Xiaoyu Yao created HDFS-10539:
---------------------------------
Summary: DecayRpcScheduler MXBean should only report decayed
CallVolumeSummary
Key: HDFS-10539
URL: https://issues.apache.org/jira/browse/HDFS-10539
Project: Hadoop HDFS
Issue Type: Bug
Components: ipc
Reporter: Namit Maheshwari
Assignee: Xiaoyu Yao
HADOOP-13197 added non-decayed call metrics in metrics2 source for
DecayedRpcScheduler. However, CallVolumeSummary in MXBean was affected
unexpectedly to include both decayed and non-decayed call volume. The root
cause is Jackson ObjectMapper simply serialize all the content of the
callCounts map which contains both non-decayed and decayed counter after
HADOOP-13197. This ticket is opened to fix the CallVolumeSummary in MXBean to
include only decayed call volume for backward compatibility and add unit test
for DecayRpcScheduler MXBean to catch this in future.
CallVolumeSummary JMX example before HADOOP-13197
{code}
"CallVolumeSummary" : "{\"hbase\":1,\"mapred\":1}"
{code}
CallVolumeSummary JMX example after HADOOP-13197
{code}
"CallVolumeSummary" : "{\"hrt_qa\":[1,2]}"
{code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
