HP OpenVMS Systems

Tracking down cause of cluster RWSCS?

» close window

The Question is:

 
We have a 2 node cluster comprising 2 4705a machines.
 
Our aim is spread our workload evenly accross both nodes.
 
We have an application that is used by about 70
users. We have about half of the users on each node.
 
On one of the nodes, the processes frequently
go in to the RWSCS state, consequently they run a lot slower
that the same job running on the other node.
 
What would cuase this to happen?
SYSGEN parameter(s)? DSSI configuration?
 
Many thanks.
 
 

The Answer is :

 
  The RWSCS state can indicate applications are waiting for a lock manager
  response from another cluster member, this can be caused by a variety of
  factors ranging from FDDI ring reinitializations to heavy I/O activity
  (VIOC and global buffers can potentially help here) to insufficient
  SCSBUFFCNT to periodic remastering of the lock trees to insufficient
  bandwidth between the systems to an inappropriate value in the LOCKDIRWT
  system parameter.
 
  As a first step, clean out any accumulated cruft in MODPARAMS.DAT, and
  replace absolute parameter value assignments with relative (eg: MIN_,
  MAX_, ADD_) whenever appropriate, and then run AUTOGEN with FEEDBACK
  followed by a system reboot.
 
  DECamds can be used to track locking activity.
 

  

     

     
answer written or last revised on ( 25-MAY-1999 )
     
» close window