Cluster-wide Alert Configuration

LiveView supports a cluster-wide alerting service for backend LiveView servers. Multiple alerts can belong to an alert group, and multiple alert groups can be specified in a LiveView cluster. In contrast to release 10.4, alerts processing is synchronized cluster-wide as of 10.5.0.

Each server supports exactly one alert group. On startup, the first node becomes the master (this is normal server behavior even with no clustering) for alerts configured in its alert group. When other LiveView nodes in the same alert group start up, they join the cluster and become hot spares. From an alert group perspective, the current master remains in that role until it is removed from the cluster. All servers issue the alert queries for all alerts in its group, and hot spares put the alerts into a local queue and expect to see the alerts being published from the master. Additionally, since configuration of which servers are in alert groups is done at install time, there can be multiple masters — each with their own alert group and behaving as described above.

Alerts are executed by one LiveView server (the master) in an alert group. Other servers in the cluster (that are in the alert group) serve as hot spares and do not execute alert actions; they simply verify the master's presence, waiting to take over if the master goes down.

Configuration of LiveView cluster-wide alerting is bound by the following behavior:

If an alert is not assigned to a specific alert group, it is assigned to the default alert group (which is configured with the empty string: null, &&, or "").

TIBCO recommends using the default alert group for the vast majority of use cases. The default alert group requires no additional configuration; all backend servers in the cluster will simply process alerts.

For the non-default alert group case, you can specify multiple alert groups. In this scenario, you designate which nodes belong to each alert group via the node name. The default alert group is where you do not add the separator in the node name. However, suppose you name your node foo-ag_bar.cluster. In this example bar is the alert group the node is in, as it is between the separator and the first dot ".". If you name your node foo.cluster, your node is in the default alert group. Note that alert group behavior described above still applies.

You must specify:

Place your configuration files in the src/main/configurations folder of your node's LiveView project.

name = "myldmengine"
version = "1.0.0"
type = "com.tibco.ep.ldm.configuration.ldmengine"
configuration = {
  LDMEngine = {
    ldm = {
      metadataStore = {
        storeType = "JDBC" 
        jdbcDataSource = "myJDBCSource"        
        jdbcMetadataTableNamePrefix = "LV_CLUSTER1_"       
      }
    }
  }

name = "myldmengine"
version = "1.0.0"
type = "com.tibco.ep.ldm.configuration.ldmengine"
configuration = {
  LDMEngine = {
    ldm = {
      metadataStore = {
        storeType = "TRANSACTIONAL_MEMORY"        
      }
    }
  }

name = "myldmengine"
version = "1.0.0"
type = "com.tibco.ep.ldm.configuration.ldmengine"
configuration = {
  LDMEngine = {
    systemProperties = { 
      "liveview.alerts.voting.dynamicpartition.pattern.suffix" = "-ag_" 
    }
  }
}

name = "mydatasource"
version = "1.0.0"
type = "com.tibco.ep.streambase.configuration.jdbcdatasource"
configuration = {
  JDBCDataSourceGroup = {
    associatedWithEngines = [ "javaengine", "otherengine[0-9]" ]
    jdbcDataSources = {    
      myJDBCSource = {   
        ...
      }
    }
  }
} 

name = "my-role-mappings"
version = "1.0.0"
type = "com.tibco.ep.dtm.configuration.security"
configuration = {
  RoleToPrivilegeMappings = {
    privileges = {
      admin = [
        {
          privilege = "LiveViewAll"
        }
      ]
    }
  }
}

This section discusses LiveView client-side alert rule configuration, which is required when setting up alerts across a cluster of LiveView nodes.

As mentioned previously, you can always set an alert rule to run in the default alert group, which is TIBCO's recommendation. For all other cases, you must have permissions to set the specific alert group.

If you provide an invalid node name or alert group name, LiveView assumes the name is correct (for example, for a future node of that name that may join the cluster). See Server-Disabled Alert Rules for more information.

If you configure an alert rule alert group with a star "*" the alert runs the alert on every node in the cluster (as it did in before release 10.5.0, when alert groups were introduced).

Clients must be 10.5.0 or above in order to configure alert groups. Otherwise, clients default to the default alert group.

The "other special alert group name" is the node name itself. Configuring the alert rule to run on the node name runs the alert on that one node only.

The following shows alert rule configuration for an alert group using lv-client:

lv-client
...
addalertrule --alertgroup myalertgroup