Periodic system slowdown/stalls

[erikwestlund]

System details

Omarchy 3.23 + AMD 9950x3d/Intel B580 GPU & AMD AI 370 Framework Laptop

What's wrong?

I run two systems, both are on the latest Omarchy (3.2.3 at the time of writing).

• Desktop with AMD 9950x3d, Intel B580 GPU
• Framework Laptop with recent AI series 370 chip and integrated graphics

I setup my machines using Ansible. Code here: https://github.com/erikwestlund/omarchy

Both machines are freshly installed from latest ISO within last week.

On both machines, I will occasionally experience short periods of the following symptoms:

• Input lag where inputs don't immediately show up (although most typing, for example, eventually makes its way through)
• Notices from applications that it is slow to respond, offering to wait or terminate
• General unresponsiveness

When this happens, btop will show little "dropouts" in the CPU core monitoring things.

Usually it lasts for a minute or two and then goes away and things are back to normal.

This is not a resource contention issue. It happens with 95% free memory and almost no CPU contention.

At first, I thought this was related perhaps to sleep/wake on my laptop. Then I figured it might be thermal throttling on my laptop. Then I thought it might be a framework laptop thing with new AMD CPUs/APUs, since others had similar issues.

But then I noticed it happening on my desktop, when temps are in the 40C range with tons of free ram/CPU. With it happening on my desktop, this also seems to rule out AMD GPU issues or issues related to AI-series AMD CPUs.

I do have some customizations to Omarchy, but mostly just a ton of theme/keybinding things which I do not think should affect this issue. You can look into my repository to get an idea of how things are set up; I am otherwise happy to answer questions.

I have done things like toggling CPU governors, which I've read can sometimes help. Nothing seems to have an immediate impact. Muddling through for a couple minutes usually ends up in a "happy" state again.

[erikwestlund]

My working theory after doing a quick real-time debug when the issue occurred last is that my more aggressive snapshotting behavior causes this, particularly upon snapshot cleanup. I was snapshotting every 5 minutes, which is nearly "free," but the hourly cleanup of snapshots is what causes a lot of IO activity and ocassionally stuttering. So I have disabled snapshots on my laptop and reduced the frequency on desktop, scheduling cleanup for nighttime.

The reason it would feel related to sleep/wake, but didn't happen instantly upon resuming, is that the timer daemon would notice it missed a cleanup, and then do it. So this would explain explain the small delay before performances fell through the floor.

I would request we keep this open for a week or so while I test and then I will close it if I see the issue resolved.

[NicolasDorier]

I never touched snapshot behavior, but I am also experiencing the same.
Lot's of RAM available, mostly idle CPU and GPU, but periodic slutters as you describe. I have a thinkpad carbon.

Logs of the snapper cleanup via journalctl -u snapper-cleanup.service indicate that it tooks only a second to achieve.

[erikwestlund]

I am reasonably convinced the snapshot cleanup caused this issue and will open this if it recurs. I have had no similar lagging since making adjustments.