(commands)
-    out = subprocess.check_output(
-        commands,
-        cwd=working_directory,
-        stderr=subprocess.STDOUT,
-        universal_newlines=True,
-        shell=not isinstance(commands, list),
-    )
-    if out is not None and split_output:
-        return out.split("\n")
-    else:
-        return out