parsl 2024.1.1__tar.gz → 2024.1.15__tar.gz

{parsl-2024.1.1/parsl.egg-info → parsl-2024.1.15}/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: parsl
-Version: 2024.1.1
+Version: 2024.1.15
 Summary: Simple data dependent workflows in Python
 Home-page: https://github.com/Parsl/parsl
-Download-URL: https://github.com/Parsl/parsl/archive/2024.01.01.tar.gz
+Download-URL: https://github.com/Parsl/parsl/archive/2024.01.15.tar.gz
 Author: The Parsl Team
 Author-email: parsl@googlegroups.com
 License: Apache 2.0

{parsl-2024.1.1 → parsl-2024.1.15}/parsl/addresses.py

@@ -72,8 +72,9 @@ def address_by_hostname() -> str:
     """
     logger.debug("Finding address by using local hostname")
     addr = platform.node()
-    logger.debug("Address found: {}".format(addr))
-    return addr
+    ip_addr = socket.gethostbyname(addr)
+    logger.debug("Address found: {}".format(ip_addr))
+    return ip_addr
 
 
 @typeguard.typechecked

{parsl-2024.1.1 → parsl-2024.1.15}/parsl/dataflow/dflow.py

@@ -177,7 +177,9 @@ class DataFlowKernel:
 
         # this must be set before executors are added since add_executors calls
         # job_status_poller.add_executors.
-        self.job_status_poller = JobStatusPoller(self)
+        self.job_status_poller = JobStatusPoller(strategy=self.config.strategy,
+                                                 max_idletime=self.config.max_idletime,
+                                                 dfk=self)
 
         self.executors: Dict[str, ParslExecutor] = {}
 
@@ -526,9 +528,7 @@ class DataFlowKernel:
         # or do nothing?
         if self.checkpoint_mode == 'task_exit':
             self.checkpoint(tasks=[task_record])
-        elif self.checkpoint_mode == 'manual' or \
-                self.checkpoint_mode == 'periodic' or \
-                self.checkpoint_mode == 'dfk_exit':
+        elif self.checkpoint_mode in ('manual', 'periodic', 'dfk_exit'):
             with self.checkpoint_lock:
                 self.checkpointable_tasks.append(task_record)
         elif self.checkpoint_mode is None:
@@ -1298,11 +1298,7 @@ class DataFlowKernel:
                         hashsum = task_record['hashsum']
                         if not hashsum:
                             continue
-                        t = {'hash': hashsum,
-                             'exception': None,
-                             'result': None}
-
-                        t['result'] = app_fu.result()
+                        t = {'hash': hashsum, 'exception': None, 'result': app_fu.result()}
 
                         # We are using pickle here since pickle dumps to a file in 'ab'
                         # mode behave like a incremental log.

{parsl-2024.1.1 → parsl-2024.1.15}/parsl/executors/high_throughput/executor.py

@@ -205,7 +205,6 @@ class HighThroughputExecutor(BlockProviderExecutor, RepresentationMixin):
 
         BlockProviderExecutor.__init__(self, provider=provider, block_error_handler=block_error_handler)
         self.label = label
-        self.launch_cmd = launch_cmd
         self.worker_debug = worker_debug
         self.storage_access = storage_access
         self.working_dir = working_dir
@@ -259,21 +258,25 @@ class HighThroughputExecutor(BlockProviderExecutor, RepresentationMixin):
         self.cpu_affinity = cpu_affinity
 
         if not launch_cmd:
-            self.launch_cmd = ("process_worker_pool.py {debug} {max_workers} "
-                               "-a {addresses} "
-                               "-p {prefetch_capacity} "
-                               "-c {cores_per_worker} "
-                               "-m {mem_per_worker} "
-                               "--poll {poll_period} "
-                               "--task_port={task_port} "
-                               "--result_port={result_port} "
-                               "--logdir={logdir} "
-                               "--block_id={{block_id}} "
-                               "--hb_period={heartbeat_period} "
-                               "{address_probe_timeout_string} "
-                               "--hb_threshold={heartbeat_threshold} "
-                               "--cpu-affinity {cpu_affinity} "
-                               "--available-accelerators {accelerators}")
+            launch_cmd = (
+                "process_worker_pool.py {debug} {max_workers} "
+                "-a {addresses} "
+                "-p {prefetch_capacity} "
+                "-c {cores_per_worker} "
+                "-m {mem_per_worker} "
+                "--poll {poll_period} "
+                "--task_port={task_port} "
+                "--result_port={result_port} "
+                "--logdir={logdir} "
+                "--block_id={{block_id}} "
+                "--hb_period={heartbeat_period} "
+                "{address_probe_timeout_string} "
+                "--hb_threshold={heartbeat_threshold} "
+                "--cpu-affinity {cpu_affinity} "
+                "--available-accelerators {accelerators}"
+            )
+
+        self.launch_cmd = launch_cmd
 
     radio_mode = "htex"
 

{parsl-2024.1.1 → parsl-2024.1.15}/parsl/executors/workqueue/executor.py

@@ -849,7 +849,7 @@ def _work_queue_submit_wait(*,
             break
 
         # Submit tasks
-        while task_queue.qsize() > 0 and not should_stop.value:
+        while task_queue.qsize() > 0 or q.empty() and not should_stop.value:
             # Obtain task from task_queue
             try:
                 task = task_queue.get(timeout=1)

{parsl-2024.1.1 → parsl-2024.1.15}/parsl/jobs/job_status_poller.py

@@ -2,7 +2,7 @@ import logging
 import parsl
 import time
 import zmq
-from typing import Dict, List, Sequence
+from typing import Dict, List, Sequence, Optional
 
 from parsl.jobs.states import JobStatus, JobState
 from parsl.jobs.strategy import Strategy
@@ -17,7 +17,7 @@ logger = logging.getLogger(__name__)
 
 
 class PollItem:
-    def __init__(self, executor: BlockProviderExecutor, dfk: "parsl.dataflow.dflow.DataFlowKernel"):
+    def __init__(self, executor: BlockProviderExecutor, dfk: Optional["parsl.dataflow.dflow.DataFlowKernel"] = None):
         self._executor = executor
         self._dfk = dfk
         self._interval = executor.status_polling_interval
@@ -26,7 +26,7 @@ class PollItem:
 
         # Create a ZMQ channel to send poll status to monitoring
         self.monitoring_enabled = False
-        if self._dfk.monitoring is not None:
+        if self._dfk and self._dfk.monitoring is not None:
             self.monitoring_enabled = True
             hub_address = self._dfk.hub_address
             hub_port = self._dfk.hub_interchange_port
@@ -100,11 +100,12 @@ class PollItem:
 
 
 class JobStatusPoller(Timer):
-    def __init__(self, dfk: "parsl.dataflow.dflow.DataFlowKernel") -> None:
+    def __init__(self, strategy: Optional[str] = None, max_idletime: float = 0.0,
+                 dfk: Optional["parsl.dataflow.dflow.DataFlowKernel"] = None) -> None:
         self._poll_items = []  # type: List[PollItem]
         self.dfk = dfk
-        self._strategy = Strategy(strategy=dfk.config.strategy,
-                                  max_idletime=dfk.config.max_idletime)
+        self._strategy = Strategy(strategy=strategy,
+                                  max_idletime=max_idletime)
         super().__init__(self.poll, interval=5, name="JobStatusPoller")
 
     def poll(self) -> None:

{parsl-2024.1.1 → parsl-2024.1.15}/parsl/providers/errors.py

@@ -1,5 +1,7 @@
 from parsl.errors import ParslError
 
+import warnings
+
 
 class ExecutionProviderException(ParslError):
     """ Base class for all exceptions
@@ -46,18 +48,23 @@ class ScriptPathError(ExecutionProviderException):
 
 
 class SubmitException(ExecutionProviderException):
-    '''Raised by the submit() method of a provider if there is an error in launching a task.
+    '''Raised by the submit() method of a provider if there is an error in launching a job.
     '''
 
-    def __init__(self, task_name, message, stdout=None, stderr=None):
-        self.task_name = task_name
+    def __init__(self, job_name, message, stdout=None, stderr=None):
+        self.job_name = job_name
         self.message = message
         self.stdout = stdout
         self.stderr = stderr
 
+    @property
+    def task_name(self) -> str:
+        warnings.warn("task_name is deprecated; use .job_name instead. This will be removed after 2024-06.", DeprecationWarning)
+        return self.job_name
+
     def __str__(self):
         # TODO: make this more user-friendly
-        return "Cannot launch task {0}: {1}; stdout={2}, stderr={3}".format(self.task_name,
-                                                                            self.message,
-                                                                            self.stdout,
-                                                                            self.stderr)
+        return "Cannot launch job {0}: {1}; stdout={2}, stderr={3}".format(self.job_name,
+                                                                           self.message,
+                                                                           self.stdout,
+                                                                           self.stderr)

{parsl-2024.1.1 → parsl-2024.1.15}/parsl/tests/conftest.py

@@ -3,6 +3,8 @@ import itertools
 import logging
 import os
 import pathlib
+import re
+import shutil
 import time
 import types
 import signal
@@ -20,6 +22,7 @@ import _pytest.runner as runner
 
 import parsl
 from parsl.dataflow.dflow import DataFlowKernelLoader
+from parsl.utils import RepresentationMixin
 
 logger = logging.getLogger(__name__)
 
@@ -44,17 +47,46 @@ def pytest_sessionstart(session):
 
 
 @pytest.fixture(scope="session")
-def tmpd_cwd_session():
-    n = datetime.now().strftime('%Y%m%d.%H%I%S')
-    with tempfile.TemporaryDirectory(dir=os.getcwd(), prefix=f".pytest-{n}-") as tmpd:
-        yield pathlib.Path(tmpd)
+def tmpd_cwd_session(pytestconfig):
+    config = re.sub(r"[^A-z0-9_-]+", "_", pytestconfig.getoption('config')[0])
+    cwd = pathlib.Path(os.getcwd())
+    pytest_dir = cwd / ".pytest"
+    pytest_dir.mkdir(mode=0o700, parents=True, exist_ok=True)
+
+    test_dir_prefix = "parsltest-"
+    link = pytest_dir / f"{test_dir_prefix}current"
+    link.unlink(missing_ok=True)
+    n = datetime.now().strftime('%Y%m%d.%H%M%S')
+    tmpd = tempfile.mkdtemp(dir=pytest_dir, prefix=f"{test_dir_prefix}{n}-{config}-")
+    tmpd = pathlib.Path(tmpd)
+    link.symlink_to(tmpd.name)
+    yield link
+
+    try:
+        preserve = int(os.getenv("PARSL_TEST_PRESERVE_NUM_RUNS", "3"))
+    except Exception:
+        preserve = 3
+
+    test_runs = sorted(
+        d for d in pytest_dir.glob(f"{test_dir_prefix}*")
+        if d.is_dir() and not d.is_symlink()
+    )
+    for run_to_remove in test_runs[:-preserve]:
+        run_to_remove.chmod(0o700)
+        for root, subdirnames, fnames in os.walk(run_to_remove):
+            rpath = pathlib.Path(root)
+            for d in subdirnames:
+                (rpath / d).lchmod(0o700)
+            for f in fnames:
+                (rpath / f).lchmod(0o600)
+        shutil.rmtree(run_to_remove)
 
 
 @pytest.fixture
 def tmpd_cwd(tmpd_cwd_session, request):
     prefix = f"{request.node.name}-"
-    with tempfile.TemporaryDirectory(dir=tmpd_cwd_session, prefix=prefix) as tmpd:
-        yield pathlib.Path(tmpd)
+    tmpd = tempfile.mkdtemp(dir=tmpd_cwd_session, prefix=prefix)
+    yield pathlib.Path(tmpd)
 
 
 def pytest_addoption(parser):
@@ -118,7 +150,7 @@ def pytest_configure(config):
 
 
 @pytest.fixture(autouse=True, scope='session')
-def load_dfk_session(request, pytestconfig):
+def load_dfk_session(request, pytestconfig, tmpd_cwd_session):
     """Load a dfk around entire test suite, except in local mode.
 
     The special path `local` indicates that configuration will not come
@@ -126,6 +158,8 @@ def load_dfk_session(request, pytestconfig):
     load_dfk_local_module for module-level configuration management.
     """
 
+    RepresentationMixin._validate_repr = True
+
     config = pytestconfig.getoption('config')[0]
 
     if config != 'local':
@@ -137,12 +171,19 @@ def load_dfk_session(request, pytestconfig):
             raise RuntimeError("DFK didn't start as None - there was a DFK from somewhere already")
 
         if hasattr(module, 'config'):
-            dfk = parsl.load(module.config)
+            parsl_conf = module.config
         elif hasattr(module, 'fresh_config'):
-            dfk = parsl.load(module.fresh_config())
+            parsl_conf = module.fresh_config()
         else:
             raise RuntimeError("Config module does not define config or fresh_config")
 
+        if parsl_conf.run_dir == "runinfo":  # the default
+            parsl_conf.run_dir = tmpd_cwd_session / parsl_conf.run_dir
+        dfk = parsl.load(parsl_conf)
+
+        for ex in dfk.executors.values():
+            ex.working_dir = tmpd_cwd_session
+
         yield
 
         if parsl.dfk() != dfk:
@@ -154,7 +195,7 @@ def load_dfk_session(request, pytestconfig):
 
 
 @pytest.fixture(autouse=True, scope='module')
-def load_dfk_local_module(request, pytestconfig):
+def load_dfk_local_module(request, pytestconfig, tmpd_cwd_session):
     """Load the dfk around test modules, in local mode.
 
     If local_config is specified in the test module, it will be loaded using
@@ -165,6 +206,7 @@ def load_dfk_local_module(request, pytestconfig):
     be used to perform more interesting DFK initialisation not possible with
     local_config.
     """
+    RepresentationMixin._validate_repr = True
 
     config = pytestconfig.getoption('config')[0]
 
@@ -177,8 +219,15 @@ def load_dfk_local_module(request, pytestconfig):
             assert callable(local_config)
             c = local_config()
             assert isinstance(c, parsl.Config)
+
+            if c.run_dir == "runinfo":  # the default
+                c.run_dir = tmpd_cwd_session / c.run_dir
+
             dfk = parsl.load(c)
 
+            for ex in dfk.executors.values():
+                ex.working_dir = tmpd_cwd_session
+
         if callable(local_setup):
             local_setup()
 

parsl-2024.1.15/parsl/tests/test_checkpointing/test_python_checkpoint_1.py

@@ -0,0 +1,33 @@
+import os
+import pytest
+
+import parsl
+from parsl import python_app
+from parsl.tests.configs.local_threads import fresh_config
+
+
+def local_config():
+    config = fresh_config()
+    config.checkpoint_mode = "manual"
+    return config
+
+
+@python_app(cache=True)
+def uuid_app():
+    import uuid
+    return uuid.uuid4()
+
+
+@pytest.mark.local
+def test_initial_checkpoint_write():
+    """1. Launch a few apps and write the checkpoint once a few have completed
+    """
+    uuid_app().result()
+
+    cpt_dir = parsl.dfk().checkpoint()
+
+    cptpath = cpt_dir + '/dfk.pkl'
+    assert os.path.exists(cptpath), f"DFK checkpoint missing: {cptpath}"
+
+    cptpath = cpt_dir + '/tasks.pkl'
+    assert os.path.exists(cptpath), f"Tasks checkpoint missing: {cptpath}"

parsl-2024.1.15/parsl/tests/test_checkpointing/test_python_checkpoint_2.py

@@ -0,0 +1,42 @@
+import contextlib
+import os
+import pytest
+import parsl
+from parsl import python_app
+
+from parsl.tests.configs.local_threads_checkpoint import fresh_config
+
+
+@contextlib.contextmanager
+def parsl_configured(run_dir, **kw):
+    c = fresh_config()
+    c.run_dir = run_dir
+    for config_attr, config_val in kw.items():
+        setattr(c, config_attr, config_val)
+    dfk = parsl.load(c)
+    for ex in dfk.executors.values():
+        ex.working_dir = run_dir
+    yield dfk
+
+    parsl.dfk().cleanup()
+    parsl.clear()
+
+
+@python_app(cache=True)
+def uuid_app():
+    import uuid
+    return uuid.uuid4()
+
+
+@pytest.mark.local
+def test_loading_checkpoint(tmpd_cwd):
+    """Load memoization table from previous checkpoint
+    """
+    with parsl_configured(tmpd_cwd, checkpoint_mode="task_exit"):
+        checkpoint_files = [os.path.join(parsl.dfk().run_dir, "checkpoint")]
+        result = uuid_app().result()
+
+    with parsl_configured(tmpd_cwd, checkpoint_files=checkpoint_files):
+        relaunched = uuid_app().result()
+
+    assert result == relaunched, "Expected following call to uuid_app to return cached uuid"

parsl-2024.1.15/parsl/tests/test_providers/test_submiterror_deprecation.py

@@ -0,0 +1,21 @@
+import pytest
+import random
+import string
+
+from parsl.providers.errors import SubmitException
+
+
+@pytest.mark.local
+def test_submit_exception_task_name_deprecation():
+    """This tests the deprecation warning of task_name in SubmitException
+    """
+    j = "the_name-" + "".join(random.sample(string.ascii_lowercase, 10))
+
+    ex = SubmitException(j, "m")
+
+    # the new behaviour
+    assert ex.job_name == j
+
+    # the old behaviour
+    with pytest.deprecated_call():
+        assert ex.task_name == j

{parsl-2024.1.1 → parsl-2024.1.15}/parsl/tests/test_radical/test_mpi_funcs.py

@@ -5,7 +5,7 @@ from parsl.tests.configs.local_radical_mpi import fresh_config as local_config
 
 
 @parsl.python_app
-def test_mpi_func(msg, sleep, comm=None, parsl_resource_specification={}):
+def some_mpi_func(msg, sleep, comm=None, parsl_resource_specification={}):
     import time
     msg = 'hello %d/%d: %s' % (comm.rank, comm.size, msg)
     time.sleep(sleep)
@@ -17,11 +17,12 @@ apps = []
 
 
 @pytest.mark.local
+@pytest.mark.radical
 def test_radical_mpi(n=7):
     # rank size should be > 1 for the
     # radical runtime system to run this function in MPI env
     for i in range(2, n):
         spec = {'ranks': i}
-        t = test_mpi_func(msg='mpi.func.%06d' % i, sleep=1, comm=None, parsl_resource_specification=spec)
+        t = some_mpi_func(msg='mpi.func.%06d' % i, sleep=1, comm=None, parsl_resource_specification=spec)
         apps.append(t)
     assert [len(app.result()) for app in apps] == list(range(2, n))

parsl-2024.1.15/parsl/tests/test_staging/__init__.py

@@ -0,0 +1,7 @@
+def read_sort_write(in_path, out_path):
+    with open(in_path) as u:
+        strs = u.read().split()
+        strs.sort()
+    with open(out_path, 'w') as s:
+        for e in strs:
+            print(e, file=s)

{parsl-2024.1.1 → parsl-2024.1.15}/parsl/tests/test_staging/test_elaborate_noop_file.py

@@ -20,8 +20,8 @@ logger = logging.getLogger(__name__)
 
 
 @bash_app
-def touch(filename, inputs=[], outputs=[]):
-    return "touch {}".format(filename)
+def touch(filename, outputs=()):
+    return f"touch {filename}"
 
 
 @python_app
@@ -31,52 +31,47 @@ def app_test_in(file):
     pass
 
 
+@pytest.fixture
+def storage_access_parsl():
+    def _setup_config(*args, **kwargs):
+        tpe = ThreadPoolExecutor(
+            label='local_threads',
+            storage_access=[NoOpTestingFileStaging(*args, **kwargs)]
+        )
+        config = Config(executors=[tpe])
+        parsl.load(config)
+
+    yield _setup_config
+
+    parsl.dfk().cleanup()
+    parsl.clear()
+
+
 @pytest.mark.local
-def test_regression_stage_out_does_not_stage_in():
-    no_stageout_config = Config(
-        executors=[
-            ThreadPoolExecutor(
-                label='local_threads',
-                storage_access=[NoOpTestingFileStaging(allow_stage_in=False)]
-            )
-        ]
-    )
-
-    parsl.load(no_stageout_config)
+def test_regression_stage_out_does_not_stage_in(storage_access_parsl, tmpd_cwd):
+    storage_access_parsl(allow_stage_in=False)
 
     # Test that the helper app runs with no staging
-    touch("test.1", outputs=[]).result()
+    touch(str(tmpd_cwd / "test.1"), outputs=[]).result()
 
-    # Test with stage-out, checking that provider stage in is never
-    # invoked. If stage-in is invoked, the the NoOpTestingFileStaging
+    # Test with stage-out, checking that provider stage-in is never
+    # invoked. If stage-in is invoked, then the NoOpTestingFileStaging
     # provider will raise an exception, which should propagate to
     # .result() here.
-    touch("test.2", outputs=[File("test.2")]).result()
+    fpath = tmpd_cwd / "test.2"
+    touch(str(fpath), outputs=[File(fpath)]).result()
 
     # Test that stage-in exceptions propagate out to user code.
     with pytest.raises(NoOpError):
         touch("test.3", inputs=[File("test.3")]).result()
 
-    parsl.dfk().cleanup()
-    parsl.clear()
-
 
 @pytest.mark.local
-def test_regression_stage_in_does_not_stage_out():
-    no_stageout_config = Config(
-        executors=[
-            ThreadPoolExecutor(
-                label='local_threads',
-                storage_access=[NoOpTestingFileStaging(allow_stage_out=False)]
-            )
-        ],
-    )
-
-    parsl.load(no_stageout_config)
-
-    f = open("test.4", "a")
-    f.write("test")
-    f.close()
+def test_regression_stage_in_does_not_stage_out(storage_access_parsl, tmpd_cwd):
+    storage_access_parsl(allow_stage_out=False)
+
+    fpath = tmpd_cwd / "test.4"
+    fpath.write_text("test")
 
     # Test that stage in does not invoke stage out. If stage out is
     # attempted, then the NoOpTestingFileStaging provider will raise
@@ -86,6 +81,3 @@ def test_regression_stage_in_does_not_stage_out():
     # Test that stage out exceptions propagate to user code.
     with pytest.raises(NoOpError):
         touch("test.5", outputs=[File("test.5")]).result()
-
-    parsl.dfk().cleanup()
-    parsl.clear()