# ResponseInfo GET ignored for some Kernels

[gavrand]

Hi

I am using Apache Toree as Jupyter Kernel and ResponseInfo is being ignored for it. It is working perfectly with standard Python2 Kernel. #Get does work perfectly !

I guess it is due to "shortcut" is being used to produce response_code in /services/notebooks/handlers.py (line 199)

response_future = self.execute_code(kernel_client, kernel_id, response_code)

instead of full packing for request_code (line 180)

request = json.dumps({
                'body' : parse_body(self.request),
                'args' : parse_args(self.request.query_arguments),
                'path' : self.path_kwargs,
                'headers' : headers_to_dict(self.request.headers)

and hence remote Apache Kernel (interacting via Jupyter protocol) might not be happy with it.
Again, it is from my limited knowledge and time spent. However Apache PySpark does not work:I

Thank you
Andrey

[parente]

Thanks for the report, @gavrand . Can you confirm which mode you're running the kernel gateway in, with jupyter websockets or notebook http? Basically, what's your run command for starting the kernel gateway?

[gavrand]

Hi Peter

Thank you for the prompt reply. It is indeed notebook-http.
The run cmd is below

jupyter kernelgateway --KernelGatewayApp.api=notebook-http --KernelGatewayApp.ip=0.0.0.0 --KernelGatewayApp.seed_uri=./notebook.ipynb

Thank you
Andrey

[parente]

@Lull3rSkat3r mind taking a look at this?

[Lull3rSkat3r]

I'll check it out.

[Lull3rSkat3r]

@gavrand, are you running Apache Toree with the scala or python interpreter?

[gavrand]

Hi @Lull3rSkat3r. It is python interpreter.

Again, #Get is working perfectly. It is only ResponseInfo where I am struggling and hence could not dump application/json :(.
If you want I could try scala one also.

Thank you
Andrey

[Lull3rSkat3r]

@gavrand, would it be possible to attach your notebook?

[gavrand]

Hi @Lull3rSkat3r

I am afraid, I am unable to send it from work by compliance reasons.
But the example from the github /kernel_gateway/tests/resources/responses_2.ipynb
does not work with Apache Toree Kernel also (the simplest one # GET /json)

Andrey

[Lull3rSkat3r]

@gavrand, no worries. I will use the example notebook. Will post my results soon.

[Lull3rSkat3r]

So investigating this in more detail exposed a bug in the Toree Kernel. Essentially the PySpark interpreter in Toree is returning print messages as an execute result. The KG code currently expects the code to output the response info via a stream message. You can see this by running a code cell in Toree PySpark:

This might bring up the point of reading the text/plain value from an execute result, but is not directly the fault of KG. If we do not want to support this we can close. I have opened a bug on Toree for the real issue TOREE 318.

[gavrand]

Thank you very much for investigation, Corey !
Will see what Toree say.

Andrey

[parente]

Tracing through the issue @Lull3rSkat3r noted above, it appears this problem has been fixed by apache/incubator-toree#47.

Going to close here since the problem was in the dependency.