DB2 FOR LUW / DB2FORLUW / 980 - DB2

On an AIX RoCE configuration with network delays, it is possible 
for the Cluster Caching Facility (CF) to RDMA Write an MRB back 
to a DB2 pureScale member then timeout waiting for the 
completion event. Although the CF times out, the member does not 
and it receives the MRB. Thus, a CQE is left on the EVD on the 
CF. The next DTO operation on that EP which requests a CQE will 
inadvertently harvest the latent completion event and thus 
declare the DTO complete prematurely thus resulting in data 
being acted on while it is still being transferred. This can 
lead to data corruption.

**************************************************************** 
* USERS AFFECTED:                                              * 
* V9.8 pureScale users.                                        * 
**************************************************************** 
* PROBLEM DESCRIPTION:                                         * 
* See Error Description                                        * 
**************************************************************** 
* RECOMMENDATION:                                              * 
* Upgrade to V9.8 Fixpack 5.                                   * 
****************************************************************

See Error Description

not known / see Local fix

Date  - problem reported    :
Date  - problem closed      :
Date  - last modified       :

28.05.2012
17.01.2013
17.01.2013

9.8.FP5