Skip to content

Latest commit

 

History

History
59 lines (36 loc) · 2.56 KB

troubleshooting.md

File metadata and controls

59 lines (36 loc) · 2.56 KB

Troubleshooting web services deployments

During deployment or when calling the web service, you can use the following resources to determine the cause of the errors.

Note: If you get a "bad gateway error" when calling the web service immediately after deploying it, it is normally caused by the deployment process still going through its steps. Wait a minute, and try again. It should go away.

1. Docker logs

After using the "service create" command, a docker container is created to package the model and its dependencies for deployment. You can find this container and view its log. Look for the image that has the name of your service.

docker ps

If the above does not show your container, try:

docker ps -a

Then use the container ID in the following call:

docker logs \<containerid>

2. Kubernetes logs

If you deployed your web service to an ACS cluster, then you can view the Kubernetes logs. To access the logs, you need to use the kubectl tool to access the Kuberentes UI. In Windows, the tool is normally installed to the c:\users\<username>\bin folder.

Install Kubectl from here

To access it, type the following:

c:\users\\\<username>\bin>kubectl proxy

Then browse to the specified address (127.0.0.1:8001/ui). Once there, click on Pods, then on the hamburger icon for your service.

3. App Insights

If you use the -l flag when deploying a web service, then the service logs are written to the App Insights instance for your environment in Azure. You can search for it using the environment name you used when using the az ml env setup command.

To access it:

  • Use -l when creating service
  • Open App Insights in Azure Portal. Use your environment name to find the App Insights instance.
  • In App Insights, click on Search in the top menu to view the results
  • Alternatively, go to Analytics > Exceptions > exceptions take | 10

4. Error handling in script

You can use exception handling in your scoring.py script's run function to return the error message:

    try:
        <code to load model and score>
    except Exception as e:
        return(str(e))

5. Other known issues

  • Do not re-use the environment name for the env setup command
  • If env setup fails, make sure you have enough cores available in your subscription
  • Do not use underscore ( _ ) in the web service name (as in my_webservice)
  • Retry if you get "Bad gateway error" when calling the web service. It normally means the container hasn't been deployed to the cluster yet.