Livelink 9.5 - Workflow Notifications

[zerk12]

Hi,

We are having problems with our LL workflow notifications. When we create a new group in LL and add that new group as the assignee to the workflow, no one in that group receives a notification that an assignment has arrived (after the WF's been initiated).

The strange thing is that once we restart our Livelink services, and then initiate that same workflow with the new group assigned, the notifications work.

Has anyone else experienced this type of behavior?

Thanks,
Jung

 

[appnair]

As far as I know when the workflow is processing and reaches a step where the group is located,it queries the relevant tables to route that assignment to the people.If the receiver has a notification set for the event then he/she recieves them. I also know from my experience that a few late workflows/tasks can seriously affect the timing/delivery of the entire notifications.If you have access to the KB my colleauge John Simon has written a posting that is very eye opening.He,I and Mark Simm of OpenText has tried to pinpoint the deficiencies in the notification subsystem and tried many things.It could be a coincidence that some stuck process gets relieved when the services are re-started and appears that it works.Do notification run only on one box,or is it staggered amongst boxes.I would suggest that you open a ticket with OpenText and get them involved.



Well, if I called the wrong number, why did you answer the phone?
James Thurber, New Yorker cartoon caption, June 5, 1937

 

[zerk12]

Thanks for the info. We do have a ticket open with OT and they are looking into it, but we've had it open for some time, without any resolve.

To answer your question, we run our notifications on one box.

 

[ggriffit]

Do these new users have the correct notifications set up ?


Greg Griffiths
Livelink Certified Developer & ECM Global Star Champion 2005

http://www.greggriffiths.org/livelink/

 

[zerk12]

Yes, all the users have their task and workflow related interests set.

The users will get the notifications, but only after the LL services are restarted.

 

[Syntergy]

Here is a good approach to isolating the problem down. When the workflow step becomes ready you check the notifyevents table to see if the notification record is ready to be processed by the agent process. The agent process runs every 5 minutes and sends out the notifications.

The following query should tell you if the notification event record is there. You should try to do this before the agent process runs (every 5 minutes) because it will be deleted if it is processed. Assuming you are on Oracle:

select * from notifyevents where eventid = 8 and eventdate = sysdate;

This will give you all the workflow step ready events for the currenct date. The column eventint1 contains the user/groupid the workflow step has been assigned to. eventint2 is the workid, eventint3 is the subworkid and eventint4 is the taskid. The workid, subworkid and taskid will tell you exactly what workflow step this event is for.

If the event is in the table the next thing to check is that the notification process is processing the event. The workflow step ready event is handled first by the EventsToMessages process and then the ProcessMessagesAgent process. If you have logging turned on you will see a file in your opentext\logs directory called "notify_run102.html" or similar. This file lists the notification processes as they are executed. Check this file very closely for any errors. Errors will often stop the entire agent process and often this will cause notifications to not get processed. It will tell you exactly where the event is in the process. The file will have output similar to:

......Perform: NodeEventProcessor
......Perform: EventsToMessages
......Completed: EventsToMessages ... 1642 ms.
......Completed: NodeEventProcessor ... 1973 ms.
....Ending Agent: EventProcessor ... 1973 ms.
....Starting Agent: ProcessMessagesAgent

(A list of messages sent will show up here)

......Sent 10 Messages...
....Ending Agent: ProcessMessagesAgent ... 40 ms.
..Done Running Agents ... 2703 ms.

If you don't see all of these events being processed that would explain why the notifications are not going out. A restart of Livelink may be resetting a flag that allows events/processes further down the agent chain to finally get processed. It's hard to say without more investigation why a Livelink restart would clear up processes that are having errors.

The other thing to check is how many notifications are being sent out per 5 minutes. There is a limit on the number of notifications to process at one time. If there is a big queue it could take a while for a message to get sent out.

You should also double check how often each process in the agent process is running. If one of the processes listed above in the sample output listing is not executing every 5 minutes that will delay or prohibit notifications going out.

We have seen problems with the system time on the Livelink machine and the system time on a separate database machine cause problems. They should be relatively close (withing minutes). If they are out of sync by one or more hours this can cause delays in when things are processed.

Finally, Livelink does some cleanup of things when it starts up. It could be possible that some items are getting deleted out of the database on startup which allows the other notifications to go out.