John Poulakos
Active User
Joined: 13 Jun 2012 Posts: 178 Location: United States
|
|
|
|
I have an condition that causes a user to get locked out of CICS. It doesn't happen very often and I know how to correct it. But, I need to document what causes it and I am having a hard time with that.
The symptoms are a user is prevented from accessing CICS because they are already signed on at a different terminal. They could stay in this state until CICS is restarted, if I did not intervene. I can identify the user in the CICS tables, find the task that is being held and kill the task or terminal. No other users are impacted.
When I kill the task (or terminal) I get an abend and dump that shows the transaction is in a DB2 program and I can tell it is running under the Open API TCB (L8) as a TRUE exit. All of our CICS transactions are Psuedo-conversational, so there is no way it could have returned a screen and held on to the terminal session. There is no evidence of LOSTERM exit being driven, so the user did not break the connection before getting a response.
The only conclusion I can draw is the TCB either never got posted or it was posted under the wrong key. I can't come up with a good reason for why or how this could happen. |
|