DB2 FOR LUW / DB2FORLUW / B10 - DB2

During processing of LOB files DB2 might sent "non-aligned
memory buffer" to do GPFS Direct I/O. This requires xw lock
instead of sx lock on GPFS level. Multiple sx locks can be
obtained concurrently this is why so many threads are
contending.

As the consequence we might see slow performance together with
high GPFS CPU usage. Issue might be triggered by various DB2
functions, therefore way of identifying this is observation on
GPFS level (for locks in subject). Sample symptom is mutex
observed in the traces:

MMFS MUTEX: Waiting on condvar 0xF1000002DAB7BE10
(BrlObjCondvar): waiting for fetch-n-lock to complete
MMFS MUTEX: Waiting for mutex 0xF1000002DAB7B628
(InodeCacheObjMutex)
MMFS MUTEX: Awakened after wait for ThMutex 0xF1000002DAB7B628
(InodeCacheObjMutex) err 0
MMFS BRL: brLockM retry 4: inode 349569 snap 0 range
0x0000004A5FEA6800-0x0000004A5FEA6FFF startRangeSeq
0xFFFFFFFFFFFFFFFF mode xw
MMFS TSTM: crtTestRange enter: range
0x0000004A5FEA6800-0x0000004A5FEA6FFF mode xw


This APAR fix is to change DB2 code to send correctly aligned
buffers. As the consequence, the GFPS contention issue is not
observed with fix in place.

****************************************************************
* USERS AFFECTED:                                              *
* ALL                                                          *
****************************************************************
* PROBLEM DESCRIPTION:                                         *
* See Error Description                                        *
****************************************************************
* RECOMMENDATION:                                              *
* Upgrade to Db2 11.1 Mod 2 Fix Pack 2 or higher               *
****************************************************************

not known / see Local fix

forerunner  : IT16916 
follow-up : 

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

02.10.2017
24.10.2017
24.10.2017