Performance degradation after KDE runs for sometime

[Marko Vojinovic]

On Friday, November 12, 2010 22:52:12 Patrick Boutilier wrote:
> On 11/12/2010 06:43 PM, Marko Vojinovic wrote:
> > On Friday, November 12, 2010 22:14:34 you wrote:
> >> On Fri, Nov 12, 2010 at 3:55 PM, Marko Vojinovic<vvmarko at gmail.com>  
wrote:
> >>> I traced the issue down to swap usage --- after a couple of days of
> >>> running, *something* seems to gradually fill the memory with junk
> >> 
> >> Run top and type "M".  That will sort process by their resident memory
> >> usage.  It won't had to spot the offender.
> > 
> > top - 22:16:38 up 10 days,  5:11,  5 users,  load average: 0.38, 0.45,
> > 0.37 Tasks: 200 total,   1 running, 199 sleeping,   0 stopped,   0
> > zombie Cpu(s): 10.4%us,  8.5%sy,  0.0%ni, 79.9%id,  1.0%wa,  0.0%hi, 
> > 0.2%si,  0.0%st Mem:   2048088k total,  1657424k used,   390664k free,  
> >  18640k buffers Swap:  4192956k total,  1038432k used,  3154524k free,  
> > 566580k cached
> > 
> >    PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
> >   
> >   2528 vmarko    20   0 2029m 171m  19m S  1.3  8.6 564:19.55 firefox
> >   2397 vmarko    20   0  366m 148m 7460 S  1.3  7.4 230:40.39 skype
> >   2366 vmarko    20   0  565m  92m 3044 S  0.0  4.6  12:10.68 emerald
> >   5723 vmarko    20   0  989m  82m  17m S  0.0  4.1  10:29.49 kmail
> >   1739 root      20   0  471m  59m  11m S 19.9  3.0   3780:01 X
> >   2021 vmarko    20   0 1380m  57m 9476 S  6.3  2.9 780:37.22 ktorrent
> >   2016 vmarko    20   0  293m  42m 8176 S  0.7  2.1 137:01.00 skype
> >   2222 vmarko    20   0  789m  28m 5436 S  0.0  1.4  11:05.85 cairo-dock
> >   1998 vmarko    20   0  937m  20m 8824 S  4.3  1.0 804:03.95
> >   plasma-desktop 2603 vmarko    20   0  546m  18m 7088 S  0.0  0.9  
> >   1:39.64 kile
> > 
> > 22635 vmarko    20   0  480m  12m 6452 S  0.7  0.6  43:06.53 konsole
> > 
> >   2053 vmarko     9 -11  501m  12m  11m S  0.3  0.6 259:54.25 pulseaudio
> > 
> > 10764 vmarko    20   0  455m 8572 4608 S  0.0  0.4   1:07.09 okular
> > 32679 vmarko    20   0  285m 8528 5616 S  0.0  0.4   0:00.11 kio_pop3
> > 
> > So what do you make of it?
> 
> You have almost 400M free and almost 600M cached. Not sure why so much
> goes into swap. What are the results of the free command after running
> this command?
> 
> swapoff -a && swapon -a

Oh, if I try to turn off the swap now (just swapoff -a) the machine will grind 
to a halt for some time (half an hour or so), then the command will fail and 
the swap will get back in use (I mean without running the swapon -a part). 
Tried it several times before. Apparently the kernel eventually figures out 
there is not enough RAM for the swapoff command to complete, and instead the 
command gets killed before swap gets off. That's at least how I explain this 
behavior to myself. But the system is in a very bad state after that, and 
basically the only thing that makes sense at that point is to reboot.

I'll try again, just to make sure, but I'll probably end up having to reboot 
the machine. But then we'd have to wait for a week or so for the system to 
reach this state again, so I'd rather postpone this step for later, when all 
other ideas run out.

Speaking of turning off swap, I tried also to do a fresh boot, then do a
swapoff -a, and keep the system running without swap. After a while (several 
days) the machine would start to choke, processes randomly getting killed, 
etc., all symptoms of "not enough memory" problem. Eventually it would end up 
with a complete lockup, and I'd have to reboot.

So any ideas how to troubleshoot this memory leakage, before I do something 
that risks requiring a reboot?

Best, :-)
Marko

P.S. Apparently I inadvertently stole the thread from the OP. But if he is 
having the same issue, I guess it doesn't hurt... :-)