When an application being managed by UCP fails, you would like a summary of all requests made to the UCP API in the hours leading up to the failure. What must be configured correctly beforehand for th

UCP's audit logs specifically record API requests made to the UCP controller. Setting the audit log level to `metadata` captures information about the API request, including who made it and what resource was affected, while the `request` level includes the full request body, both crucial for reconstructing API call sequences.

UCP logging levels (info/debug) provide operational logs for UCP components, but not a structured summary of *API requests* for audit purposes.

Docker engine log drivers manage container logs, not UCP API audit logs, and 'metadata' or 'request' are audit log levels, not log driver types.

Setting the logging level for a Kubernetes API server container to `warning` or higher would *reduce* the verbosity, missing detailed request information, and refers to the Kubernetes API, not the UCP API audit logs.