Gluster users - Aug 2015 - geo-replication master is distributed-replicated, slave is distributed only?

I?d like to have a distributed-replicated master volume, and distributed-only
slave.

Can this be done?  Just beginning the research, but so far I?ve only done
geo-replication with distributed-only volumes.  Tips/caveats on this kind
architecture are welcome.

The rationale is straightforward?the master volume should be able to stay
available with all data when suffering a node loss, but the geo-replicated
volumes can be taken offline for repairs and resync as soon as possible.

Cheers,
Christian
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://www.gluster.org/pipermail/gluster-users/attachments/20150821/37713724/attachment.html>

On 21 August 2015 at 23:46, Christian Rice <crice at pandora.com> wrote:
> I?d like to have a distributed-replicated master volume, and
> distributed-only slave.
>
> Can this be done?  Just beginning the research, but so far I?ve only done
> geo-replication with distributed-only volumes.  Tips/caveats on this kind
> architecture are welcome.
>
Yes, This can be done. Both master and slave can be of different
configurations.

But make sure that your slave volume has more effective size available than
the master volume.

HTH

//MS

>
> The rationale is straightforward?the master volume should be able to stay
> available with all data when suffering a node loss, but the geo-replicated
> volumes can be taken offline for repairs and resync as soon as possible.
>
> Cheers,
> Christian
>
> _______________________________________________
> 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/20150822/a3882a24/attachment.html>

No restrictions about Volume type in Slave side. But when a Slave node
is down, some of the files already created are not accessible so data
operations/sync may fail for those files.(These errors will be logged
in Master log files and Geo-rep will continue.).

Other problem is deleting files in Slave side may get ignored silently
in Geo-replication because Geo-rep treats "ENOENT" errors as safe
errors during delete/unlink. (Safe error because, Geo-rep tried to
delete a file and that file gets "ENOENT"(No such file/directory)
error). Due to this Slave Volume may have files which are deleted in
master.

regards
Aravinda

On 08/21/2015 11:46 PM, Christian Rice wrote:
> I?d like to have a distributed-replicated master volume, and 
> distributed-only slave.
>
> Can this be done?  Just beginning the research, but so far I?ve only 
> done geo-replication with distributed-only volumes.  Tips/caveats on 
> this kind architecture are welcome.
>
> The rationale is straightforward?the master volume should be able to 
> stay available with all data when suffering a node loss, but the 
> geo-replicated volumes can be taken offline for repairs and resync as 
> soon as possible.
>
> Cheers,
> Christian
>
>
> _______________________________________________
> 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/20150824/d338d2ed/attachment.html>