1. Change the gc_grace_seconds for config daemon keyspaces to the

   default of 10 days
2. Run periodic nodetool repair -pr on the config daemon keyspaces
   from database nodemgr, default is once every 24 hours. This can
   be controlled via DEFAULT.cassandra_repair_interval config in
   /etc/contrail/contrail-database-nodemgr.conf
Partial-Bug: #1484297

Conflicts:
	src/config/common/vnc_cassandra.py
	src/config/schema-transformer/to_bgp.py
	src/config/svc-monitor/svc_monitor/db.py

Change-Id: Ief1f010d0ee3f35fc0f5ff8e1e4eac752c5267e2
(cherry picked from commit 06ac304)

src/config/common/vnc_cassandra.py

@@ -15,6 +15,8 @@
 from pysandesh.gen_py.process_info.ttypes import ConnectionStatus, \
     ConnectionType
 from pysandesh.gen_py.sandesh.ttypes import SandeshLevel
+from sandesh_common.vns.constants import API_SERVER_KEYSPACE_NAME, \
+    CASSANDRA_DEFAULT_GC_GRACE_SECONDS
 import time
 from cfgm_common import jsonutils as json
 import utils
@@ -25,7 +27,7 @@
 
 class VncCassandraClient(object):
     # Name to ID mapping keyspace + tables
-    _UUID_KEYSPACE_NAME = 'config_db_uuid'
+    _UUID_KEYSPACE_NAME = API_SERVER_KEYSPACE_NAME
 
     # TODO describe layout
     _OBJ_UUID_CF_NAME = 'obj_uuid_table'
@@ -174,9 +176,7 @@ def _cassandra_ensure_keyspace(self, server_list,
             # TODO verify only EEXISTS
             self._logger("Warning! " + str(e), level=SandeshLevel.SYS_WARN)
 
-        gc_grace_sec = 0
-        if self._num_dbnodes > 1:
-            gc_grace_sec = 60
+        gc_grace_sec = CASSANDRA_DEFAULT_GC_GRACE_SECONDS
 
         for cf_info in cf_info_list:
             try:

src/config/schema-transformer/db.py

@@ -11,10 +11,11 @@
 import cfgm_common as common
 from cfgm_common.zkclient import IndexAllocator
 from cfgm_common.vnc_cassandra import VncCassandraClient
+from sandesh_common.vns.constants import SCHEMA_KEYSPACE_NAME
 
 class SchemaTransformerDB(VncCassandraClient):
 
-    _KEYSPACE = 'to_bgp_keyspace'
+    _KEYSPACE = SCHEMA_KEYSPACE_NAME
     _RT_CF = 'route_target_table'
     _SC_IP_CF = 'service_chain_ip_address_table'
     _SERVICE_CHAIN_CF = 'service_chain_table'

src/config/svc-monitor/svc_monitor/db.py

@@ -10,11 +10,11 @@
 
 from cfgm_common import jsonutils as json
 from cfgm_common.vnc_cassandra import VncCassandraClient
-
+from sandesh_common.vns.constants import SVC_MONITOR_KEYSPACE_NAME
 
 class ServiceMonitorDB(VncCassandraClient):
 
-    _KEYSPACE = 'svc_monitor_keyspace'
+    _KEYSPACE = SVC_MONITOR_KEYSPACE_NAME
     _SVC_SI_CF = 'service_instance_table'
     _LB_CF = 'pool_table'
 

src/discovery/disc_cassdb.py

@@ -16,9 +16,11 @@
 from pycassa.system_manager import *
 from pycassa.util import *
 from pycassa.types import *
+from sandesh_common.vns.constants import DISCOVERY_SERVER_KEYSPACE_NAME, \
+    CASSANDRA_DEFAULT_GC_GRACE_SECONDS
 
 class DiscoveryCassandraClient(object):
-    _DISCOVERY_KEYSPACE_NAME = 'DISCOVERY_SERVER'
+    _DISCOVERY_KEYSPACE_NAME = DISCOVERY_SERVER_KEYSPACE_NAME
     _DISCOVERY_CF_NAME = 'discovery'
 
     @classmethod
@@ -99,7 +101,8 @@ def _cassandra_ensure_keyspace(self, server_list,
                 (cf_name, comparator_type, validator_type) = cf_info
                 sys_mgr.create_column_family(keyspace_name, cf_name, 
                         comparator_type = comparator_type, default_validation_class = validator_type)
-                sys_mgr.alter_column_family(keyspace_name, cf_name, gc_grace_seconds=0)
+                sys_mgr.alter_column_family(keyspace_name, cf_name,
+                    gc_grace_seconds=CASSANDRA_DEFAULT_GC_GRACE_SECONDS)
             except pycassa.cassandra.ttypes.InvalidRequestException as e:
                 # TODO verify only EEXISTS
                 print "Warning! " + str(e)

src/discovery/disc_server.py

@@ -8,10 +8,10 @@
 """
 
 import gevent
-from disc_cassdb import DiscoveryCassandraClient
 from gevent import monkey
 monkey.patch_all()
 from gevent import hub
+from disc_cassdb import DiscoveryCassandraClient
 
 import sys
 import time

src/nodemgr/database_nodemgr/database_event_manager.py

@@ -27,7 +27,8 @@
 from pysandesh.gen_py.sandesh_trace.ttypes import SandeshTraceRequest
 from sandesh_common.vns.ttypes import Module, NodeType
 from sandesh_common.vns.constants import ModuleNames, NodeTypeNames,\
-    Module2NodeType, INSTANCE_ID_DEFAULT, SERVICE_CONTRAIL_DATABASE
+    Module2NodeType, INSTANCE_ID_DEFAULT, SERVICE_CONTRAIL_DATABASE, \
+    RepairNeededKeyspaces
 from subprocess import Popen, PIPE
 from StringIO import StringIO
 
@@ -43,7 +44,7 @@
 class DatabaseEventManager(EventManager):
     def __init__(self, rule_file, discovery_server,
                  discovery_port, collector_addr,
-                 hostip, minimum_diskgb):
+                 hostip, minimum_diskgb, cassandra_repair_interval):
         EventManager.__init__(
             self, rule_file, discovery_server,
             discovery_port, collector_addr)
@@ -52,6 +53,7 @@ def __init__(self, rule_file, discovery_server,
         self.module_id = ModuleNames[self.module]
         self.hostip = hostip
         self.minimum_diskgb = minimum_diskgb
+        self.cassandra_repair_interval = cassandra_repair_interval
         self.supervisor_serverurl = "unix:///tmp/supervisord_database.sock"
         self.add_current_process()
     # end __init__
@@ -193,6 +195,14 @@ def database_periodic(self):
 
     # end database_periodic
 
+    def cassandra_repair(self):
+        for keyspace in RepairNeededKeyspaces:
+            repair_file_name = '/var/log/cassandra/repair-' + keyspace + '.log'
+            with open(repair_file_name, "a") as repair_file:
+                subprocess.Popen(["nodetool", "repair", "-pr", keyspace],
+                                 stdout=repair_file, stderr=repair_file)
+    #end cassandra_repair
+
     def send_disk_usage_info(self):
         self.send_disk_usage_info_base(
             NodeStatusUVE, NodeStatus, DiskPartitionUsageStats)
@@ -222,4 +232,7 @@ def runforever(self, test=False):
             if headers['eventname'].startswith("TICK_60"):
                 self.database_periodic()
                 prev_current_time = self.event_tick_60(prev_current_time)
+                # Perform nodetool repair every cassandra_repair_interval hours
+                if self.tick_count % (60 * self.cassandra_repair_interval) == 0:
+                    self.cassandra_repair()
             self.listener_nodemgr.ok(self.stdout)

src/nodemgr/main.py

@@ -67,7 +67,8 @@ def main(args_str=' '.join(sys.argv[1:])):
     default = {'rules': '',
                'collectors': [],
                'hostip': '127.0.0.1',
-               'minimum_diskgb': 256
+               'minimum_diskgb': 256,
+               'cassandra_repair_interval': 24,
               }
     node_type = args.nodetype
     if (node_type == 'contrail-analytics'):
@@ -119,6 +120,9 @@ def main(args_str=' '.join(sys.argv[1:])):
                             help="Minimum disk space in GB's")
         parser.add_argument("--hostip",
                             help="IP address of host")
+        parser.add_argument("--cassandra_repair_interval", type=int,
+                            help="Time in hours to periodically run "
+                            "nodetool repair for cassandra maintenance")
     try:
         _args = parser.parse_args(remaining_argv)
     except:
@@ -157,10 +161,11 @@ def main(args_str=' '.join(sys.argv[1:])):
     elif (node_type == 'contrail-database'):
         hostip = _args.hostip
         minimum_diskgb = _args.minimum_diskgb
+        cassandra_repair_interval = _args.cassandra_repair_interval
         prog = DatabaseEventManager(
             rule_file, discovery_server,
             discovery_port, collector_addr,
-            hostip, minimum_diskgb)
+            hostip, minimum_diskgb, cassandra_repair_interval)
     else:
         sys.stderr.write("Node type" + str(node_type) + "is incorrect" + "\n")
         return

src/sandesh/common/vns.sandesh

@@ -432,3 +432,19 @@ const string ALARM_GENERATOR_SERVICE_NAME = "AlarmGenerator"
 const string IFMAP_SERVER_DISCOVERY_SERVICE_NAME = "IfmapServer"
 const string XMPP_SERVER_DISCOVERY_SERVICE_NAME = "xmpp-server"
 const string DNS_SERVER_DISCOVERY_SERVICE_NAME = "dns-server"
+
+// Cassandra Keyspace Names
+const string SVC_MONITOR_KEYSPACE_NAME = "svc_monitor_keyspace"
+const string SCHEMA_KEYSPACE_NAME = "to_bgp_keyspace"
+const string API_SERVER_KEYSPACE_NAME = "config_db_uuid"
+const string DISCOVERY_SERVER_KEYSPACE_NAME = "DISCOVERY_SERVER"
+
+const list<string> RepairNeededKeyspaces = [
+    API_SERVER_KEYSPACE_NAME,
+    SCHEMA_KEYSPACE_NAME,
+    SVC_MONITOR_KEYSPACE_NAME,
+    DISCOVERY_SERVER_KEYSPACE_NAME,
+]
+
+// Set to default of 10 days
+const u32 CASSANDRA_DEFAULT_GC_GRACE_SECONDS = 864000