<html> <head> <meta content="text/html; charset=ISO-8859-1" http-equiv="Content-Type"> </head> <body bgcolor="#FFFFFF" text="#000000"> On 22/04/12 18:11, Bas Roufs - message in English wrote: <blockquote cite="mid:1528726.1y26pFjnjq@viaconsensus-travel" type="cite"> <meta name="qrichtext" content="1"> <style type="text/css"> p, li { white-space: pre-wrap; } </style> Hello everybody Today, I have been tweeking with a view to limiting the use of CPU by the system Kubuntu 11.10. I have followed most of the advises at the tutorial "How to make Kubuntu (KDE) blazing fast and optimise it for performance". See: <a class="moz-txt-link-freetext" href="http://bit.ly/tRJzap">http://bit.ly/tRJzap</a> This has helped to effectively deal with the worst CPU problems I have experienced recently. However, the consumption of CPU is still 60 up to and including 100% in certain situations - notably when syncing IMAP-folders in KMAIL/ Kontact. As a matter of consequence, it is very difficult to multitask with one or a few other packages, when doing so. At one of the two laptops I am working with, I even need to use some workarounds via the system activity monitor to get Kontact started at all: "killing" one or two processes, while prioritising others. Be so so kind to provide me with feedback and ideas to handle this problem. Respectfully yours, Bas Roufs. </blockquote> Hi bas, This sounds very annoying. The system being non-responsive effective prevents you from working properly. Did you try isolating the culprit in a cgroup? That way you can maximize the processor power it is allowed to use. I didnt try it yet, but I plan to if I get problems like this. Below a summary of what I read in the Dutch "Linux magazine" 2012-02. Hope it helps. Remember to change names to your own situation. mkdir -p /cgroup/cpu mount -t cgroup -o cpu none /cgroup/cpu (check with ls -l /cgroup/cpu it works, you should get something like: root@mahatma:/# mkdir -p /cgroup/cpu root@mahatma:/# mount -t cgroup -o cpu none /cgroup/cpu root@mahatma:/# ls -l /cgroup/cpu total 0 -rw-r--r-- 1 root root 0 2012-04-23 05:24 cgroup.clone_children --w--w--w- 1 root root 0 2012-04-23 05:24 cgroup.event_control -rw-r--r-- 1 root root 0 2012-04-23 05:24 cgroup.procs -rw-r--r-- 1 root root 0 2012-04-23 05:24 cpu.rt_period_us -rw-r--r-- 1 root root 0 2012-04-23 05:24 cpu.rt_runtime_us -rw-r--r-- 1 root root 0 2012-04-23 05:24 cpu.shares -rw-r--r-- 1 root root 0 2012-04-23 05:24 notify_on_release -rw-r--r-- 1 root root 0 2012-04-23 05:24 release_agent -rw-r--r-- 1 root root 0 2012-04-23 05:24 tasks root@mahatma:/# cat /cgroup/cpu/cpu.shares 1024 The 1024 is equal to 100% of your CPU. All processes in the file "tasks" share these 1024 cycles. Now suppose you have a certain set of processes (or only one) that hog the cpu. Then you create a new directory and direct these processes to that directory. Only this time you maximize the number of cycles it is allowed to use. First you create a subdirectory under CPU, this doubles the number of cycles to 2x1024. Then you limit the number of cycles by overwriting the cpu.shares file with a smaller number. Then you copy the processes that you want limited to the tasks file in the subdirectory. From that time on, these processes should be limited in CPU use to the shares you specified. I copied from my machine, but dont have any processes to put in tasks: root@mahatma:/cgroup/cpu# mkdir vreters root@mahatma:/cgroup/cpu# ls -l vreters/ total 0 -rw-r--r-- 1 root root 0 2012-04-23 09:27 cgroup.clone_children --w--w--w- 1 root root 0 2012-04-23 09:27 cgroup.event_control -rw-r--r-- 1 root root 0 2012-04-23 09:27 cgroup.procs -rw-r--r-- 1 root root 0 2012-04-23 09:27 cpu.rt_period_us -rw-r--r-- 1 root root 0 2012-04-23 09:27 cpu.rt_runtime_us -rw-r--r-- 1 root root 0 2012-04-23 09:27 cpu.shares -rw-r--r-- 1 root root 0 2012-04-23 09:27 notify_on_release -rw-r--r-- 1 root root 0 2012-04-23 09:27 tasks root@mahatma:/cgroup/cpu# cd vreters root@mahatma:/cgroup/cpu/vreters# cat cpu.shares 1024 root@mahatma:/cgroup/cpu/vreters# echo 256 > cpu.shares Now you have 1024 + 256 cycles: about 20% for all processes in this tasks file. The tasks file is currently empty. All you have to do is copy 199001 to tasks (if that is the process that is hogging the CPU): echo 19901 > tasks From then on, you should have a more responsive system. You can find out what is wrong with the process later. I really hope this helps. <div class="moz-signature">-- Kind regards, Guus Bonnema.</div> </body> </html>