When scheduling large scheduled tasks to start within seconds of each other the core server becomes un-responsive and the ldpgp.exe application throws an application exception.
- No one process takes up CPU or memory
- Memory does spike collectively.
- The above error occurs
When building the server the /3GB switch was enabled in the boot.ini. The Core Servers should never have this switch enabled as the LDMS Core Server components were not written with the "Large Address Space Aware" flag set. Because LANDesk Management Suite is not Large Address Space Aware the application will still be only use 2GB of user memory while the Kernel memory is still only allocated 1GB of memory. By enabling the /3GB switch Kernel memory does not have enough resources to properly map the calls and the above error occurs.
Reducing the system space to 1GB with /3GB causes the maximum paged pool (drivers, heap, paged and non-paged memory) size to drop in some cases below 50% of normal resource allocated. Due to the nature of how memory works, by cutting the system resources to the kernel memory space new memory calls have limited or no ability to write new calls. This is particularly true in I/O intensive applications such as LANDesk.
Remove the /3GB switch from the Boot.ini