On Mittwoch 24 Juni 2009, Eric wrote: > Dushan Tcholich <dusanc <at> gmail.com> writes: > > On Wed, Jun 24, 2009 at 5:53 AM, Eric<eric225125 <at> yahoo.com> wrote: > > > I have been having problems for the past few months where when I'm > > > doing moderately disk intensive activities (such as downloading a git > > > tree, using a virtual machine), I will get 1-2min periods where > > > basically nothing can access the disk. > > > > > > By watching /proc/meminfo, I can see my dirty memory gradually growing > > > during these disk intensive activities until it gets very > > > large (several 100MB), then my system will be unresponsive until it all > > > goes through writeback. This same behavior can be reproduced by > > > manually running "sync" when the dirty memory is fairly large. > > > > > > Has anyone else experienced these problems? I noticed that some major > > > changes were made to how Reiser4 uses dirty memory > > > (http://lkml.org/lkml/2009/2/13/106). I believe I started seeing > > > this problem around that time. Could that have caused it? > > > > Similar simptoms were noticed a year or so ago, but when using atime. > > Can you reproduce it if you mount your partition with noatime? > > Can you make some way to reliably reproduce it? > > > > > (Sorry for my vague and probably inaccurate description of the problem; > > > I know very little about virtual memory and file systems). > > > > Welcome to the club :) > > > > Thanks > > > > Dushan > > -- > > To unsubscribe from this list: send the line "unsubscribe reiserfs-devel" > > in the body of a message to majordomo <at> vger.kernel.org > > More majordomo info at http://vger.kernel.org/majordomo-info.html > > My partition is already mounted with noatime. I also forgot to mention > that I have LZO compression enabled, could that be a problem? > > As far as reproducing this problem, I am currently opening a Windows guest > in Virtualbox, loading Rosetta Stone software, and paging through about 5 > pages of a lesson. The dirty memory would have now approached 50MB. By > running "sync" to force writeback, the system will become unresponsive > for about 30-60 seconds. > > Is there a way to manually create dirty memory that needs to be written > back? That would be a better way to reproduce. > > Eric > > -- > To unsubscribe from this list: send the line "unsubscribe reiserfs-devel" > in the body of a message to majordomo@xxxxxxxxxxxxxxx > More majordomo info at http://vger.kernel.org/majordomo-info.html hm, I can't reproduce that. I was seeing similar stuff in former times - without voluntary preemption and high timer rates ... but today the worst thing that is happening is a short jerk in the mouse... -- To unsubscribe from this list: send the line "unsubscribe reiserfs-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
