Gluster users - Sep 2015 - Keeping it Simple replication for HA

Gluster users,

I am looking to implement GlusterFS on my network for large, expandable,
and redundant storage.

I have 5 servers with 1 brick each.  All I want is a simple replication
that requires at least 3 of the 5 bricks have a copy of the data so I can
lose any 2 bricks without data loss.  I have tried replica 3 with 5 bricks
but it seems to complain that my # of bricks must be a multiple of of my
replica count.

There a simple replication method like replica=3 with 5 bricks with
glusterfs?  Is there a better or different technology I can use for this?

Thanks,
Aaron

On 09/14/2015 10:47 AM, aaron at ajserver.com wrote:
> simple replication
> that requires at least 3 of the 5 bricks have a copy of the data so I can
> lose any 2 bricks without data loss.
This is not possible with GlusterFS, you have to specify up front which 
brick has which replica.
> I have tried replica 3 with 5 bricks
> but it seems to complain that my # of bricks must be a multiple of of my
> replica count.
The old docs say "should" but I guess the new software says
"must".

 > Is there a better or different technology I can use for this?

I think CephFS will allow for "at least 3 of 5" but I don't know
the
details.

Regards,
-- 
Alex Chekholko chekh at stanford.edu

Have you considered the disperse volume? We'd normally advocate 6 servers
for a +2 redundancy factor though.

Paul C

On Tue, Sep 15, 2015 at 5:47 AM, <aaron at ajserver.com> wrote:
> Gluster users,
>
> I am looking to implement GlusterFS on my network for large, expandable,
> and redundant storage.
>
> I have 5 servers with 1 brick each.  All I want is a simple replication
> that requires at least 3 of the 5 bricks have a copy of the data so I can
> lose any 2 bricks without data loss.  I have tried replica 3 with 5 bricks
> but it seems to complain that my # of bricks must be a multiple of of my
> replica count.
>
> There a simple replication method like replica=3 with 5 bricks with
> glusterfs?  Is there a better or different technology I can use for this?
>
> Thanks,
> Aaron
>
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://www.gluster.org/mailman/listinfo/gluster-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://www.gluster.org/pipermail/gluster-users/attachments/20150915/39cd86df/attachment.html>