High utilization, slow performance on Linux OES running Novell Cluster Services.

  • 3335151
  • 25-Feb-2008
  • 31-May-2012

Environment

Novell Open Enterprise Server (OES)
Novell Cluster Services (NCS)

Situation

High utilization, slow performance, or slow drive mappings on Linux OES running Novell Cluster Services.
Usually seen with six or more virtual NCP servers per physical server. (ie. six or more secondary IP adresses per physical server)

Resolution

There is an eDirectory tunable parameter that needs to be increased.
Issue the ndsconfig command to set eDirectory's max-threads higher, default is 64. 128 is the recommended amount to fix this problem:
ndsconfig set n4u.server.max-threads=128

NDS needs to be restarted after this change; the most straightforward way of doing so is to reboot the server.

Additional Information

The problem was due to the way NDS allocates threads for NCP services. With clustering involved, each virtual NCP server was taking a finite number of threads, and the default tuning for eDir is too low to support more than about 6 virtual NCP servers per physical server.

Formerly known as TID# 10099403

Feedback service temporarily unavailable. For content questions or problems, please contact Support.