Await kernel.ready in _async_shutdown_kernel
#740
Conversation
| @@ -260,10 +260,11 @@ async def _async_shutdown_kernel( | |||
| if self._using_pending_kernels() and kernel_id in self._pending_kernels: | |||
| raise RuntimeError("Kernel is in a pending state. Cannot shutdown.") | |||
| # If the kernel is still starting, wait for it to be ready. | |||
| elif kernel_id in self._starting_kernels: | |||
| kernel = self._starting_kernels[kernel_id] | |||
| elif kernel_id in self._pending_kernels: | |||
There was a problem hiding this comment.
Since _starting_kernels is a shim for _pending_kernels defined here:
jupyter_client/jupyter_client/multikernelmanager.py
Lines 102 to 105 in 2631730
We might as well use _pending_kernels directly.
|
Looks like the CI checks have been stuck for a couple of hours. Gonna close and reopen in a couple of minutes. |
|
I just kicked a recent build and it passes, so the CI failures are relevant. |
|
I can confirm that these hang locally too. Looking into it myself. |
jupyter_client/multikernelmanager.py
Outdated
| try: | ||
| await kernel | ||
| await kernel.ready |
There was a problem hiding this comment.
I think I see the issue—you are on the right track. The naming of things is getting a little messy.
In this line, kernel is not the kernel manager, like you'd expect. It's the Future returned by _add_kernel_when_ready. Waiting this Future doesn't ensure that kernel has started, it just ensures that the start_kernel method returned. When pending kernels are enabled, start_kernel returns immediately, but the kernel might not be ready. To check that the kernel is ready, we need to call await ... .ready on the actual kernel manager.
Calling .ready on this future is redundant. It might be falsely helping the JupyterLab tests pass because it slows the method enough to dodge the race condition.
|
Thank you, @jtpio, for helping with this! This saved me a ton of time! |
|
Thank you @Zsailer for finishing this one! |
Attempt at fixing jupyterlab/jupyterlab#11886.
According to the comment above:
jupyter_client/jupyter_client/multikernelmanager.py
Line 262 in 2631730
This branch of the code should be waiting for the kernel to be ready.
This change seems to be fixing the issue noticed in JupyterLab in jupyterlab/jupyterlab#11886.
Example run: https://github.com/jupyterlab/jupyterlab/runs/4894046802?check_suite_focus=true from jupyterlab/jupyterlab#11887