Workflow deadlock exception

[arajagopal]

Sync on ECN states robot (included just before the end of ECA workflow process) getting terminated in some cases with exception attached (wt.pom.DeadlockDetectedException: Deadlock is detected on the Workflow Process) here. Workflow process monitor shows same error. This error doesn't stop parent ECN process. Only discrepancy is particular ECA process been kept in terminated state, but no impact on release. The exception is bothering us. Frequency is mostly one per day.

TS and RD working with us on this issue. Not sure this is an OOTB bug. If anyone resolved such exception, please share your experience of fixing it. We reviewed multiple times the workflow design, not seen any issue. Let me know if you have any specific question.

Ayyappan

[RandyJones]

On 07/16/14 10:57, Ayyappan Rajagopal wrote:
>
> We are in 10.1 M010.
>
> Sync on ECN states robot (included just before the end of ECA workflow process) getting terminated
> in some cases with exception attached (wt.pom.DeadlockDetectedException: Deadlock is detected on
> the Workflow Process) here. Workflow process monitor shows same error. This error doesn't stop
> parent ECN process. Only discrepancy is particular ECA process been kept in terminated state, but
> no impact on release. The exception is bothering us. Frequency is mostly one per day.
>
> TS and RD working with us on this issue. Not sure this is an OOTB bug. If anyone resolved such
> exception, please share your experience of fixing it.
>

Had a similar issue in 10.2 M010 with a workflow that worked perfect in 9.1 M050 and 9.1 M070.
Turned out to be custom code that was directly modifying oracle tables using java jdbc vs using a
Windchill API. This was in a customized promotion request in which we are changing the revision from
a numeric to '-' by using the Windchill API's. Which changes the lastmodified of all iterations of
the object to now. To fix this we were doing the following:
1. using Windchill API to get the lastmodified of all the iterations (before they were munged)
2. changing the revision to '-' with Windchill API
3. using custom java code to put back the pre-munged last modified timestamps to the proper oracle
table (epmdocument)

The custom code and the Windchill code would get in an Oracle deadlock (in 10.2 but not 9.1). I see
you are also getting oracle deadlock in the Deadlock_logs.txt you attached.

Our workaround was to now do this:
1. use Windchill API to get the lastmodified of all the iterations (before they are munged)
2. change the revision to '-' with Windchill API
3. use custom java code to store the pre-munged last modified timestamps into another oracle table
(timestamps_epmdocument) created just for this purpose
4. every morning run a manual query that resets the epmdocument lastmodified timestamps to objects
that were munged yesterday
5. then periodically check w/PTC to see if there has been any action on getting a fix for the broken
Windchill API that munges the timestamps.

> We reviewed multiple times the workflow design, not seen any issue. Let me know if you have any
> specific question.
>
> Thanks
>
> Ayyappan
>
>
>
> ----------


--
------------------------------------------------------------------------
Randy Jones
Systems Administrator
Great Plains Mfg., Inc.
1525 E North St
PO Box 5060
Salina, KS USA 67401
email: -
Phone: 785-823-3276
   Fax: 785-667-2695
------------------------------------------------------------------------

[arajagopal]

Jones,

Thanks for your reply.

We don't have any custom code and it is standard OOTB API is in place. Also, it is not happening for all ECA gets completed.

[maheshkl83]

Hello Ayyappan,

Is there some additional code executing right after the ECN is resolved? Is this issue happenning when there are more than 1 CA for an ECN?

I think we can find out long executing query and analyze the query.

If it is an SQL Server DB, go to the DB --> Right Click --> Activity Monitor --> Most Expensive Query.

I am also not sure why you are keeping CAs open even after they are "Resolved"?

Thanks,

Mahesh KL

[arajagopal]

Mahesh,

No Additional code executing after resolved except that Sync.

We had situation to see the issue only on 1 CA type scenario

We are not on SQL. It is Oracle.

Definitely we are looking back why CA need to be brought back to rework after it is been resolved.

Ayyappan