procstat allows to monitor per process statistics including:
- CPU usage
- Scheduling
- Memory
- TCP/UDP rx/tx queues
You would normally call this script from a monitoring system that allows to collect metrics over time (e.g. Zabbix, collectd etc.)
See the usage section for more details.
The PID of the process to monitor can be passed directly on the command line (--pid ) or by reading a file containing the Process ID (--pid-file )
The following metrics are currently supported:
| Metric | Description |
|---|---|
| cpu usage | Process CPU usage since it has started, calculated as 100 * ((total_time / tick rate) / elapsed time) |
| sched sum_exec_runtime | Accumulated amount of time spent on the CPU |
| sched nr_involuntary_switches | Total number of involuntary context switches |
| sched nr_voluntary_switches | Total number of voluntary context switches |
| sched nr_switches | Total number of context switches |
| sched avg_exec_runtime | Average time a task has spent executing on the CPU between two context-switches, calculated as sum_exec_runtime/nr_switches. The lower this value, the more context-switching-happy a task is |
| sched utime | Amount of time that this process has been scheduled in user mode, measured in clock ticks |
| sched stime | Amount of time that this process has been scheduled in kernel mode, measured in clock ticks |
| sched nr_migrations | Total number of times the scheduler migrated the process from one CPU core to another |
| sched sleep_max | Maximum time the task spent sleeping voluntarily |
| sched block_max | Maximum involutary delay the task experienced (e.g. waiting for disk IO) |
| sched wait_max | Maximum delay that task saw from the point it got on the run queue to the point it actually started executing its first instruction |
| sched wait_sum | Accumulated amount of time the process had to wait from run queue to execution |
| sched wait_count | Total number of times the process had to wait from run queue to execution |
| sched wait_avg | Average waiting time (calculated as wait_sum/wait_count) |
| sched iowait_sum | Accumulated amount of time spent waiting for IO |
| sched iowait_count | Total number of times the process had to wait for IO |
| sched iowait_avg | Average IO waiting time (calculated as iowait_sum/iowait_count) |
| sched nr_wakeups | Total number of times the process was woken up by the scheduler |
| mem minflt | The number of minor faults the process has made which have not required loading a memory page from disk |
| mem majflt | The number of major faults the process has made which have required loading a memory page from disk |
| mem vmpeak | Peak virtual memory size |
| mem vmsize | Virtual memory size in bytes |
| mem vmrss | Resident Set Size: number of pages the process has in real memory |
| net tcp recv_q | Sum of the outstanding data in the TCP receive queue (for all TCP sockets) in bytes |
| net tcp send_q | Sum of the outstanding data in the TCP send queue (for all TCP sockets) in bytes |
| net udp recv_q | Sum of the outstanding data in the UDP receive queue (for all UDP sockets) in bytes |
| net udp send_q | Sum of the outstanding data in the UDP send queue (for all UDP sockets) in bytes |
Example:
./procstat.sh --pid 31337 net tcp recv_q
Matteo Cerutti - matteo.cerutti@hotmail.co.uk