On Tue May 19 2009, Alexander Krause wrote: > Hi, > > > Alexander Krause: > > > > What I menat is something like this. > > > > # mkdir /mnt/aufs/overlay > > > > # mount /dev/sdb2 /mnt/aufs/overlay > > > > # while true; do find /mnt/aufs/overlay/ -ls > /dev/null; done & > > > > # seq 100 | while read i; do yes | dd of=3D/mnt/aufs/overlay/tmp/$i > > > > count=$i; > # while true; do find /mnt/aufs/overlay/ -ls > /dev/null; done > # while true; do seq 100 | while read i; do yes | dd > of=3D/mnt/aufs/overlay/tmp/$i; done > > I got both lines running for over 5 minutes - no problem. > > > > > The issue is still there with ext3 but ext2 works fine. > > Ok, that may be important. > It's weird that this only happens with kde4 and only with ext3/4. I'll try to > find the program/service which causes the problem and will run an strace or > something. >
Try something with a big executable memory footprint (which kde4 has) - - There is an on-going discussion on LKML about how/why under some conditions executable pages of memory are getting evicted from the page cache while they are still in use (which just causes them to be re-loaded). That problem has the same type of symptom - works fine for awhile (while pages are in the page cache) - and then everything slows way down. The problem reported on LKML can be seen when copying very large files. Mike > > Cheers Alex > ------------------------------------------------------------------------------ Crystal Reports - New Free Runtime and 30 Day Trial Check out the new simplified licensing option that enables unlimited royalty-free distribution of the report engine for externally facing server and web deployment. http://p.sf.net/sfu/businessobjects