Operations settings

What this group controls

Operations collects Control Center access, refresh behavior, Application Insights, debug logging, file-processing logs, health checks, and Swagger documentation.

Why it matters

Operational settings determine what admins can observe during incidents. Enable enough telemetry to diagnose problems without leaving verbose logging or unauthenticated endpoints exposed longer than needed.

Video planned

Operations settings walkthrough

Recording planned. Walk through each tab in the Operations group and explain when to change each setting.

To publish this video:

1. Upload the recording to YouTube or Microsoft Stream.
2. Save a poster frame as docs/images/video-posters/admin-operations.png.
3. Add the watch URL to this slot in docs/_data/media.yml.

Before you change anything

• Create operational roles before restricting Control Center views.
• Confirm Application Insights configuration before relying on telemetry.
• Agree with security teams before exposing health-check or Swagger endpoints.

Control Center

Automatic Data Refresh

The Automatic Data Refresh section belongs to the Control Center tab. Use it with the adjacent settings in this group so related rollout, access, and operational choices stay aligned.

Control Center Access

The Control Center Access section belongs to the Control Center tab. Use it with the adjacent settings in this group so related rollout, access, and operational choices stay aligned.

Settings

Setting	What it does	Default	Notes
Enable Daily Control Center Refresh	Defines behavior for the related admin workflow; verify the affected feature after saving.	On	control_center_auto_refresh_enabled
Refresh Time ()	Defines behavior for the related admin workflow; verify the affected feature after saving.	02:00	control_center_auto_refresh_time
Require ControlCenterAdmin App Role	Requires the ControlCenterAdmin app role before users can use this capability or view.	Off	require_member_of_control_center_admin
Allow ControlCenterDashboardReader App Role	Requires the Allow ControlCenterDashboardReader app role before users can use this capability or view.	Off	require_member_of_control_center_dashboard_reader

Logging & Health

Application Insights

The Application Insights section belongs to the Logging & Health tab. Use it with the adjacent settings in this group so related rollout, access, and operational choices stay aligned.

Debug Logging

The Debug Logging section belongs to the Logging & Health tab. Use it with the adjacent settings in this group so related rollout, access, and operational choices stay aligned.

File Process Logging

The File Process Logging section belongs to the Logging & Health tab. Use it with the adjacent settings in this group so related rollout, access, and operational choices stay aligned.

Health Check

The Health Check section belongs to the Logging & Health tab. Use it with the adjacent settings in this group so related rollout, access, and operational choices stay aligned.

API Documentation

The API Documentation section belongs to the Logging & Health tab. Use it with the adjacent settings in this group so related rollout, access, and operational choices stay aligned.

Settings

Setting	What it does	Default	Notes
Enable /external/healthcheck	Exposes the capability after required services, permissions, and rollout policy are ready.	Off	enable_external_healthcheck; capability toggle
Enable /external/healthcheckz	Exposes the capability after required services, permissions, and rollout policy are ready.	Off	enable_no_auth_external_healthcheck; capability toggle
Enable Swagger/OpenAPI Documentation (/swagger)	Exposes the capability after required services, permissions, and rollout policy are ready.	On	enable_swagger; capability toggle
Enable Application Insights Global Logging	Sends global application, agent, and orchestration logging events to Application Insights.	Off	enable_appinsights_global_logging; capability toggle
Enable Debug Logging	Captures verbose diagnostic logs for troubleshooting until disabled or the timer turns it off.	Off	enable_debug_logging; capability toggle
Enable Time-based Auto Turnoff	Defines behavior for the related admin workflow; verify the affected feature after saving.	Off	debug_logging_timer_enabled
Duration	Defines behavior for the related admin workflow; verify the affected feature after saving.	1	debug_timer_value
Time Unit	Defines behavior for the related admin workflow; verify the affected feature after saving.	hours	debug_timer_unit
Enable File Processing Logs	Records upload, extraction, indexing, and file-processing events for administrative troubleshooting.	On	enable_file_processing_logs; capability toggle
Enable Time-based Auto Turnoff	Defines behavior for the related admin workflow; verify the affected feature after saving.	Off	file_processing_logs_timer_enabled
Duration	Defines behavior for the related admin workflow; verify the affected feature after saving.	1	file_timer_value
Time Unit	Defines behavior for the related admin workflow; verify the affected feature after saving.	hours	file_timer_unit
Delete logs older than	Defines behavior for the related admin workflow; verify the affected feature after saving.	Not specified in defaults	Runtime UI control
Age unit	Defines behavior for the related admin workflow; verify the affected feature after saving.	Not specified in defaults	Runtime UI control

Common tasks

1. Configure Control Center. Set refresh behavior and access requirements, then open Control Center. Outcome to verify: Only intended admins can view refreshed operational data.
2. Enable useful logging. Enable telemetry needed for a small operation and inspect the result. Outcome to verify: The operation leaves the expected trace.
3. Expose health endpoints deliberately. Enable the needed route and test monitoring. Outcome to verify: Monitoring probes the expected endpoint only.

Troubleshooting

Symptom	Likely cause	Fix
No telemetry appears	Application Insights or file-processing logs are disabled or misconfigured.	Enable the relevant logging path and rerun a reproducible operation.

Diagnose slow tabular model requests

An AutoFunctionInvocationLoop dependency measures the enclosing Semantic Kernel operation, not each HTTP request to the model. To distinguish model HTTP waits and SDK retries from fast spreadsheet tools, temporarily enable the diagnostic client for the foreground tabular-analysis Azure OpenAI service. This does not change ordinary chat clients, durable export clients, Custom endpoints, or Anthropic clients.

For App Service code deployments, deploy the updated application, then add the following environment settings and restart the app:

Environment setting	Diagnostic default	Accepted values
SIMPLECHAT_TABULAR_HTTP_DIAGNOSTICS	Off	true, 1, yes, or on enables diagnostics
SIMPLECHAT_TABULAR_HTTP_TIMEOUT_SECONDS	60	Finite seconds from 1 through 600 per HTTP I/O timeout
SIMPLECHAT_TABULAR_HTTP_MAX_RETRIES	0	Integer from 0 through 2; retries are in addition to the first attempt
SIMPLECHAT_TABULAR_HTTP_BUDGET_SECONDS	180	Finite seconds from 1 through 1800, shared across HTTP attempts and tool rounds in one diagnostic client

Timeout, retry, and budget overrides apply only when diagnostics is enabled. Invalid numeric values use the diagnostic defaults. Set the retry count to 1 to observe a controlled retry; leave it at 0 for the first fast-failure test. The shared budget starts at the first HTTP attempt, includes elapsed time between attempts, and limits an active HTTP send to the remaining budget. It does not cancel spreadsheet loading, tool execution, SDK retry backoff, or final chat synthesis outside this client. This is not a deadline for the entire chat turn.

Keep the existing Application Insights connection string configured. The new events use log_event and do not require SimpleChat debug logging or OPENAI_LOG=debug. Reproduce the request and run:

traces
| where timestamp > ago(1h)
| where tostring(customDimensions.sc_message) startswith "[TABULAR_MODEL_HTTP]"
| project timestamp,
  event = tostring(customDimensions.sc_operation),
  run_id = tostring(customDimensions.sc_diagnostic_run_id),
  attempt_id = tostring(customDimensions.sc_http_attempt_id),
  attempt = toint(customDimensions.sc_attempt_sequence),
  sdk_retry = toint(customDimensions.sc_retry_count),
  duration_ms = todouble(customDimensions.sc_duration_ms),
  status_code = toint(customDimensions.sc_status_code),
  error_type = tostring(customDimensions.sc_error_type),
  underlying_error = tostring(customDimensions.sc_reason),
  budget_exhausted = tobool(customDimensions.sc_budget_exhausted),
  provider_request_id = tostring(customDimensions.sc_provider_request_id),
  endpoint_fingerprint = tostring(customDimensions.sc_endpoint_fingerprint),
  operation_Id
| order by timestamp asc

Each SDK HTTP attempt has an attempt_started event followed by attempt_completed or attempt_failed. Completion does not necessarily mean a successful HTTP status: inspect status_code. All attempts for one diagnostic client share run_id, including SDK retries and subsequent tool-result calls. An endpoint fingerprint hashes the host, port, and request path to distinguish deployment destinations without publishing their names or URLs. Provider request IDs are recorded only when an allowlisted response header is available; failures before response headers arrive have none. Events never contain request bodies, response bodies, tool payloads, authorization headers, or exception messages.

After testing, remove these settings or set SIMPLECHAT_TABULAR_HTTP_DIAGNOSTICS=false and restart. This restores the existing SDK defaults. The focused regression command is python -m pytest functional_tests/test_tabular_model_http_diagnostics.py -q.

Related

• Administration settings overview
• Help settings
• Scale settings