fractal-server 2.13.0__py3-none-any.whl → 2.14.0__py3-none-any.whl

fractal_server/app/runner/v2/__init__.py

@@ -1,395 +0,0 @@
-"""
-Runner backend subsystem root V2
-
-This module is the single entry point to the runner backend subsystem V2.
-Other subystems should only import this module and not its submodules or
-the individual backends.
-"""
-import os
-import traceback
-from pathlib import Path
-from typing import Optional
-
-from sqlalchemy.orm import Session as DBSyncSession
-
-from ....config import get_settings
-from ....logger import get_logger
-from ....logger import reset_logger_handlers
-from ....logger import set_logger
-from ....ssh._fabric import FractalSSH
-from ....syringe import Inject
-from ....utils import get_timestamp
-from ....zip_tools import _zip_folder_to_file_and_remove
-from ...db import DB
-from ...models.v2 import DatasetV2
-from ...models.v2 import JobV2
-from ...models.v2 import WorkflowV2
-from ...schemas.v2 import JobStatusTypeV2
-from ..exceptions import JobExecutionError
-from ..exceptions import TaskExecutionError
-from ..executors.slurm.sudo._subprocess_run_as_user import _mkdir_as_user
-from ..filenames import WORKFLOW_LOG_FILENAME
-from ..task_files import task_subfolder_name
-from ._local import process_workflow as local_process_workflow
-from ._slurm_ssh import process_workflow as slurm_ssh_process_workflow
-from ._slurm_sudo import process_workflow as slurm_sudo_process_workflow
-from .handle_failed_job import mark_last_wftask_as_failed
-from fractal_server import __VERSION__
-from fractal_server.app.models import UserSettings
-
-
-_backends = {}
-_backends["local"] = local_process_workflow
-_backends["slurm"] = slurm_sudo_process_workflow
-_backends["slurm_ssh"] = slurm_ssh_process_workflow
-
-
-def fail_job(
-    *,
-    db: DBSyncSession,
-    job: JobV2,
-    log_msg: str,
-    logger_name: str,
-    emit_log: bool = False,
-) -> None:
-    logger = get_logger(logger_name=logger_name)
-    if emit_log:
-        logger.error(log_msg)
-    reset_logger_handlers(logger)
-    job.status = JobStatusTypeV2.FAILED
-    job.end_timestamp = get_timestamp()
-    job.log = log_msg
-    db.merge(job)
-    db.commit()
-    db.close()
-    return
-
-
-def submit_workflow(
-    *,
-    workflow_id: int,
-    dataset_id: int,
-    job_id: int,
-    user_settings: UserSettings,
-    worker_init: Optional[str] = None,
-    slurm_user: Optional[str] = None,
-    user_cache_dir: Optional[str] = None,
-    fractal_ssh: Optional[FractalSSH] = None,
-) -> None:
-    """
-    Prepares a workflow and applies it to a dataset
-
-    This function wraps the process_workflow one, which is different for each
-    backend (e.g. local or slurm backend).
-
-    Args:
-        workflow_id:
-            ID of the workflow being applied
-        dataset_id:
-            Dataset ID
-        job_id:
-            Id of the job record which stores the state for the current
-            workflow application.
-        worker_init:
-            Custom executor parameters that get parsed before the execution of
-            each task.
-        user_cache_dir:
-            Cache directory (namely a path where the user can write); for the
-            slurm backend, this is used as a base directory for
-            `job.working_dir_user`.
-        slurm_user:
-            The username to impersonate for the workflow execution, for the
-            slurm backend.
-    """
-    # Declare runner backend and set `process_workflow` function
-    settings = Inject(get_settings)
-    FRACTAL_RUNNER_BACKEND = settings.FRACTAL_RUNNER_BACKEND
-    logger_name = f"WF{workflow_id}_job{job_id}"
-    logger = set_logger(logger_name=logger_name)
-
-    with next(DB.get_sync_db()) as db_sync:
-        try:
-            job: Optional[JobV2] = db_sync.get(JobV2, job_id)
-            dataset: Optional[DatasetV2] = db_sync.get(DatasetV2, dataset_id)
-            workflow: Optional[WorkflowV2] = db_sync.get(
-                WorkflowV2, workflow_id
-            )
-        except Exception as e:
-            logger.error(
-                f"Error conneting to the database. Original error: {str(e)}"
-            )
-            reset_logger_handlers(logger)
-            return
-
-        if job is None:
-            logger.error(f"JobV2 {job_id} does not exist")
-            reset_logger_handlers(logger)
-            return
-        if dataset is None or workflow is None:
-            log_msg = ""
-            if not dataset:
-                log_msg += f"Cannot fetch dataset {dataset_id} from database\n"
-            if not workflow:
-                log_msg += (
-                    f"Cannot fetch workflow {workflow_id} from database\n"
-                )
-            fail_job(
-                db=db_sync, job=job, log_msg=log_msg, logger_name=logger_name
-            )
-            return
-
-        # Declare runner backend and set `process_workflow` function
-        settings = Inject(get_settings)
-        FRACTAL_RUNNER_BACKEND = settings.FRACTAL_RUNNER_BACKEND
-        try:
-            process_workflow = _backends[settings.FRACTAL_RUNNER_BACKEND]
-        except KeyError as e:
-            fail_job(
-                db=db_sync,
-                job=job,
-                log_msg=(
-                    f"Invalid {FRACTAL_RUNNER_BACKEND=}.\n"
-                    f"Original KeyError: {str(e)}"
-                ),
-                logger_name=logger_name,
-                emit_log=True,
-            )
-            return
-
-        # Define and create server-side working folder
-        WORKFLOW_DIR_LOCAL = Path(job.working_dir)
-        if WORKFLOW_DIR_LOCAL.exists():
-            fail_job(
-                db=db_sync,
-                job=job,
-                log_msg=f"Workflow dir {WORKFLOW_DIR_LOCAL} already exists.",
-                logger_name=logger_name,
-                emit_log=True,
-            )
-            return
-
-        try:
-            # Create WORKFLOW_DIR_LOCAL
-            if FRACTAL_RUNNER_BACKEND == "slurm":
-                original_umask = os.umask(0)
-                WORKFLOW_DIR_LOCAL.mkdir(parents=True, mode=0o755)
-                os.umask(original_umask)
-            else:
-                WORKFLOW_DIR_LOCAL.mkdir(parents=True)
-
-            # Define and create WORKFLOW_DIR_REMOTE
-            if FRACTAL_RUNNER_BACKEND == "local":
-                WORKFLOW_DIR_REMOTE = WORKFLOW_DIR_LOCAL
-            elif FRACTAL_RUNNER_BACKEND == "slurm":
-                WORKFLOW_DIR_REMOTE = (
-                    Path(user_cache_dir) / WORKFLOW_DIR_LOCAL.name
-                )
-                _mkdir_as_user(
-                    folder=str(WORKFLOW_DIR_REMOTE), user=slurm_user
-                )
-            elif FRACTAL_RUNNER_BACKEND == "slurm_ssh":
-                # Folder creation is deferred to _process_workflow
-                WORKFLOW_DIR_REMOTE = (
-                    Path(user_settings.ssh_jobs_dir) / WORKFLOW_DIR_LOCAL.name
-                )
-            else:
-                logger.error(
-                    "Invalid FRACTAL_RUNNER_BACKEND="
-                    f"{settings.FRACTAL_RUNNER_BACKEND}."
-                )
-
-            # Create all tasks subfolders
-            for order in range(job.first_task_index, job.last_task_index + 1):
-                this_wftask = workflow.task_list[order]
-                task_name = this_wftask.task.name
-                subfolder_name = task_subfolder_name(
-                    order=order,
-                    task_name=task_name,
-                )
-                if FRACTAL_RUNNER_BACKEND == "slurm":
-                    # Create local subfolder (with 755) and remote one
-                    # (via `sudo -u`)
-                    original_umask = os.umask(0)
-                    (WORKFLOW_DIR_LOCAL / subfolder_name).mkdir(mode=0o755)
-                    os.umask(original_umask)
-                    _mkdir_as_user(
-                        folder=str(WORKFLOW_DIR_REMOTE / subfolder_name),
-                        user=slurm_user,
-                    )
-                else:
-                    # Create local subfolder (with standard permission set)
-                    (WORKFLOW_DIR_LOCAL / subfolder_name).mkdir()
-                    logger.info("Skip remote-subfolder creation")
-        except Exception as e:
-            error_type = type(e).__name__
-            fail_job(
-                db=db_sync,
-                job=job,
-                log_msg=(
-                    f"{error_type} error occurred while creating job folder "
-                    f"and subfolders.\nOriginal error: {str(e)}"
-                ),
-                logger_name=logger_name,
-                emit_log=True,
-            )
-            return
-
-        # After Session.commit() is called, either explicitly or when using a
-        # context manager, all objects associated with the Session are expired.
-        # https://docs.sqlalchemy.org/en/14/orm/
-        #   session_basics.html#opening-and-closing-a-session
-        # https://docs.sqlalchemy.org/en/14/orm/
-        #   session_state_management.html#refreshing-expiring
-
-        # See issue #928:
-        # https://github.com/fractal-analytics-platform/
-        #   fractal-server/issues/928
-
-        db_sync.refresh(dataset)
-        db_sync.refresh(workflow)
-        for wftask in workflow.task_list:
-            db_sync.refresh(wftask)
-
-        # Write logs
-        log_file_path = WORKFLOW_DIR_LOCAL / WORKFLOW_LOG_FILENAME
-        logger = set_logger(
-            logger_name=logger_name,
-            log_file_path=log_file_path,
-        )
-        logger.info(
-            f'Start execution of workflow "{workflow.name}"; '
-            f"more logs at {str(log_file_path)}"
-        )
-        logger.debug(f"fractal_server.__VERSION__: {__VERSION__}")
-        logger.debug(f"FRACTAL_RUNNER_BACKEND: {FRACTAL_RUNNER_BACKEND}")
-        if FRACTAL_RUNNER_BACKEND == "slurm":
-            logger.debug(f"slurm_user: {slurm_user}")
-            logger.debug(f"slurm_account: {job.slurm_account}")
-            logger.debug(f"worker_init: {worker_init}")
-        elif FRACTAL_RUNNER_BACKEND == "slurm_ssh":
-            logger.debug(f"ssh_user: {user_settings.ssh_username}")
-            logger.debug(f"base dir: {user_settings.ssh_tasks_dir}")
-            logger.debug(f"worker_init: {worker_init}")
-        logger.debug(f"job.id: {job.id}")
-        logger.debug(f"job.working_dir: {job.working_dir}")
-        logger.debug(f"job.working_dir_user: {job.working_dir_user}")
-        logger.debug(f"job.first_task_index: {job.first_task_index}")
-        logger.debug(f"job.last_task_index: {job.last_task_index}")
-        logger.debug(f'START workflow "{workflow.name}"')
-
-    try:
-        if FRACTAL_RUNNER_BACKEND == "local":
-            process_workflow = local_process_workflow
-            backend_specific_kwargs = {}
-        elif FRACTAL_RUNNER_BACKEND == "slurm":
-            process_workflow = slurm_sudo_process_workflow
-            backend_specific_kwargs = dict(
-                slurm_user=slurm_user,
-                slurm_account=job.slurm_account,
-                user_cache_dir=user_cache_dir,
-            )
-        elif FRACTAL_RUNNER_BACKEND == "slurm_ssh":
-            process_workflow = slurm_ssh_process_workflow
-            backend_specific_kwargs = dict(fractal_ssh=fractal_ssh)
-        else:
-            raise RuntimeError(
-                f"Invalid runner backend {FRACTAL_RUNNER_BACKEND=}"
-            )
-
-        # "The Session.close() method does not prevent the Session from being
-        # used again. The Session itself does not actually have a distinct
-        # “closed” state; it merely means the Session will release all database
-        # connections and ORM objects."
-        # (https://docs.sqlalchemy.org/en/20/orm/session_api.html#sqlalchemy.orm.Session.close).
-        #
-        # We close the session before the (possibly long) process_workflow
-        # call, to make sure all DB connections are released. The reason why we
-        # are not using a context manager within the try block is that we also
-        # need access to db_sync in the except branches.
-        db_sync = next(DB.get_sync_db())
-        db_sync.close()
-
-        process_workflow(
-            workflow=workflow,
-            dataset=dataset,
-            workflow_dir_local=WORKFLOW_DIR_LOCAL,
-            workflow_dir_remote=WORKFLOW_DIR_REMOTE,
-            logger_name=logger_name,
-            worker_init=worker_init,
-            first_task_index=job.first_task_index,
-            last_task_index=job.last_task_index,
-            job_attribute_filters=job.attribute_filters,
-            **backend_specific_kwargs,
-        )
-
-        logger.info(
-            f'End execution of workflow "{workflow.name}"; '
-            f"more logs at {str(log_file_path)}"
-        )
-        logger.debug(f'END workflow "{workflow.name}"')
-
-        # Update job DB entry
-        job.status = JobStatusTypeV2.DONE
-        job.end_timestamp = get_timestamp()
-        with log_file_path.open("r") as f:
-            logs = f.read()
-        job.log = logs
-        db_sync.merge(job)
-        db_sync.commit()
-
-    except TaskExecutionError as e:
-        logger.debug(f'FAILED workflow "{workflow.name}", TaskExecutionError.')
-        logger.info(f'Workflow "{workflow.name}" failed (TaskExecutionError).')
-
-        mark_last_wftask_as_failed(
-            dataset_id=dataset_id,
-            logger_name=logger_name,
-        )
-        exception_args_string = "\n".join(e.args)
-        log_msg = (
-            f"TASK ERROR: "
-            f"Task name: {e.task_name}, "
-            f"position in Workflow: {e.workflow_task_order}\n"
-            f"TRACEBACK:\n{exception_args_string}"
-        )
-        fail_job(db=db_sync, job=job, log_msg=log_msg, logger_name=logger_name)
-
-    except JobExecutionError as e:
-        logger.debug(f'FAILED workflow "{workflow.name}", JobExecutionError.')
-        logger.info(f'Workflow "{workflow.name}" failed (JobExecutionError).')
-        mark_last_wftask_as_failed(
-            dataset_id=dataset_id,
-            logger_name=logger_name,
-        )
-        fail_job(
-            db=db_sync,
-            job=job,
-            log_msg=(
-                f"JOB ERROR in Fractal job {job.id}:\n"
-                f"TRACEBACK:\n{e.assemble_error()}"
-            ),
-            logger_name=logger_name,
-        )
-
-    except Exception:
-        logger.debug(f'FAILED workflow "{workflow.name}", unknown error.')
-        logger.info(f'Workflow "{workflow.name}" failed (unkwnon error).')
-        mark_last_wftask_as_failed(
-            dataset_id=dataset_id,
-            logger_name=logger_name,
-        )
-        current_traceback = traceback.format_exc()
-        fail_job(
-            db=db_sync,
-            job=job,
-            log_msg=(
-                f"UNKNOWN ERROR in Fractal job {job.id}\n"
-                f"TRACEBACK:\n{current_traceback}"
-            ),
-            logger_name=logger_name,
-        )
-
-    finally:
-        reset_logger_handlers(logger)
-        db_sync.close()
-        _zip_folder_to_file_and_remove(folder=job.working_dir)

fractal_server/app/runner/v2/_local.py

@@ -0,0 +1,88 @@
+from pathlib import Path
+from typing import Optional
+
+from ...models.v2 import DatasetV2
+from ...models.v2 import WorkflowV2
+from ..executors.local.get_local_config import get_local_backend_config
+from ..executors.local.runner import LocalRunner
+from ..set_start_and_last_task_index import set_start_and_last_task_index
+from .runner import execute_tasks_v2
+from fractal_server.images.models import AttributeFiltersType
+
+
+def process_workflow(
+    *,
+    workflow: WorkflowV2,
+    dataset: DatasetV2,
+    workflow_dir_local: Path,
+    job_id: int,
+    workflow_dir_remote: Optional[Path] = None,
+    first_task_index: Optional[int] = None,
+    last_task_index: Optional[int] = None,
+    logger_name: str,
+    job_attribute_filters: AttributeFiltersType,
+    job_type_filters: dict[str, bool],
+    user_id: int,
+    **kwargs,
+) -> None:
+    """
+    Run a workflow through
+
+    Args:
+        workflow:
+            The workflow to be run
+        dataset:
+            Initial dataset.
+        workflow_dir_local:
+            Working directory for this run.
+        workflow_dir_remote:
+            Working directory for this run, on the user side. This argument is
+            present for compatibility with the standard backend interface, but
+            for the `local` backend it cannot be different from
+            `workflow_dir_local`.
+        first_task_index:
+            Positional index of the first task to execute; if `None`, start
+            from `0`.
+        last_task_index:
+            Positional index of the last task to execute; if `None`, proceed
+            until the last task.
+        logger_name: Logger name
+        user_id:
+
+    Raises:
+        TaskExecutionError: wrapper for errors raised during tasks' execution
+                            (positive exit codes).
+        JobExecutionError: wrapper for errors raised by the tasks' executors
+                           (negative exit codes).
+    """
+
+    if workflow_dir_remote and (workflow_dir_remote != workflow_dir_local):
+        raise NotImplementedError(
+            "Local backend does not support different directories "
+            f"{workflow_dir_local=} and {workflow_dir_remote=}"
+        )
+
+    # Set values of first_task_index and last_task_index
+    num_tasks = len(workflow.task_list)
+    first_task_index, last_task_index = set_start_and_last_task_index(
+        num_tasks,
+        first_task_index=first_task_index,
+        last_task_index=last_task_index,
+    )
+
+    with LocalRunner(root_dir_local=workflow_dir_local) as runner:
+        execute_tasks_v2(
+            wf_task_list=workflow.task_list[
+                first_task_index : (last_task_index + 1)
+            ],
+            dataset=dataset,
+            job_id=job_id,
+            runner=runner,
+            workflow_dir_local=workflow_dir_local,
+            workflow_dir_remote=workflow_dir_local,
+            logger_name=logger_name,
+            get_runner_config=get_local_backend_config,
+            job_attribute_filters=job_attribute_filters,
+            job_type_filters=job_type_filters,
+            user_id=user_id,
+        )

fractal_server/app/runner/v2/{_slurm_ssh/__init__.py → _slurm_ssh.py}

@@ -11,7 +11,7 @@
 # Institute for Biomedical Research and Pelkmans Lab from the University of
 # Zurich.
 """
-Slurm Bakend
+Slurm Backend
 
 This backend runs fractal workflows in a SLURM cluster using Clusterfutures
 Executor objects.
@@ -19,14 +19,14 @@ Executor objects.
 from pathlib import Path
 from typing import Optional
 
-from .....ssh._fabric import FractalSSH
-from ....models.v2 import DatasetV2
-from ....models.v2 import WorkflowV2
-from ...exceptions import JobExecutionError
-from ...executors.slurm.ssh.executor import FractalSlurmSSHExecutor
-from ...set_start_and_last_task_index import set_start_and_last_task_index
-from ..runner import execute_tasks_v2
-from ._submit_setup import _slurm_submit_setup
+from ....ssh._fabric import FractalSSH
+from ...models.v2 import DatasetV2
+from ...models.v2 import WorkflowV2
+from ..exceptions import JobExecutionError
+from ..executors.slurm_common.get_slurm_config import get_slurm_config
+from ..executors.slurm_ssh.runner import SlurmSSHRunner
+from ..set_start_and_last_task_index import set_start_and_last_task_index
+from .runner import execute_tasks_v2
 from fractal_server.images.models import AttributeFiltersType
 from fractal_server.logger import set_logger
 
@@ -38,17 +38,17 @@ def process_workflow(
     workflow: WorkflowV2,
     dataset: DatasetV2,
     workflow_dir_local: Path,
+    job_id: int,
     workflow_dir_remote: Optional[Path] = None,
     first_task_index: Optional[int] = None,
     last_task_index: Optional[int] = None,
     logger_name: str,
     job_attribute_filters: AttributeFiltersType,
+    job_type_filters: dict[str, bool],
     fractal_ssh: FractalSSH,
     worker_init: Optional[str] = None,
-    # Not used
-    user_cache_dir: Optional[str] = None,
-    slurm_user: Optional[str] = None,
-    slurm_account: Optional[str] = None,
+    user_id: int,
+    **kwargs,  # not used
 ) -> None:
     """
     Process workflow (SLURM backend public interface)
@@ -77,21 +77,24 @@ def process_workflow(
         logger.error(error_msg)
         raise JobExecutionError(info=error_msg)
 
-    with FractalSlurmSSHExecutor(
+    with SlurmSSHRunner(
         fractal_ssh=fractal_ssh,
-        workflow_dir_local=workflow_dir_local,
-        workflow_dir_remote=workflow_dir_remote,
+        root_dir_local=workflow_dir_local,
+        root_dir_remote=workflow_dir_remote,
         common_script_lines=worker_init,
-    ) as executor:
+    ) as runner:
         execute_tasks_v2(
             wf_task_list=workflow.task_list[
                 first_task_index : (last_task_index + 1)
             ],
             dataset=dataset,
-            executor=executor,
+            job_id=job_id,
+            runner=runner,
             workflow_dir_local=workflow_dir_local,
             workflow_dir_remote=workflow_dir_remote,
             logger_name=logger_name,
-            submit_setup_call=_slurm_submit_setup,
+            get_runner_config=get_slurm_config,
             job_attribute_filters=job_attribute_filters,
+            job_type_filters=job_type_filters,
+            user_id=user_id,
         )

fractal_server/app/runner/v2/{_slurm_sudo/__init__.py → _slurm_sudo.py}

@@ -11,7 +11,7 @@
 # Institute for Biomedical Research and Pelkmans Lab from the University of
 # Zurich.
 """
-Slurm Bakend
+Slurm Backend
 
 This backend runs fractal workflows in a SLURM cluster using Clusterfutures
 Executor objects.
@@ -19,12 +19,12 @@ Executor objects.
 from pathlib import Path
 from typing import Optional
 
-from ....models.v2 import DatasetV2
-from ....models.v2 import WorkflowV2
-from ...executors.slurm.sudo.executor import FractalSlurmSudoExecutor
-from ...set_start_and_last_task_index import set_start_and_last_task_index
-from ..runner import execute_tasks_v2
-from ._submit_setup import _slurm_submit_setup
+from ...models.v2 import DatasetV2
+from ...models.v2 import WorkflowV2
+from ..executors.slurm_common.get_slurm_config import get_slurm_config
+from ..executors.slurm_sudo.runner import SudoSlurmRunner
+from ..set_start_and_last_task_index import set_start_and_last_task_index
+from .runner import execute_tasks_v2
 from fractal_server.images.models import AttributeFiltersType
 
 
@@ -33,11 +33,14 @@ def process_workflow(
     workflow: WorkflowV2,
     dataset: DatasetV2,
     workflow_dir_local: Path,
+    job_id: int,
     workflow_dir_remote: Optional[Path] = None,
     first_task_index: Optional[int] = None,
     last_task_index: Optional[int] = None,
     logger_name: str,
     job_attribute_filters: AttributeFiltersType,
+    job_type_filters: dict[str, bool],
+    user_id: int,
     # Slurm-specific
     user_cache_dir: Optional[str] = None,
     slurm_user: Optional[str] = None,
@@ -64,25 +67,26 @@ def process_workflow(
     if isinstance(worker_init, str):
         worker_init = worker_init.split("\n")
 
-    with FractalSlurmSudoExecutor(
-        debug=True,
-        keep_logs=True,
+    with SudoSlurmRunner(
         slurm_user=slurm_user,
         user_cache_dir=user_cache_dir,
-        workflow_dir_local=workflow_dir_local,
-        workflow_dir_remote=workflow_dir_remote,
+        root_dir_local=workflow_dir_local,
+        root_dir_remote=workflow_dir_remote,
         common_script_lines=worker_init,
         slurm_account=slurm_account,
-    ) as executor:
+    ) as runner:
         execute_tasks_v2(
             wf_task_list=workflow.task_list[
                 first_task_index : (last_task_index + 1)
             ],
             dataset=dataset,
-            executor=executor,
+            job_id=job_id,
+            runner=runner,
             workflow_dir_local=workflow_dir_local,
             workflow_dir_remote=workflow_dir_remote,
             logger_name=logger_name,
-            submit_setup_call=_slurm_submit_setup,
+            get_runner_config=get_slurm_config,
             job_attribute_filters=job_attribute_filters,
+            job_type_filters=job_type_filters,
+            user_id=user_id,
         )