Spectrum SpectroSERVER memory leak using trap director on a remote SpectroSERVER that is geographically far from the Main Location Server (MLS).

Document ID : KB000020550
Last Modified Date : 14/02/2018
Show Technical Document Details

Description:

Remote SpectroSERVERs (SS) will run a remote lookup call to the MLS when the trap is received. When the MLS is geographically far from the SS that is actually configured to use Trap Director, a latency builds, causing the throughput of the TrapDirector to fall substantially. Memory utilization will startincreasing and will not stop until the SpectroSERVER eventually crashes due to memory exhaustion.

Solution:

The SOLUTION needs to be removed and changed to this:

This issue is resolved in Hotfix 11 and above for Spectrum 9.2.3. The Software Release Notice will state:

Symptom:Historically, there have been two main problem areas with Trap Director functionality:

  1. Trap Director may not be able to forward traps to remote servers as fast as they are received. This can result in the Trap Director SpectroSERVER terminating due memory exhaustion.
  2. The Trap Director does not immediately recognize when models are moved from one SpectroSERVER to another. These changes aren't accounted for until either the Trap Director cache ages out or the Trap Director SpectroSERVER is restarted.

Resolution:The following enhancements have been made to Trap Director functionality.

  1. Trap forwarding to remote landscapes is now done in an asynchronous manner. This can greatly increase forwarding throughput in a geographically distributed environment.
  2. Trap Director cache accuracy has been improved. Remote SpectroSERVERs now notify the Trap Director SpectroSERVER of events that require the cache to be updated.
  3. A new Trap Management sub-view has been added to the VNM model Information tab, which contains control options and informational fields related to trap processing.
    (H11, 168540)
    (H11, 110598, 21048835-1)
    (H11, 170858, 21258291-1)