Fluentd Dropping log messages occasionally

[vettom]

Hi
We have fluentd configured to send logs to stack driver. At times we see error message [1] "Dropping 111 log messages" We do have file buffer configured and it is never full.

Seeking help to find answers to following

• Does this error result in lose of logs even when buffer is configured?
• If above error does not result in data lose, what is best metric to monitor Fluentd for any failures? (my reasoning below[2])
• Is it possible to monitor buffer usage percentages?
• What error message when buffer gets full?

[1] [warn]: #0 Dropping 195 log message(s) error="2:Stream removed" error_code="2"
[2] Having looked at the metrics in Prometheus, I am not able to decide, may be I am not interpreting right?

• fluentd_status_buffer_total_bytes : Reports bytes use, not percentage use. Potential to use this by knowing buffer size
• fluentd_status_buffer_queue_length : THis is 0 or 1, not sure what to make of it
• stackdriver_failed_requests_count : Unsure this results in data loss or not since there is unlimited retry.
• fluentd_status_retry_count : same as above.

I am new to fluentd, thanks in advance.