NMDA: IBM DB2 snapshot backups are unresponsive

NMDA IBM DB2 snapshot backups are unresponsive.
DB2 database is write suspend mode.
Error in db2diag.log

Warning:
----------------------------------------
2016-05-20-15.05.16.771114+000 E7423877A808       LEVEL: Warning
PID     : 7733668              TID  : 47390       PROC : db2sysc 0
INSTANCE: <instance_name>      NODE : 000         DB   :
APPHDL  : 0-57647              APPID: <app_id>
AUTHID  :  
EDUID   : 47390                EDUNAME: db2agent (instance) 0
FUNCTION: DB2 UDB, database monitor, sqm___sqlmonssagnt, probe:1234
DATA #1 : <preformatted>
 
This occurs when any snapshot other than list applications is attempted
on a database that is write suspended. This is necessary due to latching
conflicts. This database is skipped in the snapshot, but other databases
should still be processed. Skipping database '<db_name>'.
For more information on write suspend, please see the
'db2 set write suspend for database' command.
------------------------------------------

DB2 workflow works as:

DB2 calls db2ACSPrepare() in the NMDA DB2 library to do initial preparation to get ready for the hardware snapshot. Once this is done, DB2 Database manager suspends the database and calls NMDA DB2 function db2ACSSnapshot(). In this case, DB2 does not call db2ACSSnapshot(). The workflow stops responding for the DB2 engine to continue the backup. IBM support PMR: 29538,004,000 is opened. IBM DB2 engineering needs more debug to determine the root cause of this hanging.

Workaround:
 

1. Cancel current backup
2. Take database out of write suspend mode

In order to discover the root cause, IBM DB2 support requests following info for further investigation,
 

1. Delete or relocate any stacks or dumps in the db2dump directory. It creates new stacks or dumps without mixing old ones. (UNIX/Linux path: /$INSTHOME/sqllib/db2dump)
2. DB2 trace for a short while:

   db2trc on -f trace.dmp
   wait about 30 seconds
   db2trc off
   db2trc fmt trace.dmp trace.fmt
   db2trc flw trace.dmp trace.flw

3. Gather stacks:

   db2pd -stack all -rep 5 5

4. After the above step completes, run,

   procstack <PID> > db2vend.pstack.out

   The <PID> is the Process ID of the db2vend process for the database.
5. Then get the snapshot backup session,

   db2 get snapshot for applications on <dbname> > db2snap.out

6. Now you can set write resume, and then try the snapshot backup again and see if it finishes.
7. Gather a db2support.zip as follows:

   db2support . -d <dbname> -c -s

8. Zip and send to IBM PMR:

• trace.dmp
• trace.fmt
• trace.flw
• db2vend.pstack.out
• db2snap.out
• db2support.zip