3247382 - On rare occasions a user transaction running in a multiplex appears to hang indefinitely -- SAP IQ

• On rare occasions a particular user transaction may appear to become stuck or hung when running on a secondary node in a multiplex environment.
  
  
• The transaction holds an uncommitted version that does not get applied.  Overtime this can cause a version build-up  and eventually could lead to an Out Of Space condition in the  IQ_SYSTEM_MAIN dbspace.
  
  
• Attempts to DROP CONNECTION or hard kill the client connection process does not terminate the connection or transaction in IQ.
  
  
• The connection involved is for a transaction that opens a NOSCROLL READONLY with HOLD cursor. The details from store procedure sp_iqconnection for the problem session shows ReqType="CLOSE"  and IQCmdType="DESTROYCURSOR".
  
  
• A thread stack dump contains a thread similar to below for a DestroyCursor that is not progressing.

  ===== Thread Number nnnn ... =====
   .. 
   .. hos_condvar::Wait  ..        
   .. hos_teamBase::Wait  ..       
   .. hc_dmfrun::ShutDown  ..      
   .. dfo_Combiner::ShutdownSink  ..
   .. dfo::PrepareToComplete  ..                
   .. dfo_Root::CompleteAllDFOsBelowRoot  ..    
   .. dfo_Root::Complete  ..                    
   .. st_cursorIQtoSA::Complete  ..             
   .. df_DestroyCursor::Execute  ..             
   .. st_command::DoCmdThroughResourceGate  ..  
   .. st_iqdml::DestroyCursor  ..               
   .. UIQDML_DestroyCursor+0x18
   .. 
   .. DB_Fini_db_cursor_and_heap  .. 
   .. a_plan_cache::CloseCursor  ..  
   .. a_plan_cache::CloseACursor  .. 
   .. Connection::CloseACursor  ..   
   .. do_close  ..                   
   

• The last messages in the secondary node iqmsg file for the connection shows a start transaction and version number communicated to the Coordinator node (CN) that does not  get applied.  After a period of time this version is listed in the CN "Warning: Coordinator version list does not match secondary servers!" messages in the iqmsg file and does not go away.   Here is an illustrative example ....

Last message seen in the secondary node for the hung connection 

I. 07/29 16:02:51. 0000765635 Txn 9048402915 0 9046464478

After a lengthy period of time the version is still being reported on the CN mismatch list

I. 08/01 16:47:34. 0001165224 Warning: Coordinator version list does not match secondary servers!

I. 08/01 16:47:34. 0001165224 W: 9046464478 9047997867 9048055648 9048062974 9048122010

• The output from store procedure sp_iqmpxversionchaininfo shows the version reported as the LastActiveVersions and OARV.  And the LastAppliedVersion number does not progress over time.

• SAP IQ 16.1 SP04.x
• SAP IQ 16.1 SP05.x

DestroyCursor, ShutdownSink, hang, unresponsive, CR827064, 827064, sybase, proc,  , KBA , BC-SYB-IQ , Sybase IQ , Problem