fractal-server 1.4.9__py3-none-any.whl → 2.0.0a0__py3-none-any.whl

fractal_server/app/routes/api/v2/workflowtask.py

@@ -0,0 +1,269 @@
+from copy import deepcopy
+from typing import Optional
+
+from fastapi import APIRouter
+from fastapi import Depends
+from fastapi import HTTPException
+from fastapi import Response
+from fastapi import status
+
+from ....db import AsyncSession
+from ....db import get_async_db
+from ....models.v1 import Task
+from ....models.v2 import TaskV2
+from ....schemas.v2 import WorkflowTaskCreateV2
+from ....schemas.v2 import WorkflowTaskReadV2
+from ....schemas.v2 import WorkflowTaskUpdateV2
+from ....security import current_active_user
+from ....security import User
+from ._aux_functions import _get_workflow_check_owner
+from ._aux_functions import _get_workflow_task_check_owner
+from ._aux_functions import _workflow_insert_task
+
+router = APIRouter()
+
+
+@router.post(
+    "/project/{project_id}/workflow/{workflow_id}/wftask/",
+    response_model=WorkflowTaskReadV2,
+    status_code=status.HTTP_201_CREATED,
+)
+async def create_workflowtask(
+    project_id: int,
+    workflow_id: int,
+    task_id: int,
+    new_task: WorkflowTaskCreateV2,
+    user: User = Depends(current_active_user),
+    db: AsyncSession = Depends(get_async_db),
+) -> Optional[WorkflowTaskReadV2]:
+    """
+    Add a WorkflowTask to a Workflow
+    """
+
+    workflow = await _get_workflow_check_owner(
+        project_id=project_id, workflow_id=workflow_id, user_id=user.id, db=db
+    )
+
+    if new_task.is_legacy_task is True:
+        task = await db.get(Task, task_id)
+        if not task.is_v2_compatible:
+            raise HTTPException(
+                status_code=status.HTTP_422_UNPROCESSABLE_ENTITY,
+                detail=f"Task {task_id} is not V2-compatible.",
+            )
+    else:
+        task = await db.get(TaskV2, task_id)
+
+    if not task:
+        if new_task.is_legacy_task:
+            error = f"Task {task_id} not found."
+        else:
+            error = f"TaskV2 {task_id} not found."
+
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail=error
+        )
+
+    if new_task.is_legacy_task is True or task.type == "parallel":
+        if (
+            new_task.meta_non_parallel is not None
+            or new_task.args_non_parallel is not None
+        ):
+            raise HTTPException(
+                status_code=status.HTTP_422_UNPROCESSABLE_ENTITY,
+                detail=(
+                    "Cannot set `WorkflowTaskV2.meta_non_parallel` or "
+                    "`WorkflowTask.args_non_parallel` if the associated Task "
+                    "is `parallel` (or legacy)."
+                ),
+            )
+    elif task.type == "non_parallel":
+        if (
+            new_task.meta_parallel is not None
+            or new_task.args_parallel is not None
+        ):
+            raise HTTPException(
+                status_code=status.HTTP_422_UNPROCESSABLE_ENTITY,
+                detail=(
+                    "Cannot set `WorkflowTaskV2.meta_parallel` or "
+                    "`WorkflowTask.args_parallel` if the associated Task "
+                    "is `non_parallel`."
+                ),
+            )
+
+    async with db:
+        workflow_task = await _workflow_insert_task(
+            workflow_id=workflow.id,
+            is_legacy_task=new_task.is_legacy_task,
+            task_id=task_id,
+            order=new_task.order,
+            meta_non_parallel=new_task.meta_non_parallel,
+            meta_parallel=new_task.meta_parallel,
+            args_non_parallel=new_task.args_non_parallel,
+            args_parallel=new_task.args_parallel,
+            input_filters=new_task.input_filters,
+            db=db,
+        )
+
+    await db.close()
+
+    return workflow_task
+
+
+@router.get(
+    "/project/{project_id}/workflow/{workflow_id}/wftask/{workflow_task_id}/",
+    response_model=WorkflowTaskReadV2,
+)
+async def read_workflowtask(
+    project_id: int,
+    workflow_id: int,
+    workflow_task_id: int,
+    user: User = Depends(current_active_user),
+    db: AsyncSession = Depends(get_async_db),
+):
+    workflow_task, _ = await _get_workflow_task_check_owner(
+        project_id=project_id,
+        workflow_task_id=workflow_task_id,
+        workflow_id=workflow_id,
+        user_id=user.id,
+        db=db,
+    )
+    return workflow_task
+
+
+@router.patch(
+    "/project/{project_id}/workflow/{workflow_id}/wftask/{workflow_task_id}/",
+    response_model=WorkflowTaskReadV2,
+)
+async def update_workflowtask(
+    project_id: int,
+    workflow_id: int,
+    workflow_task_id: int,
+    workflow_task_update: WorkflowTaskUpdateV2,
+    user: User = Depends(current_active_user),
+    db: AsyncSession = Depends(get_async_db),
+) -> Optional[WorkflowTaskReadV2]:
+    """
+    Edit a WorkflowTask of a Workflow
+    """
+
+    db_wf_task, db_workflow = await _get_workflow_task_check_owner(
+        project_id=project_id,
+        workflow_task_id=workflow_task_id,
+        workflow_id=workflow_id,
+        user_id=user.id,
+        db=db,
+    )
+
+    if db_wf_task.task_type == "parallel" and (
+        workflow_task_update.args_non_parallel is not None
+        or workflow_task_update.meta_non_parallel is not None
+    ):
+        raise HTTPException(
+            status_code=status.HTTP_422_UNPROCESSABLE_ENTITY,
+            detail=(
+                "Cannot patch `WorkflowTaskV2.args_non_parallel` or "
+                "`WorkflowTask.meta_non_parallel` if the associated Task is "
+                "parallel."
+            ),
+        )
+    elif db_wf_task.task_type == "non_parallel" and (
+        workflow_task_update.args_parallel is not None
+        or workflow_task_update.meta_parallel is not None
+    ):
+        raise HTTPException(
+            status_code=status.HTTP_422_UNPROCESSABLE_ENTITY,
+            detail=(
+                "Cannot patch `WorkflowTaskV2.args_parallel` or "
+                "`WorkflowTask.meta_parallel` if the associated Task is "
+                "non parallel."
+            ),
+        )
+
+    for key, value in workflow_task_update.dict(exclude_unset=True).items():
+        if key == "args_parallel":
+            # Get default arguments via a Task property method
+            if db_wf_task.is_legacy_task:
+                default_args = (
+                    db_wf_task.task_legacy.default_args_from_args_schema
+                )
+            else:
+                default_args = (
+                    db_wf_task.task.default_args_parallel_from_args_schema
+                )
+            # Override default_args with args value items
+            actual_args = deepcopy(default_args)
+            if value is not None:
+                for k, v in value.items():
+                    actual_args[k] = v
+            if not actual_args:
+                actual_args = None
+            setattr(db_wf_task, key, actual_args)
+        elif key == "args_non_parallel":
+            # Get default arguments via a Task property method
+            default_args = deepcopy(
+                db_wf_task.task.default_args_non_parallel_from_args_schema
+            )
+            # Override default_args with args value items
+            actual_args = default_args.copy()
+            if value is not None:
+                for k, v in value.items():
+                    actual_args[k] = v
+            if not actual_args:
+                actual_args = None
+            setattr(db_wf_task, key, actual_args)
+        elif key == "meta_parallel":
+            current_meta_parallel = deepcopy(db_wf_task.meta_parallel) or {}
+            current_meta_parallel.update(value)
+            setattr(db_wf_task, key, current_meta_parallel)
+        elif key == "meta_non_parallel":
+            current_meta_non_parallel = (
+                deepcopy(db_wf_task.meta_non_parallel) or {}
+            )
+            current_meta_non_parallel.update(value)
+            setattr(db_wf_task, key, current_meta_non_parallel)
+        # FIXME handle `input_filters`
+        else:
+            raise HTTPException(
+                status_code=status.HTTP_422_UNPROCESSABLE_ENTITY,
+                detail=f"patch_workflow_task endpoint cannot set {key=}",
+            )
+
+    await db.commit()
+    await db.refresh(db_wf_task)
+    await db.close()
+
+    return db_wf_task
+
+
+@router.delete(
+    "/project/{project_id}/workflow/{workflow_id}/wftask/{workflow_task_id}/",
+    status_code=status.HTTP_204_NO_CONTENT,
+)
+async def delete_workflowtask(
+    project_id: int,
+    workflow_id: int,
+    workflow_task_id: int,
+    user: User = Depends(current_active_user),
+    db: AsyncSession = Depends(get_async_db),
+) -> Response:
+    """
+    Delete a WorkflowTask of a Workflow
+    """
+
+    db_workflow_task, db_workflow = await _get_workflow_task_check_owner(
+        project_id=project_id,
+        workflow_task_id=workflow_task_id,
+        workflow_id=workflow_id,
+        user_id=user.id,
+        db=db,
+    )
+
+    await db.delete(db_workflow_task)
+    await db.commit()
+
+    await db.refresh(db_workflow)
+    db_workflow.task_list.reorder()
+    await db.commit()
+
+    return Response(status_code=status.HTTP_204_NO_CONTENT)

fractal_server/app/routes/aux/_job.py

@@ -4,7 +4,7 @@ from zipfile import ZIP_DEFLATED
 from zipfile import ZipFile
 
 from ...models import ApplyWorkflow
-from ...runner._common import SHUTDOWN_FILENAME
+from ...runner.filenames import SHUTDOWN_FILENAME
 
 
 def _write_shutdown_file(*, job: ApplyWorkflow):

fractal_server/app/runner/async_wrap.py

@@ -0,0 +1,27 @@
+import asyncio
+from functools import partial
+from functools import wraps
+from typing import Callable
+
+
+def async_wrap(func: Callable) -> Callable:
+    """
+    Wrap a synchronous callable in an async task
+
+    Ref: [issue #140](https://github.com/fractal-analytics-platform/fractal-server/issues/140)
+    and [this StackOverflow answer](https://stackoverflow.com/q/43241221/19085332).
+
+    Returns:
+        async_wrapper:
+            A factory that allows wrapping a blocking callable within a
+            coroutine.
+    """  # noqa: E501
+
+    @wraps(func)
+    async def async_wrapper(*args, loop=None, executor=None, **kwargs):
+        if loop is None:
+            loop = asyncio.get_event_loop()
+        pfunc = partial(func, *args, **kwargs)
+        return await loop.run_in_executor(executor, pfunc)
+
+    return async_wrapper

fractal_server/app/runner/exceptions.py

@@ -0,0 +1,129 @@
+import os
+from typing import Optional
+
+
+class TaskExecutionError(RuntimeError):
+    """
+    Forwards errors occurred during the execution of a task
+
+    This error wraps and forwards errors occurred during the execution of
+    tasks, when the exit code is larger than 0 (i.e. the error took place
+    within the task). This error also adds information that is useful to track
+    down and debug the failing task within a workflow.
+
+    Attributes:
+        workflow_task_id:
+            ID of the workflow task that failed.
+        workflow_task_order:
+            Order of the task within the workflow.
+        task_name:
+            Human readable name of the failing task.
+    """
+
+    workflow_task_id: Optional[int] = None
+    workflow_task_order: Optional[int] = None
+    task_name: Optional[str] = None
+
+    def __init__(
+        self,
+        *args,
+        workflow_task_id: Optional[int] = None,
+        workflow_task_order: Optional[int] = None,
+        task_name: Optional[str] = None,
+    ):
+        super().__init__(*args)
+        self.workflow_task_id = workflow_task_id
+        self.workflow_task_order = workflow_task_order
+        self.task_name = task_name
+
+
+class JobExecutionError(RuntimeError):
+    """
+    Forwards errors in the execution of a task that are due to external factors
+
+    This error wraps and forwards errors occurred during the execution of
+    tasks, but related to external factors like:
+
+    1. A negative exit code (e.g. because the task received a TERM or KILL
+       signal);
+    2. An error on the executor side (e.g. the SLURM executor could not
+       find the pickled file with task output).
+
+    This error also adds information that is useful to track down and debug the
+    failing task within a workflow.
+
+    Attributes:
+        info:
+            A free field for additional information
+        cmd_file:
+            Path to the file of the command that was executed (e.g. a SLURM
+            submission script).
+        stdout_file:
+            Path to the file with the command stdout
+        stderr_file:
+            Path to the file with the command stderr
+    """
+
+    cmd_file: Optional[str] = None
+    stdout_file: Optional[str] = None
+    stderr_file: Optional[str] = None
+    info: Optional[str] = None
+
+    def __init__(
+        self,
+        *args,
+        cmd_file: Optional[str] = None,
+        stdout_file: Optional[str] = None,
+        stderr_file: Optional[str] = None,
+        info: Optional[str] = None,
+    ):
+        super().__init__(*args)
+        self.cmd_file = cmd_file
+        self.stdout_file = stdout_file
+        self.stderr_file = stderr_file
+        self.info = info
+
+    def _read_file(self, filepath: str) -> str:
+        """
+        Return the content of a text file, and handle the cases where it is
+        empty or missing
+        """
+        if os.path.exists(filepath):
+            with open(filepath, "r") as f:
+                content = f.read()
+                if content:
+                    return f"Content of {filepath}:\n{content}"
+                else:
+                    return f"File {filepath} is empty\n"
+        else:
+            return f"File {filepath} is missing\n"
+
+    def assemble_error(self) -> str:
+        """
+        Read the files that are specified in attributes, and combine them in an
+        error message.
+        """
+        if self.cmd_file:
+            content = self._read_file(self.cmd_file)
+            cmd_content = f"COMMAND:\n{content}\n\n"
+        else:
+            cmd_content = ""
+        if self.stdout_file:
+            content = self._read_file(self.stdout_file)
+            out_content = f"STDOUT:\n{content}\n\n"
+        else:
+            out_content = ""
+        if self.stderr_file:
+            content = self._read_file(self.stderr_file)
+            err_content = f"STDERR:\n{content}\n\n"
+        else:
+            err_content = ""
+
+        content = f"{cmd_content}{out_content}{err_content}"
+        if self.info:
+            content = f"{content}ADDITIONAL INFO:\n{self.info}\n\n"
+
+        if not content:
+            content = str(self)
+        message = f"JobExecutionError\n\n{content}"
+        return message

fractal_server/app/runner/executors/local/__init__.py

@@ -0,0 +1,3 @@
+from .executor import FractalThreadPoolExecutor
+
+__all__ = ["FractalThreadPoolExecutor"]

fractal_server/app/runner/{_local → executors/local}/executor.py

@@ -18,8 +18,8 @@ from typing import Iterable
 from typing import Optional
 from typing import Sequence
 
-from ._local_config import get_default_local_backend_config
-from ._local_config import LocalBackendConfig
+from ...v1._local._local_config import get_default_local_backend_config
+from ...v1._local._local_config import LocalBackendConfig
 
 
 class FractalThreadPoolExecutor(ThreadPoolExecutor):

fractal_server/app/runner/executors/slurm/__init__.py

@@ -0,0 +1,3 @@
+from .executor import SlurmExecutor
+
+__all__ = ["SlurmExecutor"]

fractal_server/app/runner/{_slurm → executors/slurm}/_batching.py

@@ -14,7 +14,7 @@ Submodule to determine the number of total/parallel tasks per SLURM job.
 import math
 from typing import Optional
 
-from ....logger import set_logger
+from .....logger import set_logger
 
 logger = set_logger(__name__)
 

fractal_server/app/runner/executors/slurm/_check_jobs_status.py

@@ -0,0 +1,72 @@
+from subprocess import run  # nosec
+
+from cfut.slurm import STATES_FINISHED
+
+from .....logger import set_logger
+
+
+logger = set_logger(__name__)
+
+
+def run_squeue(job_ids):
+    res = run(  # nosec
+        [
+            "squeue",
+            "--noheader",
+            "--format=%i %T",
+            "--jobs",
+            ",".join([str(j) for j in job_ids]),
+            "--states=all",
+        ],
+        capture_output=True,
+        encoding="utf-8",
+        check=False,
+    )
+    if res.returncode != 0:
+        logger.warning(
+            f"squeue command with {job_ids}"
+            f" failed with:\n{res.stderr=}\n{res.stdout=}"
+        )
+
+    return res
+
+
+def _jobs_finished(job_ids) -> set[str]:
+    """
+    Check which ones of the given Slurm jobs already finished
+
+    The function is based on the `_jobs_finished` function from
+    clusterfutures (version 0.5).
+    Original Copyright: 2022 Adrian Sampson
+    (released under the MIT licence)
+    """
+
+    # If there is no Slurm job to check, return right away
+    if not job_ids:
+        return set()
+    id_to_state = dict()
+
+    res = run_squeue(job_ids)
+    if res.returncode == 0:
+        id_to_state = {
+            out.split()[0]: out.split()[1] for out in res.stdout.splitlines()
+        }
+    else:
+        id_to_state = dict()
+        for j in job_ids:
+            res = run_squeue([j])
+            if res.returncode != 0:
+                logger.info(f"Job {j} not found. Marked it as completed")
+                id_to_state.update({str(j): "COMPLETED"})
+            else:
+                id_to_state.update(
+                    {res.stdout.split()[0]: res.stdout.split()[1]}
+                )
+
+    # Finished jobs only stay in squeue for a few mins (configurable). If
+    # a job ID isn't there, we'll assume it's finished.
+    return {
+        j
+        for j in job_ids
+        if id_to_state.get(j, "COMPLETED") in STATES_FINISHED
+    }

fractal_server/app/runner/{_slurm → executors/slurm}/_executor_wait_thread.py

@@ -6,10 +6,9 @@ from typing import Callable
 from typing import Optional
 
 from cfut import FileWaitThread
-from cfut import slurm
-
-from ....logger import set_logger
 
+from .....logger import set_logger
+from ._check_jobs_status import _jobs_finished
 
 logger = set_logger(__name__)
 
@@ -121,7 +120,7 @@ class FractalSlurmWaitThread(FractalFileWaitThread):
         super().check(i)
         if i % (self.slurm_poll_interval // self.interval) == 0:
             try:
-                finished_jobs = slurm.jobs_finished(self.waiting.values())
+                finished_jobs = _jobs_finished(self.waiting.values())
             except Exception:
                 # Don't abandon completion checking if jobs_finished errors
                 traceback.print_exc()