Details
Description
Now we use "totalRequestCount"(RSRpcServices#requestCount) to calculate requests per second.
After HBASE-18469, "totalRequestCount" count only once for multi request.(Includes requests that are not serviced by regions.)
When we have a large number of read and write requests, the value of "Requests Per Second" is very small which does not reflect the load of the cluster.
Maybe it is more reasonable to use "totalRowActionRequestCount" to calculate RPS?
Attachments
Attachments
Issue Links
- is related to
-
AMBARI-24256 Ambari Grafana HBase metrics wrong - readRequestCount > totalRequestCount!
- Open
- relates to
-
HBASE-18469 Correct RegionServer metric of totalRequestCount
- Resolved