db2diag.log file filling

There are a number of PIDs - which have been causing high amounts of CPU usage since last Sunday

which are repeatedly throwing the following error

LEVEL: Error (OS)
PID     : 7564   
 TID  : 3086759616 
 PROC : db2pfchr 0

  INSTANCE: db2myInstance              
  NODE : 000

  FUNCTION: DB2 UDB, oper system services, sqloCSemP, probe:1

  CALLED  : OS, -, unspecified_system_function     
   SERR: EIDRM (43)

2008-12-15-10.59.40.404497+000 E2190G306  

Looking at these processes , they are taking about 70% of the CPU continiously.
These are errors , and as I have the Notify Level set at Level 2 they are reported into the db2diag.log file for
this particular instance.

Every other instance is OK .

The schedule of work to solve the problem was :
1)Stop Websphere Portal
2)stop MY_STANDBY_SERVER\db2myInstance
3)stop MY_PRIMARY_SERVER\db2MyInstance -

As instance
cd sqllib/bin
db2_kill <db2myInstance>
ipclean <db2myInstance>
As root
ps -ef | grep <db2myInstance>   (This will  check for any remaining instance processes which may still be running)
kill -9  <PID>   (terminate any remaining processes belonging to the instance)
4)start MY_STANDBY_SERVER\db2myInstance . This will allow these servers to be picked up when the Primary Server starts
5)start MY_PRIMARY_SERVER\db2MyInstance as Primary servers
6)Start Web sphere portal -
7)Confirm everything is OK

The cause of the problen was : on Sunday ,the instance  "db2myInstance" was being backed up via TSM , through a daily backup maintenance
schedule.The TSM server was not contactable -  this caused the prefetcher to attempt buffer allocation.

Ref:Jack Vamvas(http://www.dba-db2.com)