I would like to ask for help to further debug the following issue. I’m stuck and don’ t know which commands / tools I could use to track its cause.
Background: over time the responsiveness of my F31 degraded severely (mouse response, visual window switch, etc). This happened a few times and I thought it was just an personal mess-up, broken component install, custom driver, etc.
This week it reappeared (after a system update) and I decided to perform a fresh reinstall, but even in a fresh / default F31 Workstation install it is present (not extra component, backup, config, etc changed – apart from a few system tools added like htop, nmon, etc). Could it be a regression bug?
It appears to be some sort of IO issue that degrade user input / system response. Even typing delay is noticeable in almost zero system load.
Notebook Acer Aspire E5-574-592S Kernel 5.5.5-200.fc31.x86_64 F31 Workstation Installed with default options rpmfusion free/nonfree added minimal extra components added than default (htop, atop, nmon, glances sysstat...) full dnf system update / reboot Memory: 8 GB CPU: Intel® Core™ i5-6200U CPU @ 2.30GHz × 4 GPU: Intel® HD Graphics 520 (Skylake GT2) GNOME Version 3.34.4 SSD 500GB 860 Evo Samsung Dmesg output: https://hastebin.com/ipigozabuv.coffeescript Installed packages: https://hastebin.com/segarepaxu.pl
Since the issue appeared an disappeared a few times (over updates, reinstalls, etc) and the hardware works fine and fast in other OSses, I don’t think that’s any hardware issue.
How can I find out which process exactly is causing the kernel thread high usage showed in htop (red bar)?. System was fully rebooted and just terminal/htop opened – nothing beyond the default is running.
The system has a SSD. Given the fresh reinstall, could this be related to background ssd trim done by the kernel? Even so, could it impact in user input delay (mouse/keyboard)
Kernel threads: https://hastebin.com/guxerijulo.sql
I found ways to list kernel threads but don’t how to monitor CPU usage / io_wait of each one (to find out which is causing the high usage)
Glances shows a low (0.1% IO wait), there is no user load… so I can’t figure out that high red bar in htop or the system slow responsiveness.