HADOOP-16435
(2019) Hadoop provides a high level Server class for other Apache libraries like Hive and Tez to use. These Server instances are used as sessions for remote connections, and are created and deleted multiple times within single processes. Each Server instantiates RpcMetrics and RpcDetailedMetrics and registers them to the static DefaultMetricsSystem. However, on Server shutdown, RpcMetrics are not deregistered from the system.
Workload: Create many servers and monitor increases in heap usage.
Impact/symptom: Memory leak
Fix: in Rpc(Detailed)Metrics shutdown() method, deregister from DefaultMetricsSystem.
Jira: https://issues.apache.org/jira/browse/HADOOP-16435
Launching this artifact will open it within Chameleon’s shared Jupyter experiment environment, which is accessible to all Chameleon users with an active allocation.
Download ArchiveDownload an archive containing the files of this artifact.