Find *where* the executable is for a currently running process

Question

If I have 2 executables (/path/to/gcc/myapp.exe, and /path/to/intel/myapp.exe) where one is compiled with an intel compiler and one is compiled with standard gcc, running both will show up in top as just myapp. If I run both of those and have processes running as myapp, is there a way I can find out which process is coming from which executable?

We have a bunch of apps installed on our cluster and I am interested, not only in which apps are being used the most, but which versions of the apps are being used. If I see 500 occurences of R being run, I would like to know if 100 of those are from /path/to/R-3.2.1/intel/R and the other 400 are from /path/to/R-3.1.3/gcc/R, etc. How can I do this?

Answer 1

You can do this with the /proc filesystem, as noted in the proc(5) manual page:

/proc/[pid]/exe

Under Linux 2.2 and later, this file is a symbolic link containing the actual pathname of the executed command.

Because it is a symbolic link, you can read the target (do an ls -l to see the pathname).

The [pid] of course means a process-ID for the particular executable you are interested in, at runtime.

(It is unlikely that you have a system running a Linux kernel that old, so "always works" should be good enough).

To get a count, you simply need a script that walks through /proc, for each directory which you own (since you are running the programs, you can see their /proc data), and matching the pathnames against the ones you are interested int. The readlink program, for instance, is useful in bash scripts (and since it is part of coreutils, you likely have it on a Linux system).

Answer 2

To do that, you can launch top, and then press c on your keyboard to toggle the full path view.

Answer 3

If these processes are executed with different command line (for example using full path) you can press c to show/hide full command line with which the process was executed.