[IPython-User] Questions regarding notebook

MinRK benjaminrk@gmail....
Thu Jan 12 21:04:22 CST 2012


On Thu, Jan 12, 2012 at 13:25, Fernando Perez <fperez.net@gmail.com> wrote:

> On Thu, Jan 12, 2012 at 4:37 AM, Thomas Wiecki
> <thomas.wiecki@googlemail.com> wrote:
> >>
> >> # Kernel heartbeat interval in seconds.
> >> c.MappingKernelManager.time_to_dead = 3.0
> >
> > Yup, I looked there too but then remembered that I overwrote it with
> > the example config from the ipython notebook tutorial. I recreated the
> > config and the parameter is there. I set the value to 120, which does
> > the trick, and will pray the kernel server never crashes ;).
>

Of course, you are right - an extremely long heartbeat is essentially
disabling it.  To really disable it, you can set the heartbeat to 1e8 (~3
years).  If your code doesn't crash, this is not a problem.


>
> It's odd that you had to set it so high; as Thomas K. said, the
> heartbeat runs in its own C++ thread by zeromq, so the fact that your
> machine is busy should have little impact on this, unless your
> computations are hitting your box so hard and causing so much virtual
> memor paging that the thread-switching latency at the OS level is
> worse than 3s.
>

I found the probable cause, detailed in this
issue<https://github.com/ipython/ipython/issues/1260>.
 It is indeed possible to for code to get in the way of the heartbeat via
the GIL.

-MinRK



>
> Cheers,
>
> f
> _______________________________________________
> IPython-User mailing list
> IPython-User@scipy.org
> http://mail.scipy.org/mailman/listinfo/ipython-user
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.scipy.org/pipermail/ipython-user/attachments/20120112/96219e96/attachment.html 


More information about the IPython-User mailing list