Configuration Settings
Cluster coordination is controlled by a number of configuration settings to control behavior and tune failure detection using timeouts.
Global Settings
The following parameters are set in the server.cnf file and apply globally to all clustered notifications for all adapters:
Parameter Name | Description |
watt.art.
clusteredPollingNotification.
keepAliveInterval | Interval in milliseconds, with which a secondary instance will check to see if an executing instance is still alive. If not set, the secondary instance will change to the default maxLockDuration value of 180000 for the shared cache. |
watt.art.
clusteredPollingNotification.
keepAliveExpireTimeout | Time in milliseconds, that an executing node can be late before it is assumed to have failed. In general, this setting must be equal to the amount of drift anticipated on the server clocks. If not set, the secondary instance will change to the default maxLockDuration value of 180000 for the shared cache. |
Note:
The parameters can also be set using Integration Server Administrator. For more information, see webMethods Integration Server Administrator’s Guide.
Adapter-Specific Settings
In the configuration directory of the adapter's package, the clusterProperties.cnf provides settings that specify a callback scheme, and place limits on which coordination modes can be applied to notification nodes for the adapter. The clusterProperties.cnf file is an XML file in which settings may be provided globally for the adapter or specifically to a particular notification template. Template-specific settings use the template class name to set the scope of the setting. The clusterProperties.cnf file is located in the Integration Server_directory \instances\instance_name\packages\AdapterName\config folder.
The following example includes all of the major constructs of a clusterProperties.cnf file:
The outer <clusterProps> wrapper contains the following three elements:
<pollingNotifications> wraps settings for clustered polling notifications. The
<pollingNotifications> wrapper must contain the following global settings that provide the adapter's default values:
A
<callbackScheme>A
<runtimeModeLimit>These can be followed by any number of template wrappers, which contain the <callbackScheme> and <runtimeModeLimit> settings specific to each template. Template specific settings are applied to notification nodes created from the associated template name.
The <callbackScheme> setting controls how callback coordination is performed, while <runtimeModeLimit> constrains the coordination mode setting that can be set for a notification node. Valid values for these settings are included in the following tables.
Values for callbackScheme Setting
| Coordination Modes |
callbackScheme | Enable/Disable | Startup/Shutdown | Resume/Suspend |
0 | No Coordination | No Coordination | No Coordination |
1. Default | Coordinated | No Coordination | No Coordination |
2 | No Coordination | Coordinated | No Coordination |
3 | Coordinated | Coordinated | No Coordination |
4 | No Coordination | No Coordination | Coordinated |
5 | No Coordination | Coordinated | Coordinated |
6 | Coordinated | No Coordination | Coordinated |
7 | Coordinated | Coordinated | Coordinated |
Values for runtimeModeLimit Setting
runtimeModelLimit | Result |
disable | Nodes created using this template cannot be coordinated across a cluster. Nodes are forced into the disabled coordination mode. |
standby. Default | Nodes can be coordinated in standby mode or coordination may be disabled. |
distribute | Nodes can be coordinated in distributed or standby mode, or coordination may be disabled. |
When a polling notification is created or registered on a cluster-aware Integration Server, the Integration Server looks for a clusterProperties.cnf file in the adapter's config directory. If the file contains no entry for the notification's template, a new <template> entry is created using the settings specified globally for all polling notifications. If the file is completely absent or unreadable, a new file is created using the default settings identified earlier.
<listenerNotifications> is for future use. Copy wrapper and contents from the example.
<listeners> is for future use. Copy wrapper and contents from the example.
Node-Specific Settings
The polling notification schedule page in Integration Server Administrator includes cluster settings that are only editable in a clustered environment. On this page, the coordination mode may be set to one of the values supported by its template. For more information, see runtimeModeLimit earlier. In addition, timeouts may be separately configured for polling and setup callback operations.