sar -m CPU returns 0 and interval is invalid

[bugproxy]

== Comment: #0 - SEETEENA THOUFEEK <email address hidden> - 2017-04-07 00:44:28 ==
+++ This bug was initially created as a clone of Bug #140363 +++

=========
Issue
=========

'sar -m CPU <interval> <count>' command does not return valid CPU FREQ values on powerkvm. I have the result pasted below captured on powerkvm vs x86 fedora system.

On PowerKVM:

[root@c158f2u07os ~]# sar -m CPU 1 5
Linux 4.4.4-1.el7_1.pkvm3_1_1.1200.0.ppc64le (c158f2u07os.pok.stglabs.ibm.com) 06/17/2016 _ppc64le_ (160 CPU)

06:30:28 AM CPU MHz
06:30:36 AM all 0.00
06:30:44 AM all 0.00
06:30:53 AM all 0.00
06:31:01 AM all 0.00
06:31:09 AM all 0.00
Average: all 0.00

>>Though I have given 1 sec interval, the real interval seems to be different as shown above. And FREQ MHz is shown 0.

Same run on fedora x86 server:
[srikanth@bssrikanth-tp ~]$ sar -m CPU 1 5
Linux 4.3.4-300.fc23.x86_64 04/15/2016 _x86_64_ (4 CPU)

12:08:47 PM CPU MHz
12:08:48 PM all 1793.18
12:08:49 PM all 2214.71
12:08:50 PM all 2170.32
12:08:51 PM all 1758.41
12:08:52 PM all 2444.70
Average: all 2076.26

This is the commit id:

commit 93662b1dbe3eed926d67d418cd4996ede9686679
Author: Breno Leitao <email address hidden>
Date: Mon Oct 31 09:23:59 2016 -0400

    sar: Improve cpuinfo read for POWER architecture

    Currently CPU frequency is not being read properly on POWER archicture.
    It shows the frequency as:
      # sar -m CPU 1 5
    ...
      06:30:28 AM CPU MHz
      06:30:36 AM all 0.00

    This is caused because /proc/cpuinfo is differently between Intel and
    POWER.
    This patch simply fix the parsing function (read_cpuinfo).

    Signed-off-by: Breno Leitao <email address hidden>

[Michael Hohnbaum (hohnbaum)]

Jon,

A patch for the Server team to evaluate.

                      Michael

On 04/06/2017 10:59 PM, Launchpad Bug Tracker wrote:
> bugproxy (bugproxy) has assigned this bug to you for Ubuntu:
>
> == Comment: #0 - SEETEENA THOUFEEK <email address hidden> - 2017-04-07 00:44:28 ==
> +++ This bug was initially created as a clone of Bug #140363 +++
>
>
> =========
> Issue
> =========
>
> 'sar -m CPU <interval> <count>' command does not return valid CPU FREQ
> values on powerkvm. I have the result pasted below captured on powerkvm
> vs x86 fedora system.
>
>
> On PowerKVM:
>
> [root@c158f2u07os ~]# sar -m CPU 1 5
> Linux 4.4.4-1.el7_1.pkvm3_1_1.1200.0.ppc64le (c158f2u07os.pok.stglabs.ibm.com) 06/17/2016 _ppc64le_ (160 CPU)
>
> 06:30:28 AM CPU MHz
> 06:30:36 AM all 0.00
> 06:30:44 AM all 0.00
> 06:30:53 AM all 0.00
> 06:31:01 AM all 0.00
> 06:31:09 AM all 0.00
> Average: all 0.00
>
>>> Though I have given 1 sec interval, the real interval seems to be
> different as shown above. And FREQ MHz is shown 0.
>
> Same run on fedora x86 server:
> [srikanth@bssrikanth-tp ~]$ sar -m CPU 1 5
> Linux 4.3.4-300.fc23.x86_64 04/15/2016 _x86_64_ (4 CPU)
>
> 12:08:47 PM CPU MHz
> 12:08:48 PM all 1793.18
> 12:08:49 PM all 2214.71
> 12:08:50 PM all 2170.32
> 12:08:51 PM all 1758.41
> 12:08:52 PM all 2444.70
> Average: all 2076.26
>
> This is the commit id:
>
> commit 93662b1dbe3eed926d67d418cd4996ede9686679
> Author: Breno Leitao <email address hidden>
> Date: Mon Oct 31 09:23:59 2016 -0400
>
> sar: Improve cpuinfo read for POWER architecture
>
> Currently CPU frequency is not being read properly on POWER archicture.
> It shows the frequency as:
> # sar -m CPU 1 5
> ...
> 06:30:28 AM CPU MHz
> 06:30:36 AM all 0.00
>
> This is caused because /proc/cpuinfo is differently between Intel and
> POWER.
> This patch simply fix the parsing function (read_cpuinfo).
>
> Signed-off-by: Breno Leitao <email address hidden>
>
> ** Affects: ubuntu
> Importance: Undecided
> Assignee: Taco Screen team (taco-screen-team)
> Status: New
>
>
> ** Tags: architecture-ppc64le bugnameltc-153240 severity-medium targetmilestone-inin---

--
Michael Hohnbaum
OIL Program Manager
Power (ppc64el) Development Project Manager
Canonical, Ltd.

[Robert Luberda (robert-debian)]

What version of sysstat does this bug refers to?

The commit 93662b1dbe3eed926d67d418cd4996ede9686679 mentioned in the bug report is part of systat 11.4.2, and Ubuntu has 11.4.3-1 already...

[Manoj Iyer (manjo)]

Please retest with sysstat version 11.4.3-1 and report back here if the bug still exists.

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2017-05-10 00:50 EDT-------
Sorry I was stuck in other bugs and not able to verify this.

root@ubuntu:dpkg -l sysstat*
Desired=Unknown/Install/Remove/Purge/Hold
| Status=Not/Inst/Conf-files/Unpacked/halF-conf/Half-inst/trig-aWait/Trig-pend
|/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad)
||/ Name Version Architecture Description
+++-==============-============-============-=================================
ii sysstat 11.4.3-1 ppc64el system performance tools for Linu
root@ubuntu:# sar -m CPU 1 5
Linux 4.10.1 (ubuntu) 05/10/2017 _ppc64le_ (2 CPU)

12:46:30 AM CPU MHz
12:46:31 AM all 3425.00
12:46:32 AM all 3425.00
12:46:33 AM all 3425.00
12:46:34 AM all 3425.00
12:46:35 AM all 3425.00
Average: all 3425.00
root@ubuntu:~/bug_data/151438/zesty/fs/jfs#

Tested and resolved in 11.4.3-1 package. .

[Christian Ehrhardt  (paelzer)]

"Tested and resolved in 11.4.3-1 package. ." which is in zesty already.
Does this affect earlier releases that it needs to be considered for SRUs?

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2017-06-02 08:44 EDT-------
Hi,

> Does this affect earlier releases that it needs to be considered for SRUs?

I don't think so. Thanks for fixing it on zesty.