az login interactive goes to a localhost link that can't be reached

[ammarasmro]

This is autogenerated. Please review and update as needed.

Describe the bug

Command Name
az login

Errors:
Chrome returning:

This site can’t be reachedlocalhost refused to connect.
Try:

Checking the connection
[Checking the proxy and the firewall](chrome-error://chromewebdata/#buttons)
ERR_CONNECTION_REFUSED

To Reproduce:

Steps to reproduce the behavior. Note that argument values have been redacted, as they may contain sensitive information.

• Put any pre-requisite steps here...
• az login

Expected Behavior

Environment Summary

Linux-5.4.0-1074-azure-x86_64-with-glibc2.27, Ubuntu 18.04.6 LTS
Python 3.8.13
Installer: DEB

azure-cli 2.36.0

Extensions:
ml 2.3.1

Dependencies:
msal 1.17.0
azure-mgmt-resource 20.0.0

Additional Context

This happens when I try to do anything azure related through an AzureML compute instance. It opens up the interactive screen. I click on my account and it goes to a localhost:... link and gives the error above. I can curl that link from a terminal in the machine and it seems to return a success but the cli doesn't get the response and keeps waiting

[yonzhan]

@jiasli for awareness

[ammarasmro]

For reference, a workaround that I used

az login --use-device-code
pip install azure-cli-core

Then loading the workspace says something about falling back to CLI credentials and avoids the interactive window

More context: I tried 3-4 different versions of azureml-sdk and tried different conda envs (including the one provided by azure, still same thing). Updated VSCode, restarted computer and VM, still not working. Today I had a coworker who had the same problem

[jiasli]

This is more of a local configuration issue and could be similar to #10426

Is your browser trying to reach localhost on HTTPS? Or do you have any config/firewall/hosts that blocks localhost?

[ammarasmro]

Oh I actually tried that issue already, didn't work. Mine is going to:

http://localhost:port/.....

I tried the same script locally (not on the azureml VM) and interactive login worked without a problem