High Utilization on OES Linux

  • 7005005
  • 10-Dec-2009
  • 27-Apr-2012

Environment

Novell Open Enterprise Server
OES Linux
Novell Cluster Services
Novell Open Enterprise Server 2 (OES 2) Linux

Situation

High Utilization on OES Linux
High Utilization after installing NSS
High Utilization after installing Cluster Services on OES Linux
top shows high load average

Resolution

On OES Linux, NSS and NCS both create threads which are usually idle, waiting for work to do.  The properties of these threads are such that Linux counts each of them as a load on the server, when in reality they are not doing anything and not using the processor.

To know if there is something else contributing to the high load average, you'll want to check the actual CPU usage (top shows this as well).

Additional Information


To see how the NSS / NCS threads are contributing to the load average, do the following from within top:

1.  Press SHIFT-F.  This brings up the possible fields top can sort by.
2.  Press the W key, then return.  This sorts by Process Status.
3.  Press SHIFT-R.  This reverses the sort, so that hopefully the processes with status 'D' are at the top.

The novell processes with 'D' status (the MPK processes are novell processes) are each contributing approximately 1.0 to the load average.  So, if you have 4 processes in that status, subtract 4.0 from your load average to get your real load average. 

With OES2SP2, MPK calls were removed from the Novell Cluster Services code. The MPK calls were replaced with POSIX and Linux functions. These changes were made in support of the MPK calls being removed from the Novell Storage Services™ (NSS) file system software to achieve performance enhancements for the NSS file system.


Formerly known as TID# 10100312