QTConsole with --ssh option and kernel interrupt

[ohe]

In current master branch, when using a remote kernel, accessed via the --ssh option of ipython qtconsole. There is no way to interrupt a python call, either with CTRL+C either with the "Kernel" menu. In both case, the following message appears :

Kernel process is either remote or unspecified. Cannot interrupt.

By digging on this tracker, this seems to be the "normal" behavior when someone tries to send an interrupt/restart order to a remote kernel.

what are the sticking points related to this "feature"?
Is there any plans to fix this issue?

BTW, thanks for all the work done so far.

[minrk]

Yes, this is the correct behavior - only the frontend that started the kernel is allowed to interrupt and restart it.

The sticking point is that signaling is handled simply by sending an interrupt directly to the process (via Process.send_signal). This should be able to be done with any local kernel (possibly not Windows) via os.kill(pid, SIGINT). This is complicated slightly by the fact that the kernel's pid can be a function of time, because it is a new process every time it is restarted, but is still identified as the same kernel.

It cannot be done remotely until we have a mechanism for sending signals over the wire. In order for this to work, every kernel must have a nanny process that would receive messages, and signal the kernel. This should not be difficult (a similar, but much more complicated idea was done in the daemon branch of IPython 0.10, back on Launchpad), but I don't think it's on anyone's immediate todo list.

[fperez]

As a note, if you access a remote kernel via the web notebook, you can interrupt it, since the notebook server in that case acts precisely like the nanny process Min describes, and does the job of sending the signal locally for you. Not ideal, I know, but it's a limited workaround for now...

[ohe]

@fperez, this is a very interesting point. But, for many reasons, I (in fact we, in our company) can't use for now the web notebook. I fixed this issue with a big and dirty hack on top of our infrastructure as a workaround for this issue.
I see that issue is prioritized for 0.13 release ... Can't wait for it ;)

[fperez]

On Tue, Nov 22, 2011 at 9:45 AM, Olivier Hervieu
reply@reply.github.com
wrote:

@fperez, this is a very interesting point. But, for many reasons, I (in fact we, in our company) can't use for now the web notebook. I fixed this issue with a big and dirty hack on top of our infrastructure as a workaround for this issue.

OK, glad you could find a workaround. BTW, if any of the issues
you/your company have with the notebook are things we can fix (i.e.
not including supporting very old browsers without websockets or other
similar things beyond our control), by all means let us know! Issue
#977 is an 'open thread' place to list improvements you'd like to see
made to the notebook.

I see that issue is prioritized for 0.13 release ... Can't wait for it ;)

Keep in mind that mostly means 'not for 0.12'. Whether it actually
happens for 0.13 or not is more dependent on someone whipping up a
pull request for us than anything else: we have a very small team and
unfortunately this isn't the highest priority item on our list right
now...

[Carreau]

Closing as this issue is not in IPython itself and if still problematic and relevant should be opened on the right repository. This will allow to keep the number of opened issue on the IPython repo under control.

Feel free to keep commenting or reopen if necessary.

Thanks.