def test_to_repair_a_rule_with_only_1_rse_whose_transfers_failed(self):
""" JUDGE REPAIRER: Test to repair a rule with only 1 rse whose transfers failed (lock)"""
rule_repairer(once=True) # Clean out the repairer
scope = InternalScope('mock', **self.vo)
files = create_files(4, scope, self.rse4_id, bytes=100)
dataset = 'dataset_' + str(uuid())
add_did(scope, dataset, DIDType.DATASET, self.jdoe)
attach_dids(scope, dataset, files, self.jdoe)
rule_id = add_rule(dids=[{'scope': scope, 'name': dataset}], account=self.jdoe, copies=1, rse_expression=self.rse1, grouping='DATASET', weight=None, lifetime=None, locked=False, subscription_id=None)[0]
successful_transfer(scope=scope, name=files[0]['name'], rse_id=get_replica_locks(scope=files[0]['scope'], name=files[2]['name'])[0].rse_id, nowait=False)
successful_transfer(scope=scope, name=files[1]['name'], rse_id=get_replica_locks(scope=files[1]['scope'], name=files[2]['name'])[0].rse_id, nowait=False)
failed_transfer(scope=scope, name=files[2]['name'], rse_id=get_replica_locks(scope=files[2]['scope'], name=files[2]['name'])[0].rse_id)
failed_transfer(scope=scope, name=files[3]['name'], rse_id=get_replica_locks(scope=files[3]['scope'], name=files[3]['name'])[0].rse_id)
cancel_request_did(scope=scope, name=files[2]['name'], dest_rse_id=get_replica_locks(scope=files[2]['scope'], name=files[2]['name'])[0].rse_id)
cancel_request_did(scope=scope, name=files[3]['name'], dest_rse_id=get_replica_locks(scope=files[3]['scope'], name=files[2]['name'])[0].rse_id)
assert(rule_id == get_rule(rule_id)['id'].replace('-', '').lower())
assert(RuleState.STUCK == get_rule(rule_id)['state'])
rule_repairer(once=True)
# Stil assert STUCK because of delays:
assert(RuleState.STUCK == get_rule(rule_id)['state'])
assert(get_replica_locks(scope=files[2]['scope'], name=files[2]['name'])[0].rse_id == get_replica_locks(scope=files[3]['scope'], name=files[3]['name'])[0].rse_id)
def test_to_repair_a_rule_with_DATASET_grouping_whose_transfer_failed(
self):
""" JUDGE REPAIRER: Test to repair a rule with 1 failed transfer (lock)"""
rule_repairer(once=True) # Clean out the repairer
scope = InternalScope('mock')
files = create_files(4, scope, self.rse4_id, bytes=100)
dataset = 'dataset_' + str(uuid())
add_did(scope, dataset, DIDType.from_sym('DATASET'), self.jdoe)
attach_dids(scope, dataset, files, self.jdoe)
rule_id = add_rule(dids=[{
'scope': scope,
'name': dataset
}],
account=self.jdoe,
copies=1,
rse_expression=self.T1,
grouping='DATASET',
weight=None,
lifetime=None,
locked=False,
subscription_id=None,
activity='DebugJudge')[0]
successful_transfer(
scope=scope,
name=files[0]['name'],
rse_id=get_replica_locks(scope=files[0]['scope'],
name=files[2]['name'])[0].rse_id,
nowait=False)
successful_transfer(
scope=scope,
name=files[1]['name'],
rse_id=get_replica_locks(scope=files[1]['scope'],
name=files[2]['name'])[0].rse_id,
nowait=False)
failed_transfer(
scope=scope,
name=files[2]['name'],
rse_id=get_replica_locks(scope=files[2]['scope'],
name=files[2]['name'])[0].rse_id)
failed_transfer(
scope=scope,
name=files[3]['name'],
rse_id=get_replica_locks(scope=files[3]['scope'],
name=files[3]['name'])[0].rse_id)
assert (rule_id == get_rule(rule_id)['id'].replace('-', '').lower())
assert (RuleState.STUCK == get_rule(rule_id)['state'])
rule_repairer(once=True)
assert (RuleState.REPLICATING == get_rule(rule_id)['state'])
assert (get_replica_locks(
scope=files[2]['scope'],
name=files[2]['name'])[0].rse_id == get_replica_locks(
scope=files[3]['scope'], name=files[3]['name'])[0].rse_id)
assert (get_replica_locks(
scope=files[1]['scope'],
name=files[1]['name'])[0].rse_id == get_replica_locks(
scope=files[3]['scope'], name=files[3]['name'])[0].rse_id)
def test_to_repair_a_rule_with_NONE_grouping_whose_transfer_failed(self):
""" JUDGE REPAIRER: Test to repair a rule with 1 failed transfer (lock)"""
rule_repairer(once=True) # Clean out the repairer
scope = InternalScope('mock', **self.vo)
files = create_files(3, scope, self.rse4_id, bytes=100)
dataset = 'dataset_' + str(uuid())
add_did(scope, dataset, DIDType.DATASET, self.jdoe)
attach_dids(scope, dataset, files, self.jdoe)
rule_id = add_rule(dids=[{'scope': scope, 'name': dataset}], account=self.jdoe, copies=1, rse_expression=self.T1, grouping='NONE', weight=None, lifetime=None, locked=False, subscription_id=None)[0]
failed_rse_id = get_replica_locks(scope=files[2]['scope'], name=files[2]['name'])[0].rse_id
assert(get_replica(scope=files[2]['scope'], name=files[2]['name'], rse_id=failed_rse_id)['state'] == ReplicaState.COPYING)
assert(get_replica(scope=files[2]['scope'], name=files[2]['name'], rse_id=failed_rse_id)['lock_cnt'] == 1)
successful_transfer(scope=scope, name=files[0]['name'], rse_id=get_replica_locks(scope=files[0]['scope'], name=files[2]['name'])[0].rse_id, nowait=False)
successful_transfer(scope=scope, name=files[1]['name'], rse_id=get_replica_locks(scope=files[1]['scope'], name=files[2]['name'])[0].rse_id, nowait=False)
failed_transfer(scope=scope, name=files[2]['name'], rse_id=get_replica_locks(scope=files[2]['scope'], name=files[2]['name'])[0].rse_id)
assert(rule_id == get_rule(rule_id)['id'].replace('-', '').lower())
assert(RuleState.STUCK == get_rule(rule_id)['state'])
rule_repairer(once=True)
assert(RuleState.REPLICATING == get_rule(rule_id)['state'])
assert(get_replica(scope=files[2]['scope'], name=files[2]['name'], rse_id=failed_rse_id)['state'] == ReplicaState.UNAVAILABLE)
assert(get_replica(scope=files[2]['scope'], name=files[2]['name'], rse_id=failed_rse_id)['lock_cnt'] == 0)
def test_to_repair_a_rule_with_only_1_rse_whose_transfers_failed(self):
""" JUDGE REPAIRER: Test to repair a rule with only 1 rse whose transfers failed (lock)"""
rule_repairer(once=True) # Clean out the repairer
scope = 'mock'
files = create_files(4, scope, self.rse4, bytes=100)
dataset = 'dataset_' + str(uuid())
add_did(scope, dataset, DIDType.from_sym('DATASET'), 'jdoe')
attach_dids(scope, dataset, files, 'jdoe')
rule_id = add_rule(dids=[{'scope': scope, 'name': dataset}], account='jdoe', copies=1, rse_expression=self.rse1, grouping='DATASET', weight=None, lifetime=None, locked=False, subscription_id=None)[0]
successful_transfer(scope=scope, name=files[0]['name'], rse_id=get_replica_locks(scope=files[0]['scope'], name=files[2]['name'])[0].rse_id, nowait=False)
successful_transfer(scope=scope, name=files[1]['name'], rse_id=get_replica_locks(scope=files[1]['scope'], name=files[2]['name'])[0].rse_id, nowait=False)
failed_transfer(scope=scope, name=files[2]['name'], rse_id=get_replica_locks(scope=files[2]['scope'], name=files[2]['name'])[0].rse_id)
failed_transfer(scope=scope, name=files[3]['name'], rse_id=get_replica_locks(scope=files[3]['scope'], name=files[3]['name'])[0].rse_id)
cancel_request_did(scope=scope, name=files[2]['name'], dest_rse_id=get_replica_locks(scope=files[2]['scope'], name=files[2]['name'])[0].rse_id)
cancel_request_did(scope=scope, name=files[3]['name'], dest_rse_id=get_replica_locks(scope=files[3]['scope'], name=files[2]['name'])[0].rse_id)
assert(rule_id == get_rule(rule_id)['id'].replace('-', '').lower())
assert(RuleState.STUCK == get_rule(rule_id)['state'])
rule_repairer(once=True)
# Stil assert STUCK because of delays:
assert(RuleState.STUCK == get_rule(rule_id)['state'])
assert(get_replica_locks(scope=files[2]['scope'], name=files[2]['name'])[0].rse_id == get_replica_locks(scope=files[3]['scope'], name=files[3]['name'])[0].rse_id)
def test_to_repair_a_rule_with_only_1_rse_whose_site_is_blocklisted(self):
""" JUDGE REPAIRER: Test to repair a rule with only 1 rse whose site is blocklisted"""
rse = rse_name_generator()
rse_id = add_rse(rse, **self.vo)
set_local_account_limit(self.jdoe, rse_id, -1)
rule_repairer(once=True) # Clean out the repairer
region = make_region().configure('dogpile.cache.memcached',
expiration_time=900,
arguments={
'url':
config_get(
'cache', 'url', False,
'127.0.0.1:11211'),
'distributed_lock':
True
})
def change_availability(new_value):
update_rse(rse_id, {'availability_write': new_value})
# clear cache
region.delete(sha256(rse.encode()).hexdigest())
for grouping, ignore_availability in itertools.product(
["NONE", "DATASET", "ALL"], [True, False]):
scope = InternalScope('mock', **self.vo)
files = create_files(1, scope, self.rse4_id, bytes_=100)
dataset = 'dataset_' + str(uuid())
add_did(scope, dataset, DIDType.DATASET, self.jdoe)
attach_dids(scope, dataset, files, self.jdoe)
if ignore_availability:
change_availability(False)
rule_id = add_rule(dids=[{
'scope': scope,
'name': dataset
}],
account=self.jdoe,
copies=1,
rse_expression=rse,
grouping=grouping,
weight=None,
lifetime=None,
locked=False,
subscription_id=None,
ignore_availability=ignore_availability,
activity='DebugJudge')[0]
assert (RuleState.STUCK == get_rule(rule_id)['state'])
rule_repairer(once=True)
assert (RuleState.REPLICATING == get_rule(rule_id)['state'])
change_availability(True)
else:
rule_id = add_rule(dids=[{
'scope': scope,
'name': dataset
}],
account=self.jdoe,
copies=1,
rse_expression=rse,
grouping=grouping,
weight=None,
lifetime=None,
locked=False,
subscription_id=None,
ignore_availability=ignore_availability,
activity='DebugJudge')[0]
failed_transfer(scope=scope,
name=files[0]['name'],
rse_id=get_replica_locks(
scope=files[0]['scope'],
name=files[0]['name'])[0].rse_id)
change_availability(False)
assert (RuleState.STUCK == get_rule(rule_id)['state'])
rule_repairer(once=True)
assert (RuleState.STUCK == get_rule(rule_id)['state'])
change_availability(True)
rule_repairer(once=True)
assert (RuleState.REPLICATING == get_rule(rule_id)['state'])
def update_bad_request(req, dest_rse, new_state, detail, session=None):
if new_state == RequestState.FAILED:
request.set_request_state(req['request_id'], new_state, session=session)
activity = 'default'
if req['attributes']:
if type(req['attributes']) is dict:
req_attributes = json.loads(json.dumps(req['attributes']))
else:
req_attributes = json.loads(str(req['attributes']))
activity = req_attributes['activity'] if req_attributes['activity'] else 'default'
tss = time.time()
add_message('transfer-failed', {'activity': activity,
'request-id': req['request_id'],
'checksum-adler': None,
'checksum-md5': None,
'dst-rse': dest_rse,
'dst-url': None,
'name': req['name'],
'guid': None,
'file-size': None,
'previous-request-id': req['request_id'],
'protocol': None,
'reason': detail,
'transfer-link': None,
'scope': req['scope'],
'src-rse': None,
'src-url': None,
'tool-id': 'rucio-conveyor',
'transfer-endpoint': config_get('conveyor', 'ftshosts'),
'transfer-id': None},
session=session)
request.archive_request(req['request_id'], session=session)
logging.error('BAD DID %s:%s REQUEST %s details: %s' % (req['scope'],
req['name'],
req['request_id'],
detail))
try:
replica.update_replicas_states([{'rse': dest_rse,
'scope': req['scope'],
'name': req['name'],
'state': ReplicaState.UNAVAILABLE}],
session=session)
except:
logging.critical("Could not update replica state for failed transfer %s:%s at %s (%s)" % (req['scope'],
req['name'],
dest_rse,
traceback.format_exc()))
raise
tss = time.time()
try:
lock.failed_transfer(req['scope'],
req['name'],
req['dest_rse_id'],
session=session)
except:
logging.warn('Could not update lock for failed transfer %s:%s at %s (%s)' % (req['scope'],
req['name'],
dest_rse,
traceback.format_exc()))
raise
record_timer('daemons.conveyor.common.update_request_state.lock-failed_transfer', (time.time()-tss)*1000)
replica.update_replicas_states([{'rse': rse_update_name,
'scope': response['scope'],
'name': response['name'],
'state': ReplicaState.UNAVAILABLE}],
session=session)
except:
logging.critical("Could not update replica state for failed transfer %s:%s at %s (%s)" % (response['scope'],
response['name'],
rse_update_name,
traceback.format_exc()))
raise
tss = time.time()
try:
lock.failed_transfer(response['scope'],
response['name'],
rse_core.get_rse_id(rse=rse_update_name, session=session),
session=session)
except:
logging.warn('Could not update lock for failed transfer %s:%s at %s (%s)' % (response['scope'],
response['name'],
rse_update_name,
traceback.format_exc()))
raise
record_timer('daemons.conveyor.common.update_request_state.lock-failed_transfer', (time.time()-tss)*1000)
else:
logging.warn('REQUEUED DID %s:%s REQUEST %s AS %s TRY %s' % (response['scope'],
response['name'],
response['request_id'],
new_req['request_id'],
new_req['retry_count']))
