On Wed, 30 Jun 2004, Dax Kelson wrote: > On Wed, 30 Jun 2004, David J N Begley wrote: > > > Here's where my mind really gets screwed-up - if I throw data at the LVs, the > > read/write pattern tends to indicate that the striping _is_ working, which > > then contradicts the mapping that 'lvdisplay -v' shows! > > > > What's happening? Are LEs always mapped to PEs in a linear fashion whilst > > striping occurs at some lower (per-chunk-of-an-LE) layer? > > My understanding that this behaviour you are seeing is coming from the > *filesystem* (ext3, or whatever you are using). > > I bet if you used dd, such as: > > dd if=/dev/zero of=/dev/VG0/LV0 > > And watch the write pattern it would be linear. Tried that - both in sub-stripe-chunk size and in larger-than-LE/PE-size, but the pattern appeared to switch back and forth between PVs as though the striping really was happening (even started with a freshly rebooted machine in order to reset all of LVM's counters). I take it this means if you run "lvdisplay -v" on a striped LV then you would expect the LE/PE mapping table to show a different pattern (such as alternating back and forth between PVs rather than the pattern I'm currently seeing)? If it helps, all sizes (apart from the LV size, of course) are defaults, which from memory is 4MB for LE/PE size and 16KB for stripe size. Thanks.. _______________________________________________ linux-lvm mailing list linux-lvm@redhat.com https://www.redhat.com/mailman/listinfo/linux-lvm read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
