<div class="gmail_quote">On Fri, Sep 25, 2009 at 7:40 AM, Marius Gedminas <<a href="mailto:marius@pov.lt" target="_blank">marius@pov.lt</a>> wrote: <blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"> <div>On Fri, Sep 18, 2009 at 04:46:21AM -0700, J. Lennard wrote: > First I'm sorry to write what may appear as a rant, but I hope it is > not considered so. I'm writing to express huge instability problems in > Ubuntu 9.04. I migrated from ubuntu 8.04 (386) to ubuntu 9.04 (amd64) > using a simple clean install. </div>... <div>> I don't really know where to start. During past month, my machine > constantly went to trashing mode where the hard-disk light is > constantly on and I can't access anything or even swtich to linux > console for several *minutes*. This has occured more than four times > although all I usually run is a pdf viewer, an mp3 player, emacs, and > firefox with simple html pages (not even gmail, flash, etc). </div>Do you have any swap space? I've got an Asus EeePC 900 with 1 GB of RAM. Twice now I've experienced the same thing: constant disk I/O, huge latencies for any desktop task (switching windows, launching terminals). I suspect a bug in the Linux VM subsystem, since *I was not running out of memory*. Usually about 40-50% of my RAM is in disk cache---I keep track of memory usage via a GNOME panel applet. On those occasions cache size was shrinking, completely free memory was increasing, all application pages were being constantly swapped out and back in causing constant disk I/O (which is painful on an SSD). The fix was to create a temporary swap file in /tmp: sudo -s dd if=/dev/zero of=/tmp/swapfile bs=1M count=1024 mkswap /tmp/swapfile swapon /tmp/swapfile Instantly the system became responsible again, after using ~400 *kilobytes* of swap. I can only assume that having 0 swap space confused the VM somehow and kicked it into a full-on panic "let's free all the ram we can" mode. This swap file disappears after a reboot, and so far the problem hasn't recurred, so I didn't bother setting up a permanent swap partition (I'm afraid to destroy my SSD too quickly with constant writes---this already happened once thanks to ext3's journal). </blockquote><div> I have had this occur several times to me on a quad-core with 3GB ram (and 6GB swap), but only ever when resuming. If I suspend with all free memory used by cache (according to the Gnome panel applet), then occasionally on resume it spends ~5minutes unresponsive with massive disk IO before coming up. When it does come back everything is fine, except that all of that space which was cache is now completely empty. I set my vm.swappiness value to 0 (because I so rarely need it), so while I'm definitely not running out of swap space, I can see how swappiness=0 might cause the same sort of effect. I haven't found anything odd in the logs so I never filed a bug, but if it's happening to other people in other situations this bears investigating. I will open a bug on this as soon as I have time. </div></div>