If you are looking for moment by moment recovery, you need to have multiple directories, preferably on several devices, for your Namenode edit log (which is modified for each meta data change) and also multiple directories for the FS image, which is updated every few minutes by the secondary Namenode.
Having one of your directories on NFS will slow your Namenode down some, as all writes to all devices have to complete before a meta data operation is finished. I seem to recall that the writes are done in parallel. This does however give you fast failover. The secondary Namenode is a nice repository of 5 + minute old data in the event of a catastrophic failure or catastrophic user error such as a mass file removal. On Thu, Oct 1, 2009 at 6:15 AM, Stas Oskin <stas.os...@gmail.com> wrote: > Hi. > > I'm looking to spread the meta-data writing across several disks, including > NFS, to provide greater survivability. > > What make sense more - to write NameNode meta-data to NFS, or to write the > SecondaryNameNode meta-data to NFS, or a combination of them? > > Thanks. > -- Pro Hadoop, a book to guide you from beginner to hadoop mastery, http://www.amazon.com/dp/1430219424?tag=jewlerymall www.prohadoopbook.com a community for Hadoop Professionals
