Monitoring Near Real-Time Performance

Pepperdata provides the same performance monitoring capabilities as standard Linux performance commands, along with detailed data visualizations. The cluster environment—YARN or Kubernetes—determines which performance statistics are exposed.

Performance Statistics: YARN Clusters

The table shows which Linux commands correspond to which Pepperdata metrics.

Linux Command to Pepperdata Metrics Mapping
Linux Command Pepperdata Metrics Group
Task: I/O Stat group, Local FS stats group
Node: Disk Stat group
Task: Memory group
Node: Memory group

For related information, see Memory Swapping Detection and Mitigation in the Pepperdata Supervisor.
Task: CPU resource group
Node: CPU usage group, Load average group
Task: Socket stat group
Node: Network stat group, Network TCP stat group
(no equivalent Linux command) Task: Resource prioritization group, HDFS stats group, Shuffle stats group, Task JVM
Node: Dynamic allocation group

Performance Statistics: Kubernetes Clusters

For Kubernetes clusters, Pepperdata gathers the performance metrics that are exposed by Prometheus, which differ based on the operating system. For detailed collector support information, refer to the Prometheus GitHub site .