Recovery: Adjust suspicious replica recoverer creation of rules. ruci…

…o#6337

bin/rucio-replica-recoverer

@@ -228,8 +228,8 @@ Note that attempting the use the ``--vos`` argument when in single-VO mode will
   $ rucio-replica-recoverer --run-once --vos abc xyz
   2020-07-28 15:21:33,349 5488    WARNING Ignoring argument vos, this is only applicable in a multi-VO setup.
 ''', formatter_class=argparse.RawDescriptionHelpFormatter)  # NOQA: E501
-    parser.add_argument("--nattempts", action="store", default=10, help='Minimum count of suspicious file replica appearance in bad_replicas table. Default value is 10.')
-    parser.add_argument("--younger-than", action="store", default=3, help='Consider all file replicas logged in bad_replicas table since speicified number of younger-than days. Default value is 3.')
+    parser.add_argument("--nattempts", action="store", default=5, help='Minimum count of suspicious file replica appearance in bad_replicas table. Default value is 5.')
+    parser.add_argument("--younger-than", action="store", default=5, help='Consider all file replicas logged in bad_replicas table since speicified number of younger-than days. Default value is 5.')
     parser.add_argument('--vos', nargs='+', type=str, help='Optional list of VOs to consider. Only used in multi-VO mode.')
     parser.add_argument("--run-once", action="store_true", default=False, help='One iteration only.')
     parser.add_argument("--limit-suspicious-files-on-rse", action="store", default=5, help='Maximum number of suspicious replicas on an RSE before that RSE is considered problematic and the suspicious replicas on that RSE are declared "TEMPORARY_UNAVAILABLE". Default value is 5.')

etc/suspicious_replica_recoverer.json

@@ -9,4 +9,9 @@
         "datatype": ["RAW"],
         "scope": []
     }
+    {
+        "action": "dry run",
+        "datatype": [],
+        "scope": ["mc.*"]
+    }
 ]

lib/rucio/core/replica.py

@@ -3406,7 +3406,7 @@ def get_replicas_state(scope=None, name=None, *, session: "Session"):
 
 
 @read_session
-def get_suspicious_files(rse_expression, available_elsewhere, filter_=None, logger=logging.log, younger_than=10, nattempts=0, nattempts_exact=False, *, session: "Session", exclude_states=['B', 'R', 'D'], is_suspicious=False):
+def get_suspicious_files(rse_expression, available_elsewhere, filter_=None, logger=logging.log, younger_than=5, nattempts=0, nattempts_exact=False, *, session: "Session", exclude_states=['B', 'R', 'D'], is_suspicious=False):
     """
     Gets a list of replicas from bad_replicas table which are: declared more than <nattempts> times since <younger_than> date,
     present on the RSE specified by the <rse_expression> and do not have a state in <exclude_states> list.
@@ -3539,7 +3539,12 @@ def get_suspicious_reason(rse_id, scope, name, nattempts=0, logger=logging.log,
     query = session.query(bad_replicas_alias.scope, bad_replicas_alias.name, bad_replicas_alias.reason, bad_replicas_alias.rse_id)\
                    .filter(bad_replicas_alias.rse_id == rse_id,
                            bad_replicas_alias.scope == scope,
-                           bad_replicas_alias.name == name)
+                           bad_replicas_alias.name == name,
+                           bad_replicas_alias.state == 'S',
+                           ~exists(select(1).where(and_(bad_replicas_alias.rse_id == rse_id,
+                                                        bad_replicas_alias.scope == scope,
+                                                        bad_replicas_alias.name == name,
+                                                        bad_replicas_alias.state != 'S',))))
     count = query.count()
 
     query_result = query.group_by(bad_replicas_alias.rse_id, bad_replicas_alias.scope, bad_replicas_alias.name, bad_replicas_alias.reason).having(func.count() > nattempts).all()

lib/rucio/daemons/auditor/__init__.py

@@ -155,7 +155,7 @@ def process_output(output, sanity_check=True, compress=True):
     rse = os.path.basename(output[:output.rfind('_')])
     rse_id = get_rse_id(rse=rse)
     usage = get_rse_usage(rse_id=rse_id, source='rucio')[0]
-    threshold = config.config_get_float('auditor', 'threshold', False, 0.2)
+    threshold = config.config_get_float('auditor', 'threshold', False, 0.1)
 
     # Perform a basic sanity check by comparing the number of entries
     # with the total number of files on the RSE.  If the percentage is