IC60958: PROBLEM IN LOCKING CODE THAT WILL AFFECT WRITING TO THE SAME FILE FROM MUTIPLE TRANSFERS.

WebSphere MQ File Transfer Edition Fix Pack 7.0.2.1

APAR status

• Closed as program error.

Error description

• Agent with the name "AGENT_Y" was transfering files without
  problems, but later on, the agent process was still running
  but no new transfers involving this agent could be started.
  The agent appeared not to read from its command queue
  "SYSTEM.FTE.COMMAND.AGENT_Y".
  The agent also did not respond to the commands
  "fteStopAgent AGENT_Y" nor
  "fteStopAgent -i AGENT_Y"
  .
  There was 1 error message
  date:time  0027    Receiver File state: Index: 0, Filename:
  TransferMode:BINARY TransferType:FILENAME
  TransferName:/opt/home/xxxxxx03/WMQFTE/inbox/mqmpkg.COPIED3
  Attributes:GEXIST=ERROR GCHECKSUMMETHOD=MD5 GDISPOSITION=LEAVE,
  Position: 0 ], auditInformationSetArray:
  ￭com.ibm.wmqfte.statestore.impl.FTEAuditInformationSet@17e2eb9
  super.toString(): Common State
  .
  It appears that an fteCancelTransfer command was made to cancel
  a given transfer. This was sent to the source agent which
  in turn sent an internal cancel message to the destination agent
  AGENT_Y. This internal cancel message then resulted in
  the command thread sending a cancel message to the receiving
  transfer thread. The command thread is then waiting for
  the receieving thread to end. However the receiving thread
  appears to have hung which in turn is causing the command thread
  to hang.
  However it was found that there was another transfer that
  was also writing to the same file. This transfer also appears
  to be hung and  not progressing. Therefore it is possible
  that there is a locking  problem between the 2 transfers.
  

Local fix

Problem summary

• The problem was itermitent, the issue was identified in the
  WMQFTE file locking code.
  
  USERS AFFECTED:
  Users writing to the same file from multiple file transfers
  running in parallel.
  
  PLATFORMS AFFECTED:
  All supported platforms.
  

Problem conclusion

• The potential locking problem has been fixed in Websphere MQ
  File Transfer Edition V7.0.1. The problem is also fixed in the
  IFIX TFIC60958  but it was never installed by the customer
  reporting the problem.
  

Temporary fix

• IFIX available
  

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

• APAR is sysrouted TO one or more of the following:

Fix information

• Fixed component name

  WMQ FILE TRANSF

• Fixed component ID

  5724R1000

Applicable component levels

• R700 PSY

     UP