In a Google Container Engine cluster all of the standard out of the pods are aggregated and send to Google Cloud Logging. This works out fine except you don't have any control over the logging meta data like
google.logging.v1.LogSeverity
Is their a way to enrich the logging output on "standard out/err" so that the fluentd daemon on each cluster node recognizes the log severity on each log line.
But without changing something in the fluentd daemon on each node (that would defy the purpose of running as a service) or adding a custom fluentd daemon in each pod.