DB2 FOR LUW / DB2FORLUW / A10 - DB2

Graceful takeover was slow, in the minutes.  This was because 
takeover was being issued shortly after the primary activated 
and we had not internally pre-allocated log active.  So when the 
takeover request came in and on the primary we had quite a few 
transaction to rollback, undo was fighting db2loggr 
pre-allocation. 
We spent too much time in pre-allocation not enough time 
servicing read log requests for undo. 
 
This is a regression due to our recent changes to allocate 2 
files on startup and the remaining files after the fact. 
 
The solution is a cheap fix.  While takeover is in progress to 
avoid doing any file pre-allocation.  This way db2loggr is 
focused 
on rollback not taking up cycles allocating files.

**************************************************************** 
* USERS AFFECTED:                                              * 
* HADR takeover slow                                           * 
**************************************************************** 
* PROBLEM DESCRIPTION:                                         * 
* See Error Description                                        * 
**************************************************************** 
* RECOMMENDATION:                                              * 
* upgrade to v101fp5 or v105fp5.                               * 
****************************************************************

If possible, before issuing the graceful TAKEOVER, wait until 
all logs in the active log path are pre-allocated.

the fix will be delivered to v101 and v105. the slow takeover 
problem will be resolved.

not known / see Local fix

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

21.08.2014
13.07.2015
13.07.2015