question

MehdiSamsami-7668 avatar image
2 Votes"
MehdiSamsami-7668 asked JamesWatson-5515 commented

Real-time endpoint doesn't work (error 502)

Hello. I created a training pipeline using the ML designer, and also created a real-time inferencing pipeline after that. The training pipeline ran completely without any error. I also published the real-time inferencing pipeline successfully. In the end, I deployed the real-time inferencing pipeline to AKS and it resulted in a real-time endpoint with a Healthy state.

However, when I want to test the endpoint from the "Endpoints" page, it returns nothing! Just a red empty box.
51985-image.png

I tried to run the example consumption codes in the "Consume" tab in the Notebooks section of the ML Studio.
52042-image.png


But again, it didn't give any results and returned the following error description:
51986-image.png



I don't know what causes this issue. I appreciate if you could help me with this. Thanks.

azure-machine-learningazure-machine-learning-inference
image.png (19.8 KiB)
image.png (39.3 KiB)
image.png (63.4 KiB)
5 |1600 characters needed characters left characters exceeded
Toggle Comment visibility. Current Visibility: Visible to all users

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

2 Answers

ramr-msft avatar image
1 Vote"
ramr-msft answered KamranAli-0346 commented

@MehdiSamsami-7668 Thanks for the question. Can you please enable application insights as mentioned in the doc and share the logs to check. Also please share link to the sample that you are trying. Doc for creation of real-time inference endpoint authenticating to the inference cluster and doc to consume webservice Input and output module.


· 3
5 |1600 characters needed characters left characters exceeded
Toggle Comment visibility. Current Visibility: Visible to all users

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

ramr-msft avatar image ramr-msft ·

@MehdiSamsami-7668 Thanks, you can now register a trained model in Designer into the model registry and use automatically generated score.py/conda yaml to deploy it using SDK or CLI. After successful model training, publishing to inference cluster in authenticating to the AKS cluster and consume.



1 Vote 1 ·
MehdiSamsami-7668 avatar image MehdiSamsami-7668 ·

Hi @ramr-msft . Thank you for your answer. Actually, when I wanted to update the real-time inferencing pipeline and deploy it again to the same real-time endpoint, I repeatedly faced with errors! So, I deleted that real-time endpoint and deployed the pipeline (without any update) to a new real-time endpoint only with the following changes, and strangely, it works now!!!

  • I enabled application insights

  • I increased the CPU from 0.1 to 0.2

I'm not sure what is the reason, but this endpoint works correctly:
52471-image.png


1 Vote 1 ·
image.png (25.7 KiB)
KamranAli-0346 avatar image KamranAli-0346 MehdiSamsami-7668 ·

I am having same issue however, the resolution does not work for me :-(

0 Votes 0 ·
FernandoJosRibeiroJnior-4000 avatar image
0 Votes"
FernandoJosRibeiroJnior-4000 answered JamesWatson-5515 commented

I am having same issue

· 1
5 |1600 characters needed characters left characters exceeded
Toggle Comment visibility. Current Visibility: Visible to all users

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

JamesWatson-5515 avatar image JamesWatson-5515 ·

They never have any answers its very frustrating .

0 Votes 0 ·