Provide a generic sysinfo function

[giampaolo]

This is supposed to be a function returning generic system wide info designed to host information which is not cross platform. As such, the returned namedtuple will have different attributes depending on what platform we're on. The new API should be called psutil.sysinfo(). Some ideas:

Linux

psutil/psutil/_pslinux.py

Line 470 in 905030b

def sysinfo():

• procs
• procs_running
• procs_blocked
• open_files (cat /proc/sys/fs/file-nr)
• max_files (sysctl -a | grep fs.file-max)
• max_threads?
• max_pids?

FreeBSD

psutil/psutil/arch/bsd/freebsd.c

Line 997 in 905030b

PyObject *

• openfiles
• maxfiles?
• maxprocs?
• maxpid?

OpenBSD

psutil/psutil/arch/bsd/openbsd.c

Line 801 in 905030b

PyObject *

• open_files
• num_threads
• max_files?
• max_procs?

NetBSD

https://github.com/giampaolo/psutil/blob/905030b052414e66bc065fc6dff14dbd875be77a/psutil/arch/bsd/netbsd.c#L801

• max_files?
• mac_proc?

OpenBSD

psutil/psutil/_psutil_osx.c

Line 1793 in 905030b

*/

• max_proc?
• max_files?

Solaris

psutil/psutil/_psutil_sunos.c

Line 1282 in 905030b

static PyObject *

• num_threads

Windows

?

Debatable

• last_pid
  • Linux: /proc/sys/kernel/ns_last_pid
  • FreeBSD: sysctl kern.lastpid
• max_threads_perproc: (debatable because resource module can be used)
  • FreeBSD: sysctl kern.threads.max_threads_per_proc
• max_files_perproc: (debatable because resource module can be used)
  • FreeBSD: sysctl kern.maxfilesperproc

Ideas are very welcome.

[AshishNamdev]

Would like to work on it.
Let met know if someone working on this.
Will support as I am a new with python.

BR,
Ashish Namdev
Software Engineer - I
AMD

-----Original Message-----
From: "giampaolo" notifications@github.com
Sent: ‎16-‎12-‎2015 22:20
To: "giampaolo/psutil" psutil@noreply.github.com
Subject: [psutil] Provide a generic sysinfo function (#720)

This is supposed to be a function returning generic system wide info designed to host information which is not cross platform As such, the returned namedtuple will have different attributes depending on what platform we're on Some ideas:
number of open files
number of threads
number of actively running processes
number of total ctx switches
number of open sockets (FreeBSD has this)
number of open sockets (FreeBSD has this)
max files per process
max threads per process (FreeBSD has this)
—
Reply to this email directly or view it on GitHub.

[giampaolo]

I'd rather ask people to come up with ideas about what else we might provide first. Of course any implementation help / PR is welcome.

[giampaolo]

As for Windows Process Hacker is able to determine number of context switches, interrupts and system calls. Hopefully psutil should be able to do the same.

[giampaolo]

@wj32

[wj32]

What you want is SYSTEM_PERFORMANCE_INFORMATION and SYSTEM_PROCESSOR_PERFORMANCE_INFORMATION, which you can get using NtQuerySystemInformation. See phlib/include/ntexapi.h and PhSipTickCpuDialog in ProcessHacker/sysinfo.c for details.

[giampaolo]

Note: for Linux testing we can use:

~/svn/psutil$ vmstat  -s
     10124368 K total memory
      8480240 K used memory
      4781932 K active memory
      2899572 K inactive memory
      1644128 K free memory
       325008 K buffer memory
      3143460 K swap cache
      2943996 K total swap
        23944 K used swap
      2920052 K free swap
      2064780 non-nice user cpu ticks
        27900 nice user cpu ticks
      1244258 system cpu ticks
      7665170 idle cpu ticks
       186416 IO-wait cpu ticks
           18 IRQ cpu ticks
         5434 softirq cpu ticks
            0 stolen cpu ticks
      5186699 pages paged in
     18304305 pages paged out
            7 pages swapped in
         5988 pages swapped out
     87182581 interrupts
    390691624 CPU context switches
   1455448832 boot time
        85420 forks

[giampaolo]

libstatgrab has a lot of useful stuff: https://github.com/i-scream/libstatgrab/blob/master/src/libstatgrab/cpu_stats.c

[nicolargo]

Any head up concerning this development request ?

[giampaolo]

This is currently not on my radar for the immediate future.