I am not sure how many of the people on this list subscribe to LWN, but there was, for me, an eye-opening article on "Barriers and journaling filesystems": http://lwn.net/Articles/283161/
Even if you don't subscribe I'll try to explain the core of the document.... The issue is in what state is a filesystem left after a disk or Linux "crash"? Journaling filesystems can only be as good/safe/reliable as the integrity of the Journal file itself. If the data being written to a Journal file is written out-of-order, as can happen because of various disk write optimization techniques in both the Linux kernel and the disk hardware itself, then there is the possibility that the Journal "COMMIT" record get written to disk before the other Journal data records. The current solution to this is for Filesystems such as ext3 to enable "barriers", which in theory should cause a sync to occur of all pending data before the "COMMIT" record can be written. Now here's the big catch - LVM does not currently support "barriers", and so any ext3 on LVM is exposed to possible corruption. I note that RedHat defaults to putting all of their filesystems under LVM on System z right now...... The Linux Kernel defaults to not using "barriers", but SLES has a patch to enable it by default. I am curious what plans there maybe, or workaround solutions, for both SLES and RHEL? mark ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
