This is a follow-up note on my prior linux / unix kernel parameters post
Yesterday I received some feedback from oracle about why the core dump has a 4 GB limit and why you should therefore limit the shmmax to a maximum of 4GB:
” Normally on x86_64 a coredump is not limited to 4 GB, but a coredump is affected in different levels: namely memory, I/O and disk.
Most of the limitations are due to the filesystem used.
If the coredump routine in the kernel is not using the right kernel calls or the filesystem cannot handle files bigger than 4 GB (like reiserfs or some old ext2),
then the coredump will be affected.
For this reason, setting shmmax < 4GB on x86_64 is more a recommendation.”
After reading this I, closed the service request…