Posts tagged "clustered"

LiveCycle ES2: workload distribution of batch events in clustered environments

Description

 If you are producing and consuming batch events in a clustered environment in LiveCycle ES2 you may notice that all events produced are consumed on the same cluster node.  This is the expected behaviour even though it may seem as though LiveCycle is not making use of the distribution benefits of a clustered environment.

This behaviour is related to the run-time property of the process, i.e. short-lived (synchronous) versus long-lived (asynchronous).

Explanation

Producer process Consumer process Result
Short-lived Long-lived Distributed
Short-lived Short-lived Same node
Long-lived Short-lived Same node
Long-lived Long-lived Distributed

When the consumer process is short-lived, the event is created on the 1st cluster node (server1), and a work item to handle the event is created and assigned to the workmanager on server1.  The workmanager handles the work item and synchronously invokes the short-lived consumer process.  The execution of that process happens in the same workmanager thread and so everything happens only on server1.

When the consumer process is long-lived, the event is created on server1, and a work item to handle the event is created and assigned to the workmanager on server1.  The workmanager handles the work item and asynchronously sends a request to the JobManager to invoke the long-lived process.  At this point event handling is done and everything has happened on server1.  At some point after this, JobManager picks up the invocation request and starts the long-lived consumer process.  This can happen on either node (server1 or server2) in the cluster.

reference: (182264457)

VN:F [1.9.22_1171]
Was this helpful? Please rate the content.
Rating: 9.0/10 (1 vote cast)

LiveCycle ES: duplicate events being received on JBoss clustered installation

Issue

 If you are using a JBoss clustered installation of LiveCycle ES2 to handle batches of events, you may notice that some events are received twice, (i.e. 10 events thrown, 12 events received).

Solution

 This can occur if the JBoss cluster is not configured correctly.  In the cluster settings (i.e. run.bat for JBoss) it is important to give each node in the cluster a unique server name using the “Adobeidp.serverName” parameter.  For example, Adobeidp.serverName=server1 for the first node in the cluster, and Adobeidp.serverName=server2 for the second node and so on…

Additional information

 When you give the two servers in a cluster the same name, you wind up with a workmanager on each server that has the same name. Work items get assigned to a specific workmanager, but since they both have the same name it becomes a race to see which one picks up the work item first. In some specific timing cases both will pick up the work item at the same time and process it, which is the reason for the duplication of events.

reference: (182264450)

VN:F [1.9.22_1171]
Was this helpful? Please rate the content.
Rating: 0.0/10 (0 votes cast)