My Kubuntu 25.10 system tied itself in knots, to the point I couldn't see mouse movement, or reliably modify Libre-Office calc files. The shell command uptime reported load averages like 158.83 This went on for an hour or more, until I finally gave up and rebooted. I also note that the system had been up for 15 days, which might just be longer than kubuntu's tested uptime, given the frequency of minor updates.
While the KDE desktop was essentially unusable during this period, though not actually hung, I was able to ssh into the system and run top.
Really high cpu times were reported from kswapd0, snapd, systemd-jo+, plasmashell, colord-sane, and various kworker threads. (systemd-jo+ was presumably a truncated command)
"memory" usage was not notably elevated.
The "fun" started approximately when I launched a script that backs up user space to an external drive, using rsync.
---
What I want to know is how to debug this. I'd like to submit good quality bug reports to whichever upstream projects are responsible for the mess.
I'd also like to find workarounds, and possibly replace misbehaving packages with other sources of similar functionality. (Given what was on the list, that may be difficult.)
I'm an old Unix developer, and not afraid of debugging tools, profiling tools, or log files. I've even been known to look at misbehaving kernels.
What I don't know is what all these processes are or which packages they belong to. I also don't know much about current linux tools for debugging performance issues. I don't even know much about where it keeps its logs or how to read them. Is syslogd still a thing? On kubuntu?
My inclination from past experience is to suspect the desktop environment, broadly construed, but I'm a new user of kde and wayland (less than 3 months) with a corresponding lack of knowledge.
Is there a performance-debugging FAQ somewhere? Particularly one specific to Kubuntu, since I suspect involvement of both kde and snapd.
Is there a known issue?
---
Note that I don't plan to try too hard to sort this out while still on Kubuntu 25.10. The 26.04 LTS is available, and I'll probably move to it before the system's been up for another 15 days. (Why do you think I was running a backup?) But I rather doubt whatever bug(s) I hit will turn out to have been fixed in 26.04 - moved around a bit, maybe, but probably not fixed.
Also, FWIW, The system was up to date as of about 15 days ago, except for firefox, which I "pinned" after a forced update from snap was found in progress during my last Kubuntu not-quite-hang. The backup script on the other hand has been run many times without any similar incidents.
[Edited to add: I found dpkg -S. So I'm on the way to identifying the processes.
colord-sane is from the colord package, which is used by KDE. I can't really imagine why it accumulated so much cpu time.
Kswapd0 is the kernel page out demon, which should only be this busy if something is grabbing loads of memory. Maybe the backup thrashed the file system cache, particularly if it's a combined cache - executable pages and ordinary files together. Still odd that this hadn't happened other times that the backup script was run.]
While the KDE desktop was essentially unusable during this period, though not actually hung, I was able to ssh into the system and run top.
Really high cpu times were reported from kswapd0, snapd, systemd-jo+, plasmashell, colord-sane, and various kworker threads. (systemd-jo+ was presumably a truncated command)
"memory" usage was not notably elevated.
The "fun" started approximately when I launched a script that backs up user space to an external drive, using rsync.
---
What I want to know is how to debug this. I'd like to submit good quality bug reports to whichever upstream projects are responsible for the mess.
I'd also like to find workarounds, and possibly replace misbehaving packages with other sources of similar functionality. (Given what was on the list, that may be difficult.)
I'm an old Unix developer, and not afraid of debugging tools, profiling tools, or log files. I've even been known to look at misbehaving kernels.
What I don't know is what all these processes are or which packages they belong to. I also don't know much about current linux tools for debugging performance issues. I don't even know much about where it keeps its logs or how to read them. Is syslogd still a thing? On kubuntu?
My inclination from past experience is to suspect the desktop environment, broadly construed, but I'm a new user of kde and wayland (less than 3 months) with a corresponding lack of knowledge.
Is there a performance-debugging FAQ somewhere? Particularly one specific to Kubuntu, since I suspect involvement of both kde and snapd.
Is there a known issue?
---
Note that I don't plan to try too hard to sort this out while still on Kubuntu 25.10. The 26.04 LTS is available, and I'll probably move to it before the system's been up for another 15 days. (Why do you think I was running a backup?) But I rather doubt whatever bug(s) I hit will turn out to have been fixed in 26.04 - moved around a bit, maybe, but probably not fixed.
Also, FWIW, The system was up to date as of about 15 days ago, except for firefox, which I "pinned" after a forced update from snap was found in progress during my last Kubuntu not-quite-hang. The backup script on the other hand has been run many times without any similar incidents.
[Edited to add: I found dpkg -S. So I'm on the way to identifying the processes.
colord-sane is from the colord package, which is used by KDE. I can't really imagine why it accumulated so much cpu time.
Kswapd0 is the kernel page out demon, which should only be this busy if something is grabbing loads of memory. Maybe the backup thrashed the file system cache, particularly if it's a combined cache - executable pages and ordinary files together. Still odd that this hadn't happened other times that the backup script was run.]
This one was perfectly successful.
Comment