If your application experiences performance degradation or connectivity issues due to saturated sockets, you can configure Auto-Heal to monitor and restart the process before the app becomes unreachable:
Access the Portal: Go to your App Service in the Azure Portal.
Navigation: In the left-hand menu, select “Diagnose and solve problems”.
Tools: Select “Diagnostic Tools” and then click on “Auto-Heal”.
Define Conditions: Choose “Request Count”.
Example: Set it to 2000 requests within a 10-minute interval (or a value significantly higher than your average traffic).
Configure Actions: Select “Recycle”.
Finalize: Click “Save” to apply the rule.
By setting this up, Azure will monitor the traffic patterns. If it detects an anomalous spike (which often precedes a socket lockout), it will automatically recycle the worker process to clear the connection table.
Once the root cause (such as a non-singleton CosmosClient) is fixed, you may want to disable the automated recycling:
Locate the tool: Navigate back to Diagnose and solve problems > Auto-Heal.
Toggle Off: You will see a main toggle at the top (usually labeled “Enable Auto-Heal”).
Disable: Switch the toggle to Off (or Disabled). This immediately deactivates all saved rules.
Confirm: Remember to click “Save” at the bottom of the page to apply the changes.
