DB2 FOR LUW / DB2FORLUW / 970 - DB2

After upgraded the HADR Standby to v9.7 FP6 or later fixpak, a 
performance degradation of HADR Primary may occur. 
 
From db2pd -hadr on Standby side, one problem can be observed 
that Standby is using almost 0% of receive buffer, 
2013-10-09-15.52.21.000999 
   PrimaryFile  PrimaryPg  PrimaryLSN 
   S0467575.LOG 26092      0x00004758AE52CC46 
 
   StandByFile  StandByPg  StandByLSN         StandByRcvBufUsed 
   S0467575.LOG 26091      0x00004758AE52BFAA 0% 
 
And from trace collection of EDU db2hadrs on Standby side, the 
hdrAddLogFiles function will take more time to complete the log 
initialization and specially opening log file is direct i/o -If 
logging disks are slow, then in this direct i/o case, the 
performance will be impacted. 
 
581182  exit DB2 UDB oper system services sqloopenp cei 
(2.3.15.825.2) 
        pid 5636136 tid 200465 cpid 4850536 node 0 sec 14 nsec 
105142410 
        codepath = 2:11:15:16:20:46:48:49 
 
        rc = 0 
        bytes 16 
 
        Data1   (PD_TYPE_SQO_FILE_HDL,8) File handle: 
  File Handle              = 518 
  File System Block Size   = 0 bytes 
  File System Type         = UNKNOWN 
  File Handle Flags : 
    Require Sector Align   = No 
    DIO/CIO Mode           = Yes 
    Raw Block Device       = No 
    Reserved Handle        = No 
    Flush On Close         = Yes 
    Thread-Level Lock      = No 
    Write-through Mode     = No 
    File Not Tracked       = No 
 
Slower disk operation will adds up to a case that db2hadrs on 
Standby is busy with I/O manipulation for log initialization, 
during this time db2hadrs will not response any request from 
HADR Primary, therefore, HADR Primary may see performance 
degradation if db2hadrs spent longer time to initialize the log.

**************************************************************** 
* USERS AFFECTED:                                              * 
* ALL USERS                                                    * 
**************************************************************** 
* PROBLEM DESCRIPTION:                                         * 
* See Error Description                                        * 
**************************************************************** 
* RECOMMENDATION:                                              * 
* Upgrade to v9.7 FP9 or use the local fix as workaround.      * 
****************************************************************

Set a registry variable DB2_LOGGER_NON_BUFFERED_IO=OFF on HADR 
Standby to get back the  file system cache performance, however 
the side effect of DB2_LOGGER_NON_BUFFERED_IO=OFF is that it 
will enable all normal log operations to use File System Cache 
as well, this will incur overhead of File System Cache 
Management and side performance effect, specially after takeover 
the new primary will be impacted.

DB2 Version 9.7 Fix Pack 9 for Linux, UNIX, and Windows
 DB2 Version 9.7 Fix Pack 9a for Linux, UNIX, and Windows
 DB2 Version 9.7 Fix Pack 10 for Linux, UNIX, and Windows
 

This problem has been fixed in v9.7 FP9

not known / see Local fix

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

14.10.2013
17.12.2013
17.12.2013

9.7.FP9