diff --git a/swh/web/common/origin_save.py b/swh/web/common/origin_save.py index 5ce9e882..6d69f4a0 100644 --- a/swh/web/common/origin_save.py +++ b/swh/web/common/origin_save.py @@ -1,896 +1,901 @@ # Copyright (C) 2018-2021 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU Affero General Public License version 3, or any later version # See top-level LICENSE file for more information from bisect import bisect_right from datetime import datetime, timedelta, timezone from itertools import product import json import logging from typing import Any, Dict, List, Optional, Tuple from prometheus_client import Gauge import requests import sentry_sdk from django.core.exceptions import ObjectDoesNotExist, ValidationError from django.core.validators import URLValidator from django.db.models import Q, QuerySet from django.utils.html import escape from swh.scheduler.utils import create_oneshot_task_dict from swh.web import config from swh.web.common import archive from swh.web.common.exc import BadInputExc, ForbiddenExc, NotFoundExc from swh.web.common.models import ( SAVE_REQUEST_ACCEPTED, SAVE_REQUEST_PENDING, SAVE_REQUEST_REJECTED, SAVE_TASK_FAILED, SAVE_TASK_NOT_CREATED, SAVE_TASK_NOT_YET_SCHEDULED, SAVE_TASK_RUNNING, SAVE_TASK_SCHEDULED, SAVE_TASK_SUCCEEDED, + VISIT_STATUS_CREATED, + VISIT_STATUS_ONGOING, SaveAuthorizedOrigin, SaveOriginRequest, SaveUnauthorizedOrigin, ) from swh.web.common.origin_visits import get_origin_visits from swh.web.common.typing import ( OriginExistenceCheckInfo, OriginInfo, SaveOriginRequestInfo, ) from swh.web.common.utils import SWH_WEB_METRICS_REGISTRY, parse_iso8601_date_to_utc scheduler = config.scheduler() logger = logging.getLogger(__name__) # Number of days in the past to lookup for information MAX_THRESHOLD_DAYS = 30 +# Non terminal visit statuses which needs updates +NON_TERMINAL_STATUSES = [ + VISIT_STATUS_CREATED, + VISIT_STATUS_ONGOING, +] + def get_origin_save_authorized_urls() -> List[str]: """ Get the list of origin url prefixes authorized to be immediately loaded into the archive (whitelist). Returns: list: The list of authorized origin url prefix """ return [origin.url for origin in SaveAuthorizedOrigin.objects.all()] def get_origin_save_unauthorized_urls() -> List[str]: """ Get the list of origin url prefixes forbidden to be loaded into the archive (blacklist). Returns: list: the list of unauthorized origin url prefix """ return [origin.url for origin in SaveUnauthorizedOrigin.objects.all()] def can_save_origin(origin_url: str, bypass_pending_review: bool = False) -> str: """ Check if a software origin can be saved into the archive. Based on the origin url, the save request will be either: * immediately accepted if the url is whitelisted * rejected if the url is blacklisted * put in pending state for manual review otherwise Args: origin_url (str): the software origin url to check Returns: str: the origin save request status, either **accepted**, **rejected** or **pending** """ # origin url may be blacklisted for url_prefix in get_origin_save_unauthorized_urls(): if origin_url.startswith(url_prefix): return SAVE_REQUEST_REJECTED # if the origin url is in the white list, it can be immediately saved for url_prefix in get_origin_save_authorized_urls(): if origin_url.startswith(url_prefix): return SAVE_REQUEST_ACCEPTED # otherwise, the origin url needs to be manually verified if the user # that submitted it does not have special permission if bypass_pending_review: # mark the origin URL as trusted in that case SaveAuthorizedOrigin.objects.get_or_create(url=origin_url) return SAVE_REQUEST_ACCEPTED else: return SAVE_REQUEST_PENDING # map visit type to scheduler task # TODO: do not hardcode the task name here (T1157) _visit_type_task = {"git": "load-git", "hg": "load-hg", "svn": "load-svn"} _visit_type_task_privileged = { "archives": "load-archive-files", } # map scheduler task status to origin save status _save_task_status = { "next_run_not_scheduled": SAVE_TASK_NOT_YET_SCHEDULED, "next_run_scheduled": SAVE_TASK_SCHEDULED, "completed": SAVE_TASK_SUCCEEDED, "disabled": SAVE_TASK_FAILED, } # map scheduler task_run status to origin save status _save_task_run_status = { "scheduled": SAVE_TASK_SCHEDULED, "started": SAVE_TASK_RUNNING, "eventful": SAVE_TASK_SUCCEEDED, "uneventful": SAVE_TASK_SUCCEEDED, "failed": SAVE_TASK_FAILED, "permfailed": SAVE_TASK_FAILED, "lost": SAVE_TASK_FAILED, } def get_savable_visit_types_dict(privileged_user: bool = False) -> Dict: """Returned the supported task types the user has access to. Args: privileged_user: Flag to determine if all visit types should be returned or not. Default to False to only list unprivileged visit types. Returns: the dict of supported visit types for the user """ if privileged_user: task_types = {**_visit_type_task, **_visit_type_task_privileged} else: task_types = _visit_type_task return task_types def get_savable_visit_types(privileged_user: bool = False) -> List[str]: """Return the list of visit types the user can perform save requests on. Args: privileged_user: Flag to determine if all visit types should be returned or not. Default to False to only list unprivileged visit types. Returns: the list of saveable visit types """ return sorted(list(get_savable_visit_types_dict(privileged_user).keys())) def _check_visit_type_savable(visit_type: str, privileged_user: bool = False) -> None: visit_type_tasks = get_savable_visit_types(privileged_user) if visit_type not in visit_type_tasks: allowed_visit_types = ", ".join(visit_type_tasks) raise BadInputExc( f"Visit of type {visit_type} can not be saved! " f"Allowed types are the following: {allowed_visit_types}" ) _validate_url = URLValidator(schemes=["http", "https", "svn", "git"]) def _check_origin_url_valid(origin_url: str) -> None: try: _validate_url(origin_url) except ValidationError: raise BadInputExc( "The provided origin url (%s) is not valid!" % escape(origin_url) ) def origin_exists(origin_url: str) -> OriginExistenceCheckInfo: """Check the origin url for existence. If it exists, extract some more useful information on the origin. """ resp = requests.head(origin_url, allow_redirects=True) exists = resp.ok content_length: Optional[int] = None last_modified: Optional[str] = None if exists: # Also process X-Archive-Orig-* headers in case the URL targets the # Internet Archive. size_ = resp.headers.get( "Content-Length", resp.headers.get("X-Archive-Orig-Content-Length") ) content_length = int(size_) if size_ else None try: date_str = resp.headers.get( "Last-Modified", resp.headers.get("X-Archive-Orig-Last-Modified", "") ) date = datetime.strptime(date_str, "%a, %d %b %Y %H:%M:%S %Z") last_modified = date.isoformat() except ValueError: # if not provided or not parsable as per the expected format, keep it None pass return OriginExistenceCheckInfo( origin_url=origin_url, exists=exists, last_modified=last_modified, content_length=content_length, ) def _check_origin_exists(url: str) -> OriginExistenceCheckInfo: """Ensure an URL exists, if not raise an explicit message.""" metadata = origin_exists(url) if not metadata["exists"]: raise BadInputExc(f"The provided url ({escape(url)}) does not exist!") return metadata def _get_visit_info_for_save_request( save_request: SaveOriginRequest, ) -> Tuple[Optional[datetime], Optional[str]]: """Retrieve visit information out of a save request Args: save_request: Input save origin request to retrieve information for. Returns: Tuple of (visit date, optional visit status) for such save request origin """ visit_date = None visit_status = None time_now = datetime.now(tz=timezone.utc) time_delta = time_now - save_request.request_date # stop trying to find a visit date one month after save request submission # as those requests to storage are expensive and associated loading task # surely ended up with errors if time_delta.days <= MAX_THRESHOLD_DAYS: try: origin_info = archive.lookup_origin(OriginInfo(url=save_request.origin_url)) origin_visits = get_origin_visits(origin_info) visit_dates = [parse_iso8601_date_to_utc(v["date"]) for v in origin_visits] i = bisect_right(visit_dates, save_request.request_date) if i != len(visit_dates): visit_date = visit_dates[i] visit_status = origin_visits[i]["status"] - if visit_status not in ("full", "partial", "not_found"): - visit_date = None except Exception as exc: sentry_sdk.capture_exception(exc) return visit_date, visit_status def _check_visit_update_status( - save_request: SaveOriginRequest, save_task_status: str -) -> Tuple[Optional[datetime], str]: - """Given a save request and a save task status, determine whether a save request was - successful or failed. + save_request: SaveOriginRequest, +) -> Tuple[Optional[datetime], Optional[str], Optional[str]]: + """Given a save request, determine whether a save request was successful or failed. Args: save_request: Input save origin request to retrieve information for. Returns: - Tuple of (optional visit date, save task status) for such save request origin + Tuple of (optional visit date, optional visit status, optional save task status) + for such save request origin """ visit_date, visit_status = _get_visit_info_for_save_request(save_request) - save_request.visit_date = visit_date - save_request.visit_status = visit_status + loading_task_status = None if visit_date and visit_status in ("full", "partial"): # visit has been performed, mark the saving task as succeeded - save_task_status = SAVE_TASK_SUCCEEDED + loading_task_status = SAVE_TASK_SUCCEEDED elif visit_status in ("created", "ongoing"): # visit is currently running - save_task_status = SAVE_TASK_RUNNING + loading_task_status = SAVE_TASK_RUNNING elif visit_status in ("not_found", "failed"): - save_task_status = SAVE_TASK_FAILED + loading_task_status = SAVE_TASK_FAILED else: time_now = datetime.now(tz=timezone.utc) time_delta = time_now - save_request.request_date # consider the task as failed if it is still in scheduled state # 30 days after its submission if time_delta.days > MAX_THRESHOLD_DAYS: - save_task_status = SAVE_TASK_FAILED - return visit_date, save_task_status + loading_task_status = SAVE_TASK_FAILED + return visit_date, visit_status, loading_task_status + + +def _compute_task_loading_status( + task: Optional[Dict[str, Any]] = None, task_run: Optional[Dict[str, Any]] = None, +) -> Optional[str]: + loading_task_status: Optional[str] = None + # First determine the loading task status out of task information + if task: + loading_task_status = _save_task_status[task["status"]] + if task_run: + loading_task_status = _save_task_run_status[task_run["status"]] + + return loading_task_status def _update_save_request_info( save_request: SaveOriginRequest, task: Optional[Dict[str, Any]] = None, task_run: Optional[Dict[str, Any]] = None, ) -> SaveOriginRequestInfo: - """Update save request information out of task and task_run information. + """Update save request information out of the visit status and fallback to the task and + task_run information if the visit status is missing. Args: save_request: Save request task: Associated scheduler task information about the save request task_run: Most recent run occurrence of the associated task Returns: Summary of the save request information updated. """ must_save = False - visit_date = save_request.visit_date - # save task still in scheduler db - if task: - save_task_status = _save_task_status[task["status"]] - if task_run: - save_task_status = _save_task_run_status[task_run["status"]] + # To determine the save code now request's final status, the visit date must be set + # and the visit status must be a final one. Once they do, the save code now is + # definitely done. + if ( + not save_request.visit_date + or not save_request.visit_status + or save_request.visit_status in NON_TERMINAL_STATUSES + ): + visit_date, visit_status, loading_task_status = _check_visit_update_status( + save_request + ) - # Consider request from which a visit date has already been found - # as succeeded to avoid retrieving it again - if save_task_status == SAVE_TASK_SCHEDULED and visit_date: - save_task_status = SAVE_TASK_SUCCEEDED - if ( - save_task_status in (SAVE_TASK_FAILED, SAVE_TASK_SUCCEEDED) - and not visit_date - ): - visit_date, visit_status = _get_visit_info_for_save_request(save_request) - save_request.visit_date = visit_date - save_request.visit_status = visit_status - if visit_status in ("failed", "not_found"): - save_task_status = SAVE_TASK_FAILED - must_save = True - # Check tasks still marked as scheduled / not yet scheduled - if save_task_status in (SAVE_TASK_SCHEDULED, SAVE_TASK_NOT_YET_SCHEDULED): - visit_date, save_task_status = _check_visit_update_status( - save_request, save_task_status - ) + if not loading_task_status: # fallback when not provided + loading_task_status = _compute_task_loading_status(task, task_run) - # save task may have been archived - else: - save_task_status = save_request.loading_task_status - if save_task_status in (SAVE_TASK_SCHEDULED, SAVE_TASK_NOT_YET_SCHEDULED): - visit_date, save_task_status = _check_visit_update_status( - save_request, save_task_status - ) + if visit_date != save_request.visit_date: + must_save = True + save_request.visit_date = visit_date - else: - save_task_status = save_request.loading_task_status + if visit_status != save_request.visit_status: + must_save = True + save_request.visit_status = visit_status - if ( - # avoid to override final loading task status when already found - # as visit status is no longer checked once a visit date has been found - save_request.loading_task_status not in (SAVE_TASK_FAILED, SAVE_TASK_SUCCEEDED) - and save_request.loading_task_status != save_task_status - ): - save_request.loading_task_status = save_task_status - must_save = True + if ( + loading_task_status is not None + and loading_task_status != save_request.loading_task_status + ): + must_save = True + save_request.loading_task_status = loading_task_status if must_save: save_request.save() return save_request.to_dict() def create_save_origin_request( visit_type: str, origin_url: str, privileged_user: bool = False, user_id: Optional[int] = None, **kwargs, ) -> SaveOriginRequestInfo: """Create a loading task to save a software origin into the archive. This function aims to create a software origin loading task trough the use of the swh-scheduler component. First, some checks are performed to see if the visit type and origin url are valid but also if the the save request can be accepted. For the 'archives' visit type, this also ensures the artifacts actually exists. If those checks passed, the loading task is then created. Otherwise, the save request is put in pending or rejected state. All the submitted save requests are logged into the swh-web database to keep track of them. Args: visit_type: the type of visit to perform (e.g. git, hg, svn, archives, ...) origin_url: the url of the origin to save privileged: Whether the user has some more privilege than other (bypass review, access to privileged other visit types) user_id: User identifier (provided when authenticated) kwargs: Optional parameters (e.g. artifact_url, artifact_filename, artifact_version) Raises: BadInputExc: the visit type or origin url is invalid or inexistent ForbiddenExc: the provided origin url is blacklisted Returns: dict: A dict describing the save request with the following keys: * **visit_type**: the type of visit to perform * **origin_url**: the url of the origin * **save_request_date**: the date the request was submitted * **save_request_status**: the request status, either **accepted**, **rejected** or **pending** * **save_task_status**: the origin loading task status, either **not created**, **not yet scheduled**, **scheduled**, **succeed** or **failed** """ visit_type_tasks = get_savable_visit_types_dict(privileged_user) _check_visit_type_savable(visit_type, privileged_user) _check_origin_url_valid(origin_url) # if all checks passed so far, we can try and save the origin save_request_status = can_save_origin(origin_url, privileged_user) task = None # if the origin save request is accepted, create a scheduler # task to load it into the archive if save_request_status == SAVE_REQUEST_ACCEPTED: # create a task with high priority task_kwargs: Dict[str, Any] = { "priority": "high", "url": origin_url, } if visit_type == "archives": # extra arguments for that type are required archives_data = kwargs.get("archives_data", []) if not archives_data: raise BadInputExc( "Artifacts data are missing for the archives visit type." ) artifacts = [] for artifact in archives_data: artifact_url = artifact.get("artifact_url") artifact_version = artifact.get("artifact_version") if not artifact_url or not artifact_version: raise BadInputExc("Missing url or version for an artifact to load.") metadata = _check_origin_exists(artifact_url) artifacts.append( { "url": artifact_url, "version": artifact_version, "time": metadata["last_modified"], "length": metadata["content_length"], } ) task_kwargs = dict(**task_kwargs, artifacts=artifacts, snapshot_append=True) sor = None - # get list of previously sumitted save requests + # get list of previously submitted save requests (most recent first) current_sors = list( SaveOriginRequest.objects.filter( visit_type=visit_type, origin_url=origin_url - ) + ).order_by("-request_date") ) can_create_task = False # if no save requests previously submitted, create the scheduler task if not current_sors: can_create_task = True else: # get the latest submitted save request sor = current_sors[0] # if it was in pending state, we need to create the scheduler task # and update the save request info in the database if sor.status == SAVE_REQUEST_PENDING: can_create_task = True # a task has already been created to load the origin elif sor.loading_task_id != -1: # get the scheduler task and its status tasks = scheduler.get_tasks([sor.loading_task_id]) task = tasks[0] if tasks else None task_runs = scheduler.get_task_runs([sor.loading_task_id]) task_run = task_runs[0] if task_runs else None save_request_info = _update_save_request_info(sor, task, task_run) task_status = save_request_info["save_task_status"] # create a new scheduler task only if the previous one has been # already executed if ( task_status == SAVE_TASK_FAILED or task_status == SAVE_TASK_SUCCEEDED ): can_create_task = True sor = None else: can_create_task = False if can_create_task: # effectively create the scheduler task task_dict = create_oneshot_task_dict( visit_type_tasks[visit_type], **task_kwargs ) task = scheduler.create_tasks([task_dict])[0] # pending save request has been accepted if sor: sor.status = SAVE_REQUEST_ACCEPTED sor.loading_task_id = task["id"] sor.save() else: sor = SaveOriginRequest.objects.create( visit_type=visit_type, origin_url=origin_url, status=save_request_status, loading_task_id=task["id"], user_ids=f'"{user_id}"' if user_id else None, ) # save request must be manually reviewed for acceptation elif save_request_status == SAVE_REQUEST_PENDING: # check if there is already such a save request already submitted, # no need to add it to the database in that case try: sor = SaveOriginRequest.objects.get( visit_type=visit_type, origin_url=origin_url, status=save_request_status ) user_ids = sor.user_ids if sor.user_ids is not None else "" if user_id is not None and f'"{user_id}"' not in user_ids: # update user ids list sor.user_ids = f'{sor.user_ids},"{user_id}"' sor.save() # if not add it to the database except ObjectDoesNotExist: sor = SaveOriginRequest.objects.create( visit_type=visit_type, origin_url=origin_url, status=save_request_status, user_ids=f'"{user_id}"' if user_id else None, ) # origin can not be saved as its url is blacklisted, # log the request to the database anyway else: sor = SaveOriginRequest.objects.create( visit_type=visit_type, origin_url=origin_url, status=save_request_status, user_ids=f'"{user_id}"' if user_id else None, ) if save_request_status == SAVE_REQUEST_REJECTED: raise ForbiddenExc( ( 'The "save code now" request has been rejected ' "because the provided origin url is blacklisted." ) ) assert sor is not None return _update_save_request_info(sor, task) def update_save_origin_requests_from_queryset( requests_queryset: QuerySet, ) -> List[SaveOriginRequestInfo]: """Update all save requests from a SaveOriginRequest queryset, update their status in db and return the list of impacted save_requests. Args: requests_queryset: input SaveOriginRequest queryset Returns: list: A list of save origin request info dicts as described in :func:`swh.web.common.origin_save.create_save_origin_request` """ task_ids = [] for sor in requests_queryset: task_ids.append(sor.loading_task_id) save_requests = [] if task_ids: tasks = scheduler.get_tasks(task_ids) tasks = {task["id"]: task for task in tasks} task_runs = scheduler.get_task_runs(tasks) task_runs = {task_run["task"]: task_run for task_run in task_runs} for sor in requests_queryset: sr_dict = _update_save_request_info( sor, tasks.get(sor.loading_task_id), task_runs.get(sor.loading_task_id), ) save_requests.append(sr_dict) return save_requests def refresh_save_origin_request_statuses() -> List[SaveOriginRequestInfo]: """Refresh non-terminal save origin requests (SOR) in the backend. Non-terminal SOR are requests whose status is **accepted** and their task status are either **created**, **not yet scheduled**, **scheduled** or **running**. This shall compute this list of SOR, checks their status in the scheduler and optionally elasticsearch for their current status. Then update those in db. Finally, this returns the refreshed information on those SOR. """ pivot_date = datetime.now(tz=timezone.utc) - timedelta(days=MAX_THRESHOLD_DAYS) save_requests = SaveOriginRequest.objects.filter( # Retrieve accepted request statuses (all statuses) Q(status=SAVE_REQUEST_ACCEPTED), # those without the required information we need to update - Q(visit_date__isnull=True) | Q(visit_status__isnull=True), + Q(visit_date__isnull=True) + | Q(visit_status__isnull=True) + | Q(visit_status__in=NON_TERMINAL_STATUSES), # limit results to recent ones (that is roughly 30 days old at best) Q(request_date__gte=pivot_date), ) + return ( update_save_origin_requests_from_queryset(save_requests) if save_requests.count() > 0 else [] ) def get_save_origin_requests( visit_type: str, origin_url: str ) -> List[SaveOriginRequestInfo]: """ Get all save requests for a given software origin. Args: visit_type: the type of visit origin_url: the url of the origin Raises: BadInputExc: the visit type or origin url is invalid swh.web.common.exc.NotFoundExc: no save requests can be found for the given origin Returns: list: A list of save origin requests dict as described in :func:`swh.web.common.origin_save.create_save_origin_request` """ _check_visit_type_savable(visit_type) _check_origin_url_valid(origin_url) sors = SaveOriginRequest.objects.filter( visit_type=visit_type, origin_url=origin_url ) if sors.count() == 0: raise NotFoundExc( f"No save requests found for visit of type {visit_type} " f"on origin with url {origin_url}." ) return update_save_origin_requests_from_queryset(sors) def get_save_origin_task_info( save_request_id: int, full_info: bool = True ) -> Dict[str, Any]: """ Get detailed information about an accepted save origin request and its associated loading task. If the associated loading task info is archived and removed from the scheduler database, returns an empty dictionary. Args: save_request_id: identifier of a save origin request full_info: whether to return detailed info for staff users Returns: A dictionary with the following keys: - **type**: loading task type - **arguments**: loading task arguments - **id**: loading task database identifier - **backend_id**: loading task celery identifier - **scheduled**: loading task scheduling date - **ended**: loading task termination date - **status**: loading task execution status - **visit_status**: Actual visit status Depending on the availability of the task logs in the elasticsearch cluster of Software Heritage, the returned dictionary may also contain the following keys: - **name**: associated celery task name - **message**: relevant log message from task execution - **duration**: task execution time (only if it succeeded) - **worker**: name of the worker that executed the task """ try: save_request = SaveOriginRequest.objects.get(id=save_request_id) except ObjectDoesNotExist: return {} task = scheduler.get_tasks([save_request.loading_task_id]) task = task[0] if task else None if task is None: return {} task_run = scheduler.get_task_runs([task["id"]]) task_run = task_run[0] if task_run else None if task_run is None: return {} task_run["type"] = task["type"] task_run["arguments"] = task["arguments"] task_run["id"] = task_run["task"] del task_run["task"] del task_run["metadata"] # Enrich the task run with the loading visit status task_run["visit_status"] = save_request.visit_status es_workers_index_url = config.get_config()["es_workers_index_url"] if not es_workers_index_url: return task_run es_workers_index_url += "/_search" if save_request.visit_date: min_ts = save_request.visit_date max_ts = min_ts + timedelta(days=7) else: min_ts = save_request.request_date max_ts = min_ts + timedelta(days=MAX_THRESHOLD_DAYS) min_ts_unix = int(min_ts.timestamp()) * 1000 max_ts_unix = int(max_ts.timestamp()) * 1000 save_task_status = _save_task_status[task["status"]] priority = "3" if save_task_status == SAVE_TASK_FAILED else "6" query = { "bool": { "must": [ {"match_phrase": {"priority": {"query": priority}}}, {"match_phrase": {"swh_task_id": {"query": task_run["backend_id"]}}}, { "range": { "@timestamp": { "gte": min_ts_unix, "lte": max_ts_unix, "format": "epoch_millis", } } }, ] } } try: response = requests.post( es_workers_index_url, json={"query": query, "sort": ["@timestamp"]}, timeout=30, ) results = json.loads(response.text) if results["hits"]["total"]["value"] >= 1: task_run_info = results["hits"]["hits"][-1]["_source"] if "swh_logging_args_runtime" in task_run_info: duration = task_run_info["swh_logging_args_runtime"] task_run["duration"] = duration if "message" in task_run_info: task_run["message"] = task_run_info["message"] if "swh_logging_args_name" in task_run_info: task_run["name"] = task_run_info["swh_logging_args_name"] elif "swh_task_name" in task_run_info: task_run["name"] = task_run_info["swh_task_name"] if "hostname" in task_run_info: task_run["worker"] = task_run_info["hostname"] elif "host" in task_run_info: task_run["worker"] = task_run_info["host"] except Exception as exc: logger.warning("Request to Elasticsearch failed\n%s", exc) sentry_sdk.capture_exception(exc) if not full_info: for field in ("id", "backend_id", "worker"): # remove some staff only fields task_run.pop(field, None) if "message" in task_run and "Loading failure" in task_run["message"]: # hide traceback for non staff users, only display exception message_lines = task_run["message"].split("\n") message = "" for line in message_lines: if line.startswith("Traceback"): break message += f"{line}\n" message += message_lines[-1] task_run["message"] = message return task_run SUBMITTED_SAVE_REQUESTS_METRIC = "swh_web_submitted_save_requests" _submitted_save_requests_gauge = Gauge( name=SUBMITTED_SAVE_REQUESTS_METRIC, documentation="Number of submitted origin save requests", labelnames=["status", "visit_type"], registry=SWH_WEB_METRICS_REGISTRY, ) ACCEPTED_SAVE_REQUESTS_METRIC = "swh_web_accepted_save_requests" _accepted_save_requests_gauge = Gauge( name=ACCEPTED_SAVE_REQUESTS_METRIC, documentation="Number of accepted origin save requests", labelnames=["load_task_status", "visit_type"], registry=SWH_WEB_METRICS_REGISTRY, ) # Metric on the delay of save code now request per status and visit_type. This is the # time difference between the save code now is requested and the time it got ingested. ACCEPTED_SAVE_REQUESTS_DELAY_METRIC = "swh_web_save_requests_delay_seconds" _accepted_save_requests_delay_gauge = Gauge( name=ACCEPTED_SAVE_REQUESTS_DELAY_METRIC, documentation="Save Requests Duration", labelnames=["load_task_status", "visit_type"], registry=SWH_WEB_METRICS_REGISTRY, ) def compute_save_requests_metrics() -> None: """Compute Prometheus metrics related to origin save requests: - Number of submitted origin save requests - Number of accepted origin save requests - Save Code Now requests delay between request time and actual time of ingestion """ request_statuses = ( SAVE_REQUEST_ACCEPTED, SAVE_REQUEST_REJECTED, SAVE_REQUEST_PENDING, ) load_task_statuses = ( SAVE_TASK_NOT_CREATED, SAVE_TASK_NOT_YET_SCHEDULED, SAVE_TASK_SCHEDULED, SAVE_TASK_SUCCEEDED, SAVE_TASK_FAILED, SAVE_TASK_RUNNING, ) # for metrics, we want access to all visit types visit_types = get_savable_visit_types(privileged_user=True) labels_set = product(request_statuses, visit_types) for labels in labels_set: _submitted_save_requests_gauge.labels(*labels).set(0) labels_set = product(load_task_statuses, visit_types) for labels in labels_set: _accepted_save_requests_gauge.labels(*labels).set(0) duration_load_task_statuses = ( SAVE_TASK_FAILED, SAVE_TASK_SUCCEEDED, ) for labels in product(duration_load_task_statuses, visit_types): _accepted_save_requests_delay_gauge.labels(*labels).set(0) for sor in SaveOriginRequest.objects.all(): if sor.status == SAVE_REQUEST_ACCEPTED: _accepted_save_requests_gauge.labels( load_task_status=sor.loading_task_status, visit_type=sor.visit_type, ).inc() _submitted_save_requests_gauge.labels( status=sor.status, visit_type=sor.visit_type ).inc() if ( sor.loading_task_status in (SAVE_TASK_SUCCEEDED, SAVE_TASK_FAILED) and sor.visit_date is not None and sor.request_date is not None ): delay = sor.visit_date.timestamp() - sor.request_date.timestamp() _accepted_save_requests_delay_gauge.labels( load_task_status=sor.loading_task_status, visit_type=sor.visit_type, ).inc(delay) diff --git a/swh/web/tests/common/test_origin_save.py b/swh/web/tests/common/test_origin_save.py index ed211535..ea4f27da 100644 --- a/swh/web/tests/common/test_origin_save.py +++ b/swh/web/tests/common/test_origin_save.py @@ -1,709 +1,753 @@ # Copyright (C) 2019-2021 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU Affero General Public License version 3, or any later version # See top-level LICENSE file for more information from datetime import datetime, timedelta, timezone from functools import partial import re from typing import Optional import iso8601 import pytest import requests from swh.core.pytest_plugin import get_response_cb from swh.web.common.exc import BadInputExc from swh.web.common.models import ( SAVE_REQUEST_ACCEPTED, SAVE_TASK_FAILED, SAVE_TASK_RUNNING, SAVE_TASK_SCHEDULED, SAVE_TASK_SUCCEEDED, + VISIT_STATUS_CREATED, VISIT_STATUS_FULL, SaveOriginRequest, ) from swh.web.common.origin_save import ( _check_origin_exists, _check_visit_type_savable, _visit_type_task, _visit_type_task_privileged, get_savable_visit_types, get_save_origin_requests, get_save_origin_task_info, origin_exists, refresh_save_origin_request_statuses, ) from swh.web.common.typing import ( OriginExistenceCheckInfo, OriginVisitInfo, SaveOriginRequestInfo, ) from swh.web.config import get_config _es_url = "http://esnode1.internal.softwareheritage.org:9200" _es_workers_index_url = "%s/swh_workers-*" % _es_url _origin_url = "https://gitlab.com/inkscape/inkscape" _visit_type = "git" _task_id = 203525448 @pytest.fixture(autouse=True) def requests_mock_datadir(datadir, requests_mock_datadir): """Override default behavior to deal with post method""" cb = partial(get_response_cb, datadir=datadir) requests_mock_datadir.post(re.compile("https?://"), body=cb) return requests_mock_datadir @pytest.mark.django_db def test_get_save_origin_archived_task_info(mocker): _get_save_origin_task_info_test(mocker, task_archived=True) @pytest.mark.django_db def test_get_save_origin_task_full_info_with_es(mocker): _get_save_origin_task_info_test(mocker, es_available=True) @pytest.mark.django_db def test_get_save_origin_task_info_with_es(mocker): _get_save_origin_task_info_test(mocker, es_available=True, full_info=False) @pytest.mark.django_db def test_get_save_origin_task_info_without_es(mocker): _get_save_origin_task_info_test(mocker, es_available=False) def _mock_scheduler( mocker, task_status="completed", task_run_status="eventful", task_archived=False, visit_started_date=None, ): mock_scheduler = mocker.patch("swh.web.common.origin_save.scheduler") task = { "arguments": {"args": [], "kwargs": {"repo_url": _origin_url},}, "current_interval": timedelta(days=64), "id": _task_id, "next_run": datetime.now(tz=timezone.utc) + timedelta(days=64), "policy": "oneshot", "priority": "high", "retries_left": 0, "status": task_status, "type": "load-git", } mock_scheduler.get_tasks.return_value = [dict(task) if not task_archived else None] task_run = { "backend_id": "f00c712c-e820-41ce-a07c-9bf8df914205", "ended": datetime.now(tz=timezone.utc) + timedelta(minutes=5), "id": 654270631, "metadata": {}, "scheduled": datetime.now(tz=timezone.utc), "started": visit_started_date, "status": task_run_status, "task": _task_id, } mock_scheduler.get_task_runs.return_value = [ dict(task_run) if not task_archived else None ] return task, task_run @pytest.mark.parametrize( "wrong_type,privileged_user", [ ("dummy", True), ("dumb", False), ("archives", False), # when no privilege, this is rejected ], ) def test__check_visit_type_savable(wrong_type, privileged_user): with pytest.raises(BadInputExc, match="Allowed types"): _check_visit_type_savable(wrong_type, privileged_user) # when privileged_user, the following is accepted though _check_visit_type_savable("archives", True) def test_get_savable_visit_types(): default_list = list(_visit_type_task.keys()) assert set(get_savable_visit_types()) == set(default_list) privileged_list = default_list.copy() privileged_list += list(_visit_type_task_privileged.keys()) assert set(get_savable_visit_types(privileged_user=True)) == set(privileged_list) def _get_save_origin_task_info_test( mocker, task_archived=False, es_available=True, full_info=True ): swh_web_config = get_config() if es_available: swh_web_config.update({"es_workers_index_url": _es_workers_index_url}) else: swh_web_config.update({"es_workers_index_url": ""}) sor = SaveOriginRequest.objects.create( request_date=datetime.now(tz=timezone.utc), visit_type=_visit_type, origin_url="https://gitlab.com/inkscape/inkscape", status=SAVE_REQUEST_ACCEPTED, visit_date=datetime.now(tz=timezone.utc) + timedelta(hours=1), loading_task_id=_task_id, ) task, task_run = _mock_scheduler(mocker, task_archived=task_archived) es_response = requests.post("%s/_search" % _es_workers_index_url).json() task_exec_data = es_response["hits"]["hits"][-1]["_source"] sor_task_info = get_save_origin_task_info(sor.id, full_info=full_info) expected_result = ( { "type": task["type"], "arguments": task["arguments"], "id": task["id"], "backend_id": task_run["backend_id"], "scheduled": task_run["scheduled"], "started": task_run["started"], "ended": task_run["ended"], "status": task_run["status"], "visit_status": sor.visit_status, } if not task_archived else {} ) if es_available and not task_archived: expected_result.update( { "message": task_exec_data["message"], "name": task_exec_data["swh_task_name"], "worker": task_exec_data["hostname"], } ) if not full_info: expected_result.pop("id", None) expected_result.pop("backend_id", None) expected_result.pop("worker", None) if "message" in expected_result: message = "" message_lines = expected_result["message"].split("\n") for line in message_lines: if line.startswith("Traceback"): break message += f"{line}\n" message += message_lines[-1] expected_result["message"] = message assert sor_task_info == expected_result @pytest.mark.django_db def test_get_save_origin_requests_find_visit_date(mocker): # create a save request SaveOriginRequest.objects.create( request_date=datetime.now(tz=timezone.utc), visit_type=_visit_type, origin_url=_origin_url, status=SAVE_REQUEST_ACCEPTED, visit_date=None, loading_task_id=_task_id, ) # mock scheduler and archive _mock_scheduler(mocker) mock_archive = mocker.patch("swh.web.common.origin_save.archive") mock_archive.lookup_origin.return_value = {"url": _origin_url} mock_get_origin_visits = mocker.patch( "swh.web.common.origin_save.get_origin_visits" ) # create a visit for the save request visit_date = datetime.now(tz=timezone.utc).isoformat() visit_info = OriginVisitInfo( date=visit_date, formatted_date="", metadata={}, origin=_origin_url, snapshot="", status="full", type=_visit_type, url="", visit=34, ) mock_get_origin_visits.return_value = [visit_info] # check visit date has been correctly found sors = get_save_origin_requests(_visit_type, _origin_url) assert len(sors) == 1 assert sors[0]["save_task_status"] == SAVE_TASK_SUCCEEDED assert sors[0]["visit_date"] == visit_date mock_get_origin_visits.assert_called_once() # check visit is not searched again when it has been found get_save_origin_requests(_visit_type, _origin_url) mock_get_origin_visits.assert_called_once() # check visit date are not searched for save requests older than # one month sor = SaveOriginRequest.objects.create( visit_type=_visit_type, origin_url=_origin_url, status=SAVE_REQUEST_ACCEPTED, loading_task_id=_task_id, visit_date=None, ) sor.request_date = datetime.now(tz=timezone.utc) - timedelta(days=31) sor.save() _mock_scheduler(mocker, task_status="disabled", task_run_status="failed") sors = get_save_origin_requests(_visit_type, _origin_url) assert len(sors) == 2 assert sors[0]["save_task_status"] == SAVE_TASK_FAILED assert sors[0]["visit_date"] is None mock_get_origin_visits.assert_called_once() def _get_save_origin_requests( mocker, load_status, visit_status, request_date: Optional[datetime] = None ): """Wrapper around the get_origin_save_origin_request call. """ SaveOriginRequest.objects.create( request_date=datetime.now(tz=timezone.utc), visit_type=_visit_type, visit_status=visit_status, origin_url=_origin_url, status=SAVE_REQUEST_ACCEPTED, visit_date=None, loading_task_id=_task_id, ) # mock scheduler and archives _mock_scheduler( mocker, task_status="next_run_scheduled", task_run_status=load_status ) mock_archive = mocker.patch("swh.web.common.origin_save.archive") mock_archive.lookup_origin.return_value = {"url": _origin_url} mock_get_origin_visits = mocker.patch( "swh.web.common.origin_save.get_origin_visits" ) # create a visit for the save request with status created visit_date = datetime.now(tz=timezone.utc).isoformat() visit_info = OriginVisitInfo( date=visit_date, formatted_date="", metadata={}, origin=_origin_url, snapshot="", # make mypy happy status=visit_status, type=_visit_type, url="", visit=34, ) mock_get_origin_visits.return_value = [visit_info] sors = get_save_origin_requests(_visit_type, _origin_url) mock_get_origin_visits.assert_called_once() return sors @pytest.mark.parametrize("visit_date", [None, "some-date"]) def test_from_save_origin_request_to_save_request_info_dict(visit_date): """Ensure save request to json serializable dict is fine """ request_date = datetime.now(tz=timezone.utc) _visit_date = request_date + timedelta(minutes=5) if visit_date else None request_date = datetime.now(tz=timezone.utc) sor = SaveOriginRequest( request_date=request_date, visit_type=_visit_type, visit_status=VISIT_STATUS_FULL, origin_url=_origin_url, status=SAVE_REQUEST_ACCEPTED, loading_task_status=None, visit_date=_visit_date, loading_task_id=1, ) assert sor.to_dict() == SaveOriginRequestInfo( id=sor.id, origin_url=sor.origin_url, visit_type=sor.visit_type, save_request_date=sor.request_date.isoformat(), save_request_status=sor.status, save_task_status=sor.loading_task_status, visit_status=sor.visit_status, visit_date=_visit_date.isoformat() if _visit_date else None, loading_task_id=sor.loading_task_id, ) def test__check_origin_exists_404(requests_mock): url_ko = "https://example.org/some-inexistant-url" requests_mock.head(url_ko, status_code=404) with pytest.raises(BadInputExc, match="not exist"): _check_origin_exists(url_ko) def test__check_origin_exists_200(requests_mock): url = "https://example.org/url" requests_mock.head(url, status_code=200) # passes the check actual_metadata = _check_origin_exists(url) # and we actually may have retrieved some metadata on the origin assert actual_metadata == origin_exists(url) def test_origin_exists_404(requests_mock): """Origin which does not exist should be reported as inexistent""" url_ko = "https://example.org/some-inexistant-url" requests_mock.head(url_ko, status_code=404) actual_result = origin_exists(url_ko) assert actual_result == OriginExistenceCheckInfo( origin_url=url_ko, exists=False, last_modified=None, content_length=None, ) def test_origin_exists_200_no_data(requests_mock): """Existing origin should be reported as such (no extra information)""" url = "http://example.org/real-url" requests_mock.head( url, status_code=200, ) actual_result = origin_exists(url) assert actual_result == OriginExistenceCheckInfo( origin_url=url, exists=True, last_modified=None, content_length=None, ) def test_origin_exists_200_with_data(requests_mock): """Existing origin should be reported as such (+ extra information)""" url = "http://example.org/real-url" requests_mock.head( url, status_code=200, headers={ "content-length": "10", "last-modified": "Sun, 21 Aug 2011 16:26:32 GMT", }, ) actual_result = origin_exists(url) assert actual_result == OriginExistenceCheckInfo( origin_url=url, exists=True, content_length=10, last_modified="2011-08-21T16:26:32", ) def test_origin_exists_internet_archive(requests_mock): """Edge case where an artifact URL to check existence is hosted on the Internet Archive""" url = ( "https://web.archive.org/web/20100705043309/" "http://www.cs.unm.edu/~mccune/old-ftp/eqp-09e.tar.gz" ) redirect_url = ( "https://web.archive.org/web/20100610004108/" "http://www.cs.unm.edu/~mccune/old-ftp/eqp-09e.tar.gz" ) requests_mock.head( url, status_code=302, headers={"Location": redirect_url,}, ) requests_mock.head( redirect_url, status_code=200, headers={ "X-Archive-Orig-Last-Modified": "Tue, 12 May 2009 22:09:43 GMT", "X-Archive-Orig-Content-Length": "121421", }, ) actual_result = origin_exists(url) assert actual_result == OriginExistenceCheckInfo( origin_url=url, exists=True, content_length=121421, last_modified="2009-05-12T22:09:43", ) def test_origin_exists_200_with_data_unexpected_date_format(requests_mock): """Existing origin should be ok, unexpected last modif time result in no time""" url = "http://example.org/real-url2" # this is parsable but not as expected unexpected_format_date = "Sun, 21 Aug 2021 16:26:32" requests_mock.head( url, status_code=200, headers={"last-modified": unexpected_format_date,}, ) actual_result = origin_exists(url) # so the resulting date is None assert actual_result == OriginExistenceCheckInfo( origin_url=url, exists=True, content_length=None, last_modified=None, ) @pytest.mark.django_db @pytest.mark.parametrize("visit_status", ["created", "ongoing",]) def test_get_save_origin_requests_no_visit_date_found(mocker, visit_status): """Uneventful visits with failed visit status are marked as failed """ sors = _get_save_origin_requests( mocker, load_status="scheduled", visit_status=visit_status, ) # check no visit date has been found assert len(sors) == 1 assert sors[0]["save_task_status"] == SAVE_TASK_RUNNING - assert sors[0]["visit_date"] is None + assert sors[0]["visit_date"] is not None assert sors[0]["visit_status"] == visit_status @pytest.mark.django_db @pytest.mark.parametrize("visit_status", ["not_found", "failed",]) def test_get_save_origin_requests_no_failed_status_override(mocker, visit_status): """Uneventful visits with failed statuses (failed, not found) are marked as failed """ sors = _get_save_origin_requests( mocker, load_status="uneventful", visit_status=visit_status ) assert len(sors) == 1 assert sors[0]["save_task_status"] == SAVE_TASK_FAILED visit_date = sors[0]["visit_date"] - if visit_status == "failed": - assert visit_date is None - else: - assert visit_date is not None + assert visit_date is not None sors = get_save_origin_requests(_visit_type, _origin_url) assert len(sors) == 1 assert sors[0]["save_task_status"] == SAVE_TASK_FAILED assert sors[0]["visit_status"] == visit_status @pytest.mark.django_db @pytest.mark.parametrize( "load_status,visit_status", [("eventful", "full"), ("eventful", "partial"), ("uneventful", "partial"),], ) def test_get_visit_info_for_save_request_succeeded(mocker, load_status, visit_status): """Nominal scenario, below 30 days, returns something""" sors = _get_save_origin_requests( mocker, load_status=load_status, visit_status=visit_status ) assert len(sors) == 1 assert sors[0]["save_task_status"] == SAVE_TASK_SUCCEEDED assert sors[0]["visit_date"] is not None assert sors[0]["visit_status"] == visit_status sors = get_save_origin_requests(_visit_type, _origin_url) assert sors[0]["save_task_status"] == SAVE_TASK_SUCCEEDED assert sors[0]["visit_status"] == visit_status @pytest.mark.django_db @pytest.mark.parametrize("load_status", ["eventful", "uneventful",]) def test_get_visit_info_incomplete_visit_still_successful(mocker, load_status): """Incomplete visit information, yet the task is updated partially """ sors = _get_save_origin_requests( mocker, load_status=load_status, visit_status=None, ) assert len(sors) == 1 assert sors[0]["save_task_status"] == SAVE_TASK_SUCCEEDED # As the entry is missing the following information though - assert sors[0]["visit_date"] is None + assert sors[0]["visit_date"] is not None assert sors[0]["visit_status"] is None # It's still detected as to be updated by the refresh routine sors = refresh_save_origin_request_statuses() assert len(sors) == 1 assert sors[0]["save_task_status"] == SAVE_TASK_SUCCEEDED - assert sors[0]["visit_date"] is None + assert sors[0]["visit_date"] is not None assert sors[0]["visit_status"] is None @pytest.mark.django_db def test_refresh_in_progress_save_request_statuses(mocker, api_client, archive_data): """Refresh a pending save origins requests and update if the status changes """ date_now = datetime.now(tz=timezone.utc) date_pivot = date_now - timedelta(days=30) visit_started_date = date_now - timedelta(minutes=1) # returned visit status - sors = _get_save_origin_requests( - mocker, load_status=SAVE_TASK_SCHEDULED, visit_status="created", + SaveOriginRequest.objects.create( + request_date=datetime.now(tz=timezone.utc), + visit_type=_visit_type, + visit_status=VISIT_STATUS_CREATED, + origin_url=_origin_url, + status=SAVE_REQUEST_ACCEPTED, + visit_date=None, + loading_task_id=_task_id, ) - assert len(sors) == 1 + + # mock scheduler and archives + _mock_scheduler( + mocker, task_status="next_run_scheduled", task_run_status=SAVE_TASK_SCHEDULED + ) + mock_archive = mocker.patch("swh.web.common.origin_save.archive") + mock_archive.lookup_origin.return_value = {"url": _origin_url} + mock_get_origin_visits = mocker.patch( + "swh.web.common.origin_save.get_origin_visits" + ) + # create a visit for the save request with status created + visit_date = datetime.now(tz=timezone.utc).isoformat() + visit_info = OriginVisitInfo( + date=visit_date, + formatted_date="", + metadata={}, + origin=_origin_url, + snapshot="", # make mypy happy + status=VISIT_STATUS_CREATED, + type=_visit_type, + url="", + visit=34, + ) + mock_get_origin_visits.return_value = [visit_info] # make the scheduler return a running event _mock_scheduler( mocker, task_status="next_run_scheduled", task_run_status="started", visit_started_date=visit_started_date, ) # The visit is detected but still running sors = refresh_save_origin_request_statuses() + + assert mock_get_origin_visits.called and mock_get_origin_visits.call_count == 1 assert len(sors) == 1 for sor in sors: assert iso8601.parse_date(sor["save_request_date"]) >= date_pivot # The status is updated assert sor["save_task_status"] == SAVE_TASK_RUNNING # but the following entries are missing so it's not updated - assert sor["visit_date"] is None - assert sor["visit_status"] == "created" + assert sor["visit_date"] is not None + assert sor["visit_status"] == VISIT_STATUS_CREATED # make the visit status completed # make the scheduler return a running event _mock_scheduler( mocker, task_status="completed", task_run_status="eventful", visit_started_date=visit_started_date, ) - # This time around, the origin returned will have all information updated - mock_get_origin_visits = mocker.patch( - "swh.web.common.origin_save.get_origin_visits" - ) - # create a visit for the save request with status created + # This time around, the origin returned will have all required information updated + # (visit date and visit status in final state) visit_date = datetime.now(tz=timezone.utc).isoformat() - visit_info = OriginVisitInfo( - date=visit_date, - formatted_date="", - metadata={}, - origin=_origin_url, - snapshot="", # make mypy happy - status="full", - type=_visit_type, - url="", - visit=34, - ) + visit_info.update({"date": visit_date, "status": VISIT_STATUS_FULL}) mock_get_origin_visits.return_value = [visit_info] # Detected entry, this time it should be updated sors = refresh_save_origin_request_statuses() assert len(sors) == 1 + assert mock_get_origin_visits.called and mock_get_origin_visits.call_count == 1 + 1 + for sor in sors: assert iso8601.parse_date(sor["save_request_date"]) >= date_pivot # as it turns out, in this test, this won't update anything as no new status got # returned by the scheduler assert sor["save_task_status"] == SAVE_TASK_SUCCEEDED assert sor["visit_date"] == visit_date - assert sor["visit_status"] == "full" + assert sor["visit_status"] == VISIT_STATUS_FULL # Once in final state, a sor should not be updated anymore sors = refresh_save_origin_request_statuses() assert len(sors) == 0 @pytest.mark.django_db def test_refresh_save_request_statuses(mocker, api_client, archive_data): """Refresh filters save origins requests and update if changes """ date_now = datetime.now(tz=timezone.utc) date_pivot = date_now - timedelta(days=30) # returned visit status - sors = _get_save_origin_requests( - mocker, load_status=SAVE_TASK_SCHEDULED, visit_status=None, + SaveOriginRequest.objects.create( + request_date=datetime.now(tz=timezone.utc), + visit_type=_visit_type, + visit_status=None, + origin_url=_origin_url, + status=SAVE_REQUEST_ACCEPTED, + visit_date=None, + loading_task_id=_task_id, ) - assert len(sors) == 1 + + # mock scheduler and archives + _mock_scheduler( + mocker, task_status="next_run_scheduled", task_run_status=SAVE_TASK_SCHEDULED + ) + mock_archive = mocker.patch("swh.web.common.origin_save.archive") + mock_archive.lookup_origin.return_value = {"url": _origin_url} + mock_get_origin_visits = mocker.patch( + "swh.web.common.origin_save.get_origin_visits" + ) + # create a visit for the save request with status created + visit_date = datetime.now(tz=timezone.utc).isoformat() + visit_info = OriginVisitInfo( + date=visit_date, + formatted_date="", + metadata={}, + origin=_origin_url, + snapshot="", # make mypy happy + status=VISIT_STATUS_CREATED, + type=_visit_type, + url="", + visit=34, + ) + mock_get_origin_visits.return_value = [visit_info] # no changes so refresh does detect the entry but does nothing sors = refresh_save_origin_request_statuses() assert len(sors) == 1 for sor in sors: assert iso8601.parse_date(sor["save_request_date"]) >= date_pivot # as it turns out, in this test, this won't update anything as no new status got # returned by the scheduler - assert sor["save_task_status"] == SAVE_TASK_SCHEDULED + assert sor["save_task_status"] == SAVE_TASK_RUNNING # Information is empty - assert sor["visit_date"] is None - assert sor["visit_status"] is None + assert sor["visit_date"] == visit_date + assert sor["visit_status"] == VISIT_STATUS_CREATED - # make the scheduler return eventful event for that origin - _mock_scheduler(mocker) - # updates will be detected, entry should be updated but we are still missing info + # A save code now entry is detected for update, but as nothing changes, the entry + # remains in the same state sors = refresh_save_origin_request_statuses() assert len(sors) == 1 for sor in sors: assert iso8601.parse_date(sor["save_request_date"]) >= date_pivot - # The status is updated - assert sor["save_task_status"] == SAVE_TASK_SUCCEEDED - # but the following entries are missing so it's not updated - assert sor["visit_date"] is None - assert sor["visit_status"] is None + # Status is not updated as no new information is available on the visit status + # and the task status has not moved + assert sor["save_task_status"] == SAVE_TASK_RUNNING + # Information is empty + assert sor["visit_date"] == visit_date + assert sor["visit_status"] == VISIT_STATUS_CREATED # This time around, the origin returned will have all information updated - mock_get_origin_visits = mocker.patch( - "swh.web.common.origin_save.get_origin_visits" - ) # create a visit for the save request with status created visit_date = datetime.now(tz=timezone.utc).isoformat() visit_info = OriginVisitInfo( date=visit_date, formatted_date="", metadata={}, origin=_origin_url, snapshot="", # make mypy happy - status="full", + status=VISIT_STATUS_FULL, type=_visit_type, url="", visit=34, ) mock_get_origin_visits.return_value = [visit_info] # Detected entry, this time it should be updated sors = refresh_save_origin_request_statuses() assert len(sors) == 1 for sor in sors: assert iso8601.parse_date(sor["save_request_date"]) >= date_pivot # as it turns out, in this test, this won't update anything as no new status got # returned by the scheduler assert sor["save_task_status"] == SAVE_TASK_SUCCEEDED assert sor["visit_date"] == visit_date assert sor["visit_status"] == "full" # This time, nothing left to update sors = refresh_save_origin_request_statuses() assert len(sors) == 0