Zombie Hunting in JD Edwards EnterpriseOne
One new powerful feature which Oracle supplied in the 9.2.5.x tool set is the 'Zombie Kernel Process' option. This new feature was added to Server Manager as a new tick box within Monitored Events. As it is a simple looking checkmark located within the Monitor Configuration setup, it may have been overlooked by many. Don’t be fooled by the simple integration aspect of this additional checkmark, as it can save countless hours and drastically help you battle the zombie horde.
Zombie kernels within JDE may not directly infect your healthy kernels and turn them into zombies to grow the horde. Although, they can rob your system of resources, cause system instability and frustrate the user base.
Once a kernel stops responding and the parent process flags it as a zombie, the kernel can still remain active on the Enterprise server stealing RAM and CPU resources. As zombies on the system increase so do the amount of resources they keep hostage from the healthy kernels. Even clearing zombie kernels in a timely manner from the system can still rob the server, as they do not always give back 100% of the resources they took originally.
The user impact of zombies can also be very painful and frustrating. If one user in JDE triggers a business function to zombie a call object kernel, it may impact multiple users connected to that kernel. We target between 6 and 8 users per call object kernel as an ideal target of a healthy balanced system configuration. Although we often find systems running upwards of 12 to 20 users per call object kernel. If your finding say 5 zombie kernels a week, that is likely impacting around 50 users within the system. Bump that up to 5 zombies a day and they user base is not going to enjoy using EnterpriseOne.
So how does this new checkmark within the Server Manager Monitoring tool help to fight against the zombie horde? Oracle has implemented this feature to automate and simplify the process of finding the zombie origins. When an Enterprise server process (network, kernel and runbatch processes) terminates unexpectedly, the Server Manager jumps into action.
· Server Manager will first retrieve details of kernel failure from the SM Agent.
· Generates and captures the Crash Dump file with PID details.
· Parse BSFN name and problem call stack information.
· Retrieve User information from Enterprise server configuration to determine the user who last modified the BSFN.
· Format the data and build email with a very easy to read content summary.
· Notify Administrators and contacts (defined within the monitor) via email with the summary information.
This new automated summary of each zombie kernel will greatly reduce investigation into the cause behind each zombie that occurs. Which in turn will allow the team to remedy the root cause of these zombies in a timely fashion. Keeping the zombie horde at bay and allowing the system to work as intended.
If you have 9.2.5.x installed we highly recommend that this feature is configured to help the fight against zombies. If your system is not running 9.2.5.x yet, please reach out to see how we can assist.
- Todd Burton is a Senior JD Edwards Technical Architect with Main Street APPs.