On Fri, Sep 11 2009 at 2:51am -0400, Mike Snitzer <snitzer@xxxxxxxxxx> wrote: > Mikulas, > > I ran with your suggestion and it works quite well (results below). ... > The changes are detailed at the end of this mail. I still have a "TODO" > that you'll see in the patch that speaks to needing to have > snapshot_merge_process() check all origin chunks via __chunk_is_tracked() > rather than just the one check of the first chunk. Your feedback here > would be much appreciated. I'm not completely clear on the intent of > the __chunk_is_tracked() check there and what it now needs to do. I believe I've sorted this out. The TODO should've read something like: snapshot_merge_process() should delay the merging of _all_ chunks that have in-progress writes; not just the first chunk in the region that is to be merged. The revised patch (relative to Jon's unified snapshot tree) is here: http://people.redhat.com/msnitzer/patches/snapshot-merge/kernel_unified/2.6.31/dm-snapshot-merge-use-larger-io-when-merging.patch Pretty sure it _should_ work; but can't test until we get this unified snapshot patchset working properly. -- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel
