parsl 2025.5.26__tar.gz → 2025.6.2__tar.gz

{parsl-2025.5.26/parsl.egg-info → parsl-2025.6.2}/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: parsl
-Version: 2025.5.26
+Version: 2025.6.2
 Summary: Simple data dependent workflows in Python
 Home-page: https://github.com/Parsl/parsl
-Download-URL: https://github.com/Parsl/parsl/archive/2025.05.26.tar.gz
+Download-URL: https://github.com/Parsl/parsl/archive/2025.06.02.tar.gz
 Author: The Parsl Team
 Author-email: parsl@googlegroups.com
 License: Apache 2.0

{parsl-2025.5.26 → parsl-2025.6.2}/parsl/executors/high_throughput/executor.py

@@ -8,7 +8,7 @@ import warnings
 from collections import defaultdict
 from concurrent.futures import Future
 from dataclasses import dataclass
-from typing import Callable, Dict, List, Optional, Sequence, Set, Tuple, Union
+from typing import Callable, Dict, List, Optional, Sequence, Tuple, Union
 
 import typeguard
 
@@ -363,7 +363,9 @@ class HighThroughputExecutor(BlockProviderExecutor, RepresentationMixin, UsageIn
 
     def validate_resource_spec(self, resource_specification: dict):
         if resource_specification:
-            acceptable_fields: Set[str] = set()  # add new resource spec field names here to make htex accept them
+            """HTEX supports the following *Optional* resource specifications:
+            priority: lower value is higher priority"""
+            acceptable_fields = {'priority'}  # add new resource spec field names here to make htex accept them
             keys = set(resource_specification.keys())
             invalid_keys = keys - acceptable_fields
             if invalid_keys:

{parsl-2025.5.26 → parsl-2025.6.2}/parsl/executors/high_throughput/interchange.py

@@ -5,13 +5,13 @@ import logging
 import os
 import pickle
 import platform
-import queue
 import sys
 import threading
 import time
 from typing import Any, Dict, List, Optional, Sequence, Set, Tuple, cast
 
 import zmq
+from sortedcontainers import SortedList
 
 from parsl import curvezmq
 from parsl.addresses import tcp_url
@@ -131,7 +131,7 @@ class Interchange:
         self.hub_address = hub_address
         self.hub_zmq_port = hub_zmq_port
 
-        self.pending_task_queue: queue.Queue[Any] = queue.Queue(maxsize=10 ** 6)
+        self.pending_task_queue: SortedList[Any] = SortedList(key=lambda tup: (tup[0], tup[1]))
 
         # count of tasks that have been received from the submit side
         self.task_counter = 0
@@ -196,13 +196,12 @@ class Interchange:
             eg. [{'task_id':<x>, 'buffer':<buf>} ... ]
         """
         tasks = []
-        for _ in range(0, count):
-            try:
-                x = self.pending_task_queue.get(block=False)
-            except queue.Empty:
-                break
-            else:
-                tasks.append(x)
+        try:
+            for _ in range(count):
+                *_, task = self.pending_task_queue.pop()
+                tasks.append(task)
+        except IndexError:
+            pass
 
         return tasks
 
@@ -342,8 +341,15 @@ class Interchange:
         if self.task_incoming in self.socks and self.socks[self.task_incoming] == zmq.POLLIN:
             logger.debug("start task_incoming section")
             msg = self.task_incoming.recv_pyobj()
+
+            # Process priority, higher number = lower priority
+            resource_spec = msg.get('resource_spec', {})
+            priority = resource_spec.get('priority', float('inf'))
+            queue_entry = (-priority, -self.task_counter, msg)
+
             logger.debug("putting message onto pending_task_queue")
-            self.pending_task_queue.put(msg)
+
+            self.pending_task_queue.add(queue_entry)
             self.task_counter += 1
             logger.debug(f"Fetched {self.task_counter} tasks so far")
 
@@ -378,23 +384,24 @@ class Interchange:
                 return
 
             if msg['type'] == 'registration':
-                # We set up an entry only if registration works correctly
-                self._ready_managers[manager_id] = {'last_heartbeat': time.time(),
-                                                    'idle_since': time.time(),
-                                                    'block_id': None,
-                                                    'start_time': msg['start_time'],
-                                                    'max_capacity': 0,
-                                                    'worker_count': 0,
-                                                    'active': True,
-                                                    'draining': False,
-                                                    'parsl_version': msg['parsl_v'],
-                                                    'python_version': msg['python_v'],
-                                                    'tasks': []}
-                self.connected_block_history.append(msg['block_id'])
-
-                interesting_managers.add(manager_id)
-                logger.info(f"Adding manager: {manager_id!r} to ready queue")
-                m = self._ready_managers[manager_id]
+                ix_minor_py = self.current_platform['python_v'].rsplit(".", 1)[0]
+                ix_parsl_v = self.current_platform['parsl_v']
+                mgr_minor_py = msg['python_v'].rsplit(".", 1)[0]
+                mgr_parsl_v = msg['parsl_v']
+
+                m = ManagerRecord(
+                    block_id=None,
+                    start_time=msg['start_time'],
+                    tasks=[],
+                    worker_count=0,
+                    max_capacity=0,
+                    active=True,
+                    draining=False,
+                    last_heartbeat=time.time(),
+                    idle_since=time.time(),
+                    parsl_version=mgr_parsl_v,
+                    python_version=msg['python_v'],
+                )
 
                 # m is a ManagerRecord, but msg is a dict[Any,Any] and so can
                 # contain arbitrary fields beyond those in ManagerRecord (and
@@ -405,23 +412,43 @@ class Interchange:
                 logger.info(f"Registration info for manager {manager_id!r}: {msg}")
                 self._send_monitoring_info(monitoring_radio, m)
 
-                if (msg['python_v'].rsplit(".", 1)[0] != self.current_platform['python_v'].rsplit(".", 1)[0] or
-                    msg['parsl_v'] != self.current_platform['parsl_v']):
-                    logger.error(f"Manager {manager_id!r} has incompatible version info with the interchange")
-                    logger.debug("Setting kill event")
+                if (mgr_minor_py, mgr_parsl_v) != (ix_minor_py, ix_parsl_v):
                     kill_event.set()
-                    e = VersionMismatch("py.v={} parsl.v={}".format(self.current_platform['python_v'].rsplit(".", 1)[0],
-                                                                    self.current_platform['parsl_v']),
-                                        "py.v={} parsl.v={}".format(msg['python_v'].rsplit(".", 1)[0],
-                                                                    msg['parsl_v'])
-                                        )
-                    result_package = {'type': 'result', 'task_id': -1, 'exception': serialize_object(e)}
+                    e = VersionMismatch(
+                        f"py.v={ix_minor_py} parsl.v={ix_parsl_v}",
+                        f"py.v={mgr_minor_py} parsl.v={mgr_parsl_v}",
+                    )
+                    result_package = {
+                        'type': 'result',
+                        'task_id': -1,
+                        'exception': serialize_object(e),
+                    }
                     pkl_package = pickle.dumps(result_package)
                     self.results_outgoing.send(pkl_package)
-                    logger.error("Sent failure reports, shutting down interchange")
+                    logger.error(
+                        "Manager has incompatible version info with the interchange;"
+                        " sending failure reports and shutting down:"
+                        f"\n  Interchange: {e.interchange_version}"
+                        f"\n  Manager:     {e.manager_version}"
+                    )
+
                 else:
-                    logger.info(f"Manager {manager_id!r} has compatible Parsl version {msg['parsl_v']}")
-                    logger.info(f"Manager {manager_id!r} has compatible Python version {msg['python_v'].rsplit('.', 1)[0]}")
+                    # We really should update the associated data structure; but not
+                    # at this time.  *kicks can down the road*
+                    assert m['block_id'] is not None, "Verified externally currently"
+
+                    # set up entry only if we accept the registration
+                    self._ready_managers[manager_id] = m
+                    self.connected_block_history.append(m['block_id'])
+
+                    interesting_managers.add(manager_id)
+
+                    logger.info(
+                        f"Registered manager {manager_id!r} (py{mgr_minor_py},"
+                        f" {mgr_parsl_v}) and added to ready queue"
+                    )
+                    logger.debug("Manager %r -> %s", manager_id, m)
+
             elif msg['type'] == 'heartbeat':
                 manager = self._ready_managers.get(manager_id)
                 if manager:
@@ -461,10 +488,10 @@ class Interchange:
             len(self._ready_managers)
         )
 
-        if interesting_managers and not self.pending_task_queue.empty():
+        if interesting_managers and self.pending_task_queue:
             shuffled_managers = self.manager_selector.sort_managers(self._ready_managers, interesting_managers)
 
-            while shuffled_managers and not self.pending_task_queue.empty():  # cf. the if statement above...
+            while shuffled_managers and self.pending_task_queue:  # cf. the if statement above...
                 manager_id = shuffled_managers.pop()
                 m = self._ready_managers[manager_id]
                 tasks_inflight = len(m['tasks'])

{parsl-2025.5.26 → parsl-2025.6.2}/parsl/monitoring/radios/base.py

@@ -1,8 +1,5 @@
 import logging
 from abc import ABCMeta, abstractmethod
-from typing import Optional
-
-_db_manager_excepts: Optional[Exception]
 
 logger = logging.getLogger(__name__)
 

{parsl-2025.5.26 → parsl-2025.6.2}/parsl/providers/slurm/slurm.py

@@ -2,6 +2,7 @@ import logging
 import math
 import os
 import re
+import sys
 import time
 from typing import Any, Dict, Optional
 
@@ -50,6 +51,28 @@ squeue_translate_table = {
 }
 
 
+if sys.version_info < (3, 12):
+    from itertools import islice
+    from typing import Iterable
+
+    def batched(
+        iterable: Iterable[tuple[object, Any]], n: int, *, strict: bool = False
+    ):
+        """Batched
+        Turns a list into a batch of size n. This code is from
+        https://docs.python.org/3.12/library/itertools.html#itertools.batched
+        and in versions 3.12+ this can be replaced with
+        itertools.batched
+        """
+        if n < 1:
+            raise ValueError("n must be at least one")
+        iterator = iter(iterable)
+        while batch := tuple(islice(iterator, n)):
+            yield batch
+else:
+    from itertools import batched
+
+
 class SlurmProvider(ClusterProvider, RepresentationMixin):
     """Slurm Execution Provider
 
@@ -99,6 +122,12 @@ class SlurmProvider(ClusterProvider, RepresentationMixin):
         symbolic group for the job ID.
     worker_init : str
         Command to be run before starting a worker, such as 'module load Anaconda; source activate env'.
+    cmd_timeout : int (Default = 10)
+        Number of seconds to wait for slurm commands to finish. For schedulers with many this
+        may need to be increased to wait longer for scheduler information.
+    status_batch_size: int (Default = 50)
+        Number of jobs to batch together in calls to the scheduler status. For schedulers
+        with many jobs this may need to be decreased to get jobs in smaller batches.
     exclusive : bool (Default = True)
         Requests nodes which are not shared with other running jobs.
     launcher : Launcher
@@ -127,6 +156,7 @@ class SlurmProvider(ClusterProvider, RepresentationMixin):
                  regex_job_id: str = r"Submitted batch job (?P<id>\S*)",
                  worker_init: str = '',
                  cmd_timeout: int = 10,
+                 status_batch_size: int = 50,
                  exclusive: bool = True,
                  launcher: Launcher = SingleNodeLauncher()):
         label = 'slurm'
@@ -148,6 +178,8 @@ class SlurmProvider(ClusterProvider, RepresentationMixin):
         self.qos = qos
         self.constraint = constraint
         self.clusters = clusters
+        # Used to batch requests to sacct/squeue for long jobs lists
+        self.status_batch_size = status_batch_size
         self.scheduler_options = scheduler_options + '\n'
         if exclusive:
             self.scheduler_options += "#SBATCH --exclusive\n"
@@ -199,22 +231,23 @@ class SlurmProvider(ClusterProvider, RepresentationMixin):
         Returns:
               [status...] : Status list of all jobs
         '''
-        job_id_list = ','.join(
-            [jid for jid, job in self.resources.items() if not job['status'].terminal]
-        )
-        if not job_id_list:
-            logger.debug('No active jobs, skipping status update')
+        active_jobs = {jid: job for jid, job in self.resources.items() if not job["status"].terminal}
+        if len(active_jobs) == 0:
+            logger.debug("No active jobs, skipping status update")
             return
 
-        cmd = self._cmd.format(job_id_list)
-        logger.debug("Executing %s", cmd)
-        retcode, stdout, stderr = self.execute_wait(cmd)
-        logger.debug("sacct/squeue returned %s %s", stdout, stderr)
-
-        # Execute_wait failed. Do no update
-        if retcode != 0:
-            logger.warning("sacct/squeue failed with non-zero exit code {}".format(retcode))
-            return
+        stdout = ""
+        for job_batch in batched(active_jobs.items(), self.status_batch_size):
+            job_id_list = ",".join(jid for jid, job in job_batch if not job["status"].terminal)
+            cmd = self._cmd.format(job_id_list)
+            logger.debug("Executing %s", cmd)
+            retcode, batch_stdout, batch_stderr = self.execute_wait(cmd)
+            logger.debug(f"sacct/squeue returned {retcode} {batch_stdout} {batch_stderr}")
+            stdout += batch_stdout
+            # Execute_wait failed. Do no update
+            if retcode != 0:
+                logger.warning("sacct/squeue failed with non-zero exit code {}".format(retcode))
+                return
 
         jobs_missing = set(self.resources.keys())
         for line in stdout.split('\n'):

{parsl-2025.5.26 → parsl-2025.6.2}/parsl/tests/conftest.py

@@ -167,6 +167,14 @@ def pytest_configure(config):
         'markers',
         'issue_3620: Marks tests that do not work correctly on GlobusComputeExecutor (ref: issue 3620)'
     )
+    config.addinivalue_line(
+        'markers',
+        'workqueue: Marks local tests that require a working Work Queue installation'
+    )
+    config.addinivalue_line(
+        'markers',
+        'taskvine: Marks local tests that require a working Task Vine installation'
+    )
 
 
 @pytest.fixture(autouse=True, scope='session')
@@ -400,13 +408,13 @@ def try_assert():
         check_period_ms: int = 20,
     ):
         tb = create_traceback(start=1)
-        timeout_s = abs(timeout_ms) / 1000.0
         check_period_s = abs(check_period_ms) / 1000.0
         if attempts > 0:
             for _attempt_no in range(attempts):
+                time.sleep(random.random() * check_period_s)  # jitter
+                check_period_s *= 2
                 if test_func():
                     return
-                time.sleep(check_period_s)
             else:
                 att_fail = (
                     f"\n  (Still failing after attempt limit [{attempts}], testing"
@@ -415,12 +423,15 @@ def try_assert():
                 exc = AssertionError(f"{str(fail_msg)}{att_fail}".strip())
                 raise exc.with_traceback(tb)
 
-        elif timeout_s > 0:
+        elif timeout_ms > 0:
+            timeout_s = abs(timeout_ms) / 1000.0
             end = time.monotonic() + timeout_s
             while time.monotonic() < end:
+                wait_for = random.random() * check_period_s  # jitter
+                time.sleep(min(wait_for, end - time.monotonic()))
+                check_period_s *= 2
                 if test_func():
                     return
-                time.sleep(check_period_s)
             att_fail = (
                 f"\n  (Still failing after timeout [{timeout_ms}ms], with attempts "
                 f"every {check_period_ms}ms)"

{parsl-2025.5.26 → parsl-2025.6.2}/parsl/tests/test_error_handling/test_resource_spec.py

@@ -23,7 +23,7 @@ def test_resource(n=2):
             break
 
     # Specify incorrect number of resources
-    spec = {'cores': 2, 'memory': 1000}
+    spec = {'cores': 1, 'memory': 1}
     fut = double(n, parsl_resource_specification=spec)
     try:
         fut.result()
@@ -35,7 +35,7 @@ def test_resource(n=2):
 
     # Specify resources with wrong types
     # 'cpus' is incorrect, should be 'cores'
-    spec = {'cpus': 2, 'memory': 1000, 'disk': 1000}
+    spec = {'cpus': 1, 'memory': 1, 'disk': 1}
     fut = double(n, parsl_resource_specification=spec)
     try:
         fut.result()

parsl-2025.6.2/parsl/tests/test_htex/test_priority_queue.py

@@ -0,0 +1,70 @@
+import pytest
+
+import parsl
+from parsl.app.app import python_app
+from parsl.config import Config
+from parsl.executors import HighThroughputExecutor
+from parsl.executors.high_throughput.manager_selector import RandomManagerSelector
+from parsl.providers import LocalProvider
+from parsl.usage_tracking.levels import LEVEL_1
+
+
+@python_app
+def fake_task(parsl_resource_specification=None):
+    import time
+    return time.time()
+
+
+@pytest.mark.local
+def test_priority_queue():
+    provider = LocalProvider(
+        init_blocks=0,
+        max_blocks=0,
+        min_blocks=0,
+    )
+
+    htex = HighThroughputExecutor(
+        label="htex_local",
+        max_workers_per_node=1,
+        manager_selector=RandomManagerSelector(),
+        provider=provider,
+    )
+
+    config = Config(
+        executors=[htex],
+        strategy="htex_auto_scale",
+        usage_tracking=LEVEL_1,
+    )
+
+    with parsl.load(config):
+        futures = {}
+
+        # Submit tasks with mixed priorities
+        # Priorities: [10, 10, 5, 5, 1, 1] to test fallback behavior
+        for i, priority in enumerate([10, 10, 5, 5, 1, 1]):
+            spec = {'priority': priority}
+            futures[(priority, i)] = fake_task(parsl_resource_specification=spec)
+
+        provider.max_blocks = 1
+
+        # Wait for completion
+        results = {
+            key: future.result() for key, future in futures.items()
+        }
+
+        # Sort by finish time
+        sorted_by_completion = sorted(results.items(), key=lambda item: item[1])
+        execution_order = [key for key, _ in sorted_by_completion]
+
+        # check priority queue functionality
+        priorities_only = [p for (p, i) in execution_order]
+        assert priorities_only == sorted(priorities_only), "Priority execution order failed"
+
+        # check FIFO fallback
+        from collections import defaultdict
+        seen = defaultdict(list)
+        for (priority, idx) in execution_order:
+            seen[priority].append(idx)
+
+        for priority, indices in seen.items():
+            assert indices == sorted(indices), f"FIFO fallback violated for priority {priority}"

{parsl-2025.5.26 → parsl-2025.6.2}/parsl/tests/test_htex/test_resource_spec_validation.py

@@ -36,3 +36,10 @@ def test_resource_spec_validation_bad_keys():
 
     with pytest.raises(InvalidResourceSpecification):
         htex.validate_resource_spec({"num_nodes": 2})
+
+
+@pytest.mark.local
+def test_resource_spec_validation_one_key():
+    htex = HighThroughputExecutor()
+    ret_val = htex.validate_resource_spec({"priority": 2})
+    assert ret_val is None

{parsl-2025.5.26 → parsl-2025.6.2}/parsl/tests/test_monitoring/test_basic.py

@@ -6,7 +6,6 @@ import pytest
 import parsl
 from parsl import HighThroughputExecutor
 from parsl.config import Config
-from parsl.executors.taskvine import TaskVineExecutor, TaskVineManagerConfig
 from parsl.monitoring import MonitoringHub
 
 
@@ -54,9 +53,27 @@ def htex_filesystem_config():
     return c
 
 
-@pytest.mark.local
-@pytest.mark.parametrize("fresh_config", [htex_config, htex_filesystem_config, htex_udp_config])
-def test_row_counts(tmpd_cwd, fresh_config):
+def workqueue_config():
+    from parsl.tests.configs.workqueue_ex import fresh_config
+    c = fresh_config()
+    c.monitoring = MonitoringHub(
+                        hub_address="localhost",
+                        resource_monitoring_interval=1)
+    return c
+
+
+def taskvine_config():
+    from parsl.executors.taskvine import TaskVineExecutor, TaskVineManagerConfig
+    c = Config(executors=[TaskVineExecutor(manager_config=TaskVineManagerConfig(port=9000),
+                                           worker_launch_method='provider')],
+               strategy_period=0.5,
+
+               monitoring=MonitoringHub(hub_address="localhost",
+                                        resource_monitoring_interval=1))
+    return c
+
+
+def row_counts_parametrized(tmpd_cwd, fresh_config):
     # this is imported here rather than at module level because
     # it isn't available in a plain parsl install, so this module
     # would otherwise fail to import and break even a basic test
@@ -113,3 +130,23 @@ def test_row_counts(tmpd_cwd, fresh_config):
         result = connection.execute(text("SELECT COUNT(*) FROM resource"))
         (c, ) = result.first()
         assert c >= 1
+
+
+@pytest.mark.local
+@pytest.mark.parametrize("fresh_config", [htex_config, htex_filesystem_config, htex_udp_config])
+def test_row_counts_base(tmpd_cwd, fresh_config):
+    row_counts_parametrized(tmpd_cwd, fresh_config)
+
+
+@pytest.mark.workqueue
+@pytest.mark.local
+@pytest.mark.parametrize("fresh_config", [workqueue_config])
+def test_row_counts_wq(tmpd_cwd, fresh_config):
+    row_counts_parametrized(tmpd_cwd, fresh_config)
+
+
+@pytest.mark.taskvine
+@pytest.mark.local
+@pytest.mark.parametrize("fresh_config", [taskvine_config])
+def test_row_counts_tv(tmpd_cwd, fresh_config):
+    row_counts_parametrized(tmpd_cwd, fresh_config)

{parsl-2025.5.26 → parsl-2025.6.2}/parsl/tests/test_scaling/test_scale_down.py

@@ -37,6 +37,7 @@ def local_config():
         ],
         max_idletime=0.5,
         strategy='simple',
+        strategy_period=0.5,
     )
 
 

{parsl-2025.5.26 → parsl-2025.6.2}/parsl/usage_tracking/usage.py

@@ -240,5 +240,6 @@ class UsageTracker:
                 if proc.is_alive():
                     logger.warning("Usage tracking process did not end itself; sending SIGKILL")
                     proc.kill()
+                    proc.join(timeout=timeout)
 
                 proc.close()

{parsl-2025.5.26 → parsl-2025.6.2}/parsl/version.py

@@ -3,4 +3,4 @@
 Year.Month.Day[alpha/beta/..]
 Alphas will be numbered like this -> 2024.12.10a0
 """
-VERSION = '2025.05.26'
+VERSION = '2025.06.02'

{parsl-2025.5.26 → parsl-2025.6.2/parsl.egg-info}/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: parsl
-Version: 2025.5.26
+Version: 2025.6.2
 Summary: Simple data dependent workflows in Python
 Home-page: https://github.com/Parsl/parsl
-Download-URL: https://github.com/Parsl/parsl/archive/2025.05.26.tar.gz
+Download-URL: https://github.com/Parsl/parsl/archive/2025.06.02.tar.gz
 Author: The Parsl Team
 Author-email: parsl@googlegroups.com
 License: Apache 2.0

{parsl-2025.5.26 → parsl-2025.6.2}/parsl.egg-info/SOURCES.txt

@@ -343,6 +343,7 @@ parsl/tests/test_htex/test_manager_selector_by_block.py
 parsl/tests/test_htex/test_managers_command.py
 parsl/tests/test_htex/test_missing_worker.py
 parsl/tests/test_htex/test_multiple_disconnected_blocks.py
+parsl/tests/test_htex/test_priority_queue.py
 parsl/tests/test_htex/test_resource_spec_validation.py
 parsl/tests/test_htex/test_worker_failure.py
 parsl/tests/test_htex/test_zmq_binding.py

{parsl-2025.5.26 → parsl-2025.6.2}/parsl.egg-info/requires.txt

@@ -5,6 +5,7 @@ globus-sdk
 dill
 tblib
 requests
+sortedcontainers
 psutil>=5.5.1
 setproctitle
 filelock<4,>=3.13

{parsl-2025.5.26 → parsl-2025.6.2}/requirements.txt

@@ -12,6 +12,7 @@ globus-sdk
 dill
 tblib
 requests
+sortedcontainers
 psutil>=5.5.1
 setproctitle
 filelock>=3.13,<4