KDE high disk IO writes both on Leap and Tumbleweed

English Applications
1

Hello,

For some reason it seems like disk IO of KDE apps are higher on Leap and Tumbleweed compared to 2 other distro I checked. You can replicate the test: while monitoring accumulated IO writes using “sudo iotop -oa -d 3” run a ping or find in konsole (or ksysguard) and see how much “DISK WRITE” they are doing per minute. I see couple of MB per minute for each of these apps. Also, I left a Leap (on btrfs) and a Tumbleweed (on ext4) with only two windows open; a konsole running iotop and ksysguard. After 10 hours, IO writes were above 2.5GB (total system writes according to iostat).

I know these numbers are not high to cause an issue but still I think those apps should not do that much disk writes. Also same test with another distro was about 80MB vs 2.5GB disk writes. But IO Read numbers were comparable.

What could be the difference? Am I missing something or these processes are really writing that much to disk?

2

Plasma writes gigantic caches on a first session or a subsequent session that causes cache rewrite.

3

I don’t see similar behaviour in the other 2 distros. May be they disabled that by default,
is there a way to tweak it or disable it?
I don’t see why ksysguard 200MB disk writes per hour.

4

I only ever use Plasma any more in openSUSE and Fedora. They both seem to cache the same. I have no idea if their caching can be disabled or tweaked. Mostly I use KDE3 or TDE, as Plasma5 continues to exclude some KDE3 features I depend on, and is considerably more power and space hungry.

5

Hi
On GNOME here, that’s about (~2.5-2.8) what I see here via conky, never worried about it since the SSD’s will handle ~20G a day…

You should be able to check via smartctl on the device to see whats been written, keep a track of that rather then running something that will cause disk writes.

Is your journal persistent? Is journalctl chatty… (journactl -f) to see whats occurring. Push stuff to tmpfs in RAM?

6

I didn’t see any difference in journalctl. Not sure if it persistent. By the way The numbers I wrote are per process. For example on Kubuntu running ksysguard about an hour causes almost no disk writes, compared to 200MB per hour on Leap in the test I mentioned above. This happened on fresh installs too.

I dont know a simple way to check where ksysguard is writing those 200MB per hour. I tried strace but it is not easy to see what is going on from those traces.