[v7r2] fixes for ES docs, SummarizeLogsAgent and StatesAccountingAgent can also commit to Monitoring

docs/source/AdministratorGuide/Systems/MonitoringSystem/index.rst

@@ -10,32 +10,34 @@ Monitoring System
 Overview
 =========
 
-The Monitoring system is used to monitor various components of DIRAC. Currently, we have two monitoring types:
+The Monitoring system is used to monitor various components of DIRAC. Currently, we have three monitoring types:
 
   - WMSHistory: for monitoring the DIRAC WMS
-  - Component Monitoring: for monitoring  DIRAC components such as services, agents, etc.
+  - Component Monitoring: for monitoring DIRAC components such as services, agents, etc.
+  - RMS Monitoring: for monitoring the DIRAC RequestManagement System (mostly the Request Executing Agent).
 
-It is based on Elasticsearch distributed search and analytics NoSQL database. If you want to use it, you have to install the Monitoring service and
-elasticsearch db. You can use a single node, if you do not have to store lot of data, otherwise you need a cluster (more than one node).
+It is based on Elasticsearch distributed search and analytics NoSQL database.
+If you want to use it, you have to install the Monitoring service, and of course connect to a ElasticSearch instance.
 
 Install Elasticsearch
 ======================
 
-You can found in https://www.elastic.co official web site. I propose to use standard tools to install for example: yum, rpm, etc. otherwise
-you encounter some problems. If you are not familiar with managing linux packages, you have to ask your college or read some relevant documents.
+This is not covered here, as installation and administration of ES are not part of DIRAC guide.
+Just a note on the ES versions supported: ES7 and ES6 are supported, the support for ES5 is not assured,
+and the one for ES6 will be dropped in a future release.
 
 Configure the MonitoringSystem
 ===============================
 
-You can run your El cluster without authentication or using User name and password. You have to add the following parameters:
+You can run your Elastic cluster even without authentication, or using User name and password. You have to add the following parameters:
 
   - User
   - Password
   - Host
   - Port
 
-The User name and Password must be added to the local cfg file while the other can be added to the CS using the Configuration web application.
-You have to handle the EL secret information in a similar way to what is done for the other supported SQL databases, e.g. MySQL
+The *User* name and *Password* must be added to the local cfg file while the other can be added to the CS using the Configuration web application.
+You have to handle the ES secret information in a similar way to what is done for the other supported SQL databases, e.g. MySQL
 
 
 For example::
@@ -47,24 +49,54 @@ For example::
        User = test
        Password = password
      }
+   }
+
+
+The following option can be set in `Systems/Monitoring/<Setup>/Databases/MonitoringDB`:
+
+   *IndexPrefix*:  Prefix used to prepend to indexes created in the ES instance. If this
+                   is not present in the CS, the indices are prefixed with the setup name.
+
+For each monitoring types managed, the Period (how often a new index is created)
+can be defined with::
 
+   MonitoringTypes
+   {
+     ComponentMonitoring
+     {
+       # Indexing strategy. Possible values: day, week, month, year, null
+       Period = month
+     }
+     RMSMonitoring
+     {
+       # Indexing strategy. Possible values: day, week, month, year, null
+       Period = month
+     }
+     WMSHistory
+     {
+       # Indexing strategy. Possible values: day, week, month, year, null
+       Period = day
+     }
    }
 
+The given periods above are also the default periods in the code.
+
 
 Enable WMSHistory monitoring
 ============================
 
-You have to install the WorkloadManagemet/StatesMonitoringAgent. This agent is used to collect information using the JobDB and send it to the Elasticsearch database.
-If you install this agent, you can stop the StatesAccounting agent.
+You have to add ``Monitoring`` to the ``Backends`` option of WorkloadManagemet/StatesAccountingAgent.
+If you do so, this agent will collect information using the JobDB and send it to the Elasticsearch database.
+This same agent can also report to the MySQL backend of the Accounting system (which is in fact the default).
 
-Note: You can use RabbitMQ for failover. This is optional as the agent already has a failover mechanism. You can configure RabbitMQ in the local dirac.cfg file
-where the agent is running::
+Optionally, you can use an MQ system (like RabbitMQ) for failover, even though the agent already has a simple failover mechanism.
+You can configure the MQ in the local dirac.cfg file where the agent is running::
 
    Resources
    {
      MQServices
      {
-       hostname (for example lbvobox10.cern.ch)
+       hostname.some.where
        {
          MQType = Stomp
          Port = 61613
@@ -86,24 +118,59 @@ where the agent is running::
 Enable Component monitoring
 ===========================
 
-You have to set DynamicMonitoring=True in the CS::
+You have to set ``DynamicMonitoring=True`` in the CS::
 
    Systems
    {
-      Framework
-      {
-          SystemAdministrator
+     Framework
+     {
+       <instance>
+       {
+         Services
+         {
+           SystemAdministrator
            {
-              ...
-              DynamicMonitoring = True
-            }
+             ...
+             DynamicMonitoring = True
+           }
          }
        }
+     }
+   }
 
 
 .. image:: cs.png
    :align: center
 
+
+Enable RMS Monitoring
+=====================
+
+In order to enable RMSMonitoring we need to set value of ``EnableActivityMonitoring`` flag to yes/true in the CS::
+
+
+   Systems
+   {
+     RequestManagement
+     {
+       <instance>
+       {
+         Agents
+         {
+           RequestExecutingAgent
+           {
+             ...
+             EnableActivityMonitoring = True
+           }
+         }
+       }
+     }
+   }
+
+
+or inside the ``/Operations`` section as a general flag.
+
+
 Accessing the Monitoring information
 =====================================
 

src/DIRAC/MonitoringSystem/Client/MonitoringReporter.py

@@ -37,12 +37,11 @@ class MonitoringReporter(object):
   """
   .. class:: MonitoringReporter
 
-  This class is used to interact with the db using failover mechanism.
+  This class is used to interact with the ES DB, using a MQ as a failover mechanism.
 
   :param int __maxRecordsInABundle: limit the number of records to be inserted to the db.
   :param threading.RLock __documentLock: is used to lock the local store when it is being modified.
-  :param __documents: contains the recods which will be inserted to the db.
-  :type __documents: python:list
+  :param list __documents: contains the records which will be inserted to the db.
   :param str __monitoringType: type of the records which will be inserted to the db. For example: WMSHistory.
   :param str __failoverQueueName: the name of the messaging queue. For example: /queue/dirac.certification
   """
@@ -62,8 +61,8 @@ def __del__(self):
 
   def processRecords(self):
     """
-    It consumes all messages from the MQ (these are failover messages). In case of failure, the messages
-    will be inserted to the MQ again.
+    It consumes all messages from the MQ (these are failover messages).
+    In case of failure, the messages will be inserted to the MQ again.
     """
     retVal = monitoringDB.pingDB()  # if the db is not accessible, the records will be not processed from MQ
     if retVal['OK']:
@@ -74,10 +73,9 @@ def processRecords(self):
 
     result = createConsumer("Monitoring::Queues::%s" % self.__failoverQueueName)
     if not result['OK']:
-      gLogger.error("Fail to create Consumer: %s" % result['Message'])
-      return S_ERROR("Fail to create Consumer: %s" % result['Message'])
-    else:
-      mqConsumer = result['Value']
+      gLogger.error("Fail to create Consumer", result['Message'])
+      return S_ERROR("Fail to create Consumer")
+    mqConsumer = result['Value']
 
     result = S_OK()
     failedToProcess = []
@@ -128,8 +126,8 @@ def publishRecords(self, records, mqProducer=None):
 
   def commit(self):
     """
-    It inserts the accumulated data to the db. In case of failure
-    it keeps in memory/MQ
+    It inserts the accumulated data to the db.
+    In case of failure it keeps in memory/MQ
     """
     # before we try to insert the data to the db, we process all the data
     # which are already in the queue

src/DIRAC/MonitoringSystem/ConfigTemplate.cfg

@@ -6,30 +6,12 @@ Services
     Port = 9137
     Authorization
     {
-    Default = authenticated
+      Default = authenticated
       FileTransfer
       {
         Default = authenticated
       }
     }
-    MonitoringTypes
-    {
-      ComponentMonitoring
-      {
-        # Indexing strategy. Possible values: day, week, month, year, null
-        Period = month
-      }
-      RMSMonitoring
-      {
-        # Indexing strategy. Possible values: day, week, month, year, null
-        Period = month
-      }
-      WMSHistory
-      {
-        # Indexing strategy. Possible values: day, week, month, year, null
-        Period = day
-      }
-    }
   }
   ##END
-}
+}

src/DIRAC/MonitoringSystem/DB/MonitoringDB.py

@@ -3,11 +3,34 @@
 
 **Configuration Parameters**:
 
-The following options can be set in ``Systems/Monitoring/<Setup>/Databases/MonitoringDB``
+The following option can be set in `Systems/Monitoring/<Setup>/Databases/MonitoringDB`
 
 * *IndexPrefix*:  Prefix used to prepend to indexes created in the ES instance. If this
                   is not present in the CS, the indexes are prefixed with the setup name.
 
+For each monitoring types managed, the Period (how often a new index is created)
+can be defined with::
+
+   MonitoringTypes
+   {
+     ComponentMonitoring
+     {
+       # Indexing strategy. Possible values: day, week, month, year, null
+       Period = month
+     }
+     RMSMonitoring
+     {
+       # Indexing strategy. Possible values: day, week, month, year, null
+       Period = month
+     }
+     WMSHistory
+     {
+       # Indexing strategy. Possible values: day, week, month, year, null
+       Period = day
+     }
+   }
+
+
 """
 from __future__ import absolute_import
 from __future__ import division

src/DIRAC/ResourceStatusSystem/Agent/SummarizeLogsAgent.py

@@ -101,8 +101,8 @@ def execute(self):
           self.log.error(deleteResult['Message'])
           continue
 
-      if self.months:
-        self._removeOldHistoryEntries(element, self.months)
+    if self.months:
+      self._removeOldHistoryEntries(element, self.months)
 
     return S_OK()
 
@@ -252,6 +252,7 @@ def _removeOldHistoryEntries(self, element, months):
     :return: S_OK / S_ERROR
     """
     toRemove = datetime.utcnow().replace(microsecond=0) - timedelta(days=30 * months)
+    self.log.info("Removing history entries", "older than %s" % toRemove)
 
     deleteResult = self.rsClient.deleteStatusElement(element, 'History',
                                                      meta={'older': ['DateEffective', toRemove]})