[IPython-User] Help with ipengine issue

Fernando Perez fperez.net@gmail....
Fri Jan 6 12:35:18 CST 2012


Hey Bago,

On Fri, Jan 6, 2012 at 10:27 AM, Bago <mrbago@gmail.com> wrote:
>
> And a core.[pid] file (ie core.12345), I'm not sure what to do with the core
> file so I was hopping someone knew if I could try and use to it to help
> debug the issue. Also I suspect there is a memory leak involved. Has anyone
> else seen anything like this?

a core file is a sign that the engine flat out segfaulted.  You can
try to see what happened by loading the core file into gdb, but unless
you built python and the extension modules you may be using with
debugging symbols, the backtrace you'll get is kind of useless.

The thing to determine is what is causing the segfault: is it

- the engine on the same node every time, no matter what it's doing?
In that case, either that particular machine has a bad build of
something, or it could have a hardware problem such as bad ram (less
likely).

- or the one that happens to execute a particular piece of code?  This
is the more likely scenario, and if you can bracket what code is being
run when it segfaults, you should be able to reproduce the crash in
serial mode.

Note that a segfault is unlikely to be ipython's fault directly, since
we don't have any C code ourselves.  But the parallel system uses
pyzmq aggressively, and that's certainly extension code, so it could
be an issue in pyzmq.

Though given what you describe, I'm inclined to think it's a bug in
some extension code you're loading yourself...

Cheers,

f


More information about the IPython-User mailing list