Plasmashell hogs system and makes it unusable

[NoWorries]

I am having trouble with not being able to use Vivid with the desktop becoming frozen as well as the clock. This has lasted for over 20 minutes, until I consider it has more time than necessary to sort itself out, so I use the power button to close the system down.

I have attached a graphic showing what htop output gives after reboot:



If, for example, I try to exit Thunderbird, it is not removed from the taskbar. From the htop output, it appears to me that it is in a loop, which becomes worse when launching applications that lead to the desktop becoming frozen and unusable.

Has anyone else encountered this problem?

[SteveRiley]

You have multiple instances of several programs running, which is certainly unusual. In System Settings -> Startup and Shutdown -> Desktop Session, which item is selected in the On Login section? If it's one of the two Restore... options, change it to Start with an empty session. OK your way out and then restart. Does the problem go away?

[NoWorries]

You have multiple instances of several programs running, which is certainly unusual. In System Settings -> Startup and Shutdown -> Desktop Session, which item is selected in the On Login section? If it's one of the two Restore... options, change it to Start with an empty session. OK your way out and then restart. Does the problem go away?

I changed my Startup and Shutdown to Start with empty session and it made no difference. I was hoping to do this reply on Vivid but the system hung and was unusable.

[Teunis]

I've seen very similar, total lock up, and eventually got it sorted after renaming (and later removal) of .local, .config, .cache and .kde.
Must have been some stale configuration.

At least I can still shut the system down with the Magic Keys.
(Alt+SysRq+ R, E, I, S, U, B, those last six with at least a second between them)

[NoWorries]

Thanks for the guidance on this problem. I first tried renaming .cache and that stopped the system lock-up but plasmashell was still doing a lot of knitting at 100% in of my CPU's! I then renamed .config and now my desktop is back to normal after going through the process of reconfiguring it in the way that I like it. Now plasmashell is nowhere to be seen taking up CPU cycles.

I am surprised that these files contain stale configuration parameters that had this affect. All I can think is that they contained data from the early stages of Vivid's development.

I was reluctant to rename .local and .kde as they both contain "stuff" that I desperately want to keep, such as Kontact calendar files.

[Teunis]

If you want to spend the time (I didn't) you could add parts of the 'old' configurations to see where it goes wrong...

Or maybe do a diff on the old and new.

[NoWorries]

It Has Come Back

I thought that the problem had been fixed. I tried launching several applications and the system behaved itself. However, during a Skype session, it returned. It doesn't lock up the desktop, but plasmashell once again is doing a lot of knitting, as shown below.



It seems as though Skype and Plasma5 have an incompatibility problem that needs to be dealt with.

[NoWorries]

If you want to spend the time (I didn't) you could add parts of the 'old' configurations to see where it goes wrong...

Or maybe do a diff on the old and new.

Sorry to say that I have deleted the 'old' configuration files with shift+delete which totally obliterates the files.

[SteveRiley]

but plasmashell once again is doing a lot of knitting, as shown below.

Something super weird is going on here. Look at how many instances of things you have:

• 5 plasmashell
• 2 kwin_x11
• 2 kactivitymanagerd
• 2 hp-systray

In your previous screenshot:

• 3 plasmashell
• 2 kwin_x11
• 2 konsole <-- could be normal if you have more than one console window open
• 3 mysqld
• 2 zeitgeist-datahub

I suspect that one instance of Plasma Shell zooming to 100% CPU utilization is just a symptom of some more significant underlying configuration problem. Furthermore, I'm not convinced that Skype is the cause of that. Something on your machine is causing multiple instances of several programs to start, and the computer doesn't know how to handle this properly.

[NoWorries]

Something super weird is going on here. Look at how many instances of things you have:

• 5 plasmashell
• 2 kwin_x11
• 2 kactivitymanagerd
• 2 hp-systray

• I am puzzled by the numbers 5, 2, 2, 2 that you have listed and would like to know what they mean. From the htop display plasmshell processing state is S with 102% processing. The next instance of plasmashell has a processing state R ie running with 100% processing. Both X and 2 of kwin_X11 are suspended.

I suspect that one instance of Plasma Shell zooming to 100% CPU utilization is just a symptom of some more significant underlying configuration problem. Furthermore, I'm not convinced that Skype is the cause of that.

My understanding from here is that plasmashell provides the workspace for kwin and X. It is therefore involved in the running of applications. In the case of Skype, it is 32B and uses X. I therefore would think that its involvement could cause the problems I am experiencing.

Something on your machine is causing multiple instances of several programs to start, and the computer doesn't know how to handle this properly.

I would say that it is the operating system that causes this as I have in the past witnessed multiple instances of processes.

[SteveRiley]

I've learned something here.

I've not spent any time with htop and obviously didn't understand how it worked until just now. After watching it operate for a few minutes, I'm seeing the same thing -- multiple suspended instances of several processes, including plasmashell. Pressing F6 and sorting by "Command" is illuminating. I currently have 44 instances of this, all suspended. They have different, sometimes consecutive, PIDs. One has been running for 18 hours 25 minutes, another for 4 hours 15 minutes. The remaining have executed for less than 1 second.

Here's my confusion, so far -- the output of ps shows only one plasmashell. Its PID matches that of the longest-running instance in htop. But in htop, even this process shows suspended most of the time. Only occasionally does it switch to running.

Now then, pressing F5 rearranges the output in a tree format. The main plasmashell (with the same PID as in ps) has spawned all these others at some point in time. Ah hah! Pressing H, to hide user threads, reveals that all these plasmashell suspended processes are threads that belong to the main process.

Confusion cleared! Unlike top which defaults to displaying tasks, htop defaults to displaying threads. Furthermore, htop remembers your display settings.

So please excuse my ramblings here. I was chasing an incorrect idea.

[NoWorries]

I am wondering if you have used Skype with a web-cam on Vivid. The one that I have uses H.264 encoding and I am wondering if that caused problems. I have swapped back to a web-cam that uses mp4 encoding and one of these days I hope to test it after I redo my configuration - sigh!

EDIT:

I have been going through the ~/.config directory and when I changed the file plasma-org.kde.plasma.desktop-appletsrc to being a backup file the newly created version on reboot gave no plasmashell with 100% CPU load. I am not sure that I will have the time to go through each version to find the culprit entry but I hope to have some strategy worked out after a good nights sleep.

[SteveRiley]

I don't have a Skype account.

[NoWorries]

FOUND the CULPRIT

After creating a new desktop and configuring it to my liking, I found that the CPU usage was back to plasmashell hogging resources.

So I was wrong in blaming Skype for creating the problem as I only became aware of the problem during a Skype session.

So I have just finished a witch hunt for the culprit and it is the widget Network Monitor. When it is present the htop output is:


When I remove Network Monitor the output is:


Removing widgets, freezes the desktop and can take over a minute for the notification to finally go. I do like to monitor the network and I much prefer the one in Utopic which gives a noticeably different colour for sending and receiving.

I would be very interested to know if others experience this problem with Network Monitor as it will confirm that this is the culprit.

[claydoh]

https://bugs.kde.org/show_bug.cgi?id=345696
and this (probably a dupe)
https://bugs.kde.org/show_bug.cgi?id=348764