Persistent 503 Errors in West Central US - Infrastructure Failure (Logging Not Reached)

This problem requires Azure Support intervention as the root cause is outside the customer's application and configuration management. The issue is a platform-level network failure occurring before the request reaches the logging component.

DETAILS FOR AZURE SUPPORT

1. Problem Description

Our Azure Container App, 'accountservice-app' (in a Dev/Test environment), began experiencing sporadic HTTP 503 Service Unavailable errors starting abruptly ~2 days ago in the West Central US region.

The failure is intermittent and consistent with an "upstream connect error or disconnect/reset before headers" symptom at the Ingress layer, indicating a network-level connection failure/reset.

Critical Validation Summary (All App/Config Issues Ruled Out)

The following steps confirm the issue is not caused by application code, configuration, or resource limits:

Code/Config: Rolled back to the exact revision that was stable ~2 days ago. The 503 errors persisted.

Stability/Scale: Minimum replicas were scaled from 1 to 2, and the sporadic 503s persisted.

Resources: CPU usage is confirmed as extremely low (average m cores).

Ingress Config: Target Port 8080 is correct. Ingress Transport was explicitly set to HTTP/1.1.

Health Check: Readiness/Liveness probes are configured and consistently returning HTTP 200 OK.

Definitive Evidence of Infrastructure Failure

The most critical finding proving this is an infrastructure issue is the lack of logging:

The 503 errors received by the client are NOT being logged in the ContainerAppSystemLogs_CL or the ContainerAppHTTPLogs_CL tables during the time the failure occurs.

This means the failure is occurring at a point in the network stack before the request is successfully handed off to the ACA logging service—likely at the ACA Public Ingress/Load Balancer or the host machine's network interface in the West Central US region.

Requested Action

We request that the Azure infrastructure team investigate the host telemetry, network traces, and load balancer logs for the Container Apps Environment in the West Central US region to diagnose the transient connection resets.This problem requires Azure Support intervention as the root cause is outside the customer's application and configuration management. The issue is a platform-level network failure occurring before the request reaches the logging component.

DETAILS FOR AZURE SUPPORT

Problem Description

Our Azure Container App, 'accountservice-app' (in a Dev/Test environment), began experiencing sporadic HTTP 503 Service Unavailable errors starting abruptly ~2 days ago in the West Central US region.

The failure is intermittent and consistent with an "upstream connect error or disconnect/reset before headers" symptom at the Ingress layer, indicating a network-level connection failure/reset.

Critical Validation Summary (All App/Config Issues Ruled Out)

The following steps confirm the issue is not caused by application code, configuration, or resource limits:

Code/Config: Rolled back to the exact revision that was stable ~2 days ago. The 503 errors persisted.

Stability/Scale: Minimum replicas were scaled from 1 to 2, and the sporadic 503s persisted.

Resources: CPU usage is confirmed as extremely low (average m cores).

Ingress Config: Target Port 8080 is correct. Ingress Transport was explicitly set to HTTP/1.1.

Health Check: Readiness/Liveness probes are configured and consistently returning HTTP 200 OK.

Definitive Evidence of Infrastructure Failure

The most critical finding proving this is an infrastructure issue is the lack of logging:

The 503 errors received by the client are NOT being logged in the ContainerAppSystemLogs_CL or the ContainerAppHTTPLogs_CL tables during the time the failure occurs.

This means the failure is occurring at a point in the network stack before the request is successfully handed off to the ACA logging service—likely at the ACA Public Ingress/Load Balancer or the host machine's network interface in the West Central US region.

Requested Action

We request that the Azure infrastructure team investigate the host telemetry, network traces, and load balancer logs for the Container Apps Environment in the West Central US region to diagnose the transient connection resets.