Key Process Metrics

Professional

Agent gathers metrics for processes of discovered services. Currently only automatically discovered services are monitored.

If the services is running multiple processes (for example PostgreSQL, which have one process per client), the metrics will summarize all processes.

Agent gathers the following metrics:

Metric	Description
process_context_switch	Number of context switch caused by the processes
process_cpu_system	CPU used in system call by the processes in percent
process_cpu_user	CPU used in user-mode by the processes in percent
process_io_read_bytes	Disk read throughput for the processes in bytes
process_io_write_bytes	Disk write throughput for the processes in bytes
process_major_fault	Number of major fault (memory access requiring IO) caused by the processes
process_mem_bytes	Memory used by the processes in bytes
process_num_procs	Number of process in the group
process_num_threads	Number of threads in the group
process_open_filedesc	Number of files opened the processes
process_worst_fd_ratio	Worse FD ratio (open files / limit) amount the processes

Note: monitoring the processes has a small impact on agent’s CPU usage. If you don’t need those metrics and want to save some CPU cycle, you can disable gathering of processes metrics by adding the following to /etc/glouton/conf.d/99-local.conf:

agent:
  process_exporter:
    enable: false