Description
While monitoring from the connectors “Overview” tab for Connectors in Cloud UI, you may notice that Throughput - Messages (messages/sec) graph not showing the values under very less loads even though Production (bytes/sec) graph values are shown.
Applies To
Confluent Cloud
Cause
The reason for this behavior is that as the throughput is being measured in messages/sec based on granularity and round off logic, and in scenarios were it is very close to zero it is expected behavior as per below explanation.
For instance, the Last 7 days range maps to PT4H granularity (4 hours bucket) which will get mapped to 4 × 3600 = 14,400 seconds. Now with existing logic, any value below 7,200 will get rounded off to 0 only.
So with the within 4 hours buckets for Throughput (messages/sec) graph to display the details the messages/sec in that bucket should be greater than 7,200, otherwise that Throughput (messages/sec) graph will not show the expected details as per the earlier mentioned round off logic.
Resolution
In such scenarios of lower throughputs it is suggested to monitor the connector throughput directly from Metrics API. Example, monitor sent_records metrics and received_records metrics from Metrics API for source and sink connectors respectively.