Re: postgres processes spending most of their time in the kernel

"Jeffrey W. Baker" <jwbaker@acm.org> writes:
> GCC 2.95.3 on Slackware 8 i386.

Seems pretty vanilla-flavored.

> I just tried this program:

> #if defined __i386__ && defined __GNUC__
> int
> main(void)
> {
>     exit();
> }
> #endif

> which builds and runs, so the problem must lie elsewhere.

Hmm.  Perhaps the TAS support does get compiled and the problem is
something else.  It'd be worth trying to determine directly what
the code is doing --- viz, run a test case that causes all the
semop thrashing, attach to one of the backends with gdb, set a
breakpoint at semop, get a backtrace to see where it's called from.
(Repeat a few times to see if the results are consistent or not.)

You might need to configure and build with --enable-debug to get
a useful (symbolic) backtrace.

            regards, tom lane