E1: OS: Cleanup Recommendations for Zombie Processes on UNIX Servers (Doc ID 649931.1)

Last updated on DECEMBER 06, 2024

Applies to:

Goal

Users are experiencing intermittent problems where jobs go into W (Wait) or S (Submitted) status and then do not complete. This problem occurs several times on a daily basis.  It is not clear if a specific UBE is the cause of this, however there are a high number of zombie processes on the UNIX Enterprise server.  You need to know how the zombie processes are managed by the JDE parent process (all zombies are owned by jdenet_n).

You want to speed up the zombie clean up process.  The clean up process works, but with the current configuration the clean up does not happen fast enough.  The zombie processes are not ending in error. The processes end normally but are not getting cleaned up quickly.

There are about 5000 zombie processes per day on the production Enterprise server and all 5000 processes have the same parent process jdenet_n.  Overnight when the server is less busy,  2000 zombie processes are cleaned up by jdenet_n.  This leaves 3000 zombie processes per day carried into the next work day.

The maximum number of processes on UNIX has been changed to 20000 in order to compensate for the high number of zombies.  With 3000 processes a day being leftover, there is a possibility of having 6 days grace period before having to stop and start EnterpriseOne services in order to clear out the zombies. Stopping and starting E1 services once a week in order to keep the system running may not always be possible.  Is there a better solution to reduce the the number of zombies on your Enterprise Server?

Solution

My Oracle Support provides customers with access to over a million knowledge articles and a vibrant support community of peers and Oracle experts.