Bug 404119 - 100% CPU Core Usage In KSysGuard And CPU Widget, But 2% In Top And HTop And /Proc/Stat
Summary: 100% CPU Core Usage In KSysGuard And CPU Widget, But 2% In Top And HTop And /...
Status: RESOLVED WORKSFORME
Alias: None
Product: ksysguard
Classification: Unmaintained
Component: ksysguard (other bugs)
Version First Reported In: 5.14.5
Platform: Arch Linux Linux
: NOR normal
Target Milestone: ---
Assignee: KSysGuard Developers
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2019-02-08 22:25 UTC by vindicator
Modified: 2019-03-12 15:00 UTC (History)
2 users (show)

See Also:
Latest Commit:
Version Fixed/Implemented In:
Sentry Crash Report:


Attachments
screenshot of widget, ksysguard, and /proc/stat output (208.79 KB, image/png)
2019-02-08 22:25 UTC, vindicator
Details

Note You need to log in before you can comment on or make changes to this bug.
Description vindicator 2019-02-08 22:25:01 UTC
Created attachment 117938 [details]
screenshot of widget, ksysguard, and /proc/stat output

SUMMARY
 Even with everything closed, I recently found the CPU Load Monitor widget was showing 1 CPU core was at 100%.
 KSysguard "System Load" is showing that same CPU maxed.
 top, htop, and /proc/stat do not show the high usage.

STEPS TO REPRODUCE
1. Cause unknown.
2. Let me know what I need to do to get you information you need.

OBSERVED RESULT
 CPU core at 100% in ksysguard and widget, but 2% in top, htop and /proc/stat

EXPECTED RESULT
 CPU usage data show be the same across the board.

SOFTWARE/OS VERSIONS
Linux/KDE Plasma: Arch Linux
(available in About System)
KDE Plasma Version: 5.14.5
KDE Frameworks Version: 5.54.0
Qt Version: 5.12.0

ADDITIONAL INFORMATION
 My system has been up for 10 days and plasmashell crashed once without restarting so I started it back up with "kstart5 plasmashell".

 If another process is scheduled to take on that CPU core at 100%, then the 100% moves to a different core. EG, cpu1 is at 100% now. If another program grabs ahold of that core, then the 100% may move to cpu0.

 Even after killing ksysguardd, which also restarts it, 1 core is still pegged at 100%.

 Again, I don't believe that high usage is actually happening.

cpu  54455067 917166 20201837 255970922 13509992 3617901 2264309 0 9059307 0
cpu0 14057889 245901 5012493 64477240 3113544 373252 483190 0 2354530 0
cpu1 13521184 216043 4973721 63463397 3212759 1772039 533901 0 2208023 0
cpu2 12932555 218534 5148656 63653809 3920836 1069710 779397 0 2174340 0
cpu3 13943438 236685 5066965 64376474 3262852 402899 467819 0 2322412 0
Comment 1 Patrick Silva 2019-02-23 16:39:35 UTC
Here htop, ksysguard and cpuload widget also show different values of cpu/cores usage.

Operating System: Arch Linux 
KDE Plasma Version: 5.15.1
KDE Frameworks Version: 5.55.0
Qt Version: 5.12.1
Kernel Version: 4.20.11-arch2-1-ARCH
Comment 2 Richard Llom 2019-03-11 14:40:24 UTC
What happens if you create a new diagram for the cpu load, does it show the same symptoms?

To do so go File -> New Tab, OK
Open Sensor Browser (click and drag from the right side of the ksysguard window), search for cpu load, drag & drop the sensors on a row.


Also to rule out any hiccups in the system, it would be good if you could do a full system reboot.
Comment 3 vindicator 2019-03-11 21:41:41 UTC
That reply would've been more helpful sooner rather than over 1 month.

I ended up updating my system which also required a reboot and I haven't had the issue.

That was the ONLY time I've ever noticed the issue which is why time was important in getting it debugged to figure out the cause.

I don't recall if I had done the "new tab" in ksysmon, but given that the sensor widget also reflected the same issue, I'm led to believe the issue lay with whatever data source they got their information from... and THAT is what I'd like to know.
Comment 4 Richard Llom 2019-03-12 09:26:48 UTC
(In reply to naroyce from comment #3)
> That reply would've been more helpful sooner rather than over 1 month.
> 
> I ended up updating my system which also required a reboot and I haven't had
> the issue.
> 
> That was the ONLY time I've ever noticed the issue which is why time was
> important in getting it debugged to figure out the cause.
> 
If you need urgent help, the bugtracker is *really* not the place to go. KDE provides a forum, MLs, IRC and matrix channels - and archlinux as well. Which are actually intended for live support, whereas the bugtracker is not.

> I don't recall if I had done the "new tab" in ksysmon, but given that the
> sensor widget also reflected the same issue, I'm led to believe the issue
> lay with whatever data source they got their information from... and THAT is
> what I'd like to know.
> 
You can find a link to the source here:
https://kde.org/applications/system/ksysguard/
(under Development Information)

For further assistance please use one of the support channels mentioned above.
Comment 5 vindicator 2019-03-12 15:00:51 UTC
Yeah, IRC is really the first place I go, but when it comes to bugs that I chat about, no one says a word half of the time (guesstimation).

I check IRC first to see if anyone has experienced the issue and it may very well be that the issue (not this one in particular) is not a bug, or it's known and already been posted.

It wasn't "urgent" in the sense that I couldn't use my computer or anything, but I wasn't going to be able to keep my system up in perpetuity, with no knowledge of when/(if) someone would address it... in this case over 1 month.

I see stat.c using "FILE *stat = fopen("/proc/stat", "r");", so that makes it even more confounding that the output would be different.

Should it happen again, I'll just download the source and "try" to debug it myself. It sounds like that's the best advice I'd get from here. I'll just have to figure it out (if it happens again).