[Core] [Azure] Use subscription id from azure profile if not provided in config during AzureNodeProvider init #56640

marosset · 2025-09-17T21:20:11Z

Why are these changes needed?

When ray get-head-ip is called the provider_config that is passed to AzureNodeProvider's init only contains info from the cluster configuration file the user authored, which doesn't require a subscription_id to be set.
This causes a key read error when the AzureNodeProvider is getting initialized (unless the users set that).

Since other operations like ray up, ray attached, etc do not require subscription_id to be set, this patch updates the init to read the subscription_id from the currently logged in azure profile if it isn't set in the cluster config.

Related issue number

Fixes: #44254

Checks

I've signed off every commit(by using the -s flag, i.e., git commit -s) in this PR.
I've run scripts/format.sh to lint the changes in this PR.
I've included any doc changes needed for https://docs.ray.io/en/master/.
- I've added any new APIs to the API Reference. For example, if I added a
  method in Tune, I've added it in doc/source/tune/api/ under the
  corresponding .rst file.
I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
Testing Strategy
- Unit tests
- Release tests
- This PR is not tested :(

Note

AzureNodeProvider now falls back to the Azure CLI profile subscription ID if provider_config.subscription_id is missing.

Azure Autoscaler:
- python/ray/autoscaler/_private/_azure/node_provider.py
  - AzureNodeProvider.__init__: Make subscription_id optional; if absent, use get_cli_profile().get_subscription_id() and log the chosen ID.
  - Add import for get_cli_profile and use the resolved subscription_id to initialize ComputeManagementClient, NetworkManagementClient, and ResourceManagementClient.

^{Written by Cursor Bugbot for commit 13a2157. This will update automatically on new commits. Configure here.}

marosset · 2025-09-17T21:21:27Z

I tested this manually

ray get-head-ip /home/marosset/src/github.com/marosset/ray/python/ray/autoscaler/azure/tests/azure_cluster.yaml
2025-09-17 14:06:08,164 - INFO - NumExpr defaulting to 8 threads.
2025-09-17 14:06:09,373 INFO node_provider.py:74 -- subscription_id not found in provider config, but successfully read from azure profile: 48d13bcf-f3f8-4070-8f75-cd748d4152c4
2025-09-17 14:06:09,373 - INFO - No environment configuration found.
2025-09-17 14:06:09,374 - INFO - ManagedIdentityCredential will use IMDS
2025-09-17 14:06:09,377 - INFO - No environment configuration found.
2025-09-17 14:06:09,377 - INFO - ManagedIdentityCredential will use IMDS
2025-09-17 14:06:11,485 - INFO - DefaultAzureCredential acquired a token from AzureCliCredential
2025-09-17 14:06:12,331 - INFO - AzureCliCredential.get_token succeeded
2025-09-17 14:06:12,331 - INFO - DefaultAzureCredential acquired a token from AzureCliCredential
2025-09-17 14:06:14,214 - INFO - DefaultAzureCredential acquired a token from AzureCliCredential
2025-09-17 14:06:14,815 - INFO - AzureCliCredential.get_token succeeded
2025-09-17 14:06:14,815 - INFO - DefaultAzureCredential acquired a token from AzureCliCredential
20.191.82.34

my cluster config looks like

cat /home/marosset/src/github.com/marosset/ray/python/ray/autoscaler/azure/tests/azure_cluster.yaml
# Unique identifier for the head node and workers of this cluster.
cluster_name: nightly-cpu-minimal-vm
max_workers: 2
idle_timeout_minutes: 5

# Cloud-provider specific configuration.
provider:
  type: azure
  # https://azure.microsoft.com/en-us/global-infrastructure/locations
  location: westus2
  resource_group: ray-test
  cache_stopped_nodes: False

...

gemini-code-assist

Code Review

This pull request addresses an issue where ray get-head-ip would fail for Azure clusters if the subscription_id is not explicitly provided in the configuration file. The change introduces a fallback mechanism in the AzureNodeProvider's initialization to retrieve the subscription ID from the active Azure CLI profile, which aligns its behavior with other commands like ray up. This is a good fix. I have one minor suggestion for improving code consistency.

python/ray/autoscaler/_private/_azure/node_provider.py

marosset · 2025-09-17T21:21:52Z

cc @MKLepium

jackfrancis · 2025-09-19T23:43:55Z

python/ray/autoscaler/_private/_azure/node_provider.py


    def __init__(self, provider_config, cluster_name):
        NodeProvider.__init__(self, provider_config, cluster_name)
+        if "subscription_id" not in provider_config:


Let's do this the exact same way both here and in python/ray/autoscaler/_private/_azure/config.py (in _configure_resource_group).

If we do this after L80 we can avoid polluting the actual provider_config, e.g., a new addition @ L81:

if subscription_id is None: # Get subscription from logged in azure profile # if it isn't provided in the provider_config # so operations like `get-head-ip` will work subscription_id = get_cli_profile().get_subscription_id() logger.info( "subscription_id not found in provider config, falling back " f"to subscription_id from the logged in azure profile: {subscription_id}" )

let me check to see if the value is pulled out of the provider config dictionary later and if it isn't I'll update with this suggestion

Updated to not pollute the actual provider_config - PTAL

removed the line to re-add it to the dictionary and tested various ray command (up, down, status, get-head-ip) and all work as expected.

jackfrancis · 2025-10-01T18:01:49Z

python/ray/autoscaler/_private/_azure/node_provider.py

+            # if it isn't provided in the provider_config
+            # so operations like `get-head-ip` will work
+            subscription_id = get_cli_profile().get_subscription_id()
+            provider_config["subscription_id"] = subscription_id


are there any negative side-effects from updating the in-memory representation of the parsed user config in this way?

whoops, i meant to drop the extra line to add it back to the dictionary

… in config during AzureNodeProvider init Signed-off-by: Mark Rossett <[email protected]>

jackfrancis

lgtm

jackfrancis · 2025-10-01T21:09:38Z

ping @jjyao for final review

cursor · 2025-10-01T22:32:20Z

python/ray/autoscaler/_private/_azure/node_provider.py

+            # Get subscription from logged in azure profile
+            # if it isn't provided in the provider_config
+            # so operations like `get-head-ip` will work
+            subscription_id = get_cli_profile().get_subscription_id()


Bug: Azure Subscription ID Missing Causes Client Initialization Failure

If subscription_id isn't provided in the config, the code falls back to get_cli_profile().get_subscription_id(). When no default subscription is set in the Azure CLI profile, this method returns None. This None value is then used to initialize Azure management clients, which expect a valid subscription ID and will fail.

@marosset is this a valid comment?

The comment is strictly true but it is not a regression. The current implementation relies upon the subscription ID being in the config. There are no checks to deal with the config not having a subscription id in the config (hence this PR), it just fails more or less in the way that cursor is describing.

So, this PR adds another fallback layer of accomodation, but does not address the question "what's the right way to fail when no subscription ID is available in any way?" (I think this is out of scope for this PR)

… in config during AzureNodeProvider init (ray-project#56640) Signed-off-by: Mark Rossett <[email protected]> Co-authored-by: Jiajun Yao <[email protected]> Signed-off-by: Seiji Eicher <[email protected]>

… in config during AzureNodeProvider init (ray-project#56640) Signed-off-by: Mark Rossett <[email protected]> Co-authored-by: Jiajun Yao <[email protected]> Signed-off-by: Douglas Strodtman <[email protected]>

marosset requested a review from a team as a code owner September 17, 2025 21:20

gemini-code-assist bot reviewed Sep 17, 2025

View reviewed changes

python/ray/autoscaler/_private/_azure/node_provider.py Outdated Show resolved Hide resolved

marosset added the azure label Sep 17, 2025

github-project-automation bot added this to Azure-Windows and Ray Intersection (go/azure-windows-ray) Sep 17, 2025

marosset added the community-contribution Contributed by the community label Sep 17, 2025

marosset assigned jjyao Sep 17, 2025

marosset force-pushed the azure-fix-missing-subscirption-keyerror branch from f913625 to d644766 Compare September 17, 2025 21:27

jackfrancis reviewed Sep 19, 2025

View reviewed changes

marosset moved this to In Progress in Azure-Windows and Ray Intersection (go/azure-windows-ray) Sep 26, 2025

marosset force-pushed the azure-fix-missing-subscirption-keyerror branch from d644766 to 7912cb8 Compare October 1, 2025 17:58

This comment was marked as outdated.

Sign in to view

jackfrancis reviewed Oct 1, 2025

View reviewed changes

[Core] [Azure] Use subscription id from azure profile if not provided…

3ada4a5

… in config during AzureNodeProvider init Signed-off-by: Mark Rossett <[email protected]>

marosset force-pushed the azure-fix-missing-subscirption-keyerror branch from 7912cb8 to 3ada4a5 Compare October 1, 2025 18:24

jackfrancis approved these changes Oct 1, 2025

View reviewed changes

Merge branch 'master' into azure-fix-missing-subscirption-keyerror

13a2157

jjyao added the go add ONLY when ready to merge, run all tests label Oct 1, 2025

jjyao approved these changes Oct 1, 2025

View reviewed changes

cursor bot reviewed Oct 1, 2025

View reviewed changes

jjyao enabled auto-merge (squash) October 1, 2025 23:54

jjyao merged commit bef2a0e into ray-project:master Oct 2, 2025
7 of 8 checks passed

github-project-automation bot moved this from In Progress to Done in Azure-Windows and Ray Intersection (go/azure-windows-ray) Oct 2, 2025

[Core] [Azure] Use subscription id from azure profile if not provided in config during AzureNodeProvider init #56640

[Core] [Azure] Use subscription id from azure profile if not provided in config during AzureNodeProvider init #56640

Uh oh!

Conversation

marosset commented Sep 17, 2025 • edited by cursor bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Why are these changes needed?

Related issue number

Checks

Uh oh!

marosset commented Sep 17, 2025

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

Uh oh!

marosset commented Sep 17, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

This comment was marked as outdated.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jackfrancis left a comment

Choose a reason for hiding this comment

Uh oh!

jackfrancis commented Oct 1, 2025

Uh oh!

cursor bot Oct 1, 2025

Choose a reason for hiding this comment

Bug: Azure Subscription ID Missing Causes Client Initialization Failure

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

marosset commented Sep 17, 2025 •

edited by cursor bot

Loading