franco at fugro-fsi.com.au
Mon Jan 29 06:48:36 UTC 2007
I've just been looking at replacing my own FUSE based filesystem with
glusterfs and have a few questions. My filesystem appears very similar
except that it uses NFS as backing filesystems and has a separate
partition for metadata links. Although the reliance on NFS makes it
quite slow, it does provide a certain level of resilience.
I didn't see any mention in the roadmap for dealing with downed storage
servers. How do you intend to handle this given that you don't appear to
maintain any metadata? Will processes D state when a server goes down
and then recover when it reboots?
Is it always going to be necessary to restart the client filesystems
each time the configuration changes?
Do you intend to allow online resizing of filesytems?
What is the intended policy for duplicate files? I ran a simple test and
could see that the file just appeared twice in the directory listing.
This may seen an odd question but it does happen, especially if a server
goes offline for a period of time and the users want to recreate the
files with the same name. When the server comes back online you suddenly
end up of lots of duplicates.
More information about the Gluster-devel