[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: konqueror heavy swapping

On Monday 23 January 2006 10:52 am, Pol wrote:
> Sometimes, when several tens tabs are open on the desktop, konqueror
> (3.4) almost freeze the machine due to heavy swapping lasting about 10
> minutes, as i can infer from the led monitoring the hard disk activity
> By the 'top' command, one can read an average load about 20-30, while
> cpu is not intesively used.
> It is unknown to me the reason of that swapping, since there is a lot
> of swap space avalailabe on disk.
> Any help?
> How to  monitor konqueror activity more closely to figure out what is
> happening there and possibly to tell which windows are 'guilty'?

I'm just adding things I've observed in hopes that someone will recognize the 
problem(s) and point to solutions.  (I suspect the recommendation will be 
more RAM--that means a new computer as my MB is full.)

I've had similar experiences (but I run Mandriva), but not, afaicr, for 10 
minutes--more like 30 to 120 seconds.  I'm not 100% sure it's konqueror, but 
that's a reasonable supposition, as my most common mix of apps is several 
konqueror windows (with 5 to 20 tabs), a kmail window, several nedit windows 
(with a few tabs), and konsole.  (800 MHz. Pentium (III?), 512 MB RAM, 2 MB 
swap (on a dedicated hard drive)

I run top continuously in one root "konsole", and "vmstat 2" in another.

I think I have two (or more) different occurrences.

Recently, I noticed a few occasions where the load average gets very high (on 
one occasion it hit 800, and that was my longest incidence of swapping (I 
call it "sewing machine behavior").  Some other times I've seen it around 30 
(normally, all three of my load averages are below 1).  During those times, 
the wait time of the processor hits 75-95%.

Prior to updating to Mandriva 2006, I had 1 GB of swap, and noticed that when 
half the swap was in use (which corresponds to swap in use = RAM), the 
machine slowed down (swapping increased).  When I updated to Mandriva 2006, i 
doubled the size of the swap (and added a dedicated hard drive to minimize 
wear and tear on my big drive) in an effort to determine whether the slowdown 
occurred at swap in use = RAM or at swap in use = 1/2 available swap.  I 
haven't quite decided yet--but I'm tending toward swap in use = 1/2 available 

Although I strongly hope that kde gradually develops a smaller memory 
footprint, I suspect there is something about the kernel's use of swap that 
causes the behavior I see.

Randy Kramer

PS: I hate the churn that occurs when I simply close an application--doesn't 
the kernel maintain an in-memory list of RAM and swap in use that it can 
simply (with no disk access) mark as unused?

And, is this somehow related to the unused RAM is wasted philosophy?  

Reply to: