rule.py

# -*- coding: utf-8 -*-
# Copyright European Organization for Nuclear Research (CERN) since 2012
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#    http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import json
import logging
from collections.abc import Iterator, Sequence
from configparser import NoOptionError, NoSectionError
from copy import deepcopy
from datetime import datetime, timedelta
from os import path
from re import match
from string import Template
from typing import TYPE_CHECKING, Any, Callable, Literal, Optional, Type, TypeVar, Union

from dogpile.cache.api import NoValue
from sqlalchemy import delete, desc, select, update
from sqlalchemy.exc import IntegrityError, StatementError
from sqlalchemy.exc import NoResultFound  # https://pydoc.dev/sqlalchemy/latest/sqlalchemy.exc.NoResultFound.html
from sqlalchemy.sql import func
from sqlalchemy.sql.expression import and_, or_, true, null, tuple_, false

import rucio.core.did
import rucio.core.lock  # import get_replica_locks, get_files_and_replica_locks_of_dataset
import rucio.core.replica  # import get_and_lock_file_replicas, get_and_lock_file_replicas_for_dataset
from rucio.common.cache import make_region_memcached
from rucio.common.config import config_get
from rucio.common.exception import (InvalidRSEExpression, InvalidReplicationRule, InsufficientAccountLimit,
                                    DataIdentifierNotFound, RuleNotFound, InputValidationError, RSEOverQuota,
                                    ReplicationRuleCreationTemporaryFailed, InsufficientTargetRSEs, RucioException,
                                    InvalidRuleWeight, StagingAreaRuleRequiresLifetime, DuplicateRule,
                                    InvalidObject, RSEWriteBlocked, RuleReplaceFailed, RequestNotFound,
                                    ManualRuleApprovalBlocked, UnsupportedOperation, UndefinedPolicy, InvalidValueForKey,
                                    InvalidSourceReplicaExpression)
from rucio.common.plugins import PolicyPackageAlgorithms
from rucio.common.policy import policy_filter, get_scratchdisk_lifetime
from rucio.common.schema import validate_schema
from rucio.common.types import InternalScope, InternalAccount, LoggerFunction, DIDDict, RuleDict
from rucio.common.utils import str_to_date, sizefmt, chunks
from rucio.core import account_counter, rse_counter, request as request_core, transfer as transfer_core
from rucio.core.account import get_account
from rucio.core.account import has_account_attribute
from rucio.core.lifetime_exception import define_eol
from rucio.core.message import add_message
from rucio.core.monitor import MetricManager
from rucio.core.rse import get_rse_name, list_rse_attributes, get_rse, get_rse_usage
from rucio.core.rse_expression_parser import parse_expression
from rucio.core.rse_selector import RSESelector
from rucio.core.rule_grouping import apply_rule_grouping, repair_stuck_locks_and_apply_rule_grouping, create_transfer_dict, apply_rule
from rucio.db.sqla import models, filter_thread_work
from rucio.db.sqla.constants import (LockState, ReplicaState, RuleState, RuleGrouping,
                                     DIDAvailability, DIDReEvaluation, DIDType, BadFilesStatus,
                                     RequestType, RuleNotification, OBSOLETE, RSEType)
from rucio.db.sqla.session import read_session, transactional_session, stream_session

if TYPE_CHECKING:
    from sqlalchemy.orm import Session


REGION = make_region_memcached(expiration_time=900)
METRICS = MetricManager(module=__name__)
AutoApproveT = TypeVar('AutoApproveT', bound='AutoApprove')


class AutoApprove(PolicyPackageAlgorithms):
    """
    Handle automatic approval algorithms for replication rules
    """

    _algorithm_type = 'auto_approve'

    def __init__(self, rule: models.ReplicationRule, did: models.DataIdentifier, session: 'Session') -> None:
        super().__init__()
        self.rule = rule
        self.did = did
        self.session = session
        self.register("default", self.default)

    def evaluate(self) -> bool:
        """
        Evaluate the auto-approve algorithm
        """
        return self.get_configured_algorithm()(self.rule, self.did, self.session)

    @classmethod
    def get_configured_algorithm(cls: Type[AutoApproveT]) -> Callable[[models.ReplicationRule, models.DataIdentifier, 'Session'], bool]:
        """
        Get the configured auto-approve algorithm
        """
        try:
            configured_algorithm: str = str(config_get('rules', cls._algorithm_type, default='default'))
        except (NoOptionError, NoSectionError, RuntimeError):
            configured_algorithm = 'default'

        return super()._get_one_algorithm(cls._algorithm_type, configured_algorithm)

    @classmethod
    def register(cls: Type[AutoApproveT], name: str, fn_auto_approve: Callable[[models.ReplicationRule, models.DataIdentifier, 'Session'], bool]) -> None:
        """
        Register a new auto-approve algorithm
        """
        algorithm_dict = {name: fn_auto_approve}
        super()._register(cls._algorithm_type, algorithm_dict)

    @staticmethod
    def default(rule: models.ReplicationRule, did: models.DataIdentifier, session: 'Session') -> bool:
        """
        Default auto-approve algorithm
        """
        rse_expression = rule['rse_expression']
        vo = rule['account'].vo

        rses = parse_expression(rse_expression, filter_={'vo': vo}, session=session)

        auto_approve = False
        # Block manual approval for multi-rse rules
        if len(rses) > 1:
            raise InvalidReplicationRule('Ask approval is not allowed for rules with multiple RSEs')
        if len(rses) == 1 and not did.is_open and did.bytes is not None and did.length is not None:
            # This rule can be considered for auto-approval:
            rse_attr = list_rse_attributes(rse_id=rses[0]['id'], session=session)
            auto_approve = False
            if 'auto_approve_bytes' in rse_attr and 'auto_approve_files' in rse_attr:
                if did.bytes < int(rse_attr.get('auto_approve_bytes')) and did.length < int(rse_attr.get('auto_approve_files')):
                    auto_approve = True
            elif did.bytes < int(rse_attr.get('auto_approve_bytes', -1)):
                auto_approve = True
            elif did.length < int(rse_attr.get('auto_approve_files', -1)):
                auto_approve = True

        return auto_approve


@transactional_session
def add_rule(
    dids: Sequence[DIDDict],
    account: InternalAccount,
    copies: int,
    rse_expression: str,
    grouping: Literal['ALL', 'DATASET', 'NONE'],
    weight: Optional[str],
    lifetime: Optional[int],
    locked: bool,
    subscription_id: Optional[str],
    source_replica_expression: Optional[str] = None,
    activity: str = 'User Subscriptions',
    notify: Optional[Literal['Y', 'N', 'C', 'P']] = None,
    purge_replicas: bool = False,
    ignore_availability: bool = False,
    comment: Optional[str] = None,
    ask_approval: bool = False,
    asynchronous: bool = False,
    ignore_account_limit: bool = False,
    priority: int = 3,
    delay_injection: Optional[int] = None,
    split_container: bool = False,
    meta: Optional[dict[str, Any]] = None,
    *,
    session: "Session",
    logger: LoggerFunction = logging.log
) -> list[str]:
    """
    Adds a replication rule for every did in dids

    :param dids:                       List of data identifiers.
    :param account:                    Account issuing the rule.
    :param copies:                     The number of replicas.
    :param rse_expression:             RSE expression which gets resolved into a list of rses.
    :param grouping:                   ALL -  All files will be replicated to the same RSE.
                                       DATASET - All files in the same dataset will be replicated to the same RSE.
                                       NONE - Files will be completely spread over all allowed RSEs without any grouping considerations at all.
    :param weight:                     Weighting scheme to be used.
    :param lifetime:                   The lifetime of the replication rule in seconds.
    :param locked:                     If the rule is locked.
    :param subscription_id:            The subscription_id, if the rule is created by a subscription.
    :param source_replica_expression:  Only use replicas as source from this RSEs.
    :param activity:                   Activity to be passed on to the conveyor.
    :param notify:                     Notification setting of the rule ('Y', 'N', 'C'; None = 'N').
    :param purge_replicas:             Purge setting if a replica should be directly deleted after the rule is deleted.
    :param ignore_availability:        Option to ignore the availability of RSEs.
    :param comment:                    Comment about the rule.
    :param ask_approval:               Ask for approval for this rule.
    :param asynchronous:               Create replication rule asynchronously by the judge-injector.
    :param delay_injection:            Create replication after 'delay' seconds. Implies asynchronous=True.
    :param ignore_account_limit:       Ignore quota and create the rule outside of the account limits.
    :param priority:                   Priority of the rule and the transfers which should be submitted.
    :param split_container:            Should a container rule be split into individual dataset rules.
    :param meta:                       Dictionary with metadata from the WFMS.
    :param session:                    The database session in use.
    :param logger:                     Optional decorated logger that can be passed from the calling daemons or servers.
    :returns:                          A list of created replication rule ids.
    :raises:                           InvalidReplicationRule, InsufficientAccountLimit, InvalidRSEExpression, DataIdentifierNotFound, ReplicationRuleCreationTemporaryFailed, InvalidRuleWeight,
                                       StagingAreaRuleRequiresLifetime, DuplicateRule, RSEWriteBlocked, ScratchDiskLifetimeConflict, ManualRuleApprovalBlocked, RSEOverQuota
    """
    if copies <= 0:
        raise InvalidValueForKey("The number of copies for a replication rule should be greater than 0.")

    rule_ids = []

    grouping_value = {'ALL': RuleGrouping.ALL, 'NONE': RuleGrouping.NONE}.get(grouping, RuleGrouping.DATASET)

    with METRICS.timer('add_rule.total'):
        # 1. Resolve the rse_expression into a list of RSE-ids
        with METRICS.timer('add_rule.parse_rse_expression'):
            vo = account.vo
            if ignore_availability:
                rses = parse_expression(rse_expression, filter_={'vo': vo}, session=session)
            else:
                rses = parse_expression(rse_expression, filter_={'vo': vo, 'availability_write': True}, session=session)

            if lifetime is None:  # Check if one of the rses is a staging area
                if [rse for rse in rses if rse.get('staging_area', False)]:
                    raise StagingAreaRuleRequiresLifetime('Rules for a staging area must include a lifetime')

            # Check SCRATCHDISK Policy
            try:
                lifetime = get_scratch_policy(account, rses, lifetime, session=session)
            except UndefinedPolicy:
                pass

            # Auto-lock rules for TAPE rses
            if not locked and lifetime is None:
                if [rse for rse in rses if rse.get('rse_type', RSEType.DISK) == RSEType.TAPE]:
                    locked = True

            # Block manual approval if RSE does not allow it
            if ask_approval:
                for rse in rses:
                    if list_rse_attributes(rse_id=rse['id'], session=session).get('block_manual_approval', False):
                        raise ManualRuleApprovalBlocked()

            if source_replica_expression:
                try:
                    source_rses = parse_expression(source_replica_expression, filter_={'vo': vo}, session=session)
                except InvalidRSEExpression as exc:
                    raise InvalidSourceReplicaExpression from exc
            else:
                source_rses = []

        # 2. Create the rse selector
        with METRICS.timer('add_rule.create_rse_selector'):
            rseselector = RSESelector(account=account, rses=rses, weight=weight, copies=copies, ignore_account_limit=ask_approval or ignore_account_limit, session=session)

        expires_at = datetime.utcnow() + timedelta(seconds=lifetime) if lifetime is not None else None

        notify_value = {'Y': RuleNotification.YES, 'C': RuleNotification.CLOSE, 'P': RuleNotification.PROGRESS}.get(notify or '', RuleNotification.NO)

        for elem in dids:
            # 3. Get the did
            with METRICS.timer('add_rule.get_did'):
                try:
                    stmt = select(
                        models.DataIdentifier
                    ).where(
                        and_(models.DataIdentifier.scope == elem['scope'],
                             models.DataIdentifier.name == elem['name'])
                    )
                    did = session.execute(stmt).scalar_one()
                except NoResultFound as exc:
                    raise DataIdentifierNotFound('Data identifier %s:%s is not valid.' % (elem['scope'], elem['name'])) from exc
                except TypeError as error:
                    raise InvalidObject(error.args) from error  # https://pylint.readthedocs.io/en/latest/user_guide/messages/warning/raise-missing-from.html

            # 3.1 If the did is a constituent, relay the rule to the archive
            if did.did_type == DIDType.FILE and did.constituent:
                # Check if a single replica of this DID exists; Do not put rule on file if there are only replicas on TAPE
                stmt = select(
                    func.count()
                ).select_from(
                    models.RSEFileAssociation
                ).join(
                    models.RSE,
                    models.RSEFileAssociation.rse_id == models.RSE.id
                ).where(
                    and_(models.RSEFileAssociation.scope == did.scope,
                         models.RSEFileAssociation.name == did.name,
                         models.RSEFileAssociation.state == ReplicaState.AVAILABLE,
                         models.RSE.rse_type != RSEType.TAPE)
                )
                replica_cnt = session.execute(stmt).scalar()

                if replica_cnt == 0:  # Put the rule on the archive
                    stmt = select(
                        models.ConstituentAssociation
                    ).join(
                        models.RSEFileAssociation,
                        and_(models.ConstituentAssociation.scope == models.RSEFileAssociation.scope,
                             models.ConstituentAssociation.name == models.RSEFileAssociation.name)
                    ).where(
                        and_(models.ConstituentAssociation.child_scope == did.scope,
                             models.ConstituentAssociation.child_name == did.name)
                    )
                    archive = session.execute(stmt).scalars().first()
                    if archive is not None:
                        elem['name'] = archive.name
                        elem['scope'] = archive.scope
                        try:
                            stmt = select(
                                models.DataIdentifier
                            ).where(
                                and_(models.DataIdentifier.scope == elem['scope'],
                                     models.DataIdentifier.name == elem['name'])
                            )
                            did = session.execute(stmt).scalar_one()
                        except NoResultFound as exc:
                            raise DataIdentifierNotFound('Data identifier %s:%s is not valid.' % (elem['scope'], elem['name'])) from exc
                        except TypeError as error:
                            raise InvalidObject(error.args) from error
                else:  # Put the rule on the constituent directly
                    pass

            # 3.2 Get the lifetime
            eol_at = define_eol(elem['scope'], elem['name'], rses, session=session)

            # 4. Create the replication rule
            with METRICS.timer('add_rule.create_rule'):
                meta_json = None
                if meta is not None:
                    try:
                        meta_json = json.dumps(meta)
                    except Exception:
                        meta_json = None

                new_rule = models.ReplicationRule(account=account,
                                                  name=elem['name'],
                                                  scope=elem['scope'],
                                                  did_type=did.did_type,
                                                  copies=copies,
                                                  rse_expression=rse_expression,
                                                  locked=locked,
                                                  grouping=grouping_value,
                                                  expires_at=expires_at,
                                                  weight=weight,
                                                  source_replica_expression=source_replica_expression,
                                                  activity=activity,
                                                  subscription_id=subscription_id,
                                                  notification=notify_value,
                                                  purge_replicas=purge_replicas,
                                                  ignore_availability=ignore_availability,
                                                  comments=comment,
                                                  ignore_account_limit=ignore_account_limit,
                                                  priority=priority,
                                                  split_container=split_container,
                                                  meta=meta_json,
                                                  eol_at=eol_at)
                try:
                    new_rule.save(session=session)
                except IntegrityError as error:
                    if match('.*ORA-00001.*', str(error.args[0])) \
                            or match('.*IntegrityError.*UNIQUE constraint failed.*', str(error.args[0])) \
                            or match('.*1062.*Duplicate entry.*for key.*', str(error.args[0])) \
                            or match('.*IntegrityError.*duplicate key value violates unique constraint.*', error.args[0]) \
                            or match('.*UniqueViolation.*duplicate key value violates unique constraint.*', error.args[0]) \
                            or match('.*IntegrityError.*columns? .*not unique.*', error.args[0]):
                        raise DuplicateRule(error.args[0]) from error
                    raise InvalidReplicationRule(error.args[0]) from error
                rule_ids.append(new_rule.id)

            if ask_approval:
                new_rule.state = RuleState.WAITING_APPROVAL
                # Use the new rule as the argument here
                auto_approver = AutoApprove(new_rule, did, session=session)
                if auto_approver.evaluate():
                    logger(logging.DEBUG, "Auto approving rule %s", str(new_rule.id))
                    logger(logging.DEBUG, "Created rule %s for injection", str(new_rule.id))
                    approve_rule(rule_id=new_rule.id, notify_approvers=False, session=session)
                    continue

                logger(logging.DEBUG, "Created rule %s in waiting for approval", str(new_rule.id))
                __create_rule_approval_email(rule=new_rule, session=session)
                continue

            # Force ASYNC mode for large rules
            if did.length is not None and (did.length * copies) >= 10000:
                asynchronous = True
                logger(logging.DEBUG, "Forced injection of rule %s", str(new_rule.id))

            if asynchronous or delay_injection:
                # TODO: asynchronous mode only available for closed dids (on the whole tree?)
                new_rule.state = RuleState.INJECT
                logger(logging.DEBUG, "Created rule %s for injection", str(new_rule.id))
                if delay_injection:
                    new_rule.created_at = datetime.utcnow() + timedelta(seconds=delay_injection)
                    logger(logging.DEBUG, "Scheduled rule %s for injection on %s", str(new_rule.id), new_rule.created_at)
                continue

            # If Split Container is chosen, the rule will be processed ASYNC
            if split_container and did.did_type == DIDType.CONTAINER:
                new_rule.state = RuleState.INJECT
                logger(logging.DEBUG, "Created rule %s for injection due to Split Container mode", str(new_rule.id))
                continue

            # 5. Apply the rule
            with METRICS.timer('add_rule.apply_rule'):
                try:
                    apply_rule(did, new_rule, [x['id'] for x in rses], [x['id'] for x in source_rses], rseselector, session=session)
                except IntegrityError as error:
                    raise ReplicationRuleCreationTemporaryFailed(error.args[0]) from error

            if new_rule.locks_stuck_cnt > 0:
                new_rule.state = RuleState.STUCK
                new_rule.error = 'MissingSourceReplica'
                if new_rule.grouping != RuleGrouping.NONE:
                    stmt = update(
                        models.DatasetLock
                    ).where(
                        models.DatasetLock.rule_id == new_rule.id
                    ).values({
                        models.DatasetLock.state: LockState.STUCK
                    })
                    session.execute(stmt)
            elif new_rule.locks_replicating_cnt == 0:
                new_rule.state = RuleState.OK
                if new_rule.grouping != RuleGrouping.NONE:
                    stmt = update(
                        models.DatasetLock
                    ).where(
                        models.DatasetLock.rule_id == new_rule.id
                    ).values({
                        models.DatasetLock.state: LockState.OK
                    })
                    session.execute(stmt)
                    session.flush()
                if new_rule.notification == RuleNotification.YES:
                    generate_email_for_rule_ok_notification(rule=new_rule, session=session)
                generate_rule_notifications(rule=new_rule, replicating_locks_before=0, session=session)
            else:
                new_rule.state = RuleState.REPLICATING
                if new_rule.grouping != RuleGrouping.NONE:
                    stmt = update(
                        models.DatasetLock
                    ).where(
                        models.DatasetLock.rule_id == new_rule.id
                    ).values({
                        models.DatasetLock.state: LockState.REPLICATING
                    })
                    session.execute(stmt)

            # Add rule to History
            insert_rule_history(rule=new_rule, recent=True, longterm=True, session=session)

            logger(logging.INFO, "Created rule %s [%d/%d/%d] with new algorithm for did %s:%s in state %s", str(new_rule.id), new_rule.locks_ok_cnt,
                   new_rule.locks_replicating_cnt, new_rule.locks_stuck_cnt, new_rule.scope, new_rule.name, str(new_rule.state))

    return rule_ids


@transactional_session
def add_rules(
    dids: Sequence[DIDDict],
    rules: Sequence[RuleDict],
    *,
    session: "Session",
    logger: LoggerFunction = logging.log
) -> dict[tuple[InternalScope, str], list[str]]:
    """
    Adds a list of replication rules to every did in dids

    :params dids:    List of data identifiers.
    :param rules:    List of dictionaries defining replication rules.
                     {account, copies, rse_expression, grouping, weight, lifetime, locked, subscription_id, source_replica_expression, activity, notify, purge_replicas}
    :param session:  The database session in use.
    :param logger:   Optional decorated logger that can be passed from the calling daemons or servers.
    :returns:        Dictionary (scope, name) with list of created rule ids
    :raises:         InvalidReplicationRule, InsufficientAccountLimit, InvalidRSEExpression, DataIdentifierNotFound, ReplicationRuleCreationTemporaryFailed, InvalidRuleWeight,
                     StagingAreaRuleRequiresLifetime, DuplicateRule, RSEWriteBlocked, ScratchDiskLifetimeConflict, ManualRuleApprovalBlocked
    """
    if any(r.get("copies", 1) <= 0 for r in rules):
        raise InvalidValueForKey("The number of copies for a replication rule should be greater than 0.")

    with METRICS.timer('add_rules.total'):
        rule_ids = {}

        # 1. Fetch the RSEs from the RSE expression to restrict further queries just on these RSEs
        restrict_rses = []
        all_source_rses = []
        with METRICS.timer('add_rules.parse_rse_expressions'):
            for rule in rules:
                vo = rule['account'].vo
                if rule.get('ignore_availability'):
                    restrict_rses.extend(parse_expression(rule['rse_expression'], filter_={'vo': vo}, session=session))
                else:
                    restrict_rses.extend(parse_expression(rule['rse_expression'], filter_={'vo': vo, 'availability_write': True}, session=session))
            restrict_rses = list(set([rse['id'] for rse in restrict_rses]))

            for rule in rules:
                if rule.get('source_replica_expression'):
                    vo = rule['account'].vo
                    all_source_rses.extend(parse_expression(rule.get('source_replica_expression'), filter_={'vo': vo}, session=session))
            all_source_rses = list(set([rse['id'] for rse in all_source_rses]))

        for elem in dids:
            # 2. Get the did
            with METRICS.timer('add_rules.get_did'):
                try:
                    stmt = select(
                        models.DataIdentifier
                    ).where(
                        and_(models.DataIdentifier.scope == elem['scope'],
                             models.DataIdentifier.name == elem['name'])
                    )
                    did = session.execute(stmt).scalar_one()
                except NoResultFound as exc:
                    raise DataIdentifierNotFound('Data identifier %s:%s is not valid.' % (elem['scope'], elem['name'])) from exc
                except TypeError as error:
                    raise InvalidObject(error.args) from error

            # 2.1 If the did is a constituent, relay the rule to the archive
            if did.did_type == DIDType.FILE and did.constituent:  # Check if a single replica of this DID exists
                stmt = select(
                    func.count()
                ).select_from(
                    models.RSEFileAssociation
                ).join(
                    models.RSE,
                    models.RSEFileAssociation.rse_id == models.RSE.id
                ).where(
                    and_(models.RSEFileAssociation.scope == did.scope,
                         models.RSEFileAssociation.name == did.name,
                         models.RSEFileAssociation.state == ReplicaState.AVAILABLE,
                         models.RSE.rse_type != RSEType.TAPE)
                )
                replica_cnt = session.execute(stmt).scalar()
                if replica_cnt == 0:  # Put the rule on the archive
                    stmt = select(
                        models.ConstituentAssociation
                    ).join(
                        models.RSEFileAssociation,
                        and_(models.ConstituentAssociation.scope == models.RSEFileAssociation.scope,
                             models.ConstituentAssociation.name == models.RSEFileAssociation.name)
                    ).where(
                        and_(models.ConstituentAssociation.child_scope == did.scope,
                             models.ConstituentAssociation.child_name == did.name)
                    )
                    archive = session.execute(stmt).scalars().first()
                    if archive is not None:
                        elem['name'] = archive.name
                        elem['scope'] = archive.scope
                        try:
                            stmt = select(
                                models.DataIdentifier
                            ).where(
                                and_(models.DataIdentifier.scope == elem['scope'],
                                     models.DataIdentifier.name == elem['name'])
                            )
                            did = session.execute(stmt).scalar_one()
                        except NoResultFound as exc:
                            raise DataIdentifierNotFound('Data identifier %s:%s is not valid.' % (elem['scope'], elem['name'])) from exc
                        except TypeError as error:
                            raise InvalidObject(error.args) from error
                else:  # Put the rule on the constituent directly
                    pass

            rule_ids[(elem['scope'], elem['name'])] = []

            # 3. Resolve the did into its contents
            with METRICS.timer('add_rules.resolve_dids_to_locks_replicas'):
                # Get all Replicas, not only the ones interesting for the rse_expression
                datasetfiles, locks, replicas, source_replicas = __resolve_did_to_locks_and_replicas(did=did,
                                                                                                     nowait=False,
                                                                                                     restrict_rses=restrict_rses,
                                                                                                     source_rses=all_source_rses,
                                                                                                     session=session)

            for rule in rules:
                with METRICS.timer('add_rules.add_rule'):
                    # 4. Resolve the rse_expression into a list of RSE-ids
                    vo = rule['account'].vo
                    if rule.get('ignore_availability'):
                        rses = parse_expression(rule['rse_expression'], filter_={'vo': vo}, session=session)
                    else:
                        rses = parse_expression(rule['rse_expression'], filter_={'vo': vo, 'availability_write': True}, session=session)

                    if rule.get('lifetime', None) is None:  # Check if one of the rses is a staging area
                        if [rse for rse in rses if rse.get('staging_area', False)]:
                            raise StagingAreaRuleRequiresLifetime()

                    # Check SCRATCHDISK Policy
                    try:
                        lifetime = get_scratch_policy(rule.get('account'), rses, rule.get('lifetime', None), session=session)
                    except UndefinedPolicy:
                        lifetime = rule.get('lifetime', None)

                    rule['lifetime'] = lifetime

                    # 4.5 Get the lifetime
                    eol_at = define_eol(did.scope, did.name, rses, session=session)

                    # Auto-lock rules for TAPE rses
                    if not rule.get('locked', False) and rule.get('lifetime', None) is None:
                        if [rse for rse in rses if rse.get('rse_type', RSEType.DISK) == RSEType.TAPE]:
                            rule['locked'] = True

                    # Block manual approval if RSE does not allow it
                    if rule.get('ask_approval', False):
                        for rse in rses:
                            if list_rse_attributes(rse_id=rse['id'], session=session).get('block_manual_approval', False):
                                raise ManualRuleApprovalBlocked()

                    if rule.get('source_replica_expression'):
                        source_rses = parse_expression(rule.get('source_replica_expression'), filter_={'vo': vo}, session=session)
                    else:
                        source_rses = []

                    # 5. Create the RSE selector
                    with METRICS.timer('add_rules.create_rse_selector'):
                        rseselector = RSESelector(account=rule['account'], rses=rses, weight=rule.get('weight'), copies=rule['copies'], ignore_account_limit=rule.get('ask_approval', False), session=session)

                    # 4. Create the replication rule
                    with METRICS.timer('add_rules.create_rule'):
                        grouping = {'ALL': RuleGrouping.ALL, 'NONE': RuleGrouping.NONE}.get(str(rule.get('grouping')), RuleGrouping.DATASET)

                        rule_lifetime: Optional[int] = rule.get('lifetime')
                        expires_at: Optional[datetime] = datetime.utcnow() + timedelta(seconds=rule_lifetime) if rule_lifetime is not None else None

                        notify = {'Y': RuleNotification.YES, 'C': RuleNotification.CLOSE, 'P': RuleNotification.PROGRESS, None: RuleNotification.NO}.get(rule.get('notify'))

                        if rule.get('meta') is not None:
                            try:
                                meta = json.dumps(rule.get('meta'))
                            except Exception:
                                meta = None
                        else:
                            meta = None

                        new_rule = models.ReplicationRule(account=rule['account'],
                                                          name=did.name,
                                                          scope=did.scope,
                                                          did_type=did.did_type,
                                                          copies=rule['copies'],
                                                          rse_expression=rule['rse_expression'],
                                                          locked=rule.get('locked'),
                                                          grouping=grouping,
                                                          expires_at=expires_at,
                                                          weight=rule.get('weight'),
                                                          source_replica_expression=rule.get('source_replica_expression'),
                                                          activity=rule.get('activity'),
                                                          subscription_id=rule.get('subscription_id'),
                                                          notification=notify,
                                                          purge_replicas=rule.get('purge_replicas', False),
                                                          ignore_availability=rule.get('ignore_availability', False),
                                                          comments=rule.get('comment', None),
                                                          priority=rule.get('priority', 3),
                                                          split_container=rule.get('split_container', False),
                                                          meta=meta,
                                                          eol_at=eol_at)
                        try:
                            new_rule.save(session=session)
                        except IntegrityError as error:
                            if match('.*ORA-00001.*', str(error.args[0])):
                                raise DuplicateRule(error.args[0]) from error
                            elif str(error.args[0]) == '(IntegrityError) UNIQUE constraint failed: rules.scope, rules.name, rules.account, rules.rse_expression, rules.copies':
                                raise DuplicateRule(error.args[0]) from error
                            raise InvalidReplicationRule(error.args[0]) from error

                        rule_ids[(did.scope, did.name)].append(new_rule.id)

                    if rule.get('ask_approval', False):
                        new_rule.state = RuleState.WAITING_APPROVAL
                        # Block manual approval for multi-rse rules
                        if len(rses) > 1:
                            raise InvalidReplicationRule('Ask approval is not allowed for rules with multiple RSEs')
                        if len(rses) == 1 and not did.is_open and did.bytes is not None and did.length is not None:
                            # This rule can be considered for auto-approval:
                            rse_attr = list_rse_attributes(rse_id=rses[0]['id'], session=session)
                            auto_approve = False
                            if 'auto_approve_bytes' in rse_attr and 'auto_approve_files' in rse_attr:
                                if did.bytes < int(rse_attr.get('auto_approve_bytes')) and did.length < int(rse_attr.get('auto_approve_bytes')):
                                    auto_approve = True
                            elif did.bytes < int(rse_attr.get('auto_approve_bytes', -1)):
                                auto_approve = True
                            elif did.length < int(rse_attr.get('auto_approve_files', -1)):
                                auto_approve = True
                            if auto_approve:
                                logger(logging.DEBUG, "Auto approving rule %s", str(new_rule.id))
                                logger(logging.DEBUG, "Created rule %s for injection", str(new_rule.id))
                                approve_rule(rule_id=new_rule.id, notify_approvers=False, session=session)
                                continue
                        logger(logging.DEBUG, "Created rule %s in waiting for approval", str(new_rule.id))
                        __create_rule_approval_email(rule=new_rule, session=session)
                        continue

                    delay_injection = rule.get('delay_injection')
                    if rule.get('asynchronous', False) or delay_injection:
                        new_rule.state = RuleState.INJECT
                        logger(logging.DEBUG, "Created rule %s for injection", str(new_rule.id))
                        if delay_injection:
                            new_rule.created_at = datetime.utcnow() + timedelta(seconds=delay_injection)
                            logger(logging.DEBUG, "Scheduled rule %s for injection on %s", (str(new_rule.id), new_rule.created_at))
                        continue

                    if rule.get('split_container', False) and did.did_type == DIDType.CONTAINER:
                        new_rule.state = RuleState.INJECT
                        logger(logging.DEBUG, "Created rule %s for injection due to Split Container mode", str(new_rule.id))
                        continue

                    # 5. Apply the replication rule to create locks, replicas and transfers
                    with METRICS.timer('add_rules.create_locks_replicas_transfers'):
                        try:
                            __create_locks_replicas_transfers(datasetfiles=datasetfiles,
                                                              locks=locks,
                                                              replicas=replicas,
                                                              source_replicas=source_replicas,
                                                              rseselector=rseselector,
                                                              rule=new_rule,
                                                              preferred_rse_ids=[],
                                                              source_rses=[rse['id'] for rse in source_rses],
                                                              session=session)
                        except IntegrityError as error:
                            raise ReplicationRuleCreationTemporaryFailed(error.args[0]) from error

                    if new_rule.locks_stuck_cnt > 0:
                        new_rule.state = RuleState.STUCK
                        new_rule.error = 'MissingSourceReplica'
                        if new_rule.grouping != RuleGrouping.NONE:
                            stmt = update(
                                models.DatasetLock
                            ).where(
                                models.DatasetLock.rule_id == new_rule.id
                            ).values({
                                models.DatasetLock.state: LockState.STUCK
                            })
                            session.execute(stmt)
                    elif new_rule.locks_replicating_cnt == 0:
                        new_rule.state = RuleState.OK
                        if new_rule.grouping != RuleGrouping.NONE:
                            stmt = update(
                                models.DatasetLock
                            ).where(
                                models.DatasetLock.rule_id == new_rule.id
                            ).values({
                                models.DatasetLock.state: LockState.OK
                            })
                            session.execute(stmt)
                            session.flush()
                        if new_rule.notification == RuleNotification.YES:
                            generate_email_for_rule_ok_notification(rule=new_rule, session=session)
                        generate_rule_notifications(rule=new_rule, replicating_locks_before=0, session=session)
                    else:
                        new_rule.state = RuleState.REPLICATING
                        if new_rule.grouping != RuleGrouping.NONE:
                            stmt = update(
                                models.DatasetLock
                            ).where(
                                models.DatasetLock.rule_id == new_rule.id
                            ).values({
                                models.DatasetLock.state: LockState.REPLICATING
                            })
                            session.execute(stmt)

                    # Add rule to History
                    insert_rule_history(rule=new_rule, recent=True, longterm=True, session=session)

                    logger(logging.INFO, "Created rule %s [%d/%d/%d] in state %s", str(new_rule.id), new_rule.locks_ok_cnt, new_rule.locks_replicating_cnt, new_rule.locks_stuck_cnt, str(new_rule.state))

    return rule_ids


@transactional_session
def inject_rule(
    rule_id: str,
    *,
    session: "Session",
    logger: LoggerFunction = logging.log
) -> None:
    """
    Inject a replication rule.

    :param rule_id:    The id of the rule to inject.
    :param new_owner:  The new owner of the rule.
    :param session:    The database session in use.
    :param logger:     Optional decorated logger that can be passed from the calling daemons or servers.
    :raises:           InvalidReplicationRule, InsufficientAccountLimit, InvalidRSEExpression, DataId, RSEOverQuota
    """

    try:
        stmt = select(
            models.ReplicationRule
        ).where(
            models.ReplicationRule.id == rule_id
        ).with_for_update(
            nowait=True
        )
        rule = session.execute(stmt).scalar_one()
    except NoResultFound as exc:
        raise RuleNotFound('No rule with the id %s found' % (rule_id)) from exc

    # Check if rule will expire in the next 5 minutes:
    if rule.child_rule_id is None and rule.expires_at is not None and rule.expires_at < datetime.utcnow() + timedelta(seconds=300):
        logger(logging.INFO, 'Rule %s expiring soon, skipping', str(rule.id))
        return

    # Special R2D2 container handling
    if (rule.did_type == DIDType.CONTAINER and '.r2d2_request.' in rule.name) or (rule.split_container and rule.did_type == DIDType.CONTAINER):
        logger(logging.DEBUG, "Creating dataset rules for Split Container rule %s", str(rule.id))
        # Get all child datasets and put rules on them
        dids = [{'scope': dataset['scope'], 'name': dataset['name']} for dataset in rucio.core.did.list_child_datasets(scope=rule.scope, name=rule.name, session=session)]
        # Remove duplicates from the list of dictionaries
        dids = [dict(t) for t in {tuple(d.items()) for d in dids}]
        # Remove dids which already have a similar rule
        stmt = select(
            models.ReplicationRule.id
        ).where(
            and_(models.ReplicationRule.account == rule.account,
                 models.ReplicationRule.rse_expression == rule.rse_expression)
        )
        dids = [did for did in dids if session.execute(stmt.where(and_(models.ReplicationRule.scope == did['scope'], models.ReplicationRule.name == did['name']))).scalar_one_or_none() is None]
        if rule.expires_at:
            lifetime = (rule.expires_at - datetime.utcnow()).days * 24 * 3600 + (rule.expires_at - datetime.utcnow()).seconds
        else:
            lifetime = None

        notify = {RuleNotification.YES: 'Y', RuleNotification.CLOSE: 'C', RuleNotification.PROGRESS: 'P'}.get(rule.notification, 'N')

        add_rule(dids=dids,
                 account=rule.account,
                 copies=rule.copies,
                 rse_expression=rule.rse_expression,
                 grouping='DATASET',
                 weight=None,
                 lifetime=lifetime,
                 locked=False,
                 subscription_id=None,
                 activity=rule.activity,
                 notify=notify,
                 comment=rule.comments,
                 asynchronous=True,
                 ignore_availability=rule.ignore_availability,
                 ignore_account_limit=True,
                 priority=rule.priority,
                 split_container=rule.split_container,
                 session=session)
        rule.delete(session=session)
        return

    # 1. Resolve the rse_expression into a list of RSE-ids
    with METRICS.timer('inject_rule.parse_rse_expression'):
        vo = rule['account'].vo
        if rule.ignore_availability:
            rses = parse_expression(rule.rse_expression, filter_={'vo': vo}, session=session)
        else:
            rses = parse_expression(rule.rse_expression, filter_={'vo': vo, 'availability_write': True}, session=session)

        if rule.source_replica_expression:
            source_rses = parse_expression(rule.source_replica_expression, filter_={'vo': vo}, session=session)
        else:
            source_rses = []

    # 2. Create the rse selector
    with METRICS.timer('inject_rule.create_rse_selector'):
        rseselector = RSESelector(account=rule['account'], rses=rses, weight=rule.weight, copies=rule.copies, ignore_account_limit=rule.ignore_account_limit, session=session)

    # 3. Get the did
    with METRICS.timer('inject_rule.get_did'):
        try:
            stmt = select(
                models.DataIdentifier
            ).where(
                and_(models.DataIdentifier.scope == rule.scope,
                     models.DataIdentifier.name == rule.name)
            )
            did = session.execute(stmt).scalar_one()
        except NoResultFound as exc:
            raise DataIdentifierNotFound('Data identifier %s:%s is not valid.' % (rule.scope, rule.name)) from exc
        except TypeError as error:
            raise InvalidObject(error.args) from error

    # 4. Apply the rule
    with METRICS.timer('inject_rule.apply_rule'):
        try:
            apply_rule(did, rule, [x['id'] for x in rses], [x['id'] for x in source_rses], rseselector, session=session)
        except IntegrityError as error:
            raise ReplicationRuleCreationTemporaryFailed(error.args[0]) from error

    if rule.locks_stuck_cnt > 0:
        rule.state = RuleState.STUCK
        rule.error = 'MissingSourceReplica'
        if rule.grouping != RuleGrouping.NONE:
            stmt = update(
                models.DatasetLock
            ).where(
                models.DatasetLock.rule_id == rule.id
            ).values({
                models.DatasetLock.state: LockState.STUCK
            })
            session.execute(stmt)
    elif rule.locks_replicating_cnt == 0:
        rule.state = RuleState.OK
        if rule.grouping != RuleGrouping.NONE:
            stmt = update(
                models.DatasetLock
            ).where(
                models.DatasetLock.rule_id == rule.id
            ).values({
                models.DatasetLock.state: LockState.OK
            })
            session.execute(stmt)
            session.flush()
        if rule.notification == RuleNotification.YES:
            generate_email_for_rule_ok_notification(rule=rule, session=session)
        generate_rule_notifications(rule=rule, replicating_locks_before=0, session=session)
        # Try to release potential parent rules
        release_parent_rule(child_rule_id=rule.id, session=session)
    else:
        rule.state = RuleState.REPLICATING
        if rule.grouping != RuleGrouping.NONE:
            stmt = update(
                models.DatasetLock
            ).where(
                models.DatasetLock.rule_id == rule.id
            ).values({
                models.DatasetLock.state: LockState.REPLICATING
            })
            session.execute(stmt)

    # Add rule to History
    insert_rule_history(rule=rule, recent=True, longterm=True, session=session)

    logger(logging.INFO, "Created rule %s [%d/%d/%d] with new algorithm for did %s:%s in state %s", str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt, rule.scope, rule.name, str(rule.state))


@stream_session
def list_rules(
    filters: Optional[dict[str, Any]] = None,
    *,
    session: "Session"
) -> Iterator[dict[str, Any]]:
    """
    List replication rules.

    :param filters: dictionary of attributes by which the results should be filtered.
    :param session: The database session in use.
    :raises:        RucioException
    """

    stmt = select(
        models.ReplicationRule,
        models.DataIdentifier.bytes
    ).join(
        models.DataIdentifier,
        and_(
            models.ReplicationRule.scope == models.DataIdentifier.scope,
            models.ReplicationRule.name == models.DataIdentifier.name
        )
    )
    if filters is not None:
        for (key, value) in filters.items():
            if key in ['account', 'scope']:
                if '*' in value.internal:
                    value = value.internal.replace('*', '%')
                    stmt = stmt.where(getattr(models.ReplicationRule, key).like(value))
                    continue
                # else fall through
            elif key == 'created_before':
                stmt = stmt.where(models.ReplicationRule.created_at <= str_to_date(value))
                continue
            elif key == 'created_after':
                stmt = stmt.where(models.ReplicationRule.created_at >= str_to_date(value))
                continue
            elif key == 'updated_before':
                stmt = stmt.where(models.ReplicationRule.updated_at <= str_to_date(value))
                continue
            elif key == 'updated_after':
                stmt = stmt.where(models.ReplicationRule.updated_at >= str_to_date(value))
                continue
            elif key == 'state':
                if isinstance(value, str):
                    value = RuleState(value)
                else:
                    try:
                        value = RuleState[value]
                    except ValueError:
                        pass
            elif key == 'did_type' and isinstance(value, str):
                value = DIDType(value)
            elif key == 'grouping' and isinstance(value, str):
                value = RuleGrouping(value)
            stmt = stmt.where(getattr(models.ReplicationRule, key) == value)

    try:
        for rule, data_identifier_bytes in session.execute(stmt).yield_per(5):
            d = rule.to_dict()
            d['bytes'] = data_identifier_bytes
            yield d
    except StatementError as exc:
        raise RucioException('Badly formatted input (IDs?)') from exc


@stream_session
def list_rule_history(
    rule_id: str,
    *,
    session: "Session"
) -> Iterator[dict[str, Any]]:
    """
    List the rule history of a rule.

    :param rule_id: The id of the rule.
    :param session: The database session in use.
    :raises:        RucioException
    """

    stmt = select(
        models.ReplicationRuleHistory.updated_at,
        models.ReplicationRuleHistory.state,
        models.ReplicationRuleHistory.locks_ok_cnt,
        models.ReplicationRuleHistory.locks_stuck_cnt,
        models.ReplicationRuleHistory.locks_replicating_cnt
    ).where(
        models.ReplicationRuleHistory.id == rule_id
    ).order_by(
        models.ReplicationRuleHistory.updated_at
    )

    try:
        for rule in session.execute(stmt).yield_per(5):
            yield rule._asdict()
    except StatementError as exc:
        raise RucioException('Badly formatted input (IDs?)') from exc


@stream_session
def list_rule_full_history(
    scope: InternalScope,
    name: str,
    *,
    session: "Session"
) -> Iterator[dict[str, Any]]:
    """
    List the rule history of a DID.

    :param scope: The scope of the DID.
    :param name: The name of the DID.
    :param session: The database session in use.
    :raises:        RucioException
    """

    stmt = select(
        models.ReplicationRuleHistory.id.label('rule_id'),
        models.ReplicationRuleHistory.created_at,
        models.ReplicationRuleHistory.updated_at,
        models.ReplicationRuleHistory.rse_expression,
        models.ReplicationRuleHistory.state,
        models.ReplicationRuleHistory.account,
        models.ReplicationRuleHistory.locks_ok_cnt,
        models.ReplicationRuleHistory.locks_stuck_cnt,
        models.ReplicationRuleHistory.locks_replicating_cnt
    ).with_hint(
        models.ReplicationRuleHistory, 'INDEX(RULES_HISTORY_SCOPENAME_IDX)', 'oracle'
    ).where(
        and_(models.ReplicationRuleHistory.scope == scope,
             models.ReplicationRuleHistory.name == name)
    ).order_by(
        models.ReplicationRuleHistory.created_at,
        models.ReplicationRuleHistory.updated_at
    )
    for rule in session.execute(stmt).yield_per(5):
        yield rule._asdict()


@stream_session
def list_associated_rules_for_file(
    scope: InternalScope,
    name: str,
    *,
    session: "Session"
) -> Iterator[dict[str, Any]]:
    """
    List replication rules a file is affected from.

    :param scope:   Scope of the file.
    :param name:    Name of the file.
    :param session: The database session in use.
    :raises:        RucioException
    """
    rucio.core.did.get_did(scope=scope, name=name, session=session)  # Check if the did acually exists
    stmt = select(
        models.ReplicationRule,
        models.DataIdentifier.bytes
    ).distinct(
    ).join(
        models.ReplicaLock,
        models.ReplicationRule.id == models.ReplicaLock.rule_id
    ).join(
        models.DataIdentifier,
        and_(models.ReplicationRule.scope == models.DataIdentifier.scope,
             models.ReplicationRule.name == models.DataIdentifier.name)
    ).with_hint(
        models.ReplicaLock, 'INDEX(LOCKS LOCKS_PK)', 'oracle'
    ).where(
        and_(models.ReplicaLock.scope == scope,
             models.ReplicaLock.name == name)
    )
    try:
        for rule, data_identifier_bytes in session.execute(stmt).yield_per(5):
            d = rule.to_dict()
            d['bytes'] = data_identifier_bytes
            yield d
    except StatementError as exc:
        raise RucioException('Badly formatted input (IDs?)') from exc


@transactional_session
def delete_rule(
    rule_id: str,
    purge_replicas: Optional[bool] = None,
    soft: bool = False,
    delete_parent: bool = False,
    nowait: bool = False,
    *,
    session: "Session",
    ignore_rule_lock: bool = False
) -> None:
    """
    Delete a replication rule.

    :param rule_id:           The rule to delete.
    :param purge_replicas:    Purge the replicas immediately.
    :param soft:              Only perform a soft deletion.
    :param delete_parent:     Delete rules even if they have a child_rule_id set.
    :param nowait:            Nowait parameter for the FOR UPDATE statement.
    :param session:           The database session in use.
    :param ignore_rule_lock:  Ignore any locks on the rule
    :raises:                  RuleNotFound if no Rule can be found.
    :raises:                  UnsupportedOperation if the Rule is locked.
    """

    with METRICS.timer('delete_rule.total'):
        try:
            stmt = select(
                models.ReplicationRule
            ).where(
                models.ReplicationRule.id == rule_id
            ).with_for_update(
                nowait=nowait
            )
            rule = session.execute(stmt).scalar_one()
        except NoResultFound as exc:
            raise RuleNotFound('No rule with the id %s found' % rule_id) from exc
        if rule.locked and not ignore_rule_lock:
            raise UnsupportedOperation('The replication rule is locked and has to be unlocked before it can be deleted.')

        if rule.child_rule_id is not None and not delete_parent:
            raise UnsupportedOperation('The replication rule has a child rule and thus cannot be deleted.')

        if purge_replicas is not None:
            rule.purge_replicas = purge_replicas

        if soft:
            if rule.expires_at:
                rule.expires_at = min(datetime.utcnow() + timedelta(seconds=3600), rule.expires_at)
            else:
                rule.expires_at = datetime.utcnow() + timedelta(seconds=3600)
            if rule.child_rule_id is not None and delete_parent:
                rule.child_rule_id = None
            insert_rule_history(rule=rule, recent=True, longterm=False, session=session)
            return

        stmt = select(
            models.ReplicaLock
        ).where(
            models.ReplicaLock.rule_id == rule_id
        ).with_for_update(
            nowait=nowait
        )
        results = session.execute(stmt).yield_per(100)

        # Remove locks, set tombstone if applicable
        transfers_to_delete = []  # [{'scope': , 'name':, 'rse_id':}]
        account_counter_decreases = {}  # {'rse_id': [file_size, file_size, file_size]}

        for result in results:
            lock = result[0]
            if __delete_lock_and_update_replica(lock=lock, purge_replicas=rule.purge_replicas,
                                                nowait=nowait, session=session):
                transfers_to_delete.append({'scope': lock.scope, 'name': lock.name, 'rse_id': lock.rse_id})
            if lock.rse_id not in account_counter_decreases:
                account_counter_decreases[lock.rse_id] = []
            account_counter_decreases[lock.rse_id].append(lock.bytes)

        # Delete the DatasetLocks
        stmt = delete(
            models.DatasetLock
        ).where(
            models.DatasetLock.rule_id == rule_id
        ).execution_options(
            synchronize_session=False
        )
        session.execute(stmt)

        # Decrease account_counters
        for rse_id in account_counter_decreases.keys():
            account_counter.decrease(rse_id=rse_id, account=rule.account, files=len(account_counter_decreases[rse_id]),
                                     bytes_=sum(account_counter_decreases[rse_id]), session=session)

        # Try to release potential parent rules
        release_parent_rule(child_rule_id=rule.id, remove_parent_expiration=True, session=session)

        # Insert history
        insert_rule_history(rule=rule, recent=False, longterm=True, session=session)

        session.flush()
        rule.delete(session=session)

        for transfer in transfers_to_delete:
            transfers_to_cancel = request_core.cancel_request_did(scope=transfer['scope'], name=transfer['name'],
                                                                  dest_rse_id=transfer['rse_id'], session=session)
            transfer_core.cancel_transfers(transfers_to_cancel)


@transactional_session
def repair_rule(
    rule_id: str,
    *,
    session: "Session",
    logger: LoggerFunction = logging.log
) -> None:
    """
    Repair a STUCK replication rule.

    :param rule_id:   The rule to repair.
    :param session:   The database session in use.
    :param logger:    Optional decorated logger that can be passed from the calling daemons or servers.
    """

    # Rule error cases:
    # (A) A rule get's an exception on rule-creation. This can only be the MissingSourceReplica exception.
    # (B) A rule get's an error when re-evaluated: InvalidRSEExpression, InvalidRuleWeight, InsufficientTargetRSEs, RSEWriteBlocked
    #     InsufficientAccountLimit. The re-evaluation has to be done again and potential missing locks have to be
    #     created.
    # (C) Transfers fail and mark locks (and the rule) as STUCK. All STUCK locks have to be repaired.
    # (D) Files are declared as BAD.

    # start_time = time.time()
    try:
        stmt = select(
            models.ReplicationRule
        ).where(
            models.ReplicationRule.id == rule_id
        ).with_for_update(
            nowait=True
        )
        rule = session.execute(stmt).scalar_one()
        rule.updated_at = datetime.utcnow()

        # Check if rule is longer than 2 weeks in STUCK
        if rule.stuck_at is None:
            rule.stuck_at = datetime.utcnow()
        if rule.stuck_at < (datetime.utcnow() - timedelta(days=14)):
            rule.state = RuleState.SUSPENDED
            insert_rule_history(rule=rule, recent=True, longterm=False, session=session)
            logger(logging.INFO, 'Replication rule %s has been SUSPENDED', rule_id)
            return

        # Evaluate the RSE expression to see if there is an alternative RSE anyway
        try:
            vo = rule.account.vo
            rses = parse_expression(rule.rse_expression, filter_={'vo': vo}, session=session)
            if rule.ignore_availability:
                target_rses = parse_expression(rule.rse_expression, filter_={'vo': vo}, session=session)
            else:
                target_rses = parse_expression(rule.rse_expression, filter_={'vo': vo, 'availability_write': True}, session=session)
            if rule.source_replica_expression:
                source_rses = parse_expression(rule.source_replica_expression, filter_={'vo': vo}, session=session)
            else:
                source_rses = []
        except (InvalidRSEExpression, RSEWriteBlocked) as error:
            rule.state = RuleState.STUCK
            rule.error = (str(error)[:245] + '...') if len(str(error)) > 245 else str(error)
            rule.save(session=session)
            # Insert rule history
            insert_rule_history(rule=rule, recent=True, longterm=False, session=session)
            # Try to update the DatasetLocks
            if rule.grouping != RuleGrouping.NONE:
                stmt = update(
                    models.DatasetLock
                ).where(
                    models.DatasetLock.rule_id == rule.id
                ).values({
                    models.DatasetLock.state: LockState.STUCK
                })
                session.execute(stmt)
            logger(logging.DEBUG, '%s while repairing rule %s', str(error), rule_id)
            return

        # Create the RSESelector
        try:
            rseselector = RSESelector(account=rule.account,
                                      rses=target_rses,
                                      weight=rule.weight,
                                      copies=rule.copies,
                                      ignore_account_limit=rule.ignore_account_limit,
                                      session=session)
        except (InvalidRuleWeight, InsufficientTargetRSEs, InsufficientAccountLimit) as error:
            rule.state = RuleState.STUCK
            rule.error = (str(error)[:245] + '...') if len(str(error)) > 245 else str(error)
            rule.save(session=session)
            # Insert rule history
            insert_rule_history(rule=rule, recent=True, longterm=False, session=session)
            # Try to update the DatasetLocks
            if rule.grouping != RuleGrouping.NONE:
                stmt = update(
                    models.DatasetLock
                ).where(
                    models.DatasetLock.rule_id == rule.id
                ).values({
                    models.DatasetLock.state: LockState.STUCK
                })
                session.execute(stmt)
            logger(logging.DEBUG, '%s while repairing rule %s', type(error).__name__, rule_id)
            return

        # Reset the counters
        logger(logging.DEBUG, "Resetting counters for rule %s [%d/%d/%d]", str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt)
        rule.locks_ok_cnt = 0
        rule.locks_replicating_cnt = 0
        rule.locks_stuck_cnt = 0
        stmt = select(
            models.ReplicaLock.state,
            func.count(models.ReplicaLock.state).label('state_counter')
        ).where(
            models.ReplicaLock.rule_id == rule.id
        ).group_by(
            models.ReplicaLock.state
        )
        rule_counts = session.execute(stmt).all()
        for count in rule_counts:
            if count.state == LockState.OK:
                rule.locks_ok_cnt = count.state_counter
            elif count.state == LockState.REPLICATING:
                rule.locks_replicating_cnt = count.state_counter
            elif count.state == LockState.STUCK:
                rule.locks_stuck_cnt = count.state_counter
        logger(logging.DEBUG, "Finished resetting counters for rule %s [%d/%d/%d]", str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt)

        # Get the did
        stmt = select(
            models.DataIdentifier
        ).where(
            and_(models.DataIdentifier.scope == rule.scope,
                 models.DataIdentifier.name == rule.name)
        )
        did = session.execute(stmt).scalar_one()

        # Detect if there is something wrong with the dataset and
        # make the decisison on soft or hard repair.
        hard_repair = False
        if did.did_type != DIDType.FILE:
            nr_files = rucio.core.did.get_did(scope=rule.scope, name=rule.name, dynamic_depth=DIDType.FILE, session=session)['length']
        else:
            nr_files = 1
        if nr_files * rule.copies != (rule.locks_ok_cnt + rule.locks_stuck_cnt + rule.locks_replicating_cnt):
            hard_repair = True
            logger(logging.DEBUG, 'Repairing rule %s in HARD mode', str(rule.id))
        elif rule.copies > 1 and rule.grouping == RuleGrouping.NONE:
            hard_repair = True
            logger(logging.DEBUG, 'Repairing rule %s in HARD mode', str(rule.id))

        # Resolve the did to its contents
        datasetfiles, locks, replicas, source_replicas = __resolve_did_to_locks_and_replicas(did=did,
                                                                                             nowait=True,
                                                                                             restrict_rses=[rse['id'] for rse in rses],
                                                                                             source_rses=[rse['id'] for rse in source_rses],
                                                                                             only_stuck=not hard_repair,
                                                                                             session=session)

        session.flush()

        # 1. Try to find missing locks and create them based on grouping
        if did.did_type != DIDType.FILE and hard_repair:
            try:
                __find_missing_locks_and_create_them(datasetfiles=datasetfiles,
                                                     locks=locks,
                                                     replicas=replicas,
                                                     source_replicas=source_replicas,
                                                     rseselector=rseselector,
                                                     rule=rule,
                                                     source_rses=[rse['id'] for rse in source_rses],
                                                     session=session)
            except (InsufficientAccountLimit, InsufficientTargetRSEs) as error:
                rule.state = RuleState.STUCK
                rule.error = (str(error)[:245] + '...') if len(str(error)) > 245 else str(error)
                rule.save(session=session)
                # Insert rule history
                insert_rule_history(rule=rule, recent=True, longterm=False, session=session)
                # Try to update the DatasetLocks
                if rule.grouping != RuleGrouping.NONE:
                    stmt = update(
                        models.DatasetLock
                    ).where(
                        models.DatasetLock.rule_id == rule.id
                    ).values({
                        models.DatasetLock.state: LockState.STUCK
                    })
                    session.execute(stmt)
                logger(logging.DEBUG, '%s while repairing rule %s', type(error).__name__, rule_id)
                return

            session.flush()

        # 2. Try to find surplus locks and remove them
        if hard_repair:
            __find_surplus_locks_and_remove_them(datasetfiles=datasetfiles,
                                                 locks=locks,
                                                 rule=rule,
                                                 session=session)

            session.flush()

        # 3. Try to find STUCK locks and repair them based on grouping
        try:
            __find_stuck_locks_and_repair_them(datasetfiles=datasetfiles,
                                               locks=locks,
                                               replicas=replicas,
                                               source_replicas=source_replicas,
                                               rseselector=rseselector,
                                               rule=rule,
                                               source_rses=[rse['id'] for rse in source_rses],
                                               session=session)
        except (InsufficientAccountLimit, InsufficientTargetRSEs) as error:
            rule.state = RuleState.STUCK
            rule.error = (str(error)[:245] + '...') if len(str(error)) > 245 else str(error)
            rule.save(session=session)
            # Insert rule history
            insert_rule_history(rule=rule, recent=True, longterm=False, session=session)
            # Try to update the DatasetLocks
            if rule.grouping != RuleGrouping.NONE:
                stmt = update(
                    models.DatasetLock
                ).where(
                    models.DatasetLock.rule_id == rule.id
                ).values({
                    models.DatasetLock.state: LockState.STUCK
                })
                session.execute(stmt)
            logger(logging.DEBUG, '%s while repairing rule %s', type(error).__name__, rule_id)
            return

        # Delete Datasetlocks which are not relevant anymore
        stmt = select(
            models.ReplicaLock.rse_id
        ).distinct(
        ).where(
            models.ReplicaLock.rule_id == rule.id
        )
        validated_datasetlock_rse_ids = session.execute(stmt).scalars().all()

        stmt = select(
            models.DatasetLock
        ).where(
            models.DatasetLock.rule_id == rule.id
        )
        dataset_locks = session.execute(stmt).scalars().all()
        for dataset_lock in dataset_locks:
            if dataset_lock.rse_id not in validated_datasetlock_rse_ids:
                dataset_lock.delete(session=session)

        if rule.locks_stuck_cnt != 0:
            logger(logging.INFO, 'Rule %s [%d/%d/%d] state=STUCK', str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt)
            rule.state = RuleState.STUCK
            # Insert rule history
            insert_rule_history(rule=rule, recent=True, longterm=False, session=session)
            # Try to update the DatasetLocks
            if rule.grouping != RuleGrouping.NONE:
                stmt = update(
                    models.DatasetLock
                ).where(
                    models.DatasetLock.rule_id == rule.id
                ).values({
                    models.DatasetLock.state: LockState.STUCK
                })
                session.execute(stmt)
            # TODO: Increase some kind of Stuck Counter here, The rule should at some point be SUSPENDED
            return

        rule.stuck_at = None

        if rule.locks_replicating_cnt > 0:
            logger(logging.INFO, 'Rule %s [%d/%d/%d] state=REPLICATING', str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt)
            rule.state = RuleState.REPLICATING
            rule.error = None
            # Insert rule history
            insert_rule_history(rule=rule, recent=True, longterm=False, session=session)
            # Try to update the DatasetLocks
            if rule.grouping != RuleGrouping.NONE:
                stmt = update(
                    models.DatasetLock
                ).where(
                    models.DatasetLock.rule_id == rule.id
                ).values({
                    models.DatasetLock.state: LockState.REPLICATING
                })
                session.execute(stmt)
            return

        rule.state = RuleState.OK
        rule.error = None
        # Insert rule history
        insert_rule_history(rule=rule, recent=True, longterm=False, session=session)
        logger(logging.INFO, 'Rule %s [%d/%d/%d] state=OK', str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt)

        if rule.grouping != RuleGrouping.NONE:
            stmt = update(
                models.DatasetLock
            ).where(
                models.DatasetLock.rule_id == rule.id
            ).values({
                models.DatasetLock.state: LockState.OK
            })
            session.execute(stmt)
            session.flush()
        if rule.notification == RuleNotification.YES:
            generate_email_for_rule_ok_notification(rule=rule, session=session)
        generate_rule_notifications(rule=rule, replicating_locks_before=0, session=session)
        # Try to release potential parent rules
        rucio.core.rule.release_parent_rule(child_rule_id=rule.id, session=session)

        return

    except NoResultFound:
        # The rule has been deleted in the meanwhile
        return


@read_session
def get_rule(
    rule_id: str,
    *,
    session: "Session"
) -> dict[str, Any]:
    """
    Get a specific replication rule.

    :param rule_id: The rule_id to select.
    :param session: The database session in use.
    :raises:        RuleNotFound if no Rule can be found.
    """

    try:
        stmt = select(
            models.ReplicationRule
        ).where(
            models.ReplicationRule.id == rule_id
        )
        rule = session.execute(stmt).scalar_one()
        return rule.to_dict()
    except NoResultFound as exc:
        raise RuleNotFound('No rule with the id %s found' % (rule_id)) from exc
    except StatementError as exc:
        raise RucioException('Badly formatted rule id (%s)' % (rule_id)) from exc


@transactional_session
def update_rule(
    rule_id: str,
    options: dict[str, Any],
    *,
    session: "Session"
) -> None:
    """
    Update a rules options.

    :param rule_id:     The rule_id to lock.
    :param options:     Dictionary of options
    :param session:     The database session in use.
    :raises:            RuleNotFound if no Rule can be found, InputValidationError if invalid option is used, ScratchDiskLifetimeConflict if wrong ScratchDiskLifetime is used.
    """

    valid_options = ['comment', 'locked', 'lifetime', 'account', 'state',
                     'activity', 'source_replica_expression', 'cancel_requests',
                     'priority', 'child_rule_id', 'eol_at', 'meta',
                     'purge_replicas', 'boost_rule']

    for key in options:
        if key not in valid_options:
            raise InputValidationError('%s is not a valid option to set.' % key)

    try:
        query = select(
            models.ReplicationRule
        ).where(
            models.ReplicationRule.id == rule_id
        )
        rule: models.ReplicationRule = session.execute(query).scalar_one()

        for key in options:
            if key == 'lifetime':
                # Check SCRATCHDISK Policy
                vo = rule.account.vo
                rses = parse_expression(rule.rse_expression, filter_={'vo': vo}, session=session)
                try:
                    lifetime = get_scratch_policy(rule.account, rses, options['lifetime'], session=session)
                except UndefinedPolicy:
                    lifetime = options['lifetime']
                rule.expires_at = datetime.utcnow() + timedelta(seconds=lifetime) if lifetime is not None else None
            if key == 'source_replica_expression':
                rule.source_replica_expression = options['source_replica_expression']

            if key == 'comment':
                rule.comments = options['comment']

            if key == 'activity':
                validate_schema(
                    'activity', options['activity'], vo=rule.account.vo
                )
                rule.activity = options['activity']
                # Cancel transfers and re-submit them:
                query = select(
                    models.ReplicaLock
                ).where(
                    models.ReplicaLock.rule_id == rule.id,
                    models.ReplicaLock.state == LockState.REPLICATING
                )
                for lock in session.execute(query).scalars().all():
                    transfers_to_cancel = request_core.cancel_request_did(
                        scope=lock.scope,
                        name=lock.name,
                        dest_rse_id=lock.rse_id,
                        session=session
                    )
                    transfer_core.cancel_transfers(transfers_to_cancel)
                    query = select(
                        models.RSEFileAssociation.md5,
                        models.RSEFileAssociation.bytes,
                        models.RSEFileAssociation.adler32
                    ).where(
                        models.RSEFileAssociation.scope == lock.scope,
                        models.RSEFileAssociation.name == lock.name,
                        models.RSEFileAssociation.rse_id == lock.rse_id
                    )
                    md5, bytes_, adler32 = session.execute(query).one()
                    session.flush()

                    requests = create_transfer_dict(
                        dest_rse_id=lock.rse_id,
                        request_type=RequestType.TRANSFER,
                        scope=lock.scope,
                        name=lock.name,
                        rule=rule,
                        lock=lock,
                        bytes_=bytes_,
                        md5=md5,
                        adler32=adler32,
                        ds_scope=rule.scope,
                        ds_name=rule.name,
                        copy_pin_lifetime=None,
                        activity=rule.activity,
                        session=session
                    )
                    request_core.queue_requests([requests], session=session)

            elif key == 'account':
                # Check if the account exists
                get_account(options['account'], session=session)
                # Update locks
                query = select(
                    models.ReplicaLock
                ).where(
                    models.ReplicaLock.rule_id == rule.id
                )
                counter_rses = {}
                for lock in session.execute(query).scalars().all():
                    if lock.rse_id in counter_rses:
                        counter_rses[lock.rse_id].append(lock.bytes)
                    else:
                        counter_rses[lock.rse_id] = [lock.bytes]
                for locktype in (models.ReplicaLock, models.DatasetLock):
                    query = update(
                        locktype
                    ).where(
                        locktype.rule_id == rule.id
                    ).values({
                        locktype.account: options['account']
                    })
                    session.execute(query)

                # Update counters
                for rse_id in counter_rses:
                    account_counter.decrease(
                        rse_id=rse_id,
                        account=rule.account,
                        files=len(counter_rses[rse_id]),
                        bytes_=sum(counter_rses[rse_id]),
                        session=session
                    )
                    account_counter.increase(
                        rse_id=rse_id,
                        account=options['account'],
                        files=len(counter_rses[rse_id]),
                        bytes_=sum(counter_rses[rse_id]),
                        session=session
                    )
                # Update rule
                rule.account = options['account']
                session.flush()

            elif key == 'state':
                if options.get('cancel_requests', False):
                    rule_ids_to_stuck = set()
                    query = select(
                        models.ReplicaLock
                    ).where(
                        models.ReplicaLock.rule_id == rule.id,
                        models.ReplicaLock.state == LockState.REPLICATING
                    )
                    for lock in session.execute(query).scalars().all():
                        # Set locks to stuck:
                        query = select(
                            models.ReplicaLock
                        ).where(
                            models.ReplicaLock.scope == lock.scope,
                            models.ReplicaLock.name == lock.name,
                            models.ReplicaLock.rse_id == lock.rse_id,
                            models.ReplicaLock.state == LockState.REPLICATING
                        )
                        for lock2 in session.execute(query).scalars().all():
                            lock2.state = LockState.STUCK
                            rule_ids_to_stuck.add(lock2.rule_id)
                        transfers_to_cancel = request_core.cancel_request_did(
                            scope=lock.scope,
                            name=lock.name,
                            dest_rse_id=lock.rse_id,
                            session=session
                        )
                        transfer_core.cancel_transfers(transfers_to_cancel)
                        query = select(
                            models.RSEFileAssociation
                        ).where(
                            models.RSEFileAssociation.scope == lock.scope,
                            models.RSEFileAssociation.name == lock.name,
                            models.RSEFileAssociation.rse_id == lock.rse_id
                        )
                        replica = session.execute(query).scalar_one()
                        replica.state = ReplicaState.UNAVAILABLE
                    # Set rules and DATASETLOCKS to STUCK:
                    for rid in rule_ids_to_stuck:
                        query = update(
                            models.ReplicationRule
                        ).where(
                            models.ReplicationRule.id == rid,
                            models.ReplicationRule.state != RuleState.SUSPENDED
                        ).values({
                            models.ReplicationRule.state: RuleState.STUCK
                        })
                        session.execute(query)

                        query = update(
                            models.DatasetLock
                        ).where(
                            models.DatasetLock.rule_id == rid
                        ).values({
                            models.DatasetLock.state: LockState.STUCK
                        })
                        session.execute(query)

                if options['state'].lower() == 'suspended':
                    rule.state = RuleState.SUSPENDED

                elif options['state'].lower() == 'stuck':
                    rule.state = RuleState.STUCK
                    rule.stuck_at = datetime.utcnow()
                    if not options.get('cancel_requests', False):
                        query = update(
                            models.ReplicaLock
                        ).where(
                            models.ReplicaLock.rule_id == rule.id,
                            models.ReplicaLock.state == LockState.REPLICATING
                        ).values({
                            models.ReplicaLock.state: LockState.STUCK
                        })
                        session.execute(query)

                        query = update(
                            models.DatasetLock
                        ).where(
                            models.DatasetLock.rule_id == rule_id
                        ).values({
                            models.DatasetLock.state: LockState.STUCK
                        })
                        session.execute(query)

            elif key == 'cancel_requests':
                pass

            elif key == 'priority':
                try:
                    rule.priority = options[key]
                    transfers_to_update = request_core.update_requests_priority(priority=options[key], filter_={'rule_id': rule_id}, session=session)
                    transfer_core.update_transfer_priority(transfers_to_update)
                except Exception as exc:
                    raise UnsupportedOperation('The FTS Requests are already in a final state.') from exc

            elif key == 'child_rule_id':
                # Check if the child rule has the same scope/name as the parent rule
                child_id: Optional[str] = options[key]
                if child_id is None:
                    if not rule.child_rule_id:
                        raise InputValidationError('Cannot detach child when no such relationship exists')
                    # dissolve relationship
                    rule.child_rule_id = None  # type: ignore
                    # remove expiration date
                    rule.expires_at = None  # type: ignore
                else:
                    query = select(
                        models.ReplicationRule
                    ).where(
                        models.ReplicationRule.id == child_id
                    )
                    child_rule = session.execute(query).scalar_one()
                    if rule.scope != child_rule.scope or rule.name != child_rule.name:
                        raise InputValidationError('Parent and child rule must be set on the same dataset.')
                    if rule.id == options[key]:
                        raise InputValidationError('Self-referencing parent/child-relationship.')
                    if child_rule.state != RuleState.OK:
                        rule.child_rule_id = child_id  # type: ignore

            elif key == 'meta':
                # Need to json.dump the metadata
                rule.meta = json.dumps(options[key])

            else:
                setattr(rule, key, options[key])

            insert_rule_history(rule=rule, recent=True, longterm=False, session=session)

        # `boost_rule` should run after `stuck`, so lets not include it in the loop since the arguments are unordered
        if 'boost_rule' in options:
            query = select(
                models.ReplicaLock
            ).where(
                models.ReplicaLock.rule_id == rule.id,
                models.ReplicaLock.state == LockState.STUCK
            )
            for lock in session.execute(query).scalars().all():
                lock['updated_at'] -= timedelta(days=1)

            rule['updated_at'] -= timedelta(days=1)

            insert_rule_history(
                rule,
                recent=True,
                longterm=False,
                session=session
            )

    except IntegrityError as error:
        if match('.*ORA-00001.*', str(error.args[0])) \
                or match('.*IntegrityError.*UNIQUE constraint failed.*', str(error.args[0])) \
                or match('.*1062.*Duplicate entry.*for key.*', str(error.args[0])) \
                or match('.*IntegrityError.*columns? .*not unique.*', str(error.args[0])):
            raise DuplicateRule(error.args[0]) from error
        else:
            raise error
    except NoResultFound as exc:
        raise RuleNotFound('No rule with the id %s found' % (rule_id)) from exc
    except StatementError as exc:
        raise RucioException(f"A StatementError occurred while processing rule {rule_id}") from exc


@transactional_session
def reduce_rule(
    rule_id: str,
    copies: int,
    exclude_expression: Optional[str] = None,
    *,
    session: "Session"
) -> str:
    """
    Reduce the number of copies for a rule by atomically replacing the rule.

    :param rule_id:             Rule to be reduced.
    :param copies:              Number of copies of the new rule.
    :param exclude_expression:  RSE Expression of RSEs to exclude.
    :param session:             The DB Session.
    :raises:                    RuleReplaceFailed, RuleNotFound
    """
    try:
        stmt = select(
            models.ReplicationRule
        ).where(
            models.ReplicationRule.id == rule_id
        )
        rule = session.execute(stmt).scalar_one()

        if copies >= rule.copies:
            raise RuleReplaceFailed('Copies of the new rule must be smaller than the old rule.')

        if rule.state != RuleState.OK:
            raise RuleReplaceFailed('The source rule must be in state OK.')

        if exclude_expression:
            rse_expression = '(' + rule.rse_expression + ')' + '\\' + '(' + exclude_expression + ')'
        else:
            rse_expression = rule.rse_expression

        grouping = {RuleGrouping.ALL: 'ALL', RuleGrouping.NONE: 'NONE'}.get(rule.grouping, 'DATASET')

        if rule.expires_at:
            lifetime = (rule.expires_at - datetime.utcnow()).days * 24 * 3600 + (rule.expires_at - datetime.utcnow()).seconds
        else:
            lifetime = None

        notify = {RuleNotification.YES: 'Y', RuleNotification.CLOSE: 'C', RuleNotification.PROGRESS: 'P'}.get(rule.notification, 'N')

        new_rule_id = add_rule(dids=[{'scope': rule.scope, 'name': rule.name}],
                               account=rule.account,
                               copies=copies,
                               rse_expression=rse_expression,
                               grouping=grouping,
                               weight=rule.weight,
                               lifetime=lifetime,
                               locked=rule.locked,
                               subscription_id=rule.subscription_id,
                               source_replica_expression=rule.source_replica_expression,
                               activity=rule.activity,
                               notify=notify,
                               purge_replicas=rule.purge_replicas,
                               ignore_availability=rule.ignore_availability,
                               session=session)

        session.flush()

        stmt = select(
            models.ReplicationRule
        ).where(
            models.ReplicationRule.id == new_rule_id[0]
        )
        new_rule = session.execute(stmt).scalar_one()

        if new_rule.state != RuleState.OK:
            raise RuleReplaceFailed('The replacement of the rule failed.')

        delete_rule(rule_id=rule_id,
                    session=session)

        return new_rule_id[0]

    except NoResultFound as exc:
        raise RuleNotFound('No rule with the id %s found' % (rule_id)) from exc


@transactional_session
def move_rule(
    rule_id: str,
    rse_expression: str,
    override: Optional[dict[str, Any]] = None,
    *,
    session: "Session"
) -> str:
    """
    Move a replication rule to another RSE and, once done, delete the original one.

    :param rule_id:                    Rule to be moved.
    :param rse_expression:             RSE expression of the new rule.
    :param override:                   Configurations to update for the new rule.
    :param session:                    The DB Session.
    :raises:                           RuleNotFound, RuleReplaceFailed, InvalidRSEExpression
    """
    override = override or {}

    try:
        stmt = select(
            models.ReplicationRule
        ).where(
            models.ReplicationRule.id == rule_id
        )
        rule = session.execute(stmt).scalar_one()

        if rule.child_rule_id:
            raise RuleReplaceFailed('The rule must not have a child rule.')

        grouping = {RuleGrouping.ALL: 'ALL', RuleGrouping.NONE: 'NONE'}.get(rule.grouping, 'DATASET')

        if rule.expires_at:
            lifetime = (rule.expires_at - datetime.utcnow()).days * 24 * 3600 + (rule.expires_at - datetime.utcnow()).seconds
        else:
            lifetime = None

        notify = {RuleNotification.YES: 'Y', RuleNotification.CLOSE: 'C', RuleNotification.PROGRESS: 'P'}.get(rule.notification, 'N')

        options = {
            'dids': [{'scope': rule.scope, 'name': rule.name}],
            'account': rule.account,
            'copies': rule.copies,
            'rse_expression': rse_expression,
            'grouping': grouping,
            'weight': rule.weight,
            'lifetime': lifetime,
            'locked': rule.locked,
            'subscription_id': rule.subscription_id,
            'source_replica_expression': rule.source_replica_expression,
            'activity': rule.activity,
            'notify': notify,
            'purge_replicas': rule.purge_replicas,
            'ignore_availability': rule.ignore_availability,
            'comment': rule.comments,
            'session': session,
        }

        for key in override:
            if key in ['dids', 'session']:
                raise UnsupportedOperation('Not allowed to override option %s' % key)
            elif key not in options:
                raise UnsupportedOperation('Non-valid override option %s' % key)
            else:
                options[key] = override[key]

        new_rule_id = add_rule(**options)

        session.flush()

        update_rule(rule_id=rule_id, options={'child_rule_id': new_rule_id[0], 'lifetime': 0}, session=session)

        return new_rule_id[0]

    except StatementError as exc:
        raise RucioException('Badly formatted rule id (%s)' % (rule_id)) from exc
    except NoResultFound as exc:
        raise RuleNotFound('No rule with the id %s found' % (rule_id)) from exc


@transactional_session
def re_evaluate_did(
    scope: InternalScope,
    name: str,
    rule_evaluation_action: DIDReEvaluation,
    *,
    session: "Session"
) -> None:
    """
    Re-Evaluates a did.

    :param scope:                   The scope of the did to be re-evaluated.
    :param name:                    The name of the did to be re-evaluated.
    :param rule_evaluation_action:  The Rule evaluation action.
    :param session:                 The database session in use.
    :raises:                        DataIdentifierNotFound
    """

    try:
        stmt = select(
            models.DataIdentifier
        ).where(
            and_(models.DataIdentifier.scope == scope,
                 models.DataIdentifier.name == name)
        )
        did = session.execute(stmt).scalar_one()
    except NoResultFound as exc:
        raise DataIdentifierNotFound() from exc

    if rule_evaluation_action == DIDReEvaluation.ATTACH:
        __evaluate_did_attach(did, session=session)
    else:
        __evaluate_did_detach(did, session=session)

    # Update size and length of did
    if session.bind.dialect.name == 'oracle':
        stmt = select(
            func.sum(models.DataIdentifierAssociation.bytes),
            func.count(1)
        ).with_hint(
            models.DataIdentifierAssociation, 'INDEX(CONTENTS CONTENTS_PK)', 'oracle'
        ).where(
            and_(models.DataIdentifierAssociation.scope == scope,
                 models.DataIdentifierAssociation.name == name)
        )
        for bytes_, length in session.execute(stmt):
            did.bytes = bytes_
            did.length = length

    # Add an updated_col_rep
    if did.did_type == DIDType.DATASET:
        models.UpdatedCollectionReplica(scope=scope,
                                        name=name,
                                        did_type=did.did_type).save(session=session)


@read_session
def get_updated_dids(
    total_workers: int,
    worker_number: int,
    limit: int = 100,
    blocked_dids: Sequence[tuple[str, str]] = [],
    *,
    session: "Session"
) -> list[tuple[str, InternalScope, str, DIDReEvaluation]]:
    """
    Get updated dids.

    :param total_workers:      Number of total workers.
    :param worker_number:      id of the executing worker.
    :param limit:              Maximum number of dids to return.
    :param blocked_dids:       Blocked dids to filter.
    :param session:            Database session in use.
    """
    stmt = select(
        models.UpdatedDID.id,
        models.UpdatedDID.scope,
        models.UpdatedDID.name,
        models.UpdatedDID.rule_evaluation_action
    )
    stmt = filter_thread_work(session=session, query=stmt, total_threads=total_workers, thread_id=worker_number, hash_variable='name')

    # Remove blocked dids from query, but only do the first 30 ones, not to overload the query
    if blocked_dids:
        chunk = list(chunks(blocked_dids, 30))[0]
        stmt = stmt.where(tuple_(models.UpdatedDID.scope, models.UpdatedDID.name).notin_(chunk))

    if limit:
        fetched_dids = session.execute(stmt.order_by(models.UpdatedDID.created_at).limit(limit)).all()
        filtered_dids = [did._tuple() for did in fetched_dids if (did.scope, did.name) not in blocked_dids]
        if len(fetched_dids) == limit and not filtered_dids:
            return get_updated_dids(total_workers=total_workers,
                                    worker_number=worker_number,
                                    limit=None,
                                    blocked_dids=blocked_dids,
                                    session=session)
        else:
            return filtered_dids
    else:
        return [did._tuple() for did in session.execute(stmt.order_by(models.UpdatedDID.created_at)).all() if (did.scope, did.name) not in blocked_dids]


@read_session
def get_rules_beyond_eol(
    date_check: datetime,
    worker_number: int,
    total_workers: int, *,
    session: "Session"
) -> list[tuple[InternalScope,
                str,
                str,
                bool,
                str,
                Optional[datetime],
                Optional[datetime],
                InternalAccount]]:
    """
    Get rules which have eol_at before a certain date.

    :param date_check:         The reference date that should be compared to eol_at.
    :param worker_number:      id of the executing worker.
    :param total_workers:      Number of total workers.
    :param session:            Database session in use.
    """
    stmt = select(
        models.ReplicationRule.scope,
        models.ReplicationRule.name,
        models.ReplicationRule.rse_expression,
        models.ReplicationRule.locked,
        models.ReplicationRule.id,
        models.ReplicationRule.eol_at,
        models.ReplicationRule.expires_at,
        models.ReplicationRule.account
    ).where(
        models.ReplicationRule.eol_at < date_check
    )

    stmt = filter_thread_work(session=session, query=stmt, total_threads=total_workers, thread_id=worker_number, hash_variable='name')
    return [row._tuple() for row in session.execute(stmt).all()]


@read_session
def get_expired_rules(
    total_workers: int,
    worker_number: int,
    limit: int = 100,
    blocked_rules: Sequence[str] = [],
    *,
    session: "Session"
) -> list[tuple[str, str]]:
    """
    Get expired rules.

    :param total_workers:      Number of total workers.
    :param worker_number:      id of the executing worker.
    :param limit:              Maximum number of rules to return.
    :param blocked_rules:      List of blocked rules.
    :param session:            Database session in use.
    """

    stmt = select(
        models.ReplicationRule.id,
        models.ReplicationRule.rse_expression
    ).with_hint(
        models.ReplicationRule, 'INDEX(RULES RULES_EXPIRES_AT_IDX)', 'oracle'
    ).where(
        and_(models.ReplicationRule.expires_at < datetime.utcnow(),
             models.ReplicationRule.locked == false(),
             models.ReplicationRule.child_rule_id == null())
    ).order_by(
        models.ReplicationRule.expires_at
    )
    stmt = filter_thread_work(session=session, query=stmt, total_threads=total_workers, thread_id=worker_number, hash_variable='name')

    if limit:
        stmt = stmt.limit(limit)
        result = session.execute(stmt).all()
        filtered_rules = [rule._tuple() for rule in result if rule.id not in blocked_rules]
        if len(result) == limit and not filtered_rules:
            return get_expired_rules(total_workers=total_workers,
                                     worker_number=worker_number,
                                     limit=None,
                                     blocked_rules=blocked_rules,
                                     session=session)
        else:
            return filtered_rules
    else:
        return [rule._tuple() for rule in session.execute(stmt).all() if rule.id not in blocked_rules]


@read_session
def get_injected_rules(
    total_workers: int,
    worker_number: int,
    limit: int = 100,
    blocked_rules: Sequence[str] = [],
    *,
    session: "Session"
) -> list[str]:
    """
    Get rules to be injected.

    :param total_workers:      Number of total workers.
    :param worker_number:      id of the executing worker.
    :param limit:              Maximum number of rules to return.
    :param blocked_rules:      Blocked rules not to include.
    :param session:            Database session in use.
    """

    stmt = select(
        models.ReplicationRule.id
    ).with_hint(
        models.ReplicationRule, 'INDEX(RULES RULES_STATE_IDX)', 'oracle'
    ).where(
        and_(models.ReplicationRule.state == RuleState.INJECT,
             models.ReplicationRule.created_at <= datetime.utcnow())
    ).order_by(
        models.ReplicationRule.created_at
    )
    stmt = filter_thread_work(session=session, query=stmt, total_threads=total_workers, thread_id=worker_number, hash_variable='name')

    if limit:
        stmt = stmt.limit(limit)
        result = session.execute(stmt).scalars().all()
        filtered_rules = [rule for rule in result if rule not in blocked_rules]
        if len(result) == limit and not filtered_rules:
            return get_injected_rules(total_workers=total_workers,
                                      worker_number=worker_number,
                                      limit=None,
                                      blocked_rules=blocked_rules,
                                      session=session)
        else:
            return filtered_rules
    else:
        return [rule for rule in session.execute(stmt).scalars().all() if rule not in blocked_rules]


@read_session
def get_stuck_rules(
    total_workers: int,
    worker_number: int,
    delta: int = 600,
    limit: int = 10,
    blocked_rules: Sequence[str] = [],
    *,
    session: "Session"
) -> list[str]:
    """
    Get stuck rules.

    :param total_workers:      Number of total workers.
    :param worker_number:      id of the executing worker.
    :param delta:              Delta in seconds to select rules in.
    :param limit:              Maximum number of rules to select.
    :param blocked_rules:      Blocked rules to filter out.
    :param session:            Database session in use.
    """
    stmt = select(
        models.ReplicationRule.id
    ).with_hint(
        models.ReplicationRule, 'INDEX(RULES RULES_STATE_IDX)', 'oracle'
    ).where(
        and_(models.ReplicationRule.state == RuleState.STUCK,
             models.ReplicationRule.updated_at < datetime.utcnow() - timedelta(seconds=delta),
             or_(models.ReplicationRule.expires_at == null(),
                 models.ReplicationRule.expires_at > datetime.utcnow(),
                 models.ReplicationRule.locked == true()))
    ).order_by(
        models.ReplicationRule.updated_at
    )
    stmt = filter_thread_work(session=session, query=stmt, total_threads=total_workers, thread_id=worker_number, hash_variable='name')

    if limit:
        stmt = stmt.limit(limit)
        result = session.execute(stmt).scalars().all()
        filtered_rules = [rule for rule in result if rule not in blocked_rules]
        if len(result) == limit and not filtered_rules:
            return get_stuck_rules(total_workers=total_workers,
                                   worker_number=worker_number,
                                   delta=delta,
                                   limit=None,
                                   blocked_rules=blocked_rules,
                                   session=session)
        else:
            return filtered_rules
    else:
        return [rule for rule in session.execute(stmt).scalars().all() if rule not in blocked_rules]


@transactional_session
def delete_updated_did(
    id_: str,
    *,
    session: "Session"
) -> None:
    """
    Delete an updated_did by id.

    :param id_:                      Id of the row not to delete.
    :param session:                 The database session in use.
    """
    stmt = delete(
        models.UpdatedDID
    ).where(
        models.UpdatedDID.id == id_
    )
    session.execute(stmt)


@transactional_session
def update_rules_for_lost_replica(
    scope: InternalScope,
    name: str,
    rse_id: str,
    nowait: bool = False,
    *,
    session: "Session",
    logger: LoggerFunction = logging.log
) -> None:
    """
    Update rules if a file replica is lost.

    :param scope:          Scope of the replica.
    :param name:           Name of the replica.
    :param rse_id:         RSE id of the replica.
    :param nowait:         Nowait parameter for the FOR UPDATE statement.
    :param session:        The database session in use.
    :param logger:         Optional decorated logger that can be passed from the calling daemons or servers.
    """

    stmt = select(
        models.ReplicaLock
    ).where(
        and_(models.ReplicaLock.scope == scope,
             models.ReplicaLock.name == name,
             models.ReplicaLock.rse_id == rse_id)
    ).with_for_update(
        nowait=nowait
    )
    locks = session.execute(stmt).scalars().all()

    stmt = select(
        models.RSEFileAssociation
    ).where(
        and_(models.RSEFileAssociation.scope == scope,
             models.RSEFileAssociation.name == name,
             models.RSEFileAssociation.rse_id == rse_id)
    ).with_for_update(
        nowait=nowait
    )
    replica = session.execute(stmt).scalar_one()

    stmt = select(
        models.Request
    ).where(
        and_(models.Request.scope == scope,
             models.Request.name == name,
             models.Request.dest_rse_id == rse_id)
    ).with_for_update(
        nowait=nowait
    )
    requests = session.execute(stmt).scalars().all()

    rse = get_rse_name(rse_id, session=session)

    datasets = []
    parent_dids = rucio.core.did.list_parent_dids(scope=scope, name=name, session=session)
    for parent in parent_dids:
        if {'name': parent['name'], 'scope': parent['scope']} not in datasets:
            datasets.append({'name': parent['name'], 'scope': parent['scope']})

    for request in requests:
        session.delete(request)

    for lock in locks:
        stmt = select(
            models.ReplicationRule
        ).where(
            models.ReplicationRule.id == lock.rule_id
        ).with_for_update(
            nowait=nowait
        )
        rule = session.execute(stmt).scalar_one()
        rule_state_before = rule.state
        replica.lock_cnt -= 1
        if lock.state == LockState.OK:
            rule.locks_ok_cnt -= 1
        elif lock.state == LockState.REPLICATING:
            rule.locks_replicating_cnt -= 1
        elif lock.state == LockState.STUCK:
            rule.locks_stuck_cnt -= 1
        account_counter.decrease(rse_id=rse_id, account=rule.account, files=1, bytes_=lock.bytes, session=session)
        if rule.state == RuleState.SUSPENDED:
            pass
        elif rule.state == RuleState.STUCK:
            pass
        elif rule.locks_replicating_cnt == 0 and rule.locks_stuck_cnt == 0:
            rule.state = RuleState.OK
            if rule.grouping != RuleGrouping.NONE:
                stmt = update(
                    models.DatasetLock
                ).where(
                    models.DatasetLock.rule_id == rule.id
                ).values({
                    models.DatasetLock.state: LockState.OK
                })
                session.execute(stmt)
                session.flush()
            if rule_state_before != RuleState.OK:
                generate_rule_notifications(rule=rule, session=session)
                generate_email_for_rule_ok_notification(rule=rule, session=session)
            # Try to release potential parent rules
            release_parent_rule(child_rule_id=rule.id, session=session)
        # Insert rule history
        insert_rule_history(rule=rule, recent=True, longterm=False, session=session)

        session.delete(lock)

    if replica.lock_cnt != 0:
        logger(logging.ERROR, 'Replica for did %s:%s with lock_cnt = %s. This should never happen. Update lock_cnt', scope, name, replica.lock_cnt)
        replica.lock_cnt = 0

    replica.tombstone = OBSOLETE
    replica.state = ReplicaState.UNAVAILABLE
    stmt = update(
        models.DataIdentifier
    ).where(
        and_(models.DataIdentifier.scope == scope,
             models.DataIdentifier.name == name)
    ).values({
        models.DataIdentifier.availability: DIDAvailability.LOST
    })
    session.execute(stmt)

    stmt = update(
        models.BadReplicas
    ).where(
        and_(models.BadReplicas.scope == scope,
             models.BadReplicas.name == name,
             models.BadReplicas.rse_id == rse_id,
             models.BadReplicas.state == BadFilesStatus.BAD)
    ).values({
        models.BadReplicas.state: BadFilesStatus.LOST,
        models.BadReplicas.updated_at: datetime.utcnow()
    })
    session.execute(stmt)
    for dts in datasets:
        logger(logging.INFO, 'File %s:%s bad at site %s is completely lost from dataset %s:%s. Will be marked as LOST and detached', scope, name, rse, dts['scope'], dts['name'])
        rucio.core.did.detach_dids(scope=dts['scope'], name=dts['name'], dids=[{'scope': scope, 'name': name}], session=session)

        message = {'scope': scope.external,
                   'name': name,
                   'dataset_name': dts['name'],
                   'dataset_scope': dts['scope'].external}
        if scope.vo != 'def':
            message['vo'] = scope.vo

        add_message('LOST', message, session=session)


@transactional_session
def update_rules_for_bad_replica(
    scope: InternalScope,
    name: str,
    rse_id: str,
    nowait: bool = False,
    *,
    session: "Session",
    logger: LoggerFunction = logging.log
) -> None:
    """
    Update rules if a file replica is bad and has to be recreated.

    :param scope:          Scope of the replica.
    :param name:           Name of the replica.
    :param rse_id:         RSE id of the replica.
    :param nowait:         Nowait parameter for the FOR UPDATE statement.
    :param session:        The database session in use.
    :param logger:         Optional decorated logger that can be passed from the calling daemons or servers.
    """
    stmt = select(
        models.ReplicaLock
    ).where(
        and_(models.ReplicaLock.scope == scope,
             models.ReplicaLock.name == name,
             models.ReplicaLock.rse_id == rse_id)
    ).with_for_update(
        nowait=nowait
    )
    locks = session.execute(stmt).scalars().all()

    stmt = select(
        models.RSEFileAssociation
    ).where(
        and_(models.RSEFileAssociation.scope == scope,
             models.RSEFileAssociation.name == name,
             models.RSEFileAssociation.rse_id == rse_id)
    ).with_for_update(
        nowait=nowait
    )
    replica = session.execute(stmt).scalar_one()

    nlock = 0
    datasets = []
    for lock in locks:
        nlock += 1
        stmt = select(
            models.ReplicationRule
        ).where(
            models.ReplicationRule.id == lock.rule_id
        ).with_for_update(
            nowait=nowait
        )
        rule = session.execute(stmt).scalar_one()
        # If source replica expression exists, we remove it
        if rule.source_replica_expression:
            rule.source_replica_expression = None
        # Get the affected datasets
        ds_scope = rule.scope
        ds_name = rule.name
        dataset = '%s:%s' % (ds_scope, ds_name)
        if dataset not in datasets:
            datasets.append(dataset)
            logger(logging.INFO, 'Recovering file %s:%s from dataset %s:%s at site %s', scope, name, ds_scope, ds_name, get_rse_name(rse_id=rse_id, session=session))
        # Insert a new row in the UpdateCollectionReplica table
        models.UpdatedCollectionReplica(scope=ds_scope,
                                        name=ds_name,
                                        did_type=rule.did_type,
                                        rse_id=lock.rse_id).save(flush=False, session=session)
        # Set the lock counters
        if lock.state == LockState.OK:
            rule.locks_ok_cnt -= 1
        elif lock.state == LockState.REPLICATING:
            rule.locks_replicating_cnt -= 1
        elif lock.state == LockState.STUCK:
            rule.locks_stuck_cnt -= 1
        rule.locks_replicating_cnt += 1
        # Generate the request
        try:
            request_core.get_request_by_did(scope, name, rse_id, session=session)
        except RequestNotFound:
            bytes_ = replica.bytes
            md5 = replica.md5
            adler32 = replica.adler32
            request_core.queue_requests(requests=[create_transfer_dict(dest_rse_id=rse_id,
                                                                       request_type=RequestType.TRANSFER,
                                                                       scope=scope, name=name, rule=rule, lock=lock, bytes_=bytes_, md5=md5, adler32=adler32,
                                                                       ds_scope=ds_scope, ds_name=ds_name, copy_pin_lifetime=None, activity='Recovery', session=session)], session=session)
        lock.state = LockState.REPLICATING
        if rule.state == RuleState.SUSPENDED:
            pass
        elif rule.state == RuleState.STUCK:
            pass
        else:
            rule.state = RuleState.REPLICATING
            if rule.grouping != RuleGrouping.NONE:
                stmt = update(
                    models.DatasetLock
                ).where(
                    models.DatasetLock.rule_id == rule.id
                ).values({
                    models.DatasetLock.state: LockState.REPLICATING
                })
                session.execute(stmt)
        # Insert rule history
        insert_rule_history(rule=rule, recent=True, longterm=False, session=session)
    if nlock:
        stmt = update(
            models.RSEFileAssociation
        ).where(
            and_(models.RSEFileAssociation.scope == scope,
                 models.RSEFileAssociation.name == name,
                 models.RSEFileAssociation.rse_id == rse_id)
        ).values({
            models.RSEFileAssociation.state: ReplicaState.COPYING
        })
        session.execute(stmt)
    else:
        logger(logging.INFO, 'File %s:%s at site %s has no locks. Will be deleted now.', scope, name, get_rse_name(rse_id=rse_id, session=session))
        tombstone = OBSOLETE
        stmt = update(
            models.RSEFileAssociation
        ).where(
            and_(models.RSEFileAssociation.scope == scope,
                 models.RSEFileAssociation.name == name,
                 models.RSEFileAssociation.rse_id == rse_id)
        ).values({
            models.RSEFileAssociation.state: ReplicaState.UNAVAILABLE,
            models.RSEFileAssociation.tombstone: tombstone
        })
        session.execute(stmt)


@transactional_session
def generate_rule_notifications(
    rule: models.ReplicationRule,
    replicating_locks_before: Optional[int] = None,
    *,
    session: "Session"
) -> None:
    """
    Generate (If necessary) a callback for a rule (DATASETLOCK_OK, RULE_OK, DATASETLOCK_PROGRESS)

    :param rule:                       The rule object.
    :param replicating_locks_before:   Amount of replicating locks before the current state change.
    :param session:                    The Database session
    """

    session.flush()
    total_locks = rule.locks_replicating_cnt + rule.locks_ok_cnt

    if rule.state == RuleState.OK:
        # Only notify when rule is in state OK

        # RULE_OK RULE_PROGRESS NOTIFICATIONS:
        if rule.notification == RuleNotification.YES:
            payload = {'scope': rule.scope.external,
                       'name': rule.name,
                       'rule_id': rule.id}
            if rule.scope.vo != 'def':
                payload['vo'] = rule.scope.vo

            add_message(event_type='RULE_OK', payload=payload, session=session)

        elif rule.notification in [RuleNotification.CLOSE, RuleNotification.PROGRESS]:
            try:
                did = rucio.core.did.get_did(scope=rule.scope, name=rule.name, session=session)
                if not did['open']:
                    payload = {'scope': rule.scope.external,
                               'name': rule.name,
                               'rule_id': rule.id}
                    if rule.scope.vo != 'def':
                        payload['vo'] = rule.scope.vo

                    add_message(event_type='RULE_OK', payload=payload, session=session)

                    if rule.notification == RuleNotification.PROGRESS:
                        payload = {'scope': rule.scope.external,
                                   'name': rule.name,
                                   'rule_id': rule.id,
                                   'progress': __progress_class(rule.locks_replicating_cnt, total_locks)}
                        if rule.scope.vo != 'def':
                            payload['vo'] = rule.scope.vo

                        add_message(event_type='RULE_PROGRESS', payload=payload, session=session)

            except DataIdentifierNotFound:
                pass

        # DATASETLOCK_OK NOTIFICATIONS:
        if rule.grouping != RuleGrouping.NONE:
            # Only send DATASETLOCK_OK callbacks for ALL/DATASET grouped rules
            if rule.notification == RuleNotification.YES:
                stmt = select(
                    models.DatasetLock
                ).where(
                    models.DatasetLock.rule_id == rule.id
                )
                dataset_locks = session.execute(stmt).scalars().all()
                for dataset_lock in dataset_locks:
                    payload = {'scope': dataset_lock.scope.external,
                               'name': dataset_lock.name,
                               'rse': get_rse_name(rse_id=dataset_lock.rse_id, session=session),
                               'rse_id': dataset_lock.rse_id,
                               'rule_id': rule.id}
                    if dataset_lock.scope.vo != 'def':
                        payload['vo'] = dataset_lock.scope.vo

                    add_message(event_type='DATASETLOCK_OK', payload=payload, session=session)

            elif rule.notification == RuleNotification.CLOSE:
                stmt = select(
                    models.DatasetLock
                ).where(
                    models.DatasetLock.rule_id == rule.id
                )
                dataset_locks = session.execute(stmt).scalars().all()
                for dataset_lock in dataset_locks:
                    try:
                        did = rucio.core.did.get_did(scope=dataset_lock.scope, name=dataset_lock.name, session=session)
                        if not did['open']:
                            if did['length'] is None:
                                return
                            if did['length'] * rule.copies == rule.locks_ok_cnt:
                                payload = {'scope': dataset_lock.scope.external,
                                           'name': dataset_lock.name,
                                           'rse': get_rse_name(rse_id=dataset_lock.rse_id, session=session),
                                           'rse_id': dataset_lock.rse_id,
                                           'rule_id': rule.id}
                                if dataset_lock.scope.vo != 'def':
                                    payload['vo'] = dataset_lock.scope.vo

                                add_message(event_type='DATASETLOCK_OK', payload=payload, session=session)

                    except DataIdentifierNotFound:
                        pass

    elif rule.state == RuleState.REPLICATING and rule.notification == RuleNotification.PROGRESS and replicating_locks_before:
        # For RuleNotification PROGRESS rules, also notifiy when REPLICATING thresholds are passed
        if __progress_class(replicating_locks_before, total_locks) != __progress_class(rule.locks_replicating_cnt, total_locks):
            try:
                did = rucio.core.did.get_did(scope=rule.scope, name=rule.name, session=session)
                if not did['open']:
                    payload = {'scope': rule.scope.external,
                               'name': rule.name,
                               'rule_id': rule.id,
                               'progress': __progress_class(rule.locks_replicating_cnt, total_locks)}
                    if rule.scope.vo != 'def':
                        payload['vo'] = rule.scope.vo

                    add_message(event_type='RULE_PROGRESS', payload=payload, session=session)

            except DataIdentifierNotFound:
                pass


@transactional_session
def generate_email_for_rule_ok_notification(
    rule: models.ReplicationRule,
    *,
    session: "Session",
    logger: LoggerFunction = logging.log
) -> None:
    """
    Generate (If necessary) an eMail for a rule with notification mode Y.

    :param rule:     The rule object.
    :param session:  The Database session
    :param logger:   Optional decorated logger that can be passed from the calling daemons or servers.
    """

    session.flush()

    if rule.state == RuleState.OK and rule.notification == RuleNotification.YES:
        try:
            template_path = '%s/rule_ok_notification.tmpl' % config_get('common', 'mailtemplatedir')
        except NoOptionError as ex:
            logger(logging.ERROR, "Missing configuration option 'mailtemplatedir'.", exc_info=ex)
            return

        try:
            with open(template_path, 'r') as templatefile:
                template = Template(templatefile.read())
        except IOError as ex:
            logger(logging.ERROR, "Couldn't open file '%s'", template_path, exc_info=ex)
            return

        email = get_account(account=rule.account, session=session).email
        if not email:
            logger(logging.INFO, 'No email associated with rule ID %s.', rule.id)
            return

        try:
            email_body = template.safe_substitute({'rule_id': str(rule.id),
                                                   'created_at': str(rule.created_at),
                                                   'expires_at': str(rule.expires_at),
                                                   'rse_expression': rule.rse_expression,
                                                   'comment': rule.comments,
                                                   'scope': rule.scope.external,
                                                   'name': rule.name,
                                                   'did_type': rule.did_type})
        except ValueError as ex:
            logger(logging.ERROR, "Invalid mail template.", exc_info=ex)
            return

        add_message(event_type='email',
                    payload={'body': email_body,
                             'to': [email],
                             'subject': '[RUCIO] Replication rule %s has been succesfully transferred' % (str(rule.id))},
                    session=session)


@transactional_session
def insert_rule_history(
    rule: models.ReplicationRule,
    recent: bool = True,
    longterm: bool = False,
    *,
    session: "Session"
) -> None:
    """
    Insert rule history to recent/longterm history.

    :param rule:      The rule object.
    :param recent:    Insert to recent table.
    :param longterm:  Insert to longterm table.
    :param session:   The Database session.
    """
    if recent:
        models.ReplicationRuleHistoryRecent(id=rule.id, subscription_id=rule.subscription_id, account=rule.account, scope=rule.scope, name=rule.name,
                                            did_type=rule.did_type, state=rule.state, error=rule.error, rse_expression=rule.rse_expression, copies=rule.copies,
                                            expires_at=rule.expires_at, weight=rule.weight, locked=rule.locked, locks_ok_cnt=rule.locks_ok_cnt,
                                            locks_replicating_cnt=rule.locks_replicating_cnt, locks_stuck_cnt=rule.locks_stuck_cnt, source_replica_expression=rule.source_replica_expression,
                                            activity=rule.activity, grouping=rule.grouping, notification=rule.notification, stuck_at=rule.stuck_at, purge_replicas=rule.purge_replicas,
                                            ignore_availability=rule.ignore_availability, ignore_account_limit=rule.ignore_account_limit, comments=rule.comments, created_at=rule.created_at,
                                            updated_at=rule.updated_at, child_rule_id=rule.child_rule_id, eol_at=rule.eol_at,
                                            split_container=rule.split_container, meta=rule.meta).save(session=session)
    if longterm:
        models.ReplicationRuleHistory(id=rule.id, subscription_id=rule.subscription_id, account=rule.account, scope=rule.scope, name=rule.name,
                                      did_type=rule.did_type, state=rule.state, error=rule.error, rse_expression=rule.rse_expression, copies=rule.copies,
                                      expires_at=rule.expires_at, weight=rule.weight, locked=rule.locked, locks_ok_cnt=rule.locks_ok_cnt,
                                      locks_replicating_cnt=rule.locks_replicating_cnt, locks_stuck_cnt=rule.locks_stuck_cnt, source_replica_expression=rule.source_replica_expression,
                                      activity=rule.activity, grouping=rule.grouping, notification=rule.notification, stuck_at=rule.stuck_at, purge_replicas=rule.purge_replicas,
                                      ignore_availability=rule.ignore_availability, ignore_account_limit=rule.ignore_account_limit, comments=rule.comments, created_at=rule.created_at,
                                      updated_at=rule.updated_at, child_rule_id=rule.child_rule_id, eol_at=rule.eol_at,
                                      split_container=rule.split_container, meta=rule.meta).save(session=session)


@transactional_session
def approve_rule(
    rule_id: str,
    approver: str = '',
    notify_approvers: bool = True,
    *,
    session: "Session"
) -> None:
    """
    Approve a specific replication rule.

    :param rule_id:           The rule_id to approve.
    :param approver:          The account which is approving the rule.
    :param notify_approvers:  Notify the other approvers.
    :param session:           The database session in use.
    :raises:                  RuleNotFound if no Rule can be found.
    """

    try:
        stmt = select(
            models.ReplicationRule
        ).where(
            models.ReplicationRule.id == rule_id
        )
        rule = session.execute(stmt).scalar_one()
        if rule.state == RuleState.WAITING_APPROVAL:
            rule.ignore_account_limit = True
            rule.state = RuleState.INJECT
            if approver:
                approver_email = get_account(account=approver, session=session).email
                if approver_email:
                    approver = '%s (%s)' % (approver, approver_email)
            else:
                approver = 'AUTOMATIC'
            with open('%s/rule_approved_user.tmpl' % config_get('common', 'mailtemplatedir'), 'r') as templatefile:
                template = Template(templatefile.read())
                email = get_account(account=rule.account, session=session).email
                if email:
                    text = template.safe_substitute({'rule_id': str(rule.id),
                                                     'expires_at': str(rule.expires_at),
                                                     'rse_expression': rule.rse_expression,
                                                     'comment': rule.comments,
                                                     'scope': rule.scope.external,
                                                     'name': rule.name,
                                                     'did_type': rule.did_type,
                                                     'approver': approver})
                    add_message(event_type='email',
                                payload={'body': text,
                                         'to': [email],
                                         'subject': '[RUCIO] Replication rule %s has been approved' % (str(rule.id))},
                                session=session)
            # Also notify the other approvers
            if notify_approvers:
                with open('%s/rule_approved_admin.tmpl' % config_get('common', 'mailtemplatedir'), 'r') as templatefile:
                    template = Template(templatefile.read())
                text = template.safe_substitute({'rule_id': str(rule.id),
                                                 'approver': approver})
                vo = rule.account.vo
                recipients = _create_recipients_list(rse_expression=rule.rse_expression, filter_={'vo': vo}, session=session)
                for recipient in recipients:
                    add_message(event_type='email',
                                payload={'body': text,
                                         'to': [recipient[0]],
                                         'subject': 'Re: [RUCIO] Request to approve replication rule %s' % (str(rule.id))},
                                session=session)
    except NoResultFound as exc:
        raise RuleNotFound('No rule with the id %s found' % (rule_id)) from exc
    except StatementError as exc:
        raise RucioException('Badly formatted rule id (%s)' % (rule_id)) from exc


@transactional_session
def deny_rule(
    rule_id: str,
    approver: str = '',
    reason: Optional[str] = None,
    *,
    session: "Session"
) -> None:
    """
    Deny a specific replication rule.

    :param rule_id:   The rule_id to approve.
    :param approver:  The account which is denying the rule.
    :param reason:    The reason why the rule was denied.
    :param session:   The database session in use.
    :raises:          RuleNotFound if no Rule can be found.
    """

    try:
        stmt = select(
            models.ReplicationRule
        ).where(
            models.ReplicationRule.id == rule_id
        )
        rule = session.execute(stmt).scalar_one()
        if rule.state == RuleState.WAITING_APPROVAL:
            with open('%s/rule_denied_user.tmpl' % config_get('common', 'mailtemplatedir'), 'r') as templatefile:
                template = Template(templatefile.read())
            email = get_account(account=rule.account, session=session).email
            if approver:
                approver_email = get_account(account=approver, session=session).email
                if approver_email:
                    approver = '%s (%s)' % (approver, approver_email)
            else:
                approver = 'AUTOMATIC'
            if email:
                email_body = template.safe_substitute({'rule_id': str(rule.id),
                                                       'rse_expression': rule.rse_expression,
                                                       'comment': rule.comments,
                                                       'scope': rule.scope.external,
                                                       'name': rule.name,
                                                       'did_type': rule.did_type,
                                                       'approver': approver,
                                                       'reason': reason})
                add_message(event_type='email',
                            payload={'body': email_body,
                                     'to': [email],
                                     'subject': '[RUCIO] Replication rule %s has been denied' % (str(rule.id))},
                            session=session)
            delete_rule(rule_id=rule_id, ignore_rule_lock=True, session=session)
            # Also notify the other approvers
            with open('%s/rule_denied_admin.tmpl' % config_get('common', 'mailtemplatedir'), 'r') as templatefile:
                template = Template(templatefile.read())
            email_body = template.safe_substitute({'rule_id': str(rule.id),
                                                   'approver': approver,
                                                   'reason': reason})
            vo = rule.account.vo
            recipients = _create_recipients_list(rse_expression=rule.rse_expression, filter_={'vo': vo}, session=session)
            for recipient in recipients:
                add_message(event_type='email',
                            payload={'body': email_body,
                                     'to': [recipient[0]],
                                     'subject': 'Re: [RUCIO] Request to approve replication rule %s' % (str(rule.id))},
                            session=session)
    except NoResultFound as exc:
        raise RuleNotFound('No rule with the id %s found' % rule_id) from exc
    except StatementError as exc:
        raise RucioException('Badly formatted rule id (%s)' % rule_id) from exc


@transactional_session
def examine_rule(
    rule_id: str,
    *,
    session: "Session"
) -> dict[str, Any]:
    """
    Examine a replication rule for transfer errors.

    :param rule_id:            Replication rule id
    :param session:            Session of the db.
    :returns:                  Dictionary of informations
    """
    result = {'rule_error': None,
              'transfers': []}

    try:
        stmt = select(
            models.ReplicationRule
        ).where(
            models.ReplicationRule.id == rule_id
        )
        rule = session.execute(stmt).scalar_one()
        if rule.state == RuleState.OK:
            result['rule_error'] = 'This replication rule is OK'
        elif rule.state == RuleState.REPLICATING:
            result['rule_error'] = 'This replication rule is currently REPLICATING'
        elif rule.state == RuleState.SUSPENDED:
            result['rule_error'] = 'This replication rule is SUSPENDED'
        else:
            result['rule_error'] = rule.error
            # Get the stuck locks
            stmt = select(
                models.ReplicaLock
            ).where(
                and_(models.ReplicaLock.rule_id == rule_id,
                     models.ReplicaLock.state == LockState.STUCK)
            )
            stuck_locks = session.execute(stmt).scalars().all()
            for lock in stuck_locks:
                # Get the count of requests in the request_history for each lock
                stmt = select(
                    models.RequestHistory
                ).where(
                    and_(models.RequestHistory.scope == lock.scope,
                         models.RequestHistory.name == lock.name,
                         models.RequestHistory.dest_rse_id == lock.rse_id)
                ).order_by(
                    desc(models.RequestHistory.created_at)
                )
                transfers = session.execute(stmt).scalars().all()
                transfer_cnt = len(transfers)
                # Get the error of the last request that has been tried and also the SOURCE used for the last request
                last_error, last_source, last_time, sources = None, None, None, []
                if transfers:
                    last_request = transfers[0]
                    last_error = last_request.state
                    last_time = last_request.created_at
                    last_source = None if last_request.source_rse_id is None else get_rse_name(rse_id=last_request.source_rse_id, session=session)
                    stmt = select(
                        models.RSEFileAssociation
                    ).where(
                        and_(models.RSEFileAssociation.scope == lock.scope,
                             models.RSEFileAssociation.name == lock.name,
                             models.RSEFileAssociation.state == ReplicaState.AVAILABLE)
                    )
                    available_replicas = session.execute(stmt).scalars().all()

                    for replica in available_replicas:
                        sources.append((get_rse_name(rse_id=replica.rse_id, session=session),
                                        True if get_rse(rse_id=replica.rse_id, session=session)['availability_read'] else False))

                result['transfers'].append({'scope': lock.scope,
                                            'name': lock.name,
                                            'rse_id': lock.rse_id,
                                            'rse': get_rse_name(rse_id=lock.rse_id, session=session),
                                            'attempts': transfer_cnt,
                                            'last_error': str(last_error),
                                            'last_source': last_source,
                                            'sources': sources,
                                            'last_time': last_time})
        return result
    except NoResultFound as exc:
        raise RuleNotFound('No rule with the id %s found' % (rule_id)) from exc
    except StatementError as exc:
        raise RucioException('Badly formatted rule id (%s)' % (rule_id)) from exc


@transactional_session
def get_evaluation_backlog(
    expiration_time: int = 600,
    *,
    session: "Session"
) -> tuple[int, datetime]:
    """
    Counts the number of entries in the rule evaluation backlog.
    (Number of files to be evaluated)

    :returns:     Tuple (Count, Datetime of oldest entry)
    """

    cached_backlog: Union[NoValue, tuple[int, datetime]] = REGION.get('rule_evaluation_backlog', expiration_time=expiration_time)
    if isinstance(cached_backlog, NoValue):
        stmt = select(
            func.count(models.UpdatedDID.created_at),
            func.min(models.UpdatedDID.created_at)
        )
        result = session.execute(stmt).one()._tuple()
        REGION.set('rule_evaluation_backlog', result)
        return result
    return cached_backlog


@transactional_session
def release_parent_rule(
    child_rule_id: str,
    remove_parent_expiration: bool = False,
    *,
    session: "Session"
) -> None:
    """
    Release a potential parent rule, because the child_rule is OK.

    :param child_rule_id:             The child rule id.
    :param remove_parant_expiration:  If true, removes the expiration of the parent rule.
    :param session:                   The Database session
    """

    session.flush()

    stmt = select(
        models.ReplicationRule
    ).with_hint(
        models.ReplicationRule, 'INDEX(RULES RULES_CHILD_RULE_ID_IDX)', 'oracle'
    ).where(
        models.ReplicationRule.child_rule_id == child_rule_id
    )
    parent_rules = session.execute(stmt).scalars().all()
    for rule in parent_rules:
        if remove_parent_expiration:
            rule.expires_at = None
        rule.child_rule_id = None
        insert_rule_history(rule=rule, recent=True, longterm=False, session=session)


@stream_session
def list_rules_for_rse_decommissioning(
    rse_id: str,
    *,
    session: "Session"
) -> Iterator[dict[str, Any]]:
    """Return a generator of rules at the RSE that is being decommissioned.

    Decommissioning of an RSE involves deleting or moving away all rules that are
    locking the replicas that exist at the RSE. The rules can be enforcing
    dataset-level and/or file-level locks. Because rules are defined in terms of
    RSE expressions, we need to first identify the locks with the RSE id and make
    the list of rules that are enforcing such locks.
    This function has two yield statements corresponding to the two types
    (dataset-level and file-level) of locks. To avoid listing duplicates, the
    rules identified through the dataset-level locks are excluded from the
    second query using file-level locks.

    :param rse_id: Id of the RSE being decommissioned.
    :param session: The database session in use.
    :returns: A generator that yields rule dictionaries.
    """
    # Get rules with dataset locks first.
    query_rules_from_dataset_locks = select(
        models.ReplicationRule
    ).distinct(
    ).join_from(
        models.DatasetLock,
        models.ReplicationRule,
        models.DatasetLock.rule_id == models.ReplicationRule.id
    ).where(
        models.DatasetLock.rse_id == rse_id
    )

    for rule in session.execute(query_rules_from_dataset_locks).yield_per(5).scalars():
        yield rule.to_dict()

    # Make a subquery from the previous query to be excluded from the next query
    dataset_rule_ids = query_rules_from_dataset_locks.with_only_columns(models.ReplicationRule.id)

    # ReplicaLock ("locks") table is not indexed by RSE ID, so we instead go
    # through the RSEFileAssociation ("replicas") table.
    query_rules_from_replicas = select(
        models.ReplicationRule
    ).prefix_with(
        '/*+ USE_NL(locks) LEADING(replicas locks) */',
        dialect='oracle'
    ).distinct(
    ).join_from(
        models.RSEFileAssociation,
        models.ReplicaLock,
        and_(models.RSEFileAssociation.scope == models.ReplicaLock.scope,
             models.RSEFileAssociation.name == models.ReplicaLock.name,
             models.RSEFileAssociation.rse_id == models.ReplicaLock.rse_id)
    ).join(
        models.ReplicationRule,
        models.ReplicaLock.rule_id == models.ReplicationRule.id
    ).where(
        models.RSEFileAssociation.rse_id == rse_id,
        models.ReplicaLock.rule_id.not_in(dataset_rule_ids)
    )

    for rule in session.execute(query_rules_from_replicas).yield_per(5).scalars():
        yield rule.to_dict()


@transactional_session
def __find_missing_locks_and_create_them(
    datasetfiles: Sequence[dict[str, Any]],
    locks: dict[tuple[InternalScope, str], Sequence[models.ReplicaLock]],
    replicas: dict[tuple[InternalScope, str], Sequence[models.CollectionReplica]],
    source_replicas: dict[tuple[InternalScope, str], Sequence[models.CollectionReplica]],
    rseselector: RSESelector,
    rule: models.ReplicationRule,
    source_rses: Sequence[str],
    *,
    session: "Session",
    logger: LoggerFunction = logging.log
) -> None:
    """
    Find missing locks for a rule and create them.

    :param datasetfiles:       Sequence of dicts holding all datasets and files.
    :param locks:              Dict holding locks.
    :param replicas:           Dict holding replicas.
    :param source_replicas:    Dict holding source replicas.
    :param rseselector:        The RSESelector to be used.
    :param rule:               The rule.
    :param source_rses:        RSE ids for eglible source RSEs.
    :param session:            Session of the db.
    :param logger:             Optional decorated logger that can be passed from the calling daemons or servers.
    :raises:                   InsufficientAccountLimit, IntegrityError, InsufficientTargetRSEs
    :attention:                This method modifies the contents of the locks and replicas input parameters.
    """

    logger(logging.DEBUG, "Finding missing locks for rule %s [%d/%d/%d]", str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt)

    mod_datasetfiles = []    # List of Datasets and their files in the Tree [{'scope':, 'name':, 'files': []}]
    # Files are in the format [{'scope':, 'name':, 'bytes':, 'md5':, 'adler32':}]

    for dataset in datasetfiles:
        mod_files = []
        preferred_rse_ids = []
        for file in dataset['files']:
            if len([lock for lock in locks[(file['scope'], file['name'])] if lock.rule_id == rule.id]) < rule.copies:
                mod_files.append(file)
            else:
                preferred_rse_ids = [lock.rse_id for lock in locks[(file['scope'], file['name'])] if lock.rule_id == rule.id]
        if mod_files:
            logger(logging.DEBUG, 'Found missing locks for rule %s, creating them now', str(rule.id))
            mod_datasetfiles.append({'scope': dataset['scope'], 'name': dataset['name'], 'files': mod_files})
            __create_locks_replicas_transfers(datasetfiles=mod_datasetfiles,
                                              locks=locks,
                                              replicas=replicas,
                                              source_replicas=source_replicas,
                                              rseselector=rseselector,
                                              rule=rule,
                                              preferred_rse_ids=preferred_rse_ids,
                                              source_rses=source_rses,
                                              session=session)

    logger(logging.DEBUG, "Finished finding missing locks for rule %s [%d/%d/%d]", str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt)


@transactional_session
def __find_surplus_locks_and_remove_them(
    datasetfiles: Sequence[dict[str, Any]],
    locks: dict[tuple[InternalScope, str], list[models.ReplicaLock]],
    rule: models.ReplicationRule,
    *,
    session: "Session",
    logger: LoggerFunction = logging.log
) -> None:
    """
    Find surplocks locks for a rule and delete them.

    :param datasetfiles:       Dict holding all datasets and files.
    :param locks:              Dict holding locks.
    :param rule:               The rule.
    :param session:            Session of the db.
    :param logger:             Optional decorated logger that can be passed from the calling daemons or servers.
    :raises:                   InsufficientAccountLimit, IntegrityError, InsufficientTargetRSEs
    :attention:                This method modifies the contents of the locks and replicas input parameters.
    """

    logger(logging.DEBUG, "Finding surplus locks for rule %s [%d/%d/%d]", str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt)

    account_counter_decreases = {}  # {'rse_id': [file_size, file_size, file_size]}

    # Put all the files in one dictionary
    files = {}
    for ds in datasetfiles:
        for file in ds['files']:
            files[(file['scope'], file['name'])] = True

    for key in locks:
        if key not in files:
            # The lock needs to be removed
            for lock in deepcopy(locks[key]):
                if lock.rule_id == rule.id:
                    __delete_lock_and_update_replica(lock=lock, purge_replicas=rule.purge_replicas, nowait=True, session=session)
                    if lock.rse_id not in account_counter_decreases:
                        account_counter_decreases[lock.rse_id] = []
                    account_counter_decreases[lock.rse_id].append(lock.bytes)
                    if lock.state == LockState.OK:
                        rule.locks_ok_cnt -= 1
                    elif lock.state == LockState.REPLICATING:
                        rule.locks_replicating_cnt -= 1
                    elif lock.state == LockState.STUCK:
                        rule.locks_stuck_cnt -= 1
                    locks[key].remove(lock)

    logger(logging.DEBUG, "Finished finding surplus locks for rule %s [%d/%d/%d]", str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt)


@transactional_session
def __find_stuck_locks_and_repair_them(
    datasetfiles: Sequence[dict[str, Any]],
    locks: dict[tuple[InternalScope, str], Sequence[models.ReplicaLock]],
    replicas: dict[tuple[InternalScope, str], Sequence[models.CollectionReplica]],
    source_replicas: dict[tuple[InternalScope, str], Sequence[models.CollectionReplica]],
    rseselector: RSESelector,
    rule: models.ReplicationRule,
    source_rses: Sequence[str],
    *,
    session: "Session",
    logger: LoggerFunction = logging.log
) -> None:
    """
    Find stuck locks for a rule and repair them.

    :param datasetfiles:       Dict holding all datasets and files.
    :param locks:              Dict holding locks.
    :param replicas:           Dict holding replicas.
    :param source_replicas:    Dict holding source replicas.
    :param rseselector:        The RSESelector to be used.
    :param rule:               The rule.
    :param source_rses:        RSE ids of eglible source RSEs.
    :param session:            Session of the db.
    :param logger:             Optional decorated logger that can be passed from the calling daemons or servers.
    :raises:                   InsufficientAccountLimit, IntegrityError, InsufficientTargetRSEs
    :attention:                This method modifies the contents of the locks and replicas input parameters.
    """

    logger(logging.DEBUG, "Finding and repairing stuck locks for rule %s [%d/%d/%d]", str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt)

    replicas_to_create, locks_to_create, transfers_to_create, \
        locks_to_delete = repair_stuck_locks_and_apply_rule_grouping(datasetfiles=datasetfiles,
                                                                     locks=locks,
                                                                     replicas=replicas,
                                                                     source_replicas=source_replicas,
                                                                     rseselector=rseselector,
                                                                     rule=rule,
                                                                     source_rses=source_rses,
                                                                     session=session)
    # Add the replicas
    session.add_all([item for sublist in replicas_to_create.values() for item in sublist])
    session.flush()

    # Add the locks
    session.add_all([item for sublist in locks_to_create.values() for item in sublist])
    session.flush()

    # Increase rse_counters
    for rse_id in replicas_to_create.keys():
        rse_counter.increase(rse_id=rse_id, files=len(replicas_to_create[rse_id]), bytes_=sum([replica.bytes for replica in replicas_to_create[rse_id]]), session=session)

    # Increase account_counters
    for rse_id in locks_to_create.keys():
        account_counter.increase(rse_id=rse_id, account=rule.account, files=len(locks_to_create[rse_id]), bytes_=sum([lock.bytes for lock in locks_to_create[rse_id]]), session=session)

    # Decrease account_counters
    for rse_id in locks_to_delete:
        account_counter.decrease(rse_id=rse_id, account=rule.account, files=len(locks_to_delete[rse_id]), bytes_=sum([lock.bytes for lock in locks_to_delete[rse_id]]), session=session)

    # Delete the locks:
    for lock in [item for sublist in locks_to_delete.values() for item in sublist]:
        session.delete(lock)

    # Add the transfers
    request_core.queue_requests(requests=transfers_to_create, session=session)
    session.flush()
    logger(logging.DEBUG, "Finished finding and repairing stuck locks for rule %s [%d/%d/%d]", str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt)


@transactional_session
def __evaluate_did_detach(
    eval_did: models.DataIdentifier,
    *,
    session: "Session",
    logger: LoggerFunction = logging.log
) -> None:
    """
    Evaluate a parent did which has children removed.

    :param eval_did:  The did object in use.
    :param session:   The database session in use.
    :param logger:    Optional decorated logger that can be passed from the calling daemons or servers.
    """

    logger(logging.INFO, "Re-Evaluating did %s:%s for DETACH", eval_did.scope, eval_did.name)
    force_epoch = config_get('rules', 'force_epoch_when_detach', default=False, session=session)

    with METRICS.timer('evaluate_did_detach.total'):
        # Get all parent DID's
        parent_dids = rucio.core.did.list_all_parent_dids(scope=eval_did.scope, name=eval_did.name, session=session)

        # Get all RR from parents and eval_did
        stmt = select(
            models.ReplicationRule
        ).where(
            and_(models.ReplicationRule.scope == eval_did.scope,
                 models.ReplicationRule.name == eval_did.name)
        ).with_for_update(
            nowait=True
        )
        rules = list(session.execute(stmt).scalars().all())
        for did in parent_dids:
            stmt = select(
                models.ReplicationRule
            ).where(
                and_(models.ReplicationRule.scope == did['scope'],
                     models.ReplicationRule.name == did['name'])
            ).with_for_update(
                nowait=True
            )
            rules.extend(session.execute(stmt).scalars().all())

        # Iterate rules and delete locks
        transfers_to_delete = []  # [{'scope': , 'name':, 'rse_id':}]
        account_counter_decreases = {}  # {'rse_id': [file_size, file_size, file_size]}
        for rule in rules:
            # Get all the files covering this rule
            files = {}
            for file in rucio.core.did.list_files(scope=rule.scope, name=rule.name, session=session):
                files[(file['scope'], file['name'])] = True
            logger(logging.DEBUG, "Removing locks for rule %s [%d/%d/%d]", str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt)
            rule_locks_ok_cnt_before = rule.locks_ok_cnt
            stmt = select(
                models.ReplicaLock
            ).where(
                models.ReplicaLock.rule_id == rule.id
            )
            for lock in session.execute(stmt).scalars().all():
                if (lock.scope, lock.name) not in files:
                    if __delete_lock_and_update_replica(lock=lock, purge_replicas=force_epoch or rule.purge_replicas, nowait=True, session=session):
                        transfers_to_delete.append({'scope': lock.scope, 'name': lock.name, 'rse_id': lock.rse_id})
                    if lock.rse_id not in account_counter_decreases:
                        account_counter_decreases[lock.rse_id] = []
                    account_counter_decreases[lock.rse_id].append(lock.bytes)
                    if lock.state == LockState.OK:
                        rule.locks_ok_cnt -= 1
                    elif lock.state == LockState.REPLICATING:
                        rule.locks_replicating_cnt -= 1
                    elif lock.state == LockState.STUCK:
                        rule.locks_stuck_cnt -= 1
            logger(logging.DEBUG, "Finished removing locks for rule %s [%d/%d/%d]", str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt)

            if eval_did.did_type == DIDType.CONTAINER:
                # Get all datasets of eval_did
                child_datasets = {}
                for ds in rucio.core.did.list_child_datasets(scope=rule.scope, name=rule.name, session=session):
                    child_datasets[(ds['scope'], ds['name'])] = True
                logger(logging.DEBUG, "Removing dataset_locks for rule %s [%d/%d/%d]", str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt)
                stmt = select(
                    models.DatasetLock
                ).where(
                    models.DatasetLock.rule_id == rule.id
                )
                query = session.execute(stmt).scalars().all()
                for ds_lock in query:
                    if (ds_lock.scope, ds_lock.name) not in child_datasets:
                        ds_lock.delete(flush=False, session=session)
                logger(logging.DEBUG, "Finished removing dataset_locks for rule %s [%d/%d/%d]", str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt)

            if rule.state == RuleState.SUSPENDED:
                pass
            elif rule.state == RuleState.STUCK:
                pass
            elif rule.locks_replicating_cnt == 0 and rule.locks_stuck_cnt == 0:
                rule.state = RuleState.OK
                if rule.grouping != RuleGrouping.NONE:
                    stmt = update(
                        models.DatasetLock
                    ).where(
                        models.DatasetLock.rule_id == rule.id
                    ).values({
                        models.DatasetLock.state: LockState.OK
                    })
                    session.execute(stmt)
                    session.flush()
                if rule_locks_ok_cnt_before != rule.locks_ok_cnt:
                    generate_rule_notifications(rule=rule, session=session)
                    generate_email_for_rule_ok_notification(rule=rule, session=session)
                # Try to release potential parent rules
                release_parent_rule(child_rule_id=rule.id, session=session)

            # Insert rule history
            insert_rule_history(rule=rule, recent=True, longterm=False, session=session)

        session.flush()

        # Decrease account_counters
        for rse_id in account_counter_decreases:
            account_counter.decrease(rse_id=rse_id, account=rule.account, files=len(account_counter_decreases[rse_id]), bytes_=sum(account_counter_decreases[rse_id]), session=session)

        for transfer in transfers_to_delete:
            transfers_to_cancel = request_core.cancel_request_did(scope=transfer['scope'], name=transfer['name'], dest_rse_id=transfer['rse_id'], session=session)
            transfer_core.cancel_transfers(transfers_to_cancel)


@transactional_session
def __oldest_file_under(
    scope: InternalScope,
    name: str,
    *,
    session: "Session"
) -> Optional[tuple[InternalScope, str]]:
    """
    Finds oldest file in oldest container/dataset in the container or the dataset, recursively.
    Oldest means attached to its parent first.

    :param scope: dataset or container scope
    :param name: dataset or container name
    :returns: tuple (scope, name) or None
    """
    stmt = select(
        models.DataIdentifierAssociation
    ).where(
        and_(models.DataIdentifierAssociation.scope == scope,
             models.DataIdentifierAssociation.name == name)
    ).order_by(
        models.DataIdentifierAssociation.created_at
    )
    children = session.execute(stmt).scalars().all()
    for child in children:
        if child.child_type == DIDType.FILE:
            return child.child_scope, child.child_name
        elif child.child_type in (DIDType.DATASET, DIDType.CONTAINER):
            out = __oldest_file_under(child.child_scope, child.child_name, session=session)
            if out:
                return out
    return None


@transactional_session
def __evaluate_did_attach(
    eval_did: models.DataIdentifier,
    *,
    session: "Session",
    logger: LoggerFunction = logging.log
) -> None:
    """
    Evaluate a parent did which has new childs

    :param eval_did:  The did object in use.
    :param session:   The database session in use.
    :param logger:    Optional decorated logger that can be passed from the calling daemons or servers.
    :raises:          ReplicationRuleCreationTemporaryFailed
    """

    logger(logging.INFO, "Re-Evaluating did %s:%s for ATTACH", eval_did.scope, eval_did.name)

    with METRICS.timer('evaluate_did_attach.total'):
        # Get all parent DID's
        with METRICS.timer('evaluate_did_attach.list_parent_dids'):
            parent_dids = rucio.core.did.list_all_parent_dids(scope=eval_did.scope, name=eval_did.name, session=session)

        # Get immediate new child DID's
        with METRICS.timer('evaluate_did_attach.list_new_child_dids'):
            stmt = select(
                models.DataIdentifierAssociation
            ).with_hint(
                models.DataIdentifierAssociation, 'INDEX_RS_ASC(CONTENTS CONTENTS_PK)', 'oracle'
            ).where(
                and_(models.DataIdentifierAssociation.scope == eval_did.scope,
                     models.DataIdentifierAssociation.name == eval_did.name,
                     models.DataIdentifierAssociation.rule_evaluation == true())
            )
            new_child_dids = session.execute(stmt).scalars().all()
        if new_child_dids:
            # Get all unsuspended RR from parents and eval_did
            with METRICS.timer('evaluate_did_attach.get_rules'):
                rule_clauses = []
                for did in parent_dids:
                    rule_clauses.append(and_(models.ReplicationRule.scope == did['scope'],
                                             models.ReplicationRule.name == did['name']))
                rule_clauses.append(and_(models.ReplicationRule.scope == eval_did.scope,
                                         models.ReplicationRule.name == eval_did.name))
                stmt = select(
                    models.ReplicationRule
                ).where(
                    and_(or_(*rule_clauses),
                         models.ReplicationRule.state.not_in([RuleState.SUSPENDED,
                                                              RuleState.WAITING_APPROVAL,
                                                              RuleState.INJECT]))
                ).with_for_update(
                    nowait=True
                )
                rules = session.execute(stmt).scalars().all()
            if rules:
                # Resolve the new_child_dids to its locks
                with METRICS.timer('evaluate_did_attach.resolve_did_to_locks_and_replicas'):
                    # Resolve the rules to possible target rses:
                    possible_rses = []
                    source_rses = []
                    for rule in rules:
                        try:
                            vo = rule.account.vo
                            if rule.source_replica_expression:
                                source_rses.extend(parse_expression(rule.source_replica_expression, filter_={'vo': vo}, session=session))

                            # if rule.ignore_availability:
                            possible_rses.extend(parse_expression(rule.rse_expression, filter_={'vo': vo}, session=session))
                            # else:
                            #     possible_rses.extend(parse_expression(rule.rse_expression, filter={'availability_write': True}, session=session))
                        except (InvalidRSEExpression, RSEWriteBlocked):
                            possible_rses = []
                            break

                    source_rses = list(set([rse['id'] for rse in source_rses]))
                    possible_rses = list(set([rse['id'] for rse in possible_rses]))

                    datasetfiles, locks, replicas, source_replicas = __resolve_dids_to_locks_and_replicas(dids=new_child_dids,
                                                                                                          nowait=True,
                                                                                                          restrict_rses=possible_rses,
                                                                                                          source_rses=source_rses,
                                                                                                          session=session)

                # Evaluate the replication rules
                with METRICS.timer('evaluate_did_attach.evaluate_rules'):
                    for rule in rules:
                        rule_locks_ok_cnt_before = rule.locks_ok_cnt

                        # 1. Resolve the rse_expression into a list of RSE-ids
                        try:
                            vo = rule.account.vo
                            if rule.ignore_availability:
                                rses = parse_expression(rule.rse_expression, filter_={'vo': vo}, session=session)
                            else:
                                rses = parse_expression(rule.rse_expression, filter_={'vo': vo, 'availability_write': True}, session=session)
                            source_rses = []
                            if rule.source_replica_expression:
                                source_rses = parse_expression(rule.source_replica_expression, filter_={'vo': vo}, session=session)
                        except (InvalidRSEExpression, RSEWriteBlocked) as error:
                            rule.state = RuleState.STUCK
                            rule.error = (str(error)[:245] + '...') if len(str(error)) > 245 else str(error)
                            rule.save(session=session)
                            # Insert rule history
                            insert_rule_history(rule=rule, recent=True, longterm=False, session=session)
                            # Try to update the DatasetLocks
                            if rule.grouping != RuleGrouping.NONE:
                                stmt = update(
                                    models.DatasetLock
                                ).where(
                                    models.DatasetLock.rule_id == rule.id
                                ).values({
                                    models.DatasetLock.state: LockState.STUCK
                                })
                                session.execute(stmt)
                            continue

                        # 2. Create the RSE Selector
                        try:
                            rseselector = RSESelector(account=rule.account,
                                                      rses=rses,
                                                      weight=rule.weight,
                                                      copies=rule.copies,
                                                      ignore_account_limit=rule.ignore_account_limit,
                                                      session=session)
                        except (InvalidRuleWeight, InsufficientTargetRSEs, InsufficientAccountLimit, RSEOverQuota) as error:
                            rule.state = RuleState.STUCK
                            rule.error = (str(error)[:245] + '...') if len(str(error)) > 245 else str(error)
                            rule.save(session=session)
                            # Insert rule history
                            insert_rule_history(rule=rule, recent=True, longterm=False, session=session)
                            # Try to update the DatasetLocks
                            if rule.grouping != RuleGrouping.NONE:
                                stmt = update(
                                    models.DatasetLock
                                ).where(
                                    models.DatasetLock.rule_id == rule.id
                                ).values({
                                    models.DatasetLock.state: LockState.STUCK
                                })
                                session.execute(stmt)
                            continue

                        # 3. Apply the Replication rule to the Files
                        preferred_rse_ids = []
                        brother_scope_name = None

                        if rule.grouping == RuleGrouping.ALL:
                            # get oldest file, recursively, in the rule owner
                            brother_scope_name = __oldest_file_under(rule["scope"], rule["name"], session=session)

                        elif rule.grouping == RuleGrouping.DATASET and new_child_dids[0].child_type == DIDType.FILE:
                            # get oldest file in the dataset being evaluated
                            brother_scope_name = __oldest_file_under(eval_did.scope, eval_did.name, session=session)

                        if brother_scope_name:
                            scope, name = brother_scope_name
                            file_locks = rucio.core.lock.get_replica_locks(scope=scope, name=name, nowait=True, session=session)
                            preferred_rse_ids = [
                                lock['rse_id']
                                for lock in file_locks
                                if lock['rse_id'] in [rse['id'] for rse in rses] and lock['rule_id'] == rule.id
                            ]

                        locks_stuck_before = rule.locks_stuck_cnt
                        try:
                            __create_locks_replicas_transfers(datasetfiles=datasetfiles,
                                                              locks=locks,
                                                              replicas=replicas,
                                                              source_replicas=source_replicas,
                                                              rseselector=rseselector,
                                                              rule=rule,
                                                              preferred_rse_ids=preferred_rse_ids,
                                                              source_rses=[rse['id'] for rse in source_rses],
                                                              session=session)
                        except (InsufficientAccountLimit, InsufficientTargetRSEs, RSEOverQuota) as error:
                            rule.state = RuleState.STUCK
                            rule.error = (str(error)[:245] + '...') if len(str(error)) > 245 else str(error)
                            rule.save(session=session)
                            # Insert rule history
                            insert_rule_history(rule=rule, recent=True, longterm=False, session=session)
                            # Try to update the DatasetLocks
                            if rule.grouping != RuleGrouping.NONE:
                                stmt = update(
                                    models.DatasetLock
                                ).where(
                                    models.DatasetLock.rule_id == rule.id
                                ).values({
                                    models.DatasetLock.state: LockState.STUCK
                                })
                                session.execute(stmt)
                            continue

                        # 4. Update the Rule State
                        if rule.state == RuleState.STUCK:
                            pass
                        elif rule.locks_stuck_cnt > 0:
                            if locks_stuck_before != rule.locks_stuck_cnt:
                                rule.state = RuleState.STUCK
                                rule.error = 'MissingSourceReplica'
                                stmt = update(
                                    models.DatasetLock
                                ).where(
                                    models.DatasetLock.rule_id == rule.id
                                ).values({
                                    models.DatasetLock.state: LockState.STUCK
                                })
                                session.execute(stmt)
                        elif rule.locks_replicating_cnt > 0:
                            rule.state = RuleState.REPLICATING
                            if rule.grouping != RuleGrouping.NONE:
                                stmt = update(
                                    models.DatasetLock
                                ).where(
                                    models.DatasetLock.rule_id == rule.id
                                ).values({
                                    models.DatasetLock.state: LockState.REPLICATING
                                })
                                session.execute(stmt)
                        elif rule.locks_replicating_cnt == 0 and rule.locks_stuck_cnt == 0:
                            rule.state = RuleState.OK
                            if rule.grouping != RuleGrouping.NONE:
                                stmt = update(
                                    models.DatasetLock
                                ).where(
                                    models.DatasetLock.rule_id == rule.id
                                ).values({
                                    models.DatasetLock.state: LockState.OK
                                })
                                session.execute(stmt)
                                session.flush()
                            if rule_locks_ok_cnt_before < rule.locks_ok_cnt:
                                generate_rule_notifications(rule=rule, session=session)
                                generate_email_for_rule_ok_notification(rule=rule, session=session)

                        # Insert rule history
                        insert_rule_history(rule=rule, recent=True, longterm=False, session=session)

            # Unflage the dids
            with METRICS.timer('evaluate_did_attach.update_did'):
                for did in new_child_dids:
                    did.rule_evaluation = None

        session.flush()


@transactional_session
def __resolve_did_to_locks_and_replicas(
    did: models.DataIdentifier,
    nowait: bool = False,
    restrict_rses: Optional[Sequence[str]] = None,
    source_rses: Optional[Sequence[str]] = None,
    only_stuck: bool = False,
    *,
    session: "Session"
) -> tuple[list[dict[str, Any]],
           dict[tuple[str, str], models.ReplicaLock],
           dict[tuple[str, str], models.RSEFileAssociation],
           dict[tuple[str, str], str]]:
    """
    Resolves a did to its constituent childs and reads the locks and replicas of all the constituent files.

    :param did:            The db object of the did the rule is applied on.
    :param nowait:         Nowait parameter for the FOR UPDATE statement.
    :param restrict_rses:  Possible rses of the rule, so only these replica/locks should be considered.
    :param source_rses:    Source rses for this rule. These replicas are not row-locked.
    :param only_stuck:     Get results only for STUCK locks, if True.
    :param session:        Session of the db.
    :returns:              (datasetfiles, locks, replicas, source_replicas)
    """

    datasetfiles = []     # List of Datasets and their files in the Tree [{'scope':, 'name':, 'files': []}]
    # Files are in the format [{'scope':, 'name':, 'bytes':, 'md5':, 'adler32':}]
    locks = {}            # {(scope,name): [SQLAlchemy]}
    replicas = {}         # {(scope, name): [SQLAlchemy]}
    source_replicas = {}  # {(scope, name): [rse_id]

    if did.did_type == DIDType.FILE:
        datasetfiles = [{'scope': None,
                         'name': None,
                         'files': [{'scope': did.scope,
                                    'name': did.name,
                                    'bytes': did.bytes,
                                    'md5': did.md5,
                                    'adler32': did.adler32}]}]
        locks[(did.scope, did.name)] = rucio.core.lock.get_replica_locks(scope=did.scope, name=did.name, nowait=nowait, restrict_rses=restrict_rses, session=session)
        replicas[(did.scope, did.name)] = rucio.core.replica.get_and_lock_file_replicas(scope=did.scope, name=did.name, nowait=nowait, restrict_rses=restrict_rses, session=session)
        if source_rses:
            source_replicas[(did.scope, did.name)] = rucio.core.replica.get_source_replicas(scope=did.scope, name=did.name, source_rses=source_rses, session=session)

    elif did.did_type == DIDType.DATASET and only_stuck:
        files = []
        locks = rucio.core.lock.get_files_and_replica_locks_of_dataset(scope=did.scope, name=did.name, nowait=nowait, restrict_rses=restrict_rses, only_stuck=True, session=session)
        for file in locks:
            file_did = rucio.core.did.get_did(scope=file[0], name=file[1], session=session)
            files.append({'scope': file[0], 'name': file[1], 'bytes': file_did['bytes'], 'md5': file_did['md5'], 'adler32': file_did['adler32']})
            replicas[(file[0], file[1])] = rucio.core.replica.get_and_lock_file_replicas(scope=file[0], name=file[1], nowait=nowait, restrict_rses=restrict_rses, session=session)
            if source_rses:
                source_replicas[(file[0], file[1])] = rucio.core.replica.get_source_replicas(scope=file[0], name=file[1], source_rses=source_rses, session=session)
        datasetfiles = [{'scope': did.scope,
                         'name': did.name,
                         'files': files}]

    elif did.did_type == DIDType.DATASET:
        files, replicas = rucio.core.replica.get_and_lock_file_replicas_for_dataset(scope=did.scope, name=did.name, nowait=nowait, restrict_rses=restrict_rses, session=session)
        if source_rses:
            source_replicas = rucio.core.replica.get_source_replicas_for_dataset(scope=did.scope, name=did.name, source_rses=source_rses, session=session)
        datasetfiles = [{'scope': did.scope,
                         'name': did.name,
                         'files': files}]
        locks = rucio.core.lock.get_files_and_replica_locks_of_dataset(scope=did.scope, name=did.name, nowait=nowait, restrict_rses=restrict_rses, session=session)

    elif did.did_type == DIDType.CONTAINER and only_stuck:

        for dataset in rucio.core.did.list_child_datasets(scope=did.scope, name=did.name, session=session):
            files = []
            tmp_locks = rucio.core.lock.get_files_and_replica_locks_of_dataset(scope=dataset['scope'], name=dataset['name'], nowait=nowait, restrict_rses=restrict_rses, only_stuck=True, session=session)
            locks = dict(list(locks.items()) + list(tmp_locks.items()))
            for file in tmp_locks:
                file_did = rucio.core.did.get_did(scope=file[0], name=file[1], session=session)
                files.append({'scope': file[0], 'name': file[1], 'bytes': file_did['bytes'], 'md5': file_did['md5'], 'adler32': file_did['adler32']})
                replicas[(file[0], file[1])] = rucio.core.replica.get_and_lock_file_replicas(scope=file[0], name=file[1], nowait=nowait, restrict_rses=restrict_rses, session=session)
                if source_rses:
                    source_replicas[(file[0], file[1])] = rucio.core.replica.get_source_replicas(scope=file[0], name=file[1], source_rses=source_rses, session=session)
            datasetfiles.append({'scope': dataset['scope'],
                                 'name': dataset['name'],
                                 'files': files})

    elif did.did_type == DIDType.CONTAINER:

        for dataset in rucio.core.did.list_child_datasets(scope=did.scope, name=did.name, session=session):
            files, tmp_replicas = rucio.core.replica.get_and_lock_file_replicas_for_dataset(scope=dataset['scope'], name=dataset['name'], nowait=nowait, restrict_rses=restrict_rses, session=session)
            if source_rses:
                tmp_source_replicas = rucio.core.replica.get_source_replicas_for_dataset(scope=dataset['scope'], name=dataset['name'], source_rses=source_rses, session=session)
                source_replicas = dict(list(source_replicas.items()) + list(tmp_source_replicas.items()))
            tmp_locks = rucio.core.lock.get_files_and_replica_locks_of_dataset(scope=dataset['scope'], name=dataset['name'], nowait=nowait, restrict_rses=restrict_rses, session=session)
            datasetfiles.append({'scope': dataset['scope'],
                                 'name': dataset['name'],
                                 'files': files})
            replicas = dict(list(replicas.items()) + list(tmp_replicas.items()))
            locks = dict(list(locks.items()) + list(tmp_locks.items()))

        # order datasetfiles for deterministic result
        try:
            datasetfiles = sorted(datasetfiles, key=lambda x: "%s%s" % (x['scope'], x['name']))
        except Exception:
            pass

    else:
        raise InvalidReplicationRule('The did \"%s:%s\" has been deleted.' % (did.scope, did.name))

    return datasetfiles, locks, replicas, source_replicas


@transactional_session
def __resolve_dids_to_locks_and_replicas(
    dids: Sequence[models.DataIdentifierAssociation],
    nowait: bool = False,
    restrict_rses: Sequence[str] = [],
    source_rses: Optional[Sequence[str]] = None,
    *,
    session: "Session"
) -> tuple[list[dict[str, Any]],
           dict[tuple[str, str], models.ReplicaLock],
           dict[tuple[str, str], models.RSEFileAssociation],
           dict[tuple[str, str], str]]:
    """
    Resolves a list of dids to its constituent childs and reads the locks and replicas of all the constituent files.

    :param dids:           The list of DataIdentifierAssociation objects.
    :param nowait:         Nowait parameter for the FOR UPDATE statement.
    :param restrict_rses:  Possible rses of the rule, so only these replica/locks should be considered.
    :param source_rses:    Source rses for this rule. These replicas are not row-locked.
    :param session:        Session of the db.
    :returns:              (datasetfiles, locks, replicas, source_replicas)
    """

    datasetfiles = []     # List of Datasets and their files in the Tree [{'scope':, 'name':, 'files': []}]
    # Files are in the format [{'scope':, 'name':, 'bytes':, 'md5':, 'adler32':}]
    locks = {}            # {(scope,name): [SQLAlchemy]}
    replicas = {}         # {(scope, name): [SQLAlchemy]}
    source_replicas = {}  # {(scope, name): [rse_id]

    if dids[0].child_type == DIDType.FILE:
        # All the dids will be files!
        # Prepare the datasetfiles
        files = []
        for did in dids:
            files.append({'scope': did.child_scope,
                          'name': did.child_name,
                          'bytes': did.bytes,
                          'md5': did.md5,
                          'adler32': did.adler32})
            locks[(did.child_scope, did.child_name)] = []
            replicas[(did.child_scope, did.child_name)] = []
            source_replicas[(did.child_scope, did.child_name)] = []
        datasetfiles = [{'scope': dids[0].scope, 'name': dids[0].name, 'files': files}]

        # Prepare the locks and files
        lock_clauses = []
        replica_clauses = []
        for did in dids:
            lock_clauses.append(and_(models.ReplicaLock.scope == did.child_scope,
                                     models.ReplicaLock.name == did.child_name))
            replica_clauses.append(and_(models.RSEFileAssociation.scope == did.child_scope,
                                        models.RSEFileAssociation.name == did.child_name))
        lock_clause_chunks = [lock_clauses[x:x + 10] for x in range(0, len(lock_clauses), 10)]
        replica_clause_chunks = [replica_clauses[x:x + 10] for x in range(0, len(replica_clauses), 10)]

        replicas_rse_clause = []
        source_replicas_rse_clause = []
        locks_rse_clause = []
        if restrict_rses:
            for rse_id in restrict_rses:
                replicas_rse_clause.append(models.RSEFileAssociation.rse_id == rse_id)
                locks_rse_clause.append(models.ReplicaLock.rse_id == rse_id)
        if source_rses:
            for rse_id in source_rses:
                source_replicas_rse_clause.append(models.RSEFileAssociation.rse_id == rse_id)

        for lock_clause_chunk in lock_clause_chunks:
            if locks_rse_clause:
                stmt = select(
                    models.ReplicaLock
                ).with_hint(
                    models.ReplicaLock, 'INDEX(LOCKS LOCKS_PK)', 'oracle'
                ).where(
                    and_(or_(*lock_clause_chunk),
                         or_(*locks_rse_clause))
                ).with_for_update(
                    nowait=nowait
                )
                tmp_locks = session.execute(stmt).scalars().all()
            else:
                stmt = select(
                    models.ReplicaLock
                ).with_hint(
                    models.ReplicaLock, 'INDEX(LOCKS LOCKS_PK)', 'oracle'
                ).where(
                    or_(*lock_clause_chunk)
                ).with_for_update(
                    nowait=nowait
                )
                tmp_locks = session.execute(stmt).scalars().all()
            for lock in tmp_locks:
                if (lock.scope, lock.name) not in locks:
                    locks[(lock.scope, lock.name)] = [lock]
                else:
                    locks[(lock.scope, lock.name)].append(lock)

        for replica_clause_chunk in replica_clause_chunks:
            if replicas_rse_clause:
                stmt = select(
                    models.RSEFileAssociation
                ).with_hint(
                    models.RSEFileAssociation, 'INDEX(REPLICAS REPLICAS_PK)', 'oracle'
                ).where(
                    and_(or_(*replica_clause_chunk),
                         or_(*replicas_rse_clause),
                         models.RSEFileAssociation.state != ReplicaState.BEING_DELETED)
                ).with_for_update(
                    nowait=nowait
                )
                tmp_replicas = session.execute(stmt).scalars().all()
            else:
                stmt = select(
                    models.RSEFileAssociation
                ).with_hint(
                    models.RSEFileAssociation, 'INDEX(REPLICAS REPLICAS_PK)', 'oracle'
                ).where(
                    and_(or_(*replica_clause_chunk),
                         models.RSEFileAssociation.state != ReplicaState.BEING_DELETED)
                ).with_for_update(
                    nowait=nowait
                )
                tmp_replicas = session.execute(stmt).scalars().all()
            for replica in tmp_replicas:
                if (replica.scope, replica.name) not in replicas:
                    replicas[(replica.scope, replica.name)] = [replica]
                else:
                    replicas[(replica.scope, replica.name)].append(replica)

        if source_rses:
            for replica_clause_chunk in replica_clause_chunks:
                stmt = select(
                    models.RSEFileAssociation.scope,
                    models.RSEFileAssociation.name,
                    models.RSEFileAssociation.rse_id
                ).with_hint(
                    models.RSEFileAssociation, 'INDEX(REPLICAS REPLICAS_PK)', 'oracle'
                ).where(
                    and_(or_(*replica_clause_chunk),
                         or_(*source_replicas_rse_clause),
                         models.RSEFileAssociation.state == ReplicaState.AVAILABLE)
                )
                tmp_source_replicas = session.execute(stmt).all()
                for scope, name, rse_id in tmp_source_replicas:
                    if (scope, name) not in source_replicas:
                        source_replicas[(scope, name)] = [rse_id]
                    else:
                        source_replicas[(scope, name)].append(rse_id)
    else:
        # The evaluate_dids will be containers and/or datasets
        for did in dids:
            stmt = select(
                models.DataIdentifier
            ).where(
                and_(models.DataIdentifier.scope == did.child_scope,
                     models.DataIdentifier.name == did.child_name)
            )
            real_did = session.execute(stmt).scalar_one()
            tmp_datasetfiles, tmp_locks, tmp_replicas, tmp_source_replicas = __resolve_did_to_locks_and_replicas(did=real_did,
                                                                                                                 nowait=nowait,
                                                                                                                 restrict_rses=restrict_rses,
                                                                                                                 source_rses=source_rses,
                                                                                                                 session=session)
            datasetfiles.extend(tmp_datasetfiles)
            locks.update(tmp_locks)
            replicas.update(tmp_replicas)
            source_replicas.update(tmp_source_replicas)
    return datasetfiles, locks, replicas, source_replicas


@transactional_session
def __create_locks_replicas_transfers(
    datasetfiles: Sequence[dict[str, Any]],
    locks: dict[tuple[InternalScope, str], Sequence[models.ReplicaLock]],
    replicas: dict[tuple[InternalScope, str], Sequence[models.CollectionReplica]],
    source_replicas: dict[tuple[InternalScope, str], Sequence[models.CollectionReplica]],
    rseselector: RSESelector,
    rule: models.ReplicationRule,
    preferred_rse_ids: Sequence[str] = [],
    source_rses: Sequence[str] = [],
    *,
    session: "Session",
    logger: LoggerFunction = logging.log
) -> None:
    """
    Apply a created replication rule to a set of files

    :param datasetfiles:       Dict holding all datasets and files.
    :param locks:              Dict holding locks.
    :param replicas:           Dict holding replicas.
    :param source_replicas:    Dict holding source replicas.
    :param rseselector:        The RSESelector to be used.
    :param rule:               The rule.
    :param preferred_rse_ids:  Preferred RSE's to select.
    :param source_rses:        RSE ids of eglible source replicas.
    :param session:            Session of the db.
    :param logger:             Optional decorated logger that can be passed from the calling daemons or servers.
    :raises:                   InsufficientAccountLimit, IntegrityError, InsufficientTargetRSEs, RSEOverQuota
    :attention:                This method modifies the contents of the locks and replicas input parameters.
    """

    logger(logging.DEBUG, "Creating locks and replicas for rule %s [%d/%d/%d]", str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt)

    replicas_to_create, locks_to_create, transfers_to_create = apply_rule_grouping(datasetfiles=datasetfiles,
                                                                                   locks=locks,
                                                                                   replicas=replicas,
                                                                                   source_replicas=source_replicas,
                                                                                   rseselector=rseselector,
                                                                                   rule=rule,
                                                                                   preferred_rse_ids=preferred_rse_ids,
                                                                                   source_rses=source_rses,
                                                                                   session=session)
    # Add the replicas
    session.add_all([item for sublist in replicas_to_create.values() for item in sublist])
    session.flush()

    # Add the locks
    session.add_all([item for sublist in locks_to_create.values() for item in sublist])
    session.flush()

    # Increase rse_counters
    for rse_id in replicas_to_create.keys():
        rse_counter.increase(rse_id=rse_id, files=len(replicas_to_create[rse_id]), bytes_=sum([replica.bytes for replica in replicas_to_create[rse_id]]), session=session)

    # Increase account_counters
    for rse_id in locks_to_create.keys():
        account_counter.increase(rse_id=rse_id, account=rule.account, files=len(locks_to_create[rse_id]), bytes_=sum([lock.bytes for lock in locks_to_create[rse_id]]), session=session)

    # Add the transfers
    logger(logging.DEBUG, "Rule %s  [%d/%d/%d] queued %d transfers", str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt, len(transfers_to_create))
    request_core.queue_requests(requests=transfers_to_create, session=session)
    session.flush()
    logger(logging.DEBUG, "Finished creating locks and replicas for rule %s [%d/%d/%d]", str(rule.id), rule.locks_ok_cnt, rule.locks_replicating_cnt, rule.locks_stuck_cnt)


@transactional_session
def __delete_lock_and_update_replica(
    lock: models.ReplicaLock,
    purge_replicas: bool = False,
    nowait: bool = False,
    *,
    session: "Session",
    logger: LoggerFunction = logging.log
) -> bool:
    """
    Delete a lock and update the associated replica.

    :param lock:            SQLAlchemy lock object.
    :param purge_replicas:  Purge setting of the rule.
    :param nowait:          The nowait option of the FOR UPDATE statement.
    :param session:         The database session in use.
    :param logger:          Optional decorated logger that can be passed from the calling daemons or servers.
    :returns:               True, if the lock was replicating and the associated transfer should be canceled; False otherwise.
    """

    logger(logging.DEBUG, "Deleting lock %s:%s for rule %s", lock.scope, lock.name, str(lock.rule_id))
    lock.delete(session=session, flush=False)
    try:
        stmt = select(
            models.RSEFileAssociation
        ).where(
            and_(models.RSEFileAssociation.scope == lock.scope,
                 models.RSEFileAssociation.name == lock.name,
                 models.RSEFileAssociation.rse_id == lock.rse_id)
        ).with_for_update(
            nowait=nowait
        )
        replica = session.execute(stmt).scalar_one()
        replica.lock_cnt -= 1
        if replica.lock_cnt == 0:
            if purge_replicas:
                replica.tombstone = OBSOLETE
            elif replica.state == ReplicaState.UNAVAILABLE:
                replica.tombstone = OBSOLETE
            elif replica.accessed_at is not None:
                replica.tombstone = replica.accessed_at
            else:
                replica.tombstone = replica.created_at
            if lock.state == LockState.REPLICATING:
                replica.state = ReplicaState.UNAVAILABLE
                replica.tombstone = OBSOLETE
                return True
    except NoResultFound:
        logger(logging.ERROR, "Replica for lock %s:%s for rule %s on rse %s could not be found", lock.scope, lock.name, str(lock.rule_id), get_rse_name(rse_id=lock.rse_id, session=session))
    return False


@transactional_session
def __create_rule_approval_email(
    rule: models.ReplicationRule,
    *,
    session: "Session"
) -> None:
    """
    Create the rule notification email.

    :param rule:      The rule object.
    :param session:   The database session in use.
    """

    filepath = path.join(
        config_get('common', 'mailtemplatedir'), 'rule_approval_request.tmpl'  # type: ignore
    )
    with open(filepath, 'r') as templatefile:
        template = Template(templatefile.read())

    did = rucio.core.did.get_did(
        scope=rule.scope,
        name=rule.name,
        dynamic_depth=DIDType.FILE,
        session=session
    )

    reps = rucio.core.replica.list_dataset_replicas(
        scope=rule.scope, name=rule.name, session=session
    )
    rses = [rep['rse_id'] for rep in reps if rep['state'] == ReplicaState.AVAILABLE]

    # RSE occupancy
    vo = rule.account.vo
    target_rses = parse_expression(rule.rse_expression, filter_={'vo': vo}, session=session)
    if len(target_rses) > 1:
        target_rse = 'Multiple'
        free_space = 'undefined'
        free_space_after = 'undefined'
    else:
        target_rse = target_rses[0]['rse']
        target_rse_id = target_rses[0]['id']
        free_space = 'undefined'
        free_space_after = 'undefined'

        try:
            for usage in get_rse_usage(rse_id=target_rse_id, session=session):
                if usage['source'] == 'storage':
                    free_space = sizefmt(usage['free'])
                    if did['bytes'] is None:
                        free_space_after = 'undefined'
                    else:
                        free_space_after = sizefmt(usage['free'] - did['bytes'])
        except Exception:
            pass

    # Resolve recipients:
    recipients = _create_recipients_list(rse_expression=rule.rse_expression, filter_={'vo': vo}, session=session)

    for recipient in recipients:
        text = template.safe_substitute(
            {
                'rule_id': str(rule.id),
                'created_at': str(rule.created_at),
                'expires_at': str(rule.expires_at),
                'account': rule.account.external,
                'email': get_account(account=rule.account, session=session).email,
                'rse_expression': rule.rse_expression,
                'comment': rule.comments,
                'scope': rule.scope.external,
                'name': rule.name,
                'did_type': rule.did_type,
                'length': '0' if did['length'] is None else str(did['length']),
                'bytes': '0' if did['bytes'] is None else sizefmt(did['bytes']),
                'open': did.get('open', 'Not Applicable'),
                'complete_rses': ', '.join(rses),
                'approvers': ','.join([r[0] for r in recipients]),
                'approver': recipient[1],
                'target_rse': target_rse,
                'free_space': free_space,
                'free_space_after': free_space_after
            }
        )

        add_message(event_type='email',
                    payload={'body': text,
                             'to': [recipient[0]],
                             'subject': '[RUCIO] Request to approve replication rule %s' % (str(rule.id))},
                    session=session)


@transactional_session
def _create_recipients_list(
    rse_expression: str,
    filter_: Optional[str] = None,
    *,
    session: "Session"
) -> list[tuple[str, Union[str, InternalAccount]]]:
    """
    Create a list of recipients for a notification email based on rse_expression.

    :param rse_expression:  The rse_expression.
    :param session:         The database session in use.
    """

    recipients: list[tuple] = []  # (eMail, account)

    # APPROVERS-LIST
    # If there are accounts in the approvers-list of any of the RSEs only these should be used
    for rse in parse_expression(rse_expression, filter_=filter_, session=session):
        rse_attr = list_rse_attributes(rse_id=rse['id'], session=session)
        if rse_attr.get('rule_approvers'):
            for account in rse_attr.get('rule_approvers').split(','):
                account = InternalAccount(account)
                try:
                    email = get_account(account=account, session=session).email
                    if email:
                        recipients.append((email, account))
                except Exception:
                    pass

    # LOCALGROUPDISK/LOCALGROUPTAPE
    if not recipients:
        for rse in parse_expression(rse_expression, filter_=filter_, session=session):
            rse_attr = list_rse_attributes(rse_id=rse['id'], session=session)
            if rse_attr.get('type', '') in ('LOCALGROUPDISK', 'LOCALGROUPTAPE'):

                query = select(
                    models.AccountAttrAssociation.account
                ).where(
                    models.AccountAttrAssociation.key == f'country-{rse_attr.get("country", "")}',
                    models.AccountAttrAssociation.value == 'admin'
                )

                for account in session.execute(query).scalars().all():
                    try:
                        email = get_account(account=account, session=session).email
                        if email:
                            recipients.append((email, account))
                    except Exception:
                        pass

    # GROUPDISK
    if not recipients:
        for rse in parse_expression(rse_expression, filter_=filter_, session=session):
            rse_attr = list_rse_attributes(rse_id=rse['id'], session=session)
            if rse_attr.get('type', '') == 'GROUPDISK':

                query = select(
                    models.AccountAttrAssociation.account
                ).where(
                    models.AccountAttrAssociation.key == f'group-{rse_attr.get("physgroup", "")}',
                    models.AccountAttrAssociation.value == 'admin'
                )

                for account in session.execute(query).scalars().all():
                    try:
                        email = get_account(account=account, session=session).email
                        if email:
                            recipients.append((email, account))
                    except Exception:
                        pass

    # DDMADMIN as default
    if not recipients:
        default_mail_from = config_get(
            'core', 'default_mail_from', raise_exception=False, default=None
        )
        if default_mail_from:
            recipients = [(default_mail_from, 'ddmadmin')]

    return list(set(recipients))


def __progress_class(replicating_locks, total_locks):
    """
    Returns the progress class (10%, 20%, ...) of currently replicating locks.

    :param replicating_locks:   Currently replicating locks.
    :param total_locks:         Total locks.
    """

    try:
        return int(float(total_locks - replicating_locks) / float(total_locks) * 10) * 10
    except Exception:
        return 0


@policy_filter
@transactional_session
def archive_localgroupdisk_datasets(
    scope: InternalScope,
    name: str,
    *,
    session: "Session",
    logger: LoggerFunction = logging.log
) -> None:
    """
    ATLAS policy to archive a dataset which has a replica on LOCALGROUPDISK

    :param scope:    Scope of the dataset.
    :param name:     Name of the dataset.
    :param session:  The database session in use.
    :param logger:   Optional decorated logger that can be passed from the calling daemons or servers.
    """

    rses_to_rebalance = []

    archive = InternalScope('archive', vo=scope.vo)
    # Check if the archival dataset already exists
    try:
        rucio.core.did.get_did(scope=archive, name=name, session=session)
        return
    except DataIdentifierNotFound:
        pass

    # Check if the dataset has a rule on a LOCALGROUPDISK
    for lock in rucio.core.lock.get_dataset_locks(scope=scope, name=name, session=session):
        if 'LOCALGROUPDISK' in lock['rse']:
            rses_to_rebalance.append({'rse_id': lock['rse_id'], 'rse': lock['rse'], 'account': lock['account']})
    # Remove duplicates from list
    rses_to_rebalance = [dict(t) for t in set([tuple(sorted(d.items())) for d in rses_to_rebalance])]

    # There is at least one rule on LOCALGROUPDISK
    if rses_to_rebalance:
        content = [x for x in rucio.core.did.list_content(scope=scope, name=name, session=session)]
        if content:
            # Create the archival dataset
            did = rucio.core.did.get_did(scope=scope, name=name, session=session)
            meta = rucio.core.did.get_metadata(scope=scope, name=name, session=session)
            new_meta = {k: v for k, v in meta.items() if k in ['project', 'datatype', 'run_number', 'stream_name', 'prod_step', 'version', 'campaign', 'task_id', 'panda_id'] and v is not None}
            rucio.core.did.add_did(scope=archive,
                                   name=name,
                                   did_type=DIDType.DATASET,
                                   account=did['account'],
                                   statuses={},
                                   meta=new_meta,
                                   rules=[],
                                   lifetime=None,
                                   dids=[],
                                   rse_id=None,
                                   session=session)
            rucio.core.did.attach_dids(scope=archive, name=name, dids=content, account=did['account'], session=session)
            if not did['open']:
                rucio.core.did.set_status(scope=archive, name=name, open=False, session=session)

            for rse in rses_to_rebalance:
                add_rule(dids=[{'scope': archive, 'name': name}],
                         account=rse['account'],
                         copies=1,
                         rse_expression=rse['rse'],
                         grouping='DATASET',
                         weight=None,
                         lifetime=None,
                         locked=False,
                         subscription_id=None,
                         ignore_account_limit=True,
                         ignore_availability=True,
                         session=session)
            logger(logging.DEBUG, 'Re-Scoped %s:%s', scope, name)


@policy_filter
@read_session
def get_scratch_policy(
    account: InternalAccount,
    rses: Sequence[dict[str, Any]],
    lifetime: Optional[int],
    *,
    session: "Session"
) -> Optional[int]:
    """
    ATLAS policy for rules on SCRATCHDISK

    :param account:  Account of the rule.
    :param rses:     List of RSEs.
    :param lifetime: Lifetime.
    :param session:  The database session in use.
    """

    scratchdisk_lifetime = get_scratchdisk_lifetime()
    # Check SCRATCHDISK Policy
    if not has_account_attribute(account=account, key='admin', session=session) and (lifetime is None or lifetime > 60 * 60 * 24 * scratchdisk_lifetime):
        # Check if one of the rses is a SCRATCHDISK:
        if [rse for rse in rses if list_rse_attributes(rse_id=rse['id'], session=session).get('type') == 'SCRATCHDISK']:
            lifetime = 60 * 60 * 24 * scratchdisk_lifetime - 1
    return lifetime