$ Linux Command

Processes

pidstat

Report statistics for Linux processes and threads.

monitoringperformancecpumemoryioprocessthreads

Additional Notes

pidstat is a monitoring tool from the sysstat package that reports statistics for individual processes and threads. It can show CPU usage, memory usage, I/O statistics, context switches, stack reports, and more, per-process and over time.

System administrators and performance analysts use pidstat to identify resource-intensive processes, diagnose performance issues, monitor thread behavior, and track I/O bottlenecks. Unlike top which shows a dynamic screen, pidstat produces timestamped output suitable for logging, graphing, and retroactive analysis.

Syntax

pidstat [options] [interval] [count]

Parameters

  • interval: Sampling interval in seconds. If omitted, reports statistics since boot.
  • count: Number of reports to generate. If omitted, reports indefinitely (when interval is given).

Common Options

  • -p pid: Report only for the specified PID. Use -p ALL for all PIDs (default).
  • -u: Report CPU utilization (default).
  • -r: Report page faults and memory utilization.
  • -d: Report I/O statistics (reads, writes, I/O time).
  • -w: Report context switches (voluntary and involuntary).
  • -t: Show statistics for threads in addition to processes.
  • -T TASK|CHILD|ALL: Specify which processes to report: TASK (current), CHILD (child tasks), ALL.
  • -C command: Filter by command name (supports pattern matching).
  • -G process-name: Filter by process name (exact match).
  • -R: Report realtime priority and scheduling policy.
  • -S: Report scheduling policy (SCHED_OTHER, SCHED_FIFO, SCHED_RR).
  • -v: Report kernel tables (FDSIZE, threads, etc.).
  • -h: Show horizontally aligned output without headers.
  • -e program: Execute a program and monitor it until completion.
  • -l: Show the full command name with arguments.
  • -s: Report stack utilization.
  • --human: Show values in human-readable format.
  • --interval=ms: Set interval in milliseconds.

Examples

pidstat 2 5

Show CPU usage for all processes every 2 seconds, 5 times.

pidstat -u -p 1234 2

Show CPU usage for PID 1234 every 2 seconds indefinitely.

pidstat -r 2 10

Show memory and page fault statistics every 2 seconds, 10 times.

pidstat -d 1

Show I/O statistics for all processes every second.

pidstat -w -p ALL 5

Show context switch statistics every 5 seconds.

pidstat -C nginx 2 10

Show CPU statistics for processes whose command name contains nginx.

pidstat -t -p 5678 2

Show per-thread statistics for PID 5678.

pidstat -e ./myprogram

Run myprogram and monitor its resource usage until it exits.

pidstat -ruv 5

Show CPU, memory, and kernel table statistics together.

pidstat -p ALL -l 3

Show statistics with full command-line arguments.

Practical Notes

  • pidstat is part of the sysstat package. Install with sudo apt install sysstat or sudo yum install sysstat.
  • The first report shows statistics since boot. Subsequent reports show delta between intervals.
  • Use -t to see thread-level statistics. Without it, threads are grouped into their parent process.
  • For disk I/O monitoring, pidstat -d shows read/write throughput and I/O wait time per process.
  • Combine with sar (also from sysstat) for system-wide historical performance analysis.
  • Output is CSV-friendly, making it easy to import into spreadsheets or time-series databases.