Messages in this thread | | | Subject | Per process control of core file naming | Date | Thu, 13 Dec 2007 15:54:48 -0500 | From | "Sanders, Rob M." <> |
| |
Hello all, I know I can use the /proc/sys/kernel/core_pattern and /proc/sys/kernel/core_uses_pid files to control core filenaming for *all* core files on a system, but is there a way for a particular process to have control of its own core file names? The down side to the current way Linux does this now is that *all* processes are affected, requiring someone with root/sudo access (or someone with said privs to chmod these files) to change the behavior. It would be nice for a process/programmer to have finer grain control (subject to override by /proc/sys/kernel/core*) on the names of the core files. Tru64's has this, as 'enhanced core file naming', where a process can decide at runtime how a file will be named (still being written to the CWD if space/permissions allow). I'd found this feature to be quite useful several years ago on a Tru64 cluster, where I had multiple instances of the same executable running across the cluster. It was very nice to be able to turn on and off the enhanced naming ( core.program_name.host_name.numeric_tag) as I needed to, without requiring everyone *else* to use the same naming conventions.
Thanks...
Rob
| |