Howdy, I'm looking at using Gluster for one of our storage projects and decided to start by trying out Gluster 3.1 beta on a couple test virtual machines by creating a single volume of 32G. The volume is a distributed volume made up of 4 8GB bricks. As a test I attempted 3 separate copies to /mnt/glusterfs, each copying a different iso file of approx 3G from my client (Fedora 13 x86_64 mounting using the gluster client, not NFS "mount -t glusterfs 10.0.0.1:vol1 /mnt/glusterfs"). cp ~/1.iso /mnt/glusterfs/iso/ cp ~/2.iso /mnt/glusterfs/iso/ cp ~/3.iso /mnt/glusterfs/iso/ I expected that each copy would result in the ISO file ending up on a different brick, instead they all went to the same brick. Each test brick is only 8G, thus the final copy ultimately failed when the brick reached full capacity. Questions: 1. Shouldn't the glusterfs distribute the separate files across the bricks (i.e. file-01 -> brick1, file-02 -> brick30, file-03 -> brickXX)? 2. Shouldn't GlusterFS select a destination brick with the available capacity to store the transfer? 3. Is there a gluster command that will display the current capacity and utilization of each brick in a volume? For example, in Lustre you can do an 'lfs df'. I initially expected 'gluster volume info' to provide that information. I've read through the Beta doc, but didn't find the above info: http://www.gluster.com/community/documentation/index.php/GlusterFS_3.1beta#Step_3:_Mounting_a_volume Configuration for the test servers: Server/Brick Size FS srv-01 /export/lun0 -> 8G ext4 /export/lun1 -> 8G ext4 srv-02 /export/lun2 -> 8G ext4 /export/lun3 -> 8G ext4 gluster volume info Volume Name: vol1 Type: Distribute Status: Started Number of Bricks: 4 Transport-type: tcp Bricks: Brick1: 10.0.0.1:/export/lun1 Brick2: 10.0.0.1:/export/lun0 Brick3: 10.0.0.2:/export/lun2 Brick4: 10.0.0.2:/export/lun3 ================================= Mike Hanby mhanby at uab.edu UAB School of Engineering Information Systems Specialist II IT HPCS / Research Computing
