Why Do Disk Read/Write Traffic and Average Disk Read/Write Duration Fluctuate Occasionally When the Production/Consumption Traffic Are Low?

The Disk Read Speed, Disk Write Speed, Average Disk Read Time, and Average Disk Write Time metrics output average values in the recent minute. These metrics are for reference only. This usually occurs because Kafka asynchronously flushes data to disks (that is, synchronous flushing is not enabled for topics). As a result, flow control is applied for seconds when the disk write bandwidth reaches the upper limit. In that same second of reaching the limit, write operations are delayed significantly, causing instantaneous fluctuation of the disk read/write traffic and average read/write duration. However, in general, such fluctuation does not affect services.