DB2 FOR LUW / DB2FORLUW / A10 - DB2

In DB2 HADR environment, DB2 health monitor process might leak 
memory after a TAKEOVER. Once database is activated and first 
TAEKOVER is issued, the leak will be exposed only on the former 
PRIMARY (then STANDBY), after a subsequent TAKEOVER, problem 
will be visible on both instances. 
Rate of the leak is around 4kB per minute and will occur also 
with default HEALTH_MON OFF in database manager configuration. 
 
To confirm this specific problem is hit, one can: 
1. Check PID of health monitor process: 
$ db2pd -edus | grep db2acd 
 
2. Use db2pd to traverse the heap of db2acd process and list 
internal headers of memory allocations by running: 
$ db2pd -memblocks private pid=<pid_of_db2acd> 
 
3. Leak can be identified by examining the output of the command 
above. Leaking allocation will have: 
- PoolID/PoolName 'OSSe' 
- two leaking allocations, with size of 4267 bytes from LOC 
(line of code) 1207 and with size of 64 bytes from LOC 1199 (LOC 
might slightly differ between fix packs) 
e.g. 
 
Address            PoolID PoolName Size(Bytes) LOC   File 
0x00007FB0E42A5770 OSSe   OSSe     64          1199  0 
0x00007FB0E42A57E0 OSSe   OSSe     4267        1207  0 
0x00007FB0E42A68C0 OSSe   OSSe     64          1199  0 
0x00007FB0E42A6930 OSSe   OSSe     4267        1207  0 
0x00007FB0E42A7A10 OSSe   OSSe     64          1199  0 
0x00007FB0E42A7A80 OSSe   OSSe     4267        1207  0 
0x00007FB0E42A8B60 OSSe   OSSe     64          1199  0 
0x00007FB0E42A8BD0 OSSe   OSSe     4267        1207  0

**************************************************************** 
* USERS AFFECTED:                                              * 
* ALL                                                          * 
**************************************************************** 
* PROBLEM DESCRIPTION:                                         * 
* See Error Description                                        * 
**************************************************************** 
* RECOMMENDATION:                                              * 
* Upgrade to DB2 10.1 Fix Pack 6                               * 
****************************************************************

Memory can be released by restarting health monitor process. 
There is not explicit way to do it, but db2acd can be terminated 
as any other fenced process using db2fmpterm command and once 
terminated, it will be automatically restarted: 
$ db2fmpterm <pid_of_db2acd>

First fixed in DB2 10.1 Fix Pack 6

not known / see Local fix

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

07.11.2016
02.03.2017
02.03.2017