[HADOOP-14502] Confusion/name conflict between NameNodeActivity#BlockReportNumOps and RpcDetailedActivity#BlockReportNumOps

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 3.0.0-alpha4
• Component/s: metrics
• Labels: Incompatible
• Target Version/s: 3.0.0-alpha4
• Hadoop Flags: Incompatible change, Reviewed
• Release Note: Remove the BlockReport(NumOps,AvgTime) metrics emitted under the NameNodeActivity context in favor of StorageBlockReport(NumOps,AvgTime) which more accurately represent the metric. Same for the corresponding quantile metrics.
• Assignee: Erik Krogen
• Reporter: Erik Krogen
• Created: 07/Jun/17 18:23
• Updated: 27/Jul/17 22:41
• Resolved: 21/Jun/17 23:34

Description

Currently the BlockReport(NumOps|AvgTime) metrics emitted under the RpcDetailedActivity context and those emitted under the NameNodeActivity context are actually reporting different things despite having the same name. NameNodeActivity reports the count/time of per storage block reports, whereas RpcDetailedActivity reports the count/time of per datanode block reports. This makes for a confusing experience with two metrics having the same name reporting different values.

We already have the StorageBlockReportsOps metric under NameNodeActivity. Can we make StorageBlockReport a MutableRate metric and remove NameNodeActivity#BlockReport metric? Open to other suggestions about how to address this as well. The 3.0 release seems a good time to make this incompatible change.

Issue Links

• relates to: HADOOP-14492 RpcDetailedMetrics and NameNodeMetrics use different rate metrics abstraction cause the Xavgtime confused