Hi, Thanks. I think im sticking to using '/export/lvmvolumename/' as the mountpoint, and a subdir 'brick' in that for glusterfs: /export/glusterfslv1/brick This has the same advantage as you mention, and tells me what lvm2 volumes are part of the gluster volume. I think im not going to use numbers like 'brick1' 'brick2' in the directory structure: 'gluster volume info' also talks about brick numbers in its output, whicj can be confusing and lead to stuff like this : gluster volume info Volume Name: glustervol01 Type: Replicate Volume ID: 96e28207-aaa0-4a33-9973-7021790f1f93 Status: Created Number of Bricks: 1 x 2 = 2 Transport-type: tcp Bricks: Brick1: 192.168.126.128:/export/brick2/dir Brick2: 192.168.126.129:/export/brick3/dir Thanks, Regards, John Smith On Tue, May 14, 2013 at 10:05 PM, Brian Candler <B.Candler at pobox.com> wrote: > On Tue, May 14, 2013 at 08:56:02PM +0200, John Smith wrote: >> Thanks. Is there a preferred naming convention to go along with it ? > > It's up to you, but I used /exports/brick1/myvol, /exports/brick2/myvol > and it seemed logical to me. > > There's an important benefit to doing this. If the filesystem doesn't mount > for any reason at startup, then you are left with an empty /exports/brick1 > mount point. glusterd sees that /exports/brick1/myvol is not present and so > glusterd terminates. This is what you want. > > If you had used /exports/brick1 as the brick then glusterd would start up, > see the empty directory, and if it's part of a replicated volume would start > self-healing data into it - potentially filling up your root filesystem very > quickly! > > Another side benefit is that you can if you wish re-use the same filesystem > as a member of multiple volumes. For example > > /exports/brick1/myvol > might be part of a replicated volume, and > /exports/brick1/othervol > might be part of a distributed volume. > > This can be simpler and easier to manage than using something like LVM to > slice up your brick, although the volume "free space" you see may not work > quite how you expect. > > Regards, > > Brian.
