Replicating distributed volumes

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hello,


  Until yesterday all of my experiments with Gluster have
been with a vi'd vol file and the Glusterfsd process.
Michael and I have now done testing with glusterd under
Gluster version 3.2.6 and this has raised some issues.
The one I'd like to query in this thread is replicating 
distributed volumes.

  Attached is an archive containing three vol files from 
our gluster 3.2.6 (glusterfsd only) work.  They create a 
deployment as follows;

    Server 2 (vol file: instance2.vol)
       Network;
         IP: 192.168.1.2
       Storage;
         Physical;
           5 x SATA Disks
           -> Combined via distribute as "brick"
         Served;
           "brick" as per above

    Server 1 (vol file: instance1.vol)
       Network;
         IP: 192.168.1.1
       Storage;
         Physical;
           4 x SATA Disks
           -> Combined via distribute as "localbrick"
         Logical;
           1 x remote gluster (192.168.1.2:/brick)
           -> stacked as "remotebrick"
         Served;
           "brick" as built from a replicate of "localbrick"
           and "remotebrick"
            
    Client n (vol file: client.vol)
       Network;
         IP: 192.168.1.*
       Storage;
         Physical;
           Irrelevant 
         Served;
           nil
         Mounted;
           gluster volume 192.168.1.1:/brick via the
           first server (Server 1).



  In this scenario any client wanting access to this
single replicated capacity would simply mount the
"brick" offered by Server 1.  That client would not
be an active participant in the replicating or 
distributing processes.

  I can't find an equivalent configuration to the above,
using the same software (gluster version 3.2.6) but
via the "glusterd" and gluster command process.

  Is there a way to construct a replicated distributed
volume via the "gluster" command interface?  If not,
this intentional or is it simply that my use case is
a fringe/rare case?

  Is there any reason why I should not;

    1)  re-write the server side vol files to match
         the supplied vol files

  -and/or-

    2) provide a patch for the missing functionality
        in the gluster command / glusterd manage-
        ment tools? (i.e. our attempts to use volumes
        as sub-structures were all defeated by the 
        gluster management tools)


  More fundamentally, is the horizontally and 
vertically interchangeable brick structure of the
previous generations of gluster being phased out 
in exchange for a more rigid and/or fixed config 
style architecture?  


Thanks,



--
Ian Latter
Late night coder ..
http://midnightcode.org/

Attachment: vol_files.tgz
Description: application/compressed-tar


[Index of Archives]     [Gluster Users]     [Ceph Users]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Security]     [Bugtraq]     [Linux]     [Linux OMAP]     [Linux MIPS]     [eCos]     [Asterisk Internet PBX]     [Linux API]

  Powered by Linux