http://bugzilla.kernel.org/show_bug.cgi?id=14830 Theodore Tso <tytso@xxxxxxx> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |tytso@xxxxxxx --- Comment #5 from Theodore Tso <tytso@xxxxxxx> 2010-01-19 19:38:34 --- So the Linux write cache mystery is unlikely to solve the problem since that was talking about backporting some tuning parameters from 2.6.22 to the RHEL/CentOS 2.6.18 kernel, and here the problem seems to be that the sync is taking much longer on a 2.6.31 FC 12 kernel. So the first thing I notice is the fact that you have the nodelalloc mount option enabled. Any particular reason why you did that? Try removing that; one of the reasons why ext4 is generally described as being much better than ext3 with respect to this problem (of the machine becoming unresponsive during a sync) is because of delayed allocation, and you've turned it off. So try removing nodelalloc and see if that makes the performance come back. Another thing that might be worth testing is to see whether an ext3 filesystem on a 2.6.31 FC 12 kernel behaves any differently. This may be something that is some kind of VM tuning issue between RHEL 5.4 and a modern kernel; I don't know how many people try running Fedora 12 on a system with large amounts of memory and an NFS load, and maybe there is some kind of tuning issue that has been exposed. So that's a quick experiment that's worth doing just so we can figure out where we need to concentrate our diagnostic efforts. Another thing to try is to do some instrumentation using iostat to see what the system is doing, before, during, and after the sync command. -- Configure bugmail: http://bugzilla.kernel.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching the assignee of the bug. -- To unsubscribe from this list: send the line "unsubscribe linux-ext4" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html