Hi Community,
I'm experiencing persistent deployment timeouts (often 1-2+ hours) when trying to provision a VNet-integrated (internal) Azure Container Apps Environment in the West US 3 region.
We are deploying via Bicep/ARM templates. The target infrastructure subnet meets the minimum size requirements (/23
) for a non-zone-redundant environment and has no service delegations prior to deployment.
After extensive troubleshooting, including deploying a test VM into a separate subnet within the same VNet and applying the exact same Network Security Group (NSG) used by the ACA infrastructure subnet, we've narrowed down the issue.
Key Finding:
Connectivity tests run from the test VM (using Azure's default DNS resolver 168.63.129.16) show that DNS resolution consistently fails specifically for the regional Microsoft Container Registry data endpoint: westus3.data.mcr.microsoft.com.
- DNS resolution for other public endpoints (like
management.azure.com
, login.microsoftonline.com
, mcr.microsoft.com
) works correctly from the same VM.
- DNS resolution for our Private Endpoints via Private DNS Zones also works correctly.
- TCP connectivity tests pass for all endpoints except those where DNS resolution fails.
- The NSG associated with the subnet allows outbound traffic according to ACA documentation (including TCP/443 to Service Tags
MicrosoftContainerRegistry
, AzureResourceManager
, AzureActiveDirectory
, AzureMonitor
, and UDP/TCP 53 to Azure DNS).
- Crucially, the DNS failure for
westus3.data.mcr.microsoft.com
persists even when we temporarily added an NSG rule to allow ALL outbound traffic (protocol: *
, destination: Internet
). This strongly suggests the issue is not the NSG configuration itself.
Question:
Has anyone else experienced similar ACA Environment deployment timeouts in West US 3 (or potentially other regions) recently? Specifically, has anyone encountered issues resolving regional MCR data endpoints like [region].data.mcr.microsoft.com from within a VNet using Azure DNS?
We suspect a potential Azure DNS issue specific to this endpoint/region/VNet context and are planning to open a support ticket, but wanted to check with the community as well.
Thanks for any insights!