https://www.ibm.com/developerworks/community/blogs/aimsupport/entry/ibm_bpm_event_manager_common_problems?lang=en
IBM Business Process Manager Event Manager - Common symptoms and how to resolve them
Introduction
The event manager is the central component in IBM Business Process Manager that is responsible for scheduling a number of different tasks. If it's not working correctly, you might run into severe problems that need to be resolved quickly.
This blog entry describes some of the most common symptoms and shows how to resolve them.
In Part I, some common event manager problems will be shown. Part II explains how to analyze and fix these problems. Part III lists the available APARs that are related to the event manager.
Special thx to Mark Filley and Bill Wentworth for their technical input and in depth review!
Part I - Common event manager symptoms
Symptom A - Event manager is not processing any work
The event manger is responsible for the scheduling of various jobs like:
- Executing undercover agents (UCAs)
- Executing system lane tasks
- Triggering business process definition (BPD) timers
- Scheduling BPD notifications, which are essential to move the process flow forward through the business process diagram (BPMN)
If your process instances are stuck, timers not fired, and UCAs are no
longer executed, then the event manager is not running or could be
blocked.
The Process Admin Console gives you a comprehensive view that shows the status of the event manager. When the status light is red, the event manager is paused or did not start and its jobs are accumulating with a Scheduled Time much behind the current time. In the Process Admin Console, these jobs would show 'Job Status' as 'Scheduled' and jobs would not be in the 'Executing' state.
For example:
The following screen shot, which was taken from the Process Admin Console, shows the last event manager heart beat expiration time stamp of "12/10/2014 2:00:15 PM." This time stamp is normally ahead of the current time. The event manager job's (UCAs and BPD notifications) Scheduled Time shows an earlier time stamp and a job is not currently executing. In this example, the event manager is shown as inactive (red light), which explains the situation.
Note: Even if the event manager is not running, it is possible to start new process instances, but they will not move forward! As services are not scheduled by the event manager, those could also be executed.
Symptom B - Event manager shows jobs with a scheduled date of 2099
The Process Admin Console can show event manager jobs scheduled for 2099 as shown here:
Symptom C - Event manager is active, but long running system lane tasks block the event manager throughput
There can be situations where the event manager is actively working,
but you experience throughput problems. For example, the flow in the
process instances is not moving forward or the execution of timers is
delayed.
The following screen shot shows five system lane
activities being executed, but a couple of BPD notifications are waiting
to be executed. These BPD notifications are overdue as the 'Scheduled
Time' is greater than the current time. This situation can indicate that
the event manager configuration needs to be tuned and/or the execution
time for system lane tasks needs to be optimized, if possible.
See resolution section C,D,
Symptom D - UCAs are not processing at the desired rate
According to the definition in the process application, UCAs are bound
to a couple of synchronous queues or a single asynchronous queue managed
by the event manager. The capacity for these queues is defined by the
following parameters in the 80EventManager.xml configuration file:
<asy
<syn
These numbers limit the rate of UCAs that can be executed at a time.
See resolution section C,D,
Symptom E - Many BPD timers wake up at the same time
When the event manager processes a timer, it loads the applicable task
into the "BPD async queue," whose capacity is defined by the <bpd
Keep in mind that this queue is shared between timer executions, BPD notifications and the execution of system lane tasks.
See resolution section C,D,
Symptom F - Event Manager warning messages CWLLG2156W, CWLLG2236W occur
If the BPM run time detects that the database connection pool is too
small, it will dynamically reduce the queue sizes and you will see
entries in SystemOut.log like the following messages:
"CWLLG2156W: The database connection pool size xxx of the Process Server data source might be too small." and/or
"CWLLG2236W: The configured <%%%
These messages indicate that there is a mismatch between the event manager queue capacity and the JDBC data source pool size.
See resolution section C,D,E,F - Event manager is active, but throughput problems exist to analyze and resolve that problem.
Symptom G - Event manager tasks fail when LombardiEventEmitterInputQueue reached max threshold
When you have your IBM Business Process Manager environment configured
to forward monitoring events to a Business Monitor server, the execution
of event manager tasks involves sending a message to the local queue
called "Lom
If the queue depth of the Lomb
J2CA0027E: An exception occurred while invoking prepare on an XA Resource Adapter from DataSource jms/
data
at com.
at com.
at com.
at com.
at com.
at com.
at com.