2008-04-15-18.18.59.657868+330 I13193166G410      LEVEL: Severe
PID     : 17759                TID  : 3086378688  PROC : db2agent (instance) 0
INSTANCE: carsdb               NODE : 000
APPHDL  : 0-829                APPID: A0010A84.K613.009285105339
FUNCTION: DB2 UDB, routine_infrastructure, sqlerCallDL, probe:70
RETCODE : ZRC=0x81360012=-2127167470=SQLZ_RC_CMERR, SQLT_SQLJC
          "External Comm error"

I am seeing above severe error in db2diag log file.

Whats mean of this error??

Why came and how to remove?

The 0x81360012 message is only informational, and can be ignored. It
means communication failure when the authentication of a user has
failed.

Here are three possibilities that will cause 0x81360012:

• when trying to attach to the database via Control Center from a remote workstation with the Administration Client and using the incorrect userid/password, this type of error message can occur.
  

• These messages can also be caused by the health monitor being started on any (remote) machine that has the instance cataloged. By default, the health monitor tries to attach to all nodes in the node directory, and since it is not supplying a userid and password, the server will reject it, and log the 0x81360012 messages in the db2diag.log.
  

• It's can also be seen with the detaildeadlock eventmonitor switched on. So to stop these errors appearing in the diaglog ensure that the deadlock event monitor is dropped (if you don't need to monitor deadlocks). In V8 the event monitor DB2DETAILDEADLOCK is defined by default with AUTOSTART. So at database connection the event monitor will be started. When users attach to an instance from a remote admin client the need to specify the userid and password (or the correct one).

This error can also happen when you're creating a backup image and the backup image is getting bigger than 2Gb. Check if your filesystem  - where the backup is being generating - is large files enabled.