diff --git a/mypy.ini b/mypy.ini index 328b66e..070dde4 100644 --- a/mypy.ini +++ b/mypy.ini @@ -1,33 +1,36 @@ [mypy] namespace_packages = True warn_unused_ignores = True # 3rd party libraries without stubs (yet) [mypy-arrow.*] ignore_missing_imports = True [mypy-celery.*] ignore_missing_imports = True [mypy-elasticsearch.*] ignore_missing_imports = True +[mypy-humanize.*] +ignore_missing_imports = True + [mypy-kombu.*] ignore_missing_imports = True [mypy-pika.*] ignore_missing_imports = True [mypy-pkg_resources.*] ignore_missing_imports = True [mypy-psycopg2.*] ignore_missing_imports = True [mypy-pytest.*] ignore_missing_imports = True [mypy-pytest_postgresql.*] ignore_missing_imports = True diff --git a/requirements.txt b/requirements.txt index 68aa3ff..483689c 100644 --- a/requirements.txt +++ b/requirements.txt @@ -1,19 +1,19 @@ # Add here external Python modules dependencies, one per line. Module names # should match https://pypi.python.org/pypi names. For the full spec or # dependency lines, see https://pip.readthedocs.org/en/1.1/requirements.html -arrow attrs attrs-strict celery >= 4.3 Click elasticsearch > 5.4 flask +humanize pika >= 1.1.0 psycopg2 pyyaml setuptools typing-extensions # test dependencies # hypothesis diff --git a/swh/scheduler/backend.py b/swh/scheduler/backend.py index 5160269..6234f7c 100644 --- a/swh/scheduler/backend.py +++ b/swh/scheduler/backend.py @@ -1,733 +1,727 @@ # Copyright (C) 2015-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import json import logging from typing import Any, Dict, Iterable, List, Optional, Tuple, Union from uuid import UUID -from arrow import Arrow, utcnow import attr -from psycopg2.extensions import AsIs import psycopg2.extras import psycopg2.pool from swh.core.db import BaseDb from swh.core.db.common import db_transaction +from swh.scheduler.utils import utcnow from .exc import StaleData from .model import ( ListedOrigin, ListedOriginPageToken, Lister, PaginatedListedOriginList, ) logger = logging.getLogger(__name__) -def adapt_arrow(arrow): - return AsIs("'%s'::timestamptz" % arrow.isoformat()) - - psycopg2.extensions.register_adapter(dict, psycopg2.extras.Json) -psycopg2.extensions.register_adapter(Arrow, adapt_arrow) psycopg2.extras.register_uuid() def format_query(query, keys): """Format a query with the given keys""" query_keys = ", ".join(keys) placeholders = ", ".join(["%s"] * len(keys)) return query.format(keys=query_keys, placeholders=placeholders) class SchedulerBackend: """Backend for the Software Heritage scheduling database. """ def __init__(self, db, min_pool_conns=1, max_pool_conns=10): """ Args: db_conn: either a libpq connection string, or a psycopg2 connection """ if isinstance(db, psycopg2.extensions.connection): self._pool = None self._db = BaseDb(db) else: self._pool = psycopg2.pool.ThreadedConnectionPool( min_pool_conns, max_pool_conns, db, cursor_factory=psycopg2.extras.RealDictCursor, ) self._db = None def get_db(self): if self._db: return self._db return BaseDb.from_pool(self._pool) def put_db(self, db): if db is not self._db: db.put_conn() task_type_keys = [ "type", "description", "backend_name", "default_interval", "min_interval", "max_interval", "backoff_factor", "max_queue_length", "num_retries", "retry_delay", ] @db_transaction() def create_task_type(self, task_type, db=None, cur=None): """Create a new task type ready for scheduling. Args: task_type (dict): a dictionary with the following keys: - type (str): an identifier for the task type - description (str): a human-readable description of what the task does - backend_name (str): the name of the task in the job-scheduling backend - default_interval (datetime.timedelta): the default interval between two task runs - min_interval (datetime.timedelta): the minimum interval between two task runs - max_interval (datetime.timedelta): the maximum interval between two task runs - backoff_factor (float): the factor by which the interval changes at each run - max_queue_length (int): the maximum length of the task queue for this task type """ keys = [key for key in self.task_type_keys if key in task_type] query = format_query( """insert into task_type ({keys}) values ({placeholders}) on conflict do nothing""", keys, ) cur.execute(query, [task_type[key] for key in keys]) @db_transaction() def get_task_type(self, task_type_name, db=None, cur=None): """Retrieve the task type with id task_type_name""" query = format_query( "select {keys} from task_type where type=%s", self.task_type_keys, ) cur.execute(query, (task_type_name,)) return cur.fetchone() @db_transaction() def get_task_types(self, db=None, cur=None): """Retrieve all registered task types""" query = format_query("select {keys} from task_type", self.task_type_keys,) cur.execute(query) return cur.fetchall() @db_transaction() def get_or_create_lister( self, name: str, instance_name: Optional[str] = None, db=None, cur=None ) -> Lister: """Retrieve information about the given instance of the lister from the database, or create the entry if it did not exist. """ if instance_name is None: instance_name = "" select_cols = ", ".join(Lister.select_columns()) insert_cols, insert_meta = ( ", ".join(tup) for tup in Lister.insert_columns_and_metavars() ) query = f""" with added as ( insert into listers ({insert_cols}) values ({insert_meta}) on conflict do nothing returning {select_cols} ) select {select_cols} from added union all select {select_cols} from listers where (name, instance_name) = (%(name)s, %(instance_name)s); """ cur.execute(query, attr.asdict(Lister(name=name, instance_name=instance_name))) return Lister(**cur.fetchone()) @db_transaction() def update_lister(self, lister: Lister, db=None, cur=None) -> Lister: """Update the state for the given lister instance in the database. Returns: a new Lister object, with all fields updated from the database Raises: StaleData if the `updated` timestamp for the lister instance in database doesn't match the one passed by the user. """ select_cols = ", ".join(Lister.select_columns()) set_vars = ", ".join( f"{col} = {meta}" for col, meta in zip(*Lister.insert_columns_and_metavars()) ) query = f"""update listers set {set_vars} where id=%(id)s and updated=%(updated)s returning {select_cols}""" cur.execute(query, attr.asdict(lister)) updated = cur.fetchone() if not updated: raise StaleData("Stale data; Lister state not updated") return Lister(**updated) @db_transaction() def record_listed_origins( self, listed_origins: Iterable[ListedOrigin], db=None, cur=None ) -> List[ListedOrigin]: """Record a set of origins that a lister has listed. This performs an "upsert": origins with the same (lister_id, url, visit_type) values are updated with new values for extra_loader_arguments, last_update and last_seen. """ pk_cols = ListedOrigin.primary_key_columns() select_cols = ListedOrigin.select_columns() insert_cols, insert_meta = ListedOrigin.insert_columns_and_metavars() upsert_cols = [col for col in insert_cols if col not in pk_cols] upsert_set = ", ".join(f"{col} = EXCLUDED.{col}" for col in upsert_cols) query = f"""INSERT into listed_origins ({", ".join(insert_cols)}) VALUES %s ON CONFLICT ({", ".join(pk_cols)}) DO UPDATE SET {upsert_set} RETURNING {", ".join(select_cols)} """ ret = psycopg2.extras.execute_values( cur=cur, sql=query, argslist=(attr.asdict(origin) for origin in listed_origins), template=f"({', '.join(insert_meta)})", page_size=1000, fetch=True, ) return [ListedOrigin(**d) for d in ret] @db_transaction() def get_listed_origins( self, lister_id: Optional[UUID] = None, url: Optional[str] = None, limit: int = 1000, page_token: Optional[ListedOriginPageToken] = None, db=None, cur=None, ) -> PaginatedListedOriginList: """Get information on the listed origins matching either the `url` or `lister_id`, or both arguments. """ query_filters: List[str] = [] query_params: List[Union[int, str, UUID, Tuple[UUID, str]]] = [] if lister_id: query_filters.append("lister_id = %s") query_params.append(lister_id) if url is not None: query_filters.append("url = %s") query_params.append(url) if page_token is not None: query_filters.append("(lister_id, url) > %s") # the typeshed annotation for tuple() is too strict. query_params.append(tuple(page_token)) # type: ignore query_params.append(limit) select_cols = ", ".join(ListedOrigin.select_columns()) if query_filters: where_clause = "where %s" % (" and ".join(query_filters)) else: where_clause = "" query = f"""SELECT {select_cols} from listed_origins {where_clause} ORDER BY lister_id, url LIMIT %s""" cur.execute(query, tuple(query_params)) origins = [ListedOrigin(**d) for d in cur] if len(origins) == limit: page_token = (origins[-1].lister_id, origins[-1].url) else: page_token = None return PaginatedListedOriginList(origins, page_token) task_create_keys = [ "type", "arguments", "next_run", "policy", "status", "retries_left", "priority", ] task_keys = task_create_keys + ["id", "current_interval"] @db_transaction() def create_tasks(self, tasks, policy="recurring", db=None, cur=None): """Create new tasks. Args: tasks (list): each task is a dictionary with the following keys: - type (str): the task type - arguments (dict): the arguments for the task runner, keys: - args (list of str): arguments - kwargs (dict str -> str): keyword arguments - next_run (datetime.datetime): the next scheduled run for the task Returns: a list of created tasks. """ cur.execute("select swh_scheduler_mktemp_task()") db.copy_to( tasks, "tmp_task", self.task_create_keys, default_values={"policy": policy, "status": "next_run_not_scheduled"}, cur=cur, ) query = format_query( "select {keys} from swh_scheduler_create_tasks_from_temp()", self.task_keys, ) cur.execute(query) return cur.fetchall() @db_transaction() def set_status_tasks( self, task_ids, status="disabled", next_run=None, db=None, cur=None ): """Set the tasks' status whose ids are listed. If given, also set the next_run date. """ if not task_ids: return query = ["UPDATE task SET status = %s"] args = [status] if next_run: query.append(", next_run = %s") args.append(next_run) query.append(" WHERE id IN %s") args.append(tuple(task_ids)) cur.execute("".join(query), args) @db_transaction() def disable_tasks(self, task_ids, db=None, cur=None): """Disable the tasks whose ids are listed.""" return self.set_status_tasks(task_ids, db=db, cur=cur) @db_transaction() def search_tasks( self, task_id=None, task_type=None, status=None, priority=None, policy=None, before=None, after=None, limit=None, db=None, cur=None, ): """Search tasks from selected criterions""" where = [] args = [] if task_id: if isinstance(task_id, (str, int)): where.append("id = %s") else: where.append("id in %s") task_id = tuple(task_id) args.append(task_id) if task_type: if isinstance(task_type, str): where.append("type = %s") else: where.append("type in %s") task_type = tuple(task_type) args.append(task_type) if status: if isinstance(status, str): where.append("status = %s") else: where.append("status in %s") status = tuple(status) args.append(status) if priority: if isinstance(priority, str): where.append("priority = %s") else: priority = tuple(priority) where.append("priority in %s") args.append(priority) if policy: where.append("policy = %s") args.append(policy) if before: where.append("next_run <= %s") args.append(before) if after: where.append("next_run >= %s") args.append(after) query = "select * from task" if where: query += " where " + " and ".join(where) if limit: query += " limit %s :: bigint" args.append(limit) cur.execute(query, args) return cur.fetchall() @db_transaction() def get_tasks(self, task_ids, db=None, cur=None): """Retrieve the info of tasks whose ids are listed.""" query = format_query("select {keys} from task where id in %s", self.task_keys) cur.execute(query, (tuple(task_ids),)) return cur.fetchall() @db_transaction() def peek_ready_tasks( self, task_type, timestamp=None, num_tasks=None, num_tasks_priority=None, db=None, cur=None, ): """Fetch the list of ready tasks Args: task_type (str): filtering task per their type timestamp (datetime.datetime): peek tasks that need to be executed before that timestamp num_tasks (int): only peek at num_tasks tasks (with no priority) num_tasks_priority (int): only peek at num_tasks_priority tasks (with priority) Returns: a list of tasks """ if timestamp is None: timestamp = utcnow() cur.execute( """select * from swh_scheduler_peek_ready_tasks( %s, %s, %s :: bigint, %s :: bigint)""", (task_type, timestamp, num_tasks, num_tasks_priority), ) logger.debug("PEEK %s => %s" % (task_type, cur.rowcount)) return cur.fetchall() @db_transaction() def grab_ready_tasks( self, task_type, timestamp=None, num_tasks=None, num_tasks_priority=None, db=None, cur=None, ): """Fetch the list of ready tasks, and mark them as scheduled Args: task_type (str): filtering task per their type timestamp (datetime.datetime): grab tasks that need to be executed before that timestamp num_tasks (int): only grab num_tasks tasks (with no priority) num_tasks_priority (int): only grab oneshot num_tasks tasks (with priorities) Returns: a list of tasks """ if timestamp is None: timestamp = utcnow() cur.execute( """select * from swh_scheduler_grab_ready_tasks( %s, %s, %s :: bigint, %s :: bigint)""", (task_type, timestamp, num_tasks, num_tasks_priority), ) logger.debug("GRAB %s => %s" % (task_type, cur.rowcount)) return cur.fetchall() task_run_create_keys = ["task", "backend_id", "scheduled", "metadata"] @db_transaction() def schedule_task_run( self, task_id, backend_id, metadata=None, timestamp=None, db=None, cur=None ): """Mark a given task as scheduled, adding a task_run entry in the database. Args: task_id (int): the identifier for the task being scheduled backend_id (str): the identifier of the job in the backend metadata (dict): metadata to add to the task_run entry timestamp (datetime.datetime): the instant the event occurred Returns: a fresh task_run entry """ if metadata is None: metadata = {} if timestamp is None: timestamp = utcnow() cur.execute( "select * from swh_scheduler_schedule_task_run(%s, %s, %s, %s)", (task_id, backend_id, metadata, timestamp), ) return cur.fetchone() @db_transaction() def mass_schedule_task_runs(self, task_runs, db=None, cur=None): """Schedule a bunch of task runs. Args: task_runs (list): a list of dicts with keys: - task (int): the identifier for the task being scheduled - backend_id (str): the identifier of the job in the backend - metadata (dict): metadata to add to the task_run entry - scheduled (datetime.datetime): the instant the event occurred Returns: None """ cur.execute("select swh_scheduler_mktemp_task_run()") db.copy_to(task_runs, "tmp_task_run", self.task_run_create_keys, cur=cur) cur.execute("select swh_scheduler_schedule_task_run_from_temp()") @db_transaction() def start_task_run( self, backend_id, metadata=None, timestamp=None, db=None, cur=None ): """Mark a given task as started, updating the corresponding task_run entry in the database. Args: backend_id (str): the identifier of the job in the backend metadata (dict): metadata to add to the task_run entry timestamp (datetime.datetime): the instant the event occurred Returns: the updated task_run entry """ if metadata is None: metadata = {} if timestamp is None: timestamp = utcnow() cur.execute( "select * from swh_scheduler_start_task_run(%s, %s, %s)", (backend_id, metadata, timestamp), ) return cur.fetchone() @db_transaction() def end_task_run( self, backend_id, status, metadata=None, timestamp=None, result=None, db=None, cur=None, ): """Mark a given task as ended, updating the corresponding task_run entry in the database. Args: backend_id (str): the identifier of the job in the backend status (str): how the task ended; one of: 'eventful', 'uneventful', 'failed' metadata (dict): metadata to add to the task_run entry timestamp (datetime.datetime): the instant the event occurred Returns: the updated task_run entry """ if metadata is None: metadata = {} if timestamp is None: timestamp = utcnow() cur.execute( "select * from swh_scheduler_end_task_run(%s, %s, %s, %s)", (backend_id, status, metadata, timestamp), ) return cur.fetchone() @db_transaction() def filter_task_to_archive( self, after_ts: str, before_ts: str, limit: int = 10, page_token: Optional[str] = None, db=None, cur=None, ) -> Dict[str, Any]: """Compute the tasks to archive within the datetime interval [after_ts, before_ts[. The method returns a paginated result. Returns: dict with the following keys: - **next_page_token**: opaque token to be used as `page_token` to retrieve the next page of result. If absent, there is no more pages to gather. - **tasks**: list of task dictionaries with the following keys: **id** (str): origin task id **started** (Optional[datetime]): started date **scheduled** (datetime): scheduled date **arguments** (json dict): task's arguments ... """ assert not page_token or isinstance(page_token, str) last_id = -1 if page_token is None else int(page_token) tasks = [] cur.execute( "select * from swh_scheduler_task_to_archive(%s, %s, %s, %s)", (after_ts, before_ts, last_id, limit + 1), ) for row in cur: task = dict(row) # nested type index does not accept bare values # transform it as a dict to comply with this task["arguments"]["args"] = { i: v for i, v in enumerate(task["arguments"]["args"]) } kwargs = task["arguments"]["kwargs"] task["arguments"]["kwargs"] = json.dumps(kwargs) tasks.append(task) if len(tasks) >= limit + 1: # remains data, add pagination information result = { "tasks": tasks[:limit], "next_page_token": str(tasks[-1]["task_id"]), } else: result = {"tasks": tasks} return result @db_transaction() def delete_archived_tasks(self, task_ids, db=None, cur=None): """Delete archived tasks as much as possible. Only the task_ids whose complete associated task_run have been cleaned up will be. """ _task_ids = _task_run_ids = [] for task_id in task_ids: _task_ids.append(task_id["task_id"]) _task_run_ids.append(task_id["task_run_id"]) cur.execute( "select * from swh_scheduler_delete_archived_tasks(%s, %s)", (_task_ids, _task_run_ids), ) task_run_keys = [ "id", "task", "backend_id", "scheduled", "started", "ended", "metadata", "status", ] @db_transaction() def get_task_runs(self, task_ids, limit=None, db=None, cur=None): """Search task run for a task id""" where = [] args = [] if task_ids: if isinstance(task_ids, (str, int)): where.append("task = %s") else: where.append("task in %s") task_ids = tuple(task_ids) args.append(task_ids) else: return () query = "select * from task_run where " + " and ".join(where) if limit: query += " limit %s :: bigint" args.append(limit) cur.execute(query, args) return cur.fetchall() @db_transaction() def get_priority_ratios(self, db=None, cur=None): cur.execute("select id, ratio from priority_ratio") return {row["id"]: row["ratio"] for row in cur.fetchall()} diff --git a/swh/scheduler/celery_backend/listener.py b/swh/scheduler/celery_backend/listener.py index b608269..9306fa0 100644 --- a/swh/scheduler/celery_backend/listener.py +++ b/swh/scheduler/celery_backend/listener.py @@ -1,220 +1,220 @@ # Copyright (C) 2015-2018 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import datetime import logging import sys import time -from arrow import utcnow import celery from celery.events import EventReceiver import click from kombu import Queue from swh.core.statsd import statsd +from swh.scheduler.utils import utcnow class ReliableEventReceiver(EventReceiver): def __init__( self, channel, handlers=None, routing_key="#", node_id=None, app=None, queue_prefix="celeryev", accept=None, ): super(ReliableEventReceiver, self).__init__( channel, handlers, routing_key, node_id, app, queue_prefix, accept ) self.queue = Queue( ".".join([self.queue_prefix, self.node_id]), exchange=self.exchange, routing_key=self.routing_key, auto_delete=False, durable=True, ) def get_consumers(self, consumer, channel): return [ consumer( queues=[self.queue], callbacks=[self._receive], no_ack=False, accept=self.accept, ) ] def _receive(self, bodies, message): if not isinstance(bodies, list): # celery<4 returned body as element bodies = [bodies] for body in bodies: type, body = self.event_from_message(body) self.process(type, body, message) def process(self, type, event, message): """Process the received event by dispatching it to the appropriate handler.""" handler = self.handlers.get(type) or self.handlers.get("*") if handler: handler(event, message) statsd.increment( "swh_scheduler_listener_handled_event_total", tags={"event_type": type} ) ACTION_SEND_DELAY = datetime.timedelta(seconds=1.0) ACTION_QUEUE_MAX_LENGTH = 1000 def event_monitor(app, backend): logger = logging.getLogger("swh.scheduler.listener") actions = { "last_send": utcnow() - 2 * ACTION_SEND_DELAY, "queue": [], } def try_perform_actions(actions=actions): logger.debug("Try perform pending actions") if actions["queue"] and ( len(actions["queue"]) > ACTION_QUEUE_MAX_LENGTH or utcnow() - actions["last_send"] > ACTION_SEND_DELAY ): perform_actions(actions) def perform_actions(actions, backend=backend): logger.info("Perform %s pending actions" % len(actions["queue"])) action_map = { "start_task_run": backend.start_task_run, "end_task_run": backend.end_task_run, } messages = [] db = backend.get_db() try: cursor = db.cursor(None) for action in actions["queue"]: messages.append(action["message"]) function = action_map[action["action"]] args = action.get("args", ()) kwargs = action.get("kwargs", {}) kwargs["cur"] = cursor function(*args, **kwargs) except Exception: db.conn.rollback() else: db.conn.commit() finally: backend.put_db(db) for message in messages: if not message.acknowledged: message.ack() actions["queue"] = [] actions["last_send"] = utcnow() def queue_action(action, actions=actions): actions["queue"].append(action) try_perform_actions() def catchall_event(event, message): logger.debug("event: %s %s", event["type"], event.get("name", "N/A")) if not message.acknowledged: message.ack() try_perform_actions() def task_started(event, message): logger.debug("task_started: %s %s", event["type"], event.get("name", "N/A")) queue_action( { "action": "start_task_run", "args": [event["uuid"]], "kwargs": { "timestamp": utcnow(), "metadata": {"worker": event["hostname"],}, }, "message": message, } ) def task_succeeded(event, message): logger.debug("task_succeeded: event: %s" % event) logger.debug(" message: %s" % message) result = event["result"] logger.debug("task_succeeded: result: %s" % result) try: status = result.get("status") if status == "success": status = "eventful" if result.get("eventful") else "uneventful" except Exception: status = "eventful" if result else "uneventful" queue_action( { "action": "end_task_run", "args": [event["uuid"]], "kwargs": {"timestamp": utcnow(), "status": status, "result": result,}, "message": message, } ) def task_failed(event, message): logger.debug("task_failed: event: %s" % event) logger.debug(" message: %s" % message) queue_action( { "action": "end_task_run", "args": [event["uuid"]], "kwargs": {"timestamp": utcnow(), "status": "failed",}, "message": message, } ) recv = ReliableEventReceiver( celery.current_app.connection(), app=celery.current_app, handlers={ "task-started": task_started, "task-result": task_succeeded, "task-failed": task_failed, "*": catchall_event, }, node_id="listener", ) errors = 0 while True: try: recv.capture(limit=None, timeout=None, wakeup=True) errors = 0 except KeyboardInterrupt: logger.exception("Keyboard interrupt, exiting") break except Exception: logger.exception("Unexpected exception") if errors < 5: time.sleep(errors) errors += 1 else: logger.error("Too many consecutive errors, exiting") sys.exit(1) @click.command() @click.pass_context def main(ctx): click.echo("Deprecated! Use 'swh-scheduler listener' instead.", err=True) ctx.exit(1) if __name__ == "__main__": main() diff --git a/swh/scheduler/celery_backend/pika_listener.py b/swh/scheduler/celery_backend/pika_listener.py index d8047e8..7df7bec 100644 --- a/swh/scheduler/celery_backend/pika_listener.py +++ b/swh/scheduler/celery_backend/pika_listener.py @@ -1,102 +1,98 @@ # Copyright (C) 2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information -import datetime import json import logging import sys import pika from swh.core.statsd import statsd from swh.scheduler import get_scheduler +from swh.scheduler.utils import utcnow logger = logging.getLogger(__name__) -def utcnow(): - return datetime.datetime.now(tz=datetime.timezone.utc) - - def get_listener(broker_url, queue_name, scheduler_backend): connection = pika.BlockingConnection(pika.URLParameters(broker_url)) channel = connection.channel() channel.queue_declare(queue=queue_name, durable=True) exchange = "celeryev" routing_key = "#" channel.queue_bind(queue=queue_name, exchange=exchange, routing_key=routing_key) channel.basic_qos(prefetch_count=1000) channel.basic_consume( queue=queue_name, on_message_callback=get_on_message(scheduler_backend), ) return channel def get_on_message(scheduler_backend): def on_message(channel, method_frame, properties, body): try: events = json.loads(body) except Exception: logger.warning("Could not parse body %r", body) events = [] if not isinstance(events, list): events = [events] for event in events: logger.debug("Received event %r", event) process_event(event, scheduler_backend) channel.basic_ack(delivery_tag=method_frame.delivery_tag) return on_message def process_event(event, scheduler_backend): uuid = event.get("uuid") if not uuid: return event_type = event["type"] statsd.increment( "swh_scheduler_listener_handled_event_total", tags={"event_type": event_type} ) if event_type == "task-started": scheduler_backend.start_task_run( uuid, timestamp=utcnow(), metadata={"worker": event.get("hostname")}, ) elif event_type == "task-result": result = event["result"] status = None if isinstance(result, dict) and "status" in result: status = result["status"] if status == "success": status = "eventful" if result.get("eventful") else "uneventful" if status is None: status = "eventful" if result else "uneventful" scheduler_backend.end_task_run( uuid, timestamp=utcnow(), status=status, result=result ) elif event_type == "task-failed": scheduler_backend.end_task_run(uuid, timestamp=utcnow(), status="failed") if __name__ == "__main__": url = sys.argv[1] logging.basicConfig(level=logging.DEBUG) scheduler_backend = get_scheduler("local", args={"db": "service=swh-scheduler"}) channel = get_listener(url, "celeryev.test", scheduler_backend) logger.info("Start consuming") channel.start_consuming() diff --git a/swh/scheduler/celery_backend/runner.py b/swh/scheduler/celery_backend/runner.py index 957c1b7..ea469ad 100644 --- a/swh/scheduler/celery_backend/runner.py +++ b/swh/scheduler/celery_backend/runner.py @@ -1,131 +1,132 @@ # Copyright (C) 2015-2018 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import logging -import arrow from kombu.utils.uuid import uuid from swh.core.statsd import statsd from swh.scheduler import compute_nb_tasks_from, get_scheduler +from swh.scheduler.utils import utcnow logger = logging.getLogger(__name__) + # Max batch size for tasks MAX_NUM_TASKS = 10000 def run_ready_tasks(backend, app): """Run tasks that are ready Args: backend (Scheduler): backend to read tasks to schedule app (App): Celery application to send tasks to Returns: A list of dictionaries:: { 'task': the scheduler's task id, 'backend_id': Celery's task id, - 'scheduler': arrow.utcnow() + 'scheduler': utcnow() } The result can be used to block-wait for the tasks' results:: backend_tasks = run_ready_tasks(self.scheduler, app) for task in backend_tasks: AsyncResult(id=task['backend_id']).get() """ all_backend_tasks = [] while True: task_types = {} pending_tasks = [] for task_type in backend.get_task_types(): task_type_name = task_type["type"] task_types[task_type_name] = task_type max_queue_length = task_type["max_queue_length"] if max_queue_length is None: max_queue_length = 0 backend_name = task_type["backend_name"] if max_queue_length: try: queue_length = app.get_queue_length(backend_name) except ValueError: queue_length = None if queue_length is None: # Running without RabbitMQ (probably a test env). num_tasks = MAX_NUM_TASKS else: num_tasks = min(max_queue_length - queue_length, MAX_NUM_TASKS) else: num_tasks = MAX_NUM_TASKS # only pull tasks if the buffer is at least 1/5th empty (= 80% # full), to help postgresql use properly indexed queries. if num_tasks > min(MAX_NUM_TASKS, max_queue_length) // 5: num_tasks, num_tasks_priority = compute_nb_tasks_from(num_tasks) grabbed_tasks = backend.grab_ready_tasks( task_type_name, num_tasks=num_tasks, num_tasks_priority=num_tasks_priority, ) if grabbed_tasks: pending_tasks.extend(grabbed_tasks) logger.info( "Grabbed %s tasks %s", len(grabbed_tasks), task_type_name ) statsd.increment( "swh_scheduler_runner_scheduled_task_total", len(grabbed_tasks), tags={"task_type": task_type_name}, ) if not pending_tasks: return all_backend_tasks backend_tasks = [] celery_tasks = [] for task in pending_tasks: args = task["arguments"]["args"] kwargs = task["arguments"]["kwargs"] backend_name = task_types[task["type"]]["backend_name"] backend_id = uuid() celery_tasks.append((backend_name, backend_id, args, kwargs)) data = { "task": task["id"], "backend_id": backend_id, - "scheduled": arrow.utcnow(), + "scheduled": utcnow(), } backend_tasks.append(data) logger.debug("Sent %s celery tasks", len(backend_tasks)) backend.mass_schedule_task_runs(backend_tasks) for backend_name, backend_id, args, kwargs in celery_tasks: app.send_task( backend_name, task_id=backend_id, args=args, kwargs=kwargs, ) all_backend_tasks.extend(backend_tasks) def main(): from .config import app as main_app for module in main_app.conf.CELERY_IMPORTS: __import__(module) main_backend = get_scheduler("local") try: run_ready_tasks(main_backend, main_app) except Exception: main_backend.rollback() raise if __name__ == "__main__": main() diff --git a/swh/scheduler/cli/task.py b/swh/scheduler/cli/task.py index c304163..e2d2d97 100644 --- a/swh/scheduler/cli/task.py +++ b/swh/scheduler/cli/task.py @@ -1,768 +1,751 @@ # Copyright (C) 2016-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information # WARNING: do not import unnecessary things here to keep cli startup time under # control import locale import logging from typing import TYPE_CHECKING, Any, Dict, Iterator, Optional import click from . import cli if TYPE_CHECKING: # importing swh.storage.interface triggers the load of 300+ modules, so... from swh.model.model import Origin from swh.storage.interface import StorageInterface locale.setlocale(locale.LC_ALL, "") -ARROW_LOCALE = locale.getlocale(locale.LC_TIME)[0] - - -class DateTimeType(click.ParamType): - name = "time and date" - - def convert(self, value, param, ctx): - import arrow - - if not isinstance(value, arrow.Arrow): - value = arrow.get(value) - - return value - - -DATETIME = DateTimeType() CONTEXT_SETTINGS = dict(help_option_names=["-h", "--help"]) +DATETIME = click.DateTime() def format_dict(d): + """Recursively format date objects in the dict passed as argument""" import datetime - import arrow - ret = {} for k, v in d.items(): - if isinstance(v, (arrow.Arrow, datetime.date, datetime.datetime)): - v = arrow.get(v).format() + if isinstance(v, (datetime.date, datetime.datetime)): + v = v.isoformat() elif isinstance(v, dict): v = format_dict(v) ret[k] = v return ret def pretty_print_list(list, indent=0): """Pretty-print a list""" return "".join("%s%r\n" % (" " * indent, item) for item in list) def pretty_print_dict(dict, indent=0): """Pretty-print a list""" return "".join( "%s%s: %r\n" % (" " * indent, click.style(key, bold=True), value) for key, value in sorted(dict.items()) ) def pretty_print_run(run, indent=4): fmt = ( "{indent}{backend_id} [{status}]\n" "{indent} scheduled: {scheduled} [{started}:{ended}]" ) return fmt.format(indent=" " * indent, **format_dict(run)) def pretty_print_task(task, full=False): """Pretty-print a task If 'full' is True, also print the status and priority fields. >>> import datetime >>> task = { ... 'id': 1234, ... 'arguments': { ... 'args': ['foo', 'bar', True], ... 'kwargs': {'key': 'value', 'key2': 42}, ... }, ... 'current_interval': datetime.timedelta(hours=1), ... 'next_run': datetime.datetime(2019, 2, 21, 13, 52, 35, 407818), ... 'policy': 'oneshot', ... 'priority': None, ... 'status': 'next_run_not_scheduled', ... 'type': 'test_task', ... } >>> print(click.unstyle(pretty_print_task(task))) Task 1234 - Next run: ... (2019-02-21 13:52:35+00:00) + Next run: ... (2019-02-21T13:52:35.407818) Interval: 1:00:00 Type: test_task Policy: oneshot Args: 'foo' 'bar' True Keyword args: key: 'value' key2: 42 >>> print(click.unstyle(pretty_print_task(task, full=True))) Task 1234 - Next run: ... (2019-02-21 13:52:35+00:00) + Next run: ... (2019-02-21T13:52:35.407818) Interval: 1:00:00 Type: test_task Policy: oneshot Status: next_run_not_scheduled Priority:\x20 Args: 'foo' 'bar' True Keyword args: key: 'value' key2: 42 """ - import arrow + import humanize - next_run = arrow.get(task["next_run"]) + next_run = task["next_run"] lines = [ "%s %s\n" % (click.style("Task", bold=True), task["id"]), click.style(" Next run: ", bold=True), - "%s (%s)" % (next_run.humanize(locale=ARROW_LOCALE), next_run.format()), + "%s (%s)" % (humanize.naturaldate(next_run), next_run.isoformat()), "\n", click.style(" Interval: ", bold=True), str(task["current_interval"]), "\n", click.style(" Type: ", bold=True), task["type"] or "", "\n", click.style(" Policy: ", bold=True), task["policy"] or "", "\n", ] if full: lines += [ click.style(" Status: ", bold=True), task["status"] or "", "\n", click.style(" Priority: ", bold=True), task["priority"] or "", "\n", ] lines += [ click.style(" Args:\n", bold=True), pretty_print_list(task["arguments"]["args"], indent=4), click.style(" Keyword args:\n", bold=True), pretty_print_dict(task["arguments"]["kwargs"], indent=4), ] return "".join(lines) @cli.group("task") @click.pass_context def task(ctx): """Manipulate tasks.""" pass @task.command("schedule") @click.option( "--columns", "-c", multiple=True, default=["type", "args", "kwargs", "next_run"], type=click.Choice(["type", "args", "kwargs", "policy", "next_run"]), help="columns present in the CSV file", ) @click.option("--delimiter", "-d", default=",") @click.argument("file", type=click.File(encoding="utf-8")) @click.pass_context def schedule_tasks(ctx, columns, delimiter, file): """Schedule tasks from a CSV input file. The following columns are expected, and can be set through the -c option: - type: the type of the task to be scheduled (mandatory) - args: the arguments passed to the task (JSON list, defaults to an empty list) - kwargs: the keyword arguments passed to the task (JSON object, defaults to an empty dict) - next_run: the date at which the task should run (datetime, defaults to now) The CSV can be read either from a named file, or from stdin (use - as filename). Use sample: cat scheduling-task.txt | \ python3 -m swh.scheduler.cli \ --database 'service=swh-scheduler-dev' \ task schedule \ --columns type --columns kwargs --columns policy \ --delimiter ';' - """ import csv import json - import arrow + from swh.scheduler.utils import utcnow tasks = [] - now = arrow.utcnow() + now = utcnow() scheduler = ctx.obj["scheduler"] if not scheduler: raise ValueError("Scheduler class (local/remote) must be instantiated") reader = csv.reader(file, delimiter=delimiter) for line in reader: task = dict(zip(columns, line)) args = json.loads(task.pop("args", "[]")) kwargs = json.loads(task.pop("kwargs", "{}")) task["arguments"] = { "args": args, "kwargs": kwargs, } - task["next_run"] = DATETIME.convert(task.get("next_run", now), None, None) + task["next_run"] = task.get("next_run", now) tasks.append(task) created = scheduler.create_tasks(tasks) output = [ "Created %d tasks\n" % len(created), ] for task in created: output.append(pretty_print_task(task)) click.echo_via_pager("\n".join(output)) @task.command("add") @click.argument("type", nargs=1, required=True) @click.argument("options", nargs=-1) @click.option( "--policy", "-p", default="recurring", type=click.Choice(["recurring", "oneshot"]) ) @click.option( "--priority", "-P", default=None, type=click.Choice(["low", "normal", "high"]) ) @click.option("--next-run", "-n", default=None) @click.pass_context def schedule_task(ctx, type, options, policy, priority, next_run): """Schedule one task from arguments. The first argument is the name of the task type, further ones are positional and keyword argument(s) of the task, in YAML format. Keyword args are of the form key=value. Usage sample: swh-scheduler --database 'service=swh-scheduler' \ task add list-pypi swh-scheduler --database 'service=swh-scheduler' \ task add list-debian-distribution --policy=oneshot distribution=stretch Note: if the priority is not given, the task won't have the priority set, which is considered as the lowest priority level. """ - import arrow + from swh.scheduler.utils import utcnow from .utils import parse_options scheduler = ctx.obj["scheduler"] if not scheduler: raise ValueError("Scheduler class (local/remote) must be instantiated") - now = arrow.utcnow() + now = utcnow() (args, kw) = parse_options(options) task = { "type": type, "policy": policy, "priority": priority, "arguments": {"args": args, "kwargs": kw,}, - "next_run": DATETIME.convert(next_run or now, None, None), + "next_run": next_run or now, } created = scheduler.create_tasks([task]) output = [ "Created %d tasks\n" % len(created), ] for task in created: output.append(pretty_print_task(task)) click.echo("\n".join(output)) def iter_origins( # use string annotations to prevent some pkg loading storage: "StorageInterface", page_token: "Optional[str]" = None, ) -> "Iterator[Origin]": """Iterate over origins in the storage. Optionally starting from page_token. This logs regularly an info message during pagination with the page_token. This, in order to feed it back to the cli if the process interrupted. Yields origin model objects from the storage """ while True: page_result = storage.origin_list(page_token=page_token) page_token = page_result.next_page_token yield from page_result.results if not page_token: break click.echo(f"page_token: {page_token}\n") @task.command("schedule_origins") @click.argument("type", nargs=1, required=True) @click.argument("options", nargs=-1) @click.option( "--batch-size", "-b", "origin_batch_size", default=10, show_default=True, type=int, help="Number of origins per task", ) @click.option( "--page-token", default=0, show_default=True, type=str, help="Only schedule tasks for origins whose ID is greater", ) @click.option( "--limit", default=None, type=int, help="Limit the tasks scheduling up to this number of tasks", ) @click.option("--storage-url", "-g", help="URL of the (graph) storage API") @click.option( "--dry-run/--no-dry-run", is_flag=True, default=False, help="List only what would be scheduled.", ) @click.pass_context def schedule_origin_metadata_index( ctx, type, options, storage_url, origin_batch_size, page_token, limit, dry_run ): """Schedules tasks for origins that are already known. The first argument is the name of the task type, further ones are keyword argument(s) of the task in the form key=value, where value is in YAML format. Usage sample: swh-scheduler --database 'service=swh-scheduler' \ task schedule_origins index-origin-metadata """ from itertools import islice from swh.storage import get_storage from .utils import parse_options, schedule_origin_batches scheduler = ctx.obj["scheduler"] storage = get_storage("remote", url=storage_url) if dry_run: scheduler = None (args, kw) = parse_options(options) if args: raise click.ClickException("Only keywords arguments are allowed.") origins = iter_origins(storage, page_token=page_token) if limit: origins = islice(origins, limit) origin_urls = (origin.url for origin in origins) schedule_origin_batches(scheduler, type, origin_urls, origin_batch_size, kw) @task.command("list-pending") @click.argument("task-types", required=True, nargs=-1) @click.option( "--limit", "-l", required=False, type=click.INT, help="The maximum number of tasks to fetch", ) @click.option( "--before", "-b", required=False, type=DATETIME, help="List all jobs supposed to run before the given date", ) @click.pass_context def list_pending_tasks(ctx, task_types, limit, before): """List the tasks that are going to be run. You can override the number of tasks to fetch """ from swh.scheduler import compute_nb_tasks_from scheduler = ctx.obj["scheduler"] if not scheduler: raise ValueError("Scheduler class (local/remote) must be instantiated") num_tasks, num_tasks_priority = compute_nb_tasks_from(limit) output = [] for task_type in task_types: pending = scheduler.peek_ready_tasks( task_type, timestamp=before, num_tasks=num_tasks, num_tasks_priority=num_tasks_priority, ) output.append("Found %d %s tasks\n" % (len(pending), task_type)) for task in pending: output.append(pretty_print_task(task)) click.echo("\n".join(output)) @task.command("list") @click.option( "--task-id", "-i", default=None, multiple=True, metavar="ID", help="List only tasks whose id is ID.", ) @click.option( "--task-type", "-t", default=None, multiple=True, metavar="TYPE", help="List only tasks of type TYPE", ) @click.option( "--limit", "-l", required=False, type=click.INT, help="The maximum number of tasks to fetch.", ) @click.option( "--status", "-s", multiple=True, metavar="STATUS", type=click.Choice( ("next_run_not_scheduled", "next_run_scheduled", "completed", "disabled") ), default=None, help="List tasks whose status is STATUS.", ) @click.option( "--policy", "-p", default=None, type=click.Choice(["recurring", "oneshot"]), help="List tasks whose policy is POLICY.", ) @click.option( "--priority", "-P", default=None, multiple=True, type=click.Choice(["all", "low", "normal", "high"]), help="List tasks whose priority is PRIORITY.", ) @click.option( "--before", "-b", required=False, type=DATETIME, metavar="DATETIME", help="Limit to tasks supposed to run before the given date.", ) @click.option( "--after", "-a", required=False, type=DATETIME, metavar="DATETIME", help="Limit to tasks supposed to run after the given date.", ) @click.option( "--list-runs", "-r", is_flag=True, default=False, help="Also list past executions of each task.", ) @click.pass_context def list_tasks( ctx, task_id, task_type, limit, status, policy, priority, before, after, list_runs ): """List tasks. """ scheduler = ctx.obj["scheduler"] if not scheduler: raise ValueError("Scheduler class (local/remote) must be instantiated") if not task_type: task_type = [x["type"] for x in scheduler.get_task_types()] # if task_id is not given, default value for status is # 'next_run_not_scheduled' # if task_id is given, default status is 'all' if task_id is None and status is None: status = ["next_run_not_scheduled"] if status and "all" in status: status = None if priority and "all" in priority: priority = None output = [] tasks = scheduler.search_tasks( task_id=task_id, task_type=task_type, status=status, priority=priority, policy=policy, before=before, after=after, limit=limit, ) if list_runs: runs = {t["id"]: [] for t in tasks} for r in scheduler.get_task_runs([task["id"] for task in tasks]): runs[r["task"]].append(r) else: runs = {} output.append("Found %d tasks\n" % (len(tasks))) for task in tasks: output.append(pretty_print_task(task, full=True)) if runs.get(task["id"]): output.append(click.style(" Executions:", bold=True)) for run in runs[task["id"]]: output.append(pretty_print_run(run, indent=4)) click.echo("\n".join(output)) @task.command("respawn") @click.argument("task-ids", required=True, nargs=-1) @click.option( "--next-run", "-n", required=False, type=DATETIME, metavar="DATETIME", default=None, help="Re spawn the selected tasks at this date", ) @click.pass_context def respawn_tasks(ctx, task_ids, next_run): """Respawn tasks. Respawn tasks given by their ids (see the 'task list' command to find task ids) at the given date (immediately by default). Eg. swh-scheduler task respawn 1 3 12 """ - import arrow + from swh.scheduler.utils import utcnow scheduler = ctx.obj["scheduler"] if not scheduler: raise ValueError("Scheduler class (local/remote) must be instantiated") if next_run is None: - next_run = arrow.utcnow() + next_run = utcnow() output = [] scheduler.set_status_tasks( task_ids, status="next_run_not_scheduled", next_run=next_run ) output.append("Respawn tasks %s\n" % (task_ids,)) click.echo("\n".join(output)) @task.command("archive") @click.option( "--before", "-b", default=None, help="""Task whose ended date is anterior will be archived. Default to current month's first day.""", ) @click.option( "--after", "-a", default=None, help="""Task whose ended date is after the specified date will be archived. Default to prior month's first day.""", ) @click.option( "--batch-index", default=1000, type=click.INT, help="Batch size of tasks to read from db to archive", ) @click.option( "--bulk-index", default=200, type=click.INT, help="Batch size of tasks to bulk index", ) @click.option( "--batch-clean", default=1000, type=click.INT, help="Batch size of task to clean after archival", ) @click.option( "--dry-run/--no-dry-run", is_flag=True, default=False, help="Default to list only what would be archived.", ) @click.option("--verbose", is_flag=True, default=False, help="Verbose mode") @click.option( "--cleanup/--no-cleanup", is_flag=True, default=True, help="Clean up archived tasks (default)", ) @click.option( "--start-from", type=click.STRING, default=None, help="(Optional) default page to start from.", ) @click.pass_context def archive_tasks( ctx, before, after, batch_index, bulk_index, batch_clean, dry_run, verbose, cleanup, start_from, ): """Archive task/task_run whose (task_type is 'oneshot' and task_status is 'completed') or (task_type is 'recurring' and task_status is 'disabled'). With --dry-run flag set (default), only list those. """ from itertools import groupby - import arrow - from swh.core.utils import grouper from swh.scheduler.backend_es import ElasticSearchBackend + from swh.scheduler.utils import utcnow config = ctx.obj["config"] scheduler = ctx.obj["scheduler"] if not scheduler: raise ValueError("Scheduler class (local/remote) must be instantiated") logging.basicConfig(level=logging.DEBUG if verbose else logging.INFO) logger = logging.getLogger(__name__) logging.getLogger("urllib3").setLevel(logging.WARN) logging.getLogger("elasticsearch").setLevel(logging.ERROR) if dry_run: logger.info("**DRY-RUN** (only reading db)") if not cleanup: logger.info("**NO CLEANUP**") es_storage = ElasticSearchBackend(**config) - now = arrow.utcnow() + now = utcnow() # Default to archive tasks from a rolling month starting the week # prior to the current one if not before: before = now.shift(weeks=-1).format("YYYY-MM-DD") if not after: after = now.shift(weeks=-1).shift(months=-1).format("YYYY-MM-DD") logger.debug( "index: %s; cleanup: %s; period: [%s ; %s]" % (not dry_run, not dry_run and cleanup, after, before) ) def get_index_name( data: Dict[str, Any], es_storage: ElasticSearchBackend = es_storage ) -> str: """Given a data record, determine the index's name through its ending date. This varies greatly depending on the task_run's status. """ date = data.get("started") if not date: date = data["scheduled"] return es_storage.compute_index_name(date.year, date.month) def index_data(before, page_token, batch_index): while True: result = scheduler.filter_task_to_archive( after, before, page_token=page_token, limit=batch_index ) tasks_sorted = sorted(result["tasks"], key=get_index_name) groups = groupby(tasks_sorted, key=get_index_name) for index_name, tasks_group in groups: logger.debug("Index tasks to %s" % index_name) if dry_run: for task in tasks_group: yield task continue yield from es_storage.streaming_bulk( index_name, tasks_group, source=["task_id", "task_run_id"], chunk_size=bulk_index, ) page_token = result.get("next_page_token") if page_token is None: break gen = index_data(before, page_token=start_from, batch_index=batch_index) if cleanup: for task_ids in grouper(gen, n=batch_clean): task_ids = list(task_ids) logger.info("Clean up %s tasks: [%s, ...]" % (len(task_ids), task_ids[0])) if dry_run: # no clean up continue ctx.obj["scheduler"].delete_archived_tasks(task_ids) else: for task_ids in grouper(gen, n=batch_index): task_ids = list(task_ids) logger.info("Indexed %s tasks: [%s, ...]" % (len(task_ids), task_ids[0])) logger.debug("Done!") diff --git a/swh/scheduler/tests/es/test_cli_task.py b/swh/scheduler/tests/es/test_cli_task.py index d0182fd..f426198 100644 --- a/swh/scheduler/tests/es/test_cli_task.py +++ b/swh/scheduler/tests/es/test_cli_task.py @@ -1,111 +1,111 @@ # Copyright (C) 2019 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import datetime import logging import random import uuid -import arrow from click.testing import CliRunner import pytest from swh.scheduler.cli import cli +from swh.scheduler.utils import utcnow from ..common import TASK_TYPES, TEMPLATES, tasks_from_template logger = logging.getLogger(__name__) @pytest.mark.usefixtures("swh_elasticsearch_backend") def test_cli_archive_tasks(swh_sched, swh_sched_config_file): scheduler = swh_sched template_git = TEMPLATES["git"] template_hg = TEMPLATES["hg"] # first initialize scheduler's db (is this still needed?) for tt in TASK_TYPES.values(): scheduler.create_task_type(tt) - next_run_start = arrow.utcnow().datetime - datetime.timedelta(days=1) + next_run_start = utcnow() - datetime.timedelta(days=1) recurring = tasks_from_template(template_git, next_run_start, 100) oneshots = tasks_from_template( template_hg, next_run_start - datetime.timedelta(days=1), 50 ) past_time = next_run_start - datetime.timedelta(days=7) all_tasks = recurring + oneshots result = scheduler.create_tasks(all_tasks) assert len(result) == len(all_tasks) # simulate task run backend_tasks = [ { "task": task["id"], "backend_id": str(uuid.uuid4()), "scheduled": next_run_start - datetime.timedelta(minutes=i % 60), } for i, task in enumerate(result) ] scheduler.mass_schedule_task_runs(backend_tasks) # Disable some tasks tasks_to_disable = set() for task in result: status = random.choice(["disabled", "completed"]) if status == "disabled": tasks_to_disable.add(task["id"]) scheduler.disable_tasks(tasks_to_disable) git_tasks = scheduler.search_tasks(task_type=template_git["type"]) hg_tasks = scheduler.search_tasks(task_type=template_hg["type"]) assert len(git_tasks) + len(hg_tasks) == len(all_tasks) # Ensure the task_run are in expected state task_runs = scheduler.get_task_runs([t["id"] for t in git_tasks + hg_tasks]) # Same for the tasks for t in git_tasks + hg_tasks: if t["id"] in tasks_to_disable: assert t["status"] == "disabled" future_time = next_run_start + datetime.timedelta(days=1) for tr in task_runs: assert past_time <= tr["scheduled"] assert tr["scheduled"] < future_time runner = CliRunner() result = runner.invoke( cli, [ "--config-file", swh_sched_config_file, "task", "archive", "--after", past_time.isoformat(), "--before", future_time.isoformat(), "--cleanup", ], obj={"log_level": logging.DEBUG,}, ) assert result.exit_code == 0, result.output # disabled tasks should no longer be in the scheduler git_tasks = scheduler.search_tasks(task_type=template_git["type"]) hg_tasks = scheduler.search_tasks(task_type=template_hg["type"]) remaining_tasks = git_tasks + hg_tasks count_disabled = 0 for task in remaining_tasks: logger.debug(f"task status: {task['status']}") if task["status"] == "disabled": count_disabled += 1 assert count_disabled == 0 assert len(remaining_tasks) == len(all_tasks) - len(tasks_to_disable) diff --git a/swh/scheduler/tests/test_cli.py b/swh/scheduler/tests/test_cli.py index 4619119..ae14d88 100644 --- a/swh/scheduler/tests/test_cli.py +++ b/swh/scheduler/tests/test_cli.py @@ -1,795 +1,795 @@ # Copyright (C) 2019-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import datetime from itertools import islice import logging import re import tempfile from unittest.mock import patch from click.testing import CliRunner import pytest from swh.core.api.classes import stream_results from swh.model.model import Origin from swh.scheduler.cli import cli -from swh.scheduler.utils import create_task_dict +from swh.scheduler.utils import create_task_dict, utcnow from swh.storage import get_storage CLI_CONFIG = """ scheduler: cls: foo args: {} """ def invoke(scheduler, catch_exceptions, args): runner = CliRunner() with patch( "swh.scheduler.get_scheduler" ) as get_scheduler_mock, tempfile.NamedTemporaryFile( "a", suffix=".yml" ) as config_fd: config_fd.write(CLI_CONFIG) config_fd.seek(0) get_scheduler_mock.return_value = scheduler args = ["-C" + config_fd.name,] + args result = runner.invoke(cli, args, obj={"log_level": logging.WARNING}) if not catch_exceptions and result.exception: print(result.output) raise result.exception return result def test_schedule_tasks(swh_scheduler): csv_data = ( b'swh-test-ping;[["arg1", "arg2"]];{"key": "value"};' - + datetime.datetime.utcnow().isoformat().encode() + + utcnow().isoformat().encode() + b"\n" + b'swh-test-ping;[["arg3", "arg4"]];{"key": "value"};' - + datetime.datetime.utcnow().isoformat().encode() + + utcnow().isoformat().encode() + b"\n" ) with tempfile.NamedTemporaryFile(suffix=".csv") as csv_fd: csv_fd.write(csv_data) csv_fd.seek(0) result = invoke( swh_scheduler, False, ["task", "schedule", "-d", ";", csv_fd.name] ) expected = r""" Created 2 tasks Task 1 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: recurring Args: \['arg1', 'arg2'\] Keyword args: key: 'value' Task 2 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: recurring Args: \['arg3', 'arg4'\] Keyword args: key: 'value' """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), result.output def test_schedule_tasks_columns(swh_scheduler): with tempfile.NamedTemporaryFile(suffix=".csv") as csv_fd: csv_fd.write(b'swh-test-ping;oneshot;["arg1", "arg2"];{"key": "value"}\n') csv_fd.seek(0) result = invoke( swh_scheduler, False, [ "task", "schedule", "-c", "type", "-c", "policy", "-c", "args", "-c", "kwargs", "-d", ";", csv_fd.name, ], ) expected = r""" Created 1 tasks Task 1 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Args: 'arg1' 'arg2' Keyword args: key: 'value' """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), result.output def test_schedule_task(swh_scheduler): result = invoke( swh_scheduler, False, ["task", "add", "swh-test-ping", "arg1", "arg2", "key=value",], ) expected = r""" Created 1 tasks Task 1 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: recurring Args: 'arg1' 'arg2' Keyword args: key: 'value' """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), result.output def test_list_pending_tasks_none(swh_scheduler): result = invoke(swh_scheduler, False, ["task", "list-pending", "swh-test-ping",]) expected = r""" Found 0 swh-test-ping tasks """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), result.output def test_list_pending_tasks(swh_scheduler): task1 = create_task_dict("swh-test-ping", "oneshot", key="value1") task2 = create_task_dict("swh-test-ping", "oneshot", key="value2") task2["next_run"] += datetime.timedelta(days=1) swh_scheduler.create_tasks([task1, task2]) result = invoke(swh_scheduler, False, ["task", "list-pending", "swh-test-ping",]) expected = r""" Found 1 swh-test-ping tasks Task 1 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Args: Keyword args: key: 'value1' """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), result.output swh_scheduler.grab_ready_tasks("swh-test-ping") result = invoke(swh_scheduler, False, ["task", "list-pending", "swh-test-ping",]) expected = r""" Found 0 swh-test-ping tasks """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), result.output def test_list_pending_tasks_filter(swh_scheduler): task = create_task_dict("swh-test-multiping", "oneshot", key="value") swh_scheduler.create_tasks([task]) result = invoke(swh_scheduler, False, ["task", "list-pending", "swh-test-ping",]) expected = r""" Found 0 swh-test-ping tasks """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), result.output def test_list_pending_tasks_filter_2(swh_scheduler): swh_scheduler.create_tasks( [ create_task_dict("swh-test-multiping", "oneshot", key="value"), create_task_dict("swh-test-ping", "oneshot", key="value2"), ] ) result = invoke(swh_scheduler, False, ["task", "list-pending", "swh-test-ping",]) expected = r""" Found 1 swh-test-ping tasks Task 2 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Args: Keyword args: key: 'value2' """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), result.output # Fails because "task list-pending --limit 3" only returns 2 tasks, because # of how compute_nb_tasks_from works. @pytest.mark.xfail def test_list_pending_tasks_limit(swh_scheduler): swh_scheduler.create_tasks( [ create_task_dict("swh-test-ping", "oneshot", key="value%d" % i) for i in range(10) ] ) result = invoke( swh_scheduler, False, ["task", "list-pending", "swh-test-ping", "--limit", "3",] ) expected = r""" Found 2 swh-test-ping tasks Task 1 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Args: Keyword args: key: 'value0' Task 2 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Args: Keyword args: key: 'value1' Task 3 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Args: Keyword args: key: 'value2' """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), result.output def test_list_pending_tasks_before(swh_scheduler): task1 = create_task_dict("swh-test-ping", "oneshot", key="value") task2 = create_task_dict("swh-test-ping", "oneshot", key="value2") task1["next_run"] += datetime.timedelta(days=3) task2["next_run"] += datetime.timedelta(days=1) swh_scheduler.create_tasks([task1, task2]) result = invoke( swh_scheduler, False, [ "task", "list-pending", "swh-test-ping", "--before", (datetime.date.today() + datetime.timedelta(days=2)).isoformat(), ], ) expected = r""" Found 1 swh-test-ping tasks Task 2 - Next run: in a day \(.*\) + Next run: tomorrow \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Args: Keyword args: key: 'value2' """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), result.output def test_list_tasks(swh_scheduler): task1 = create_task_dict("swh-test-ping", "oneshot", key="value1") task2 = create_task_dict("swh-test-ping", "oneshot", key="value2") task1["next_run"] += datetime.timedelta(days=3, hours=2) swh_scheduler.create_tasks([task1, task2]) swh_scheduler.grab_ready_tasks("swh-test-ping") result = invoke(swh_scheduler, False, ["task", "list",]) expected = r""" Found 2 tasks Task 1 - Next run: in 3 days \(.*\) + Next run: .+ \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Status: next_run_not_scheduled Priority:\x20 Args: Keyword args: key: 'value1' Task 2 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Status: next_run_scheduled Priority:\x20 Args: Keyword args: key: 'value2' """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), result.output def test_list_tasks_id(swh_scheduler): task1 = create_task_dict("swh-test-ping", "oneshot", key="value1") task2 = create_task_dict("swh-test-ping", "oneshot", key="value2") task3 = create_task_dict("swh-test-ping", "oneshot", key="value3") swh_scheduler.create_tasks([task1, task2, task3]) result = invoke(swh_scheduler, False, ["task", "list", "--task-id", "2",]) expected = r""" Found 1 tasks Task 2 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Status: next_run_not_scheduled Priority:\x20 Args: Keyword args: key: 'value2' """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), result.output def test_list_tasks_id_2(swh_scheduler): task1 = create_task_dict("swh-test-ping", "oneshot", key="value1") task2 = create_task_dict("swh-test-ping", "oneshot", key="value2") task3 = create_task_dict("swh-test-ping", "oneshot", key="value3") swh_scheduler.create_tasks([task1, task2, task3]) result = invoke( swh_scheduler, False, ["task", "list", "--task-id", "2", "--task-id", "3"] ) expected = r""" Found 2 tasks Task 2 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Status: next_run_not_scheduled Priority:\x20 Args: Keyword args: key: 'value2' Task 3 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Status: next_run_not_scheduled Priority:\x20 Args: Keyword args: key: 'value3' """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), result.output def test_list_tasks_type(swh_scheduler): task1 = create_task_dict("swh-test-ping", "oneshot", key="value1") task2 = create_task_dict("swh-test-multiping", "oneshot", key="value2") task3 = create_task_dict("swh-test-ping", "oneshot", key="value3") swh_scheduler.create_tasks([task1, task2, task3]) result = invoke( swh_scheduler, False, ["task", "list", "--task-type", "swh-test-ping"] ) expected = r""" Found 2 tasks Task 1 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Status: next_run_not_scheduled Priority:\x20 Args: Keyword args: key: 'value1' Task 3 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Status: next_run_not_scheduled Priority:\x20 Args: Keyword args: key: 'value3' """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), result.output def test_list_tasks_limit(swh_scheduler): task1 = create_task_dict("swh-test-ping", "oneshot", key="value1") task2 = create_task_dict("swh-test-ping", "oneshot", key="value2") task3 = create_task_dict("swh-test-ping", "oneshot", key="value3") swh_scheduler.create_tasks([task1, task2, task3]) result = invoke(swh_scheduler, False, ["task", "list", "--limit", "2",]) expected = r""" Found 2 tasks Task 1 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Status: next_run_not_scheduled Priority:\x20 Args: Keyword args: key: 'value1' Task 2 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Status: next_run_not_scheduled Priority:\x20 Args: Keyword args: key: 'value2' """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), result.output def test_list_tasks_before(swh_scheduler): task1 = create_task_dict("swh-test-ping", "oneshot", key="value1") task2 = create_task_dict("swh-test-ping", "oneshot", key="value2") task1["next_run"] += datetime.timedelta(days=3, hours=2) swh_scheduler.create_tasks([task1, task2]) swh_scheduler.grab_ready_tasks("swh-test-ping") result = invoke( swh_scheduler, False, [ "task", "list", "--before", (datetime.date.today() + datetime.timedelta(days=2)).isoformat(), ], ) expected = r""" Found 1 tasks Task 2 - Next run: just now \(.*\) + Next run: today \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Status: next_run_scheduled Priority:\x20 Args: Keyword args: key: 'value2' """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), result.output def test_list_tasks_after(swh_scheduler): task1 = create_task_dict("swh-test-ping", "oneshot", key="value1") task2 = create_task_dict("swh-test-ping", "oneshot", key="value2") task1["next_run"] += datetime.timedelta(days=3, hours=2) swh_scheduler.create_tasks([task1, task2]) swh_scheduler.grab_ready_tasks("swh-test-ping") result = invoke( swh_scheduler, False, [ "task", "list", "--after", (datetime.date.today() + datetime.timedelta(days=2)).isoformat(), ], ) expected = r""" Found 1 tasks Task 1 - Next run: in 3 days \(.*\) + Next run: .+ \(.*\) Interval: 1 day, 0:00:00 Type: swh-test-ping Policy: oneshot Status: next_run_not_scheduled Priority:\x20 Args: Keyword args: key: 'value1' """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), result.output def _fill_storage_with_origins(storage, nb_origins): origins = [Origin(url=f"http://example.com/{i}") for i in range(nb_origins)] storage.origin_add(origins) return origins @pytest.fixture def storage(): """An instance of in-memory storage that gets injected into the CLI functions.""" storage = get_storage(cls="memory") with patch("swh.storage.get_storage") as get_storage_mock: get_storage_mock.return_value = storage yield storage @patch("swh.scheduler.cli.utils.TASK_BATCH_SIZE", 3) def test_task_schedule_origins_dry_run(swh_scheduler, storage): """Tests the scheduling when origin_batch_size*task_batch_size is a divisor of nb_origins.""" _fill_storage_with_origins(storage, 90) result = invoke( swh_scheduler, False, ["task", "schedule_origins", "--dry-run", "swh-test-ping",], ) # Check the output expected = r""" Scheduled 3 tasks \(30 origins\). Scheduled 6 tasks \(60 origins\). Scheduled 9 tasks \(90 origins\). Done. """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), repr(result.output) # Check scheduled tasks tasks = swh_scheduler.search_tasks() assert len(tasks) == 0 def _assert_origin_tasks_contraints(tasks, max_tasks, max_task_size, expected_origins): # check there are not too many tasks assert len(tasks) <= max_tasks # check tasks are not too large assert all(len(task["arguments"]["args"][0]) <= max_task_size for task in tasks) # check the tasks are exhaustive assert sum([len(task["arguments"]["args"][0]) for task in tasks]) == len( expected_origins ) assert set.union(*(set(task["arguments"]["args"][0]) for task in tasks)) == { origin.url for origin in expected_origins } @patch("swh.scheduler.cli.utils.TASK_BATCH_SIZE", 3) def test_task_schedule_origins(swh_scheduler, storage): """Tests the scheduling when neither origin_batch_size or task_batch_size is a divisor of nb_origins.""" origins = _fill_storage_with_origins(storage, 70) result = invoke( swh_scheduler, False, ["task", "schedule_origins", "swh-test-ping", "--batch-size", "20",], ) # Check the output expected = r""" Scheduled 3 tasks \(60 origins\). Scheduled 4 tasks \(70 origins\). Done. """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), repr(result.output) # Check tasks tasks = swh_scheduler.search_tasks() _assert_origin_tasks_contraints(tasks, 4, 20, origins) assert all(task["arguments"]["kwargs"] == {} for task in tasks) def test_task_schedule_origins_kwargs(swh_scheduler, storage): """Tests support of extra keyword-arguments.""" origins = _fill_storage_with_origins(storage, 30) result = invoke( swh_scheduler, False, [ "task", "schedule_origins", "swh-test-ping", "--batch-size", "20", 'key1="value1"', 'key2="value2"', ], ) # Check the output expected = r""" Scheduled 2 tasks \(30 origins\). Done. """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), repr(result.output) # Check tasks tasks = swh_scheduler.search_tasks() _assert_origin_tasks_contraints(tasks, 2, 20, origins) assert all( task["arguments"]["kwargs"] == {"key1": "value1", "key2": "value2"} for task in tasks ) def test_task_schedule_origins_with_limit(swh_scheduler, storage): """Tests support of extra keyword-arguments.""" _fill_storage_with_origins(storage, 50) limit = 20 expected_origins = list(islice(stream_results(storage.origin_list), limit)) nb_origins = len(expected_origins) assert nb_origins == limit max_task_size = 5 nb_tasks, remainder = divmod(nb_origins, max_task_size) assert remainder == 0 # made the numbers go round result = invoke( swh_scheduler, False, [ "task", "schedule_origins", "swh-test-ping", "--batch-size", max_task_size, "--limit", limit, ], ) # Check the output expected = rf""" Scheduled {nb_tasks} tasks \({nb_origins} origins\). Done. """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), repr(result.output) tasks = swh_scheduler.search_tasks() _assert_origin_tasks_contraints(tasks, max_task_size, nb_origins, expected_origins) def test_task_schedule_origins_with_page_token(swh_scheduler, storage): """Tests support of extra keyword-arguments.""" nb_total_origins = 50 origins = _fill_storage_with_origins(storage, nb_total_origins) # prepare page_token and origins result expectancy page_result = storage.origin_list(limit=10) assert len(page_result.results) == 10 page_token = page_result.next_page_token assert page_token is not None # remove the first 10 origins listed as we won't see those in tasks expected_origins = [o for o in origins if o not in page_result.results] nb_origins = len(expected_origins) assert nb_origins == nb_total_origins - len(page_result.results) max_task_size = 10 nb_tasks, remainder = divmod(nb_origins, max_task_size) assert remainder == 0 result = invoke( swh_scheduler, False, [ "task", "schedule_origins", "swh-test-ping", "--batch-size", max_task_size, "--page-token", page_token, ], ) # Check the output expected = rf""" Scheduled {nb_tasks} tasks \({nb_origins} origins\). Done. """.lstrip() assert result.exit_code == 0, result.output assert re.fullmatch(expected, result.output, re.MULTILINE), repr(result.output) # Check tasks tasks = swh_scheduler.search_tasks() _assert_origin_tasks_contraints(tasks, max_task_size, nb_origins, expected_origins) diff --git a/swh/scheduler/tests/test_scheduler.py b/swh/scheduler/tests/test_scheduler.py index bdc8e57..9c8f6b7 100644 --- a/swh/scheduler/tests/test_scheduler.py +++ b/swh/scheduler/tests/test_scheduler.py @@ -1,729 +1,731 @@ # Copyright (C) 2017-2019 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from collections import defaultdict import copy import datetime import inspect import random from typing import Any, Dict, List, Optional import uuid -from arrow import utcnow import attr import pytest from swh.scheduler.exc import StaleData from swh.scheduler.interface import SchedulerInterface from swh.scheduler.model import ListedOrigin, ListedOriginPageToken +from swh.scheduler.utils import utcnow from .common import LISTERS, TASK_TYPES, TEMPLATES, tasks_from_template +ONEDAY = datetime.timedelta(days=1) + def subdict(d, keys=None, excl=()): if keys is None: keys = [k for k in d.keys()] return {k: d[k] for k in keys if k not in excl} class TestScheduler: def test_interface(self, swh_scheduler): """Checks all methods of SchedulerInterface are implemented by this backend, and that they have the same signature.""" # Create an instance of the protocol (which cannot be instantiated # directly, so this creates a subclass, then instantiates it) interface = type("_", (SchedulerInterface,), {})() assert "create_task_type" in dir(interface) missing_methods = [] for meth_name in dir(interface): if meth_name.startswith("_"): continue interface_meth = getattr(interface, meth_name) try: concrete_meth = getattr(swh_scheduler, meth_name) except AttributeError: if not getattr(interface_meth, "deprecated_endpoint", False): # The backend is missing a (non-deprecated) endpoint missing_methods.append(meth_name) continue expected_signature = inspect.signature(interface_meth) actual_signature = inspect.signature(concrete_meth) assert expected_signature == actual_signature, meth_name assert missing_methods == [] def test_get_priority_ratios(self, swh_scheduler): assert swh_scheduler.get_priority_ratios() == { "high": 0.5, "normal": 0.3, "low": 0.2, } def test_add_task_type(self, swh_scheduler): tt = TASK_TYPES["git"] swh_scheduler.create_task_type(tt) assert tt == swh_scheduler.get_task_type(tt["type"]) tt2 = TASK_TYPES["hg"] swh_scheduler.create_task_type(tt2) assert tt == swh_scheduler.get_task_type(tt["type"]) assert tt2 == swh_scheduler.get_task_type(tt2["type"]) def test_create_task_type_idempotence(self, swh_scheduler): tt = TASK_TYPES["git"] swh_scheduler.create_task_type(tt) swh_scheduler.create_task_type(tt) assert tt == swh_scheduler.get_task_type(tt["type"]) def test_get_task_types(self, swh_scheduler): tt, tt2 = TASK_TYPES["git"], TASK_TYPES["hg"] swh_scheduler.create_task_type(tt) swh_scheduler.create_task_type(tt2) actual_task_types = swh_scheduler.get_task_types() assert tt in actual_task_types assert tt2 in actual_task_types def test_create_tasks(self, swh_scheduler): priority_ratio = self._priority_ratio(swh_scheduler) self._create_task_types(swh_scheduler) num_tasks_priority = 100 tasks_1 = tasks_from_template(TEMPLATES["git"], utcnow(), 100) tasks_2 = tasks_from_template( TEMPLATES["hg"], utcnow(), 100, num_tasks_priority, priorities=priority_ratio, ) tasks = tasks_1 + tasks_2 # tasks are returned only once with their ids ret1 = swh_scheduler.create_tasks(tasks + tasks_1 + tasks_2) set_ret1 = set([t["id"] for t in ret1]) # creating the same set result in the same ids ret = swh_scheduler.create_tasks(tasks) set_ret = set([t["id"] for t in ret]) # Idempotence results assert set_ret == set_ret1 assert len(ret) == len(ret1) ids = set() actual_priorities = defaultdict(int) for task, orig_task in zip(ret, tasks): task = copy.deepcopy(task) task_type = TASK_TYPES[orig_task["type"].split("-")[-1]] assert task["id"] not in ids assert task["status"] == "next_run_not_scheduled" assert task["current_interval"] == task_type["default_interval"] assert task["policy"] == orig_task.get("policy", "recurring") priority = task.get("priority") if priority: actual_priorities[priority] += 1 assert task["retries_left"] == (task_type["num_retries"] or 0) ids.add(task["id"]) del task["id"] del task["status"] del task["current_interval"] del task["retries_left"] if "policy" not in orig_task: del task["policy"] if "priority" not in orig_task: del task["priority"] assert task == orig_task assert dict(actual_priorities) == { priority: int(ratio * num_tasks_priority) for priority, ratio in priority_ratio.items() } def test_peek_ready_tasks_no_priority(self, swh_scheduler): self._create_task_types(swh_scheduler) t = utcnow() task_type = TEMPLATES["git"]["type"] tasks = tasks_from_template(TEMPLATES["git"], t, 100) random.shuffle(tasks) swh_scheduler.create_tasks(tasks) ready_tasks = swh_scheduler.peek_ready_tasks(task_type) assert len(ready_tasks) == len(tasks) for i in range(len(ready_tasks) - 1): assert ready_tasks[i]["next_run"] <= ready_tasks[i + 1]["next_run"] # Only get the first few ready tasks limit = random.randrange(5, 5 + len(tasks) // 2) ready_tasks_limited = swh_scheduler.peek_ready_tasks(task_type, num_tasks=limit) assert len(ready_tasks_limited) == limit assert ready_tasks_limited == ready_tasks[:limit] # Limit by timestamp max_ts = tasks[limit - 1]["next_run"] ready_tasks_timestamped = swh_scheduler.peek_ready_tasks( task_type, timestamp=max_ts ) for ready_task in ready_tasks_timestamped: assert ready_task["next_run"] <= max_ts # Make sure we get proper behavior for the first ready tasks assert ready_tasks[: len(ready_tasks_timestamped)] == ready_tasks_timestamped # Limit by both ready_tasks_both = swh_scheduler.peek_ready_tasks( task_type, timestamp=max_ts, num_tasks=limit // 3 ) assert len(ready_tasks_both) <= limit // 3 for ready_task in ready_tasks_both: assert ready_task["next_run"] <= max_ts assert ready_task in ready_tasks[: limit // 3] def _priority_ratio(self, swh_scheduler): return swh_scheduler.get_priority_ratios() def test_peek_ready_tasks_mixed_priorities(self, swh_scheduler): priority_ratio = self._priority_ratio(swh_scheduler) self._create_task_types(swh_scheduler) t = utcnow() task_type = TEMPLATES["git"]["type"] num_tasks_priority = 100 num_tasks_no_priority = 100 # Create tasks with and without priorities tasks = tasks_from_template( TEMPLATES["git"], t, num=num_tasks_no_priority, num_priority=num_tasks_priority, priorities=priority_ratio, ) random.shuffle(tasks) swh_scheduler.create_tasks(tasks) # take all available tasks ready_tasks = swh_scheduler.peek_ready_tasks(task_type) assert len(ready_tasks) == len(tasks) assert num_tasks_priority + num_tasks_no_priority == len(ready_tasks) count_tasks_per_priority = defaultdict(int) for task in ready_tasks: priority = task.get("priority") if priority: count_tasks_per_priority[priority] += 1 assert dict(count_tasks_per_priority) == { priority: int(ratio * num_tasks_priority) for priority, ratio in priority_ratio.items() } # Only get some ready tasks num_tasks = random.randrange(5, 5 + num_tasks_no_priority // 2) num_tasks_priority = random.randrange(5, num_tasks_priority // 2) ready_tasks_limited = swh_scheduler.peek_ready_tasks( task_type, num_tasks=num_tasks, num_tasks_priority=num_tasks_priority ) count_tasks_per_priority = defaultdict(int) for task in ready_tasks_limited: priority = task.get("priority") count_tasks_per_priority[priority] += 1 import math for priority, ratio in priority_ratio.items(): expected_count = math.ceil(ratio * num_tasks_priority) actual_prio = count_tasks_per_priority[priority] assert actual_prio == expected_count or actual_prio == expected_count + 1 assert count_tasks_per_priority[None] == num_tasks def test_grab_ready_tasks(self, swh_scheduler): priority_ratio = self._priority_ratio(swh_scheduler) self._create_task_types(swh_scheduler) t = utcnow() task_type = TEMPLATES["git"]["type"] num_tasks_priority = 100 num_tasks_no_priority = 100 # Create tasks with and without priorities tasks = tasks_from_template( TEMPLATES["git"], t, num=num_tasks_no_priority, num_priority=num_tasks_priority, priorities=priority_ratio, ) random.shuffle(tasks) swh_scheduler.create_tasks(tasks) first_ready_tasks = swh_scheduler.peek_ready_tasks( task_type, num_tasks=10, num_tasks_priority=10 ) grabbed_tasks = swh_scheduler.grab_ready_tasks( task_type, num_tasks=10, num_tasks_priority=10 ) for peeked, grabbed in zip(first_ready_tasks, grabbed_tasks): assert peeked["status"] == "next_run_not_scheduled" del peeked["status"] assert grabbed["status"] == "next_run_scheduled" del grabbed["status"] assert peeked == grabbed assert peeked["priority"] == grabbed["priority"] def test_get_tasks(self, swh_scheduler): self._create_task_types(swh_scheduler) t = utcnow() tasks = tasks_from_template(TEMPLATES["git"], t, 100) tasks = swh_scheduler.create_tasks(tasks) random.shuffle(tasks) while len(tasks) > 1: length = random.randrange(1, len(tasks)) cur_tasks = sorted(tasks[:length], key=lambda x: x["id"]) tasks[:length] = [] ret = swh_scheduler.get_tasks(task["id"] for task in cur_tasks) # result is not guaranteed to be sorted ret.sort(key=lambda x: x["id"]) assert ret == cur_tasks def test_search_tasks(self, swh_scheduler): def make_real_dicts(lst): """RealDictRow is not a real dict.""" return [dict(d.items()) for d in lst] self._create_task_types(swh_scheduler) t = utcnow() tasks = tasks_from_template(TEMPLATES["git"], t, 100) tasks = swh_scheduler.create_tasks(tasks) assert make_real_dicts(swh_scheduler.search_tasks()) == make_real_dicts(tasks) def assert_filtered_task_ok( self, task: Dict[str, Any], after: datetime.datetime, before: datetime.datetime ) -> None: """Ensure filtered tasks have the right expected properties (within the range, recurring disabled, etc..) """ started = task["started"] date = started if started is not None else task["scheduled"] assert after <= date and date <= before if task["task_policy"] == "oneshot": assert task["task_status"] in ["completed", "disabled"] if task["task_policy"] == "recurring": assert task["task_status"] in ["disabled"] def test_filter_task_to_archive(self, swh_scheduler): """Filtering only list disabled recurring or completed oneshot tasks """ self._create_task_types(swh_scheduler) _time = utcnow() recurring = tasks_from_template(TEMPLATES["git"], _time, 12) oneshots = tasks_from_template(TEMPLATES["hg"], _time, 12) total_tasks = len(recurring) + len(oneshots) # simulate scheduling tasks pending_tasks = swh_scheduler.create_tasks(recurring + oneshots) backend_tasks = [ { "task": task["id"], "backend_id": str(uuid.uuid4()), "scheduled": utcnow(), } for task in pending_tasks ] swh_scheduler.mass_schedule_task_runs(backend_tasks) # we simulate the task are being done _tasks = [] for task in backend_tasks: t = swh_scheduler.end_task_run(task["backend_id"], status="eventful") _tasks.append(t) # Randomly update task's status per policy status_per_policy = {"recurring": 0, "oneshot": 0} status_choice = { # policy: [tuple (1-for-filtering, 'associated-status')] "recurring": [ (1, "disabled"), (0, "completed"), (0, "next_run_not_scheduled"), ], "oneshot": [ (0, "next_run_not_scheduled"), (1, "disabled"), (1, "completed"), ], } tasks_to_update = defaultdict(list) _task_ids = defaultdict(list) # randomize 'disabling' recurring task or 'complete' oneshot task for task in pending_tasks: policy = task["policy"] _task_ids[policy].append(task["id"]) status = random.choice(status_choice[policy]) if status[0] != 1: continue # elected for filtering status_per_policy[policy] += status[0] tasks_to_update[policy].append(task["id"]) swh_scheduler.disable_tasks(tasks_to_update["recurring"]) # hack: change the status to something else than completed/disabled swh_scheduler.set_status_tasks( _task_ids["oneshot"], status="next_run_not_scheduled" ) # complete the tasks to update swh_scheduler.set_status_tasks(tasks_to_update["oneshot"], status="completed") total_tasks_filtered = ( status_per_policy["recurring"] + status_per_policy["oneshot"] ) # no pagination scenario # retrieve tasks to archive - after = _time.shift(days=-1) - after_ts = after.format("YYYY-MM-DD") - before = utcnow().shift(days=1) - before_ts = before.format("YYYY-MM-DD") + after = _time - ONEDAY + after_ts = after.strftime("%Y-%m-%d") + before = utcnow() + ONEDAY + before_ts = before.strftime("%Y-%m-%d") tasks_result = swh_scheduler.filter_task_to_archive( after_ts=after_ts, before_ts=before_ts, limit=total_tasks ) tasks_to_archive = tasks_result["tasks"] assert len(tasks_to_archive) == total_tasks_filtered assert tasks_result.get("next_page_token") is None actual_filtered_per_status = {"recurring": 0, "oneshot": 0} for task in tasks_to_archive: self.assert_filtered_task_ok(task, after, before) actual_filtered_per_status[task["task_policy"]] += 1 assert actual_filtered_per_status == status_per_policy # pagination scenario nb_tasks = 3 tasks_result = swh_scheduler.filter_task_to_archive( after_ts=after_ts, before_ts=before_ts, limit=nb_tasks ) tasks_to_archive2 = tasks_result["tasks"] assert len(tasks_to_archive2) == nb_tasks next_page_token = tasks_result["next_page_token"] assert next_page_token is not None all_tasks = tasks_to_archive2 while next_page_token is not None: # Retrieve paginated results tasks_result = swh_scheduler.filter_task_to_archive( after_ts=after_ts, before_ts=before_ts, limit=nb_tasks, page_token=next_page_token, ) tasks_to_archive2 = tasks_result["tasks"] assert len(tasks_to_archive2) <= nb_tasks all_tasks.extend(tasks_to_archive2) next_page_token = tasks_result.get("next_page_token") actual_filtered_per_status = {"recurring": 0, "oneshot": 0} for task in all_tasks: self.assert_filtered_task_ok(task, after, before) actual_filtered_per_status[task["task_policy"]] += 1 assert actual_filtered_per_status == status_per_policy def test_delete_archived_tasks(self, swh_scheduler): self._create_task_types(swh_scheduler) _time = utcnow() recurring = tasks_from_template(TEMPLATES["git"], _time, 12) oneshots = tasks_from_template(TEMPLATES["hg"], _time, 12) total_tasks = len(recurring) + len(oneshots) pending_tasks = swh_scheduler.create_tasks(recurring + oneshots) backend_tasks = [ { "task": task["id"], "backend_id": str(uuid.uuid4()), "scheduled": utcnow(), } for task in pending_tasks ] swh_scheduler.mass_schedule_task_runs(backend_tasks) _tasks = [] percent = random.randint(0, 100) # random election removal boundary for task in backend_tasks: t = swh_scheduler.end_task_run(task["backend_id"], status="eventful") c = random.randint(0, 100) if c <= percent: _tasks.append({"task_id": t["task"], "task_run_id": t["id"]}) swh_scheduler.delete_archived_tasks(_tasks) all_tasks = [task["id"] for task in swh_scheduler.search_tasks()] tasks_count = len(all_tasks) tasks_run_count = len(swh_scheduler.get_task_runs(all_tasks)) assert tasks_count == total_tasks - len(_tasks) assert tasks_run_count == total_tasks - len(_tasks) def test_get_task_runs_no_task(self, swh_scheduler): """No task exist in the scheduler's db, get_task_runs() should always return an empty list. """ assert not swh_scheduler.get_task_runs(task_ids=()) assert not swh_scheduler.get_task_runs(task_ids=(1, 2, 3)) assert not swh_scheduler.get_task_runs(task_ids=(1, 2, 3), limit=10) def test_get_task_runs_no_task_executed(self, swh_scheduler): """No task has been executed yet, get_task_runs() should always return an empty list. """ self._create_task_types(swh_scheduler) _time = utcnow() recurring = tasks_from_template(TEMPLATES["git"], _time, 12) oneshots = tasks_from_template(TEMPLATES["hg"], _time, 12) swh_scheduler.create_tasks(recurring + oneshots) assert not swh_scheduler.get_task_runs(task_ids=()) assert not swh_scheduler.get_task_runs(task_ids=(1, 2, 3)) assert not swh_scheduler.get_task_runs(task_ids=(1, 2, 3), limit=10) def test_get_task_runs_with_scheduled(self, swh_scheduler): """Some tasks have been scheduled but not executed yet, get_task_runs() should not return an empty list. limit should behave as expected. """ self._create_task_types(swh_scheduler) _time = utcnow() recurring = tasks_from_template(TEMPLATES["git"], _time, 12) oneshots = tasks_from_template(TEMPLATES["hg"], _time, 12) total_tasks = len(recurring) + len(oneshots) pending_tasks = swh_scheduler.create_tasks(recurring + oneshots) backend_tasks = [ { "task": task["id"], "backend_id": str(uuid.uuid4()), "scheduled": utcnow(), } for task in pending_tasks ] swh_scheduler.mass_schedule_task_runs(backend_tasks) assert not swh_scheduler.get_task_runs(task_ids=[total_tasks + 1]) btask = backend_tasks[0] runs = swh_scheduler.get_task_runs(task_ids=[btask["task"]]) assert len(runs) == 1 run = runs[0] assert subdict(run, excl=("id",)) == { "task": btask["task"], "backend_id": btask["backend_id"], "scheduled": btask["scheduled"], "started": None, "ended": None, "metadata": None, "status": "scheduled", } runs = swh_scheduler.get_task_runs( task_ids=[bt["task"] for bt in backend_tasks], limit=2 ) assert len(runs) == 2 runs = swh_scheduler.get_task_runs( task_ids=[bt["task"] for bt in backend_tasks] ) assert len(runs) == total_tasks keys = ("task", "backend_id", "scheduled") assert ( sorted([subdict(x, keys) for x in runs], key=lambda x: x["task"]) == backend_tasks ) def test_get_task_runs_with_executed(self, swh_scheduler): """Some tasks have been executed, get_task_runs() should not return an empty list. limit should behave as expected. """ self._create_task_types(swh_scheduler) _time = utcnow() recurring = tasks_from_template(TEMPLATES["git"], _time, 12) oneshots = tasks_from_template(TEMPLATES["hg"], _time, 12) pending_tasks = swh_scheduler.create_tasks(recurring + oneshots) backend_tasks = [ { "task": task["id"], "backend_id": str(uuid.uuid4()), "scheduled": utcnow(), } for task in pending_tasks ] swh_scheduler.mass_schedule_task_runs(backend_tasks) btask = backend_tasks[0] ts = utcnow() swh_scheduler.start_task_run( btask["backend_id"], metadata={"something": "stupid"}, timestamp=ts ) runs = swh_scheduler.get_task_runs(task_ids=[btask["task"]]) assert len(runs) == 1 assert subdict(runs[0], excl=("id")) == { "task": btask["task"], "backend_id": btask["backend_id"], "scheduled": btask["scheduled"], "started": ts, "ended": None, "metadata": {"something": "stupid"}, "status": "started", } ts2 = utcnow() swh_scheduler.end_task_run( btask["backend_id"], metadata={"other": "stuff"}, timestamp=ts2, status="eventful", ) runs = swh_scheduler.get_task_runs(task_ids=[btask["task"]]) assert len(runs) == 1 assert subdict(runs[0], excl=("id")) == { "task": btask["task"], "backend_id": btask["backend_id"], "scheduled": btask["scheduled"], "started": ts, "ended": ts2, "metadata": {"something": "stupid", "other": "stuff"}, "status": "eventful", } def test_get_or_create_lister(self, swh_scheduler): db_listers = [] for lister_args in LISTERS: db_listers.append(swh_scheduler.get_or_create_lister(**lister_args)) for lister, lister_args in zip(db_listers, LISTERS): assert lister.name == lister_args["name"] assert lister.instance_name == lister_args.get("instance_name", "") lister_get_again = swh_scheduler.get_or_create_lister( lister.name, lister.instance_name ) assert lister == lister_get_again def test_update_lister(self, swh_scheduler, stored_lister): lister = attr.evolve(stored_lister, current_state={"updated": "now"}) updated_lister = swh_scheduler.update_lister(lister) assert updated_lister.updated > lister.updated assert updated_lister == attr.evolve(lister, updated=updated_lister.updated) def test_update_lister_stale(self, swh_scheduler, stored_lister): swh_scheduler.update_lister(stored_lister) with pytest.raises(StaleData) as exc: swh_scheduler.update_lister(stored_lister) assert "state not updated" in exc.value.args[0] def test_record_listed_origins(self, swh_scheduler, listed_origins): ret = swh_scheduler.record_listed_origins(listed_origins) assert set(returned.url for returned in ret) == set( origin.url for origin in listed_origins ) assert all(origin.first_seen == origin.last_seen for origin in ret) def test_record_listed_origins_upsert(self, swh_scheduler, listed_origins): # First, insert `cutoff` origins cutoff = 100 assert cutoff < len(listed_origins) ret = swh_scheduler.record_listed_origins(listed_origins[:cutoff]) assert len(ret) == cutoff # Then, insert all origins, including the `cutoff` first. ret = swh_scheduler.record_listed_origins(listed_origins) assert len(ret) == len(listed_origins) # Two different "first seen" values assert len(set(origin.first_seen for origin in ret)) == 2 # But a single "last seen" value assert len(set(origin.last_seen for origin in ret)) == 1 def test_get_listed_origins_exact(self, swh_scheduler, listed_origins): swh_scheduler.record_listed_origins(listed_origins) for i, origin in enumerate(listed_origins): ret = swh_scheduler.get_listed_origins( lister_id=origin.lister_id, url=origin.url ) assert ret.next_page_token is None assert len(ret.origins) == 1 assert ret.origins[0].lister_id == origin.lister_id assert ret.origins[0].url == origin.url @pytest.mark.parametrize("num_origins,limit", [(20, 6), (5, 42), (20, 20)]) def test_get_listed_origins_limit( self, swh_scheduler, listed_origins, num_origins, limit ) -> None: added_origins = sorted( listed_origins[:num_origins], key=lambda o: (o.lister_id, o.url) ) swh_scheduler.record_listed_origins(added_origins) returned_origins: List[ListedOrigin] = [] call_count = 0 next_page_token: Optional[ListedOriginPageToken] = None while True: call_count += 1 ret = swh_scheduler.get_listed_origins( lister_id=listed_origins[0].lister_id, limit=limit, page_token=next_page_token, ) returned_origins.extend(ret.origins) next_page_token = ret.next_page_token if next_page_token is None: break assert call_count == (num_origins // limit) + 1 assert len(returned_origins) == num_origins assert [(origin.lister_id, origin.url) for origin in returned_origins] == [ (origin.lister_id, origin.url) for origin in added_origins ] def test_get_listed_origins_all(self, swh_scheduler, listed_origins) -> None: swh_scheduler.record_listed_origins(listed_origins) ret = swh_scheduler.get_listed_origins(limit=len(listed_origins) + 1) assert ret.next_page_token is None assert len(ret.origins) == len(listed_origins) def _create_task_types(self, scheduler): for tt in TASK_TYPES.values(): scheduler.create_task_type(tt) diff --git a/swh/scheduler/utils.py b/swh/scheduler/utils.py index 9f6718a..f8e7935 100644 --- a/swh/scheduler/utils.py +++ b/swh/scheduler/utils.py @@ -1,76 +1,80 @@ # Copyright (C) 2017-2018 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from datetime import datetime, timezone +def utcnow(): + return datetime.now(tz=timezone.utc) + + def get_task(task_name): """Retrieve task object in our application instance by its fully qualified python name. Args: task_name (str): task's name (e.g swh.loader.git.tasks.LoadDiskGitRepository) Returns: Instance of task """ from swh.scheduler.celery_backend.config import app for module in app.conf.CELERY_IMPORTS: __import__(module) return app.tasks[task_name] def create_task_dict(type, policy, *args, **kwargs): """Create a task with type and policy, scheduled for as soon as possible. Args: type (str): Type of oneshot task as per swh-scheduler's db table task_type's column (Ex: load-git, check-deposit) policy (str): oneshot or recurring policy Returns: Expected dictionary for the one-shot task scheduling api (swh.scheduler.backend.create_tasks) """ task_extra = {} for extra_key in ["priority", "retries_left"]: if extra_key in kwargs: extra_val = kwargs.pop(extra_key) task_extra[extra_key] = extra_val task = { "policy": policy, "type": type, - "next_run": datetime.now(tz=timezone.utc), + "next_run": utcnow(), "arguments": { "args": args if args else [], "kwargs": kwargs if kwargs else {}, }, } task.update(task_extra) return task def create_oneshot_task_dict(type, *args, **kwargs): """Create a oneshot task scheduled for as soon as possible. Args: type (str): Type of oneshot task as per swh-scheduler's db table task_type's column (Ex: load-git, check-deposit) Returns: Expected dictionary for the one-shot task scheduling api (swh.scheduler.backend.create_tasks) """ return create_task_dict(type, "oneshot", *args, **kwargs)