[Gluster-users] Advice needed for geo replication

[Kotresh Hiremath Ravishankar]

Hi Felipe,

All the observations you have made are correct. AFR is a synchronous replication
where the client replicates the data which is limited by the speed by the slowest
node (in your case HDD node). AFR is the replicating each brick and is part of single
volume. At the end, you will have single volume where each brick's data is replicated
and the single volume is highly available.

But Geo-replication is a disaster recovery solution and is replicating whole volume
which is asynchronous in nature and eventually consistent. At the end you will have
two volumes master volume (in your case SSD one) and slave volume (in your case HDD one).
If performance is of concern and it's ok for the nature of replication to be eventually
consistent, you should go for geo-replication. And yes failover is manual.

Thanks and Regards,
Kotresh H R

----- Original Message -----
> From: "Felipe Arturo Polanco" <felipeapolanco at gmail.com>
> To: gluster-users at gluster.org
> Sent: Thursday, May 4, 2017 6:49:35 PM
> Subject: [Gluster-users] Advice needed for geo replication
> 
> Hi,
> 
> I would like some advice on setting up a replicated or geo replicated setup
> in Gluster.
> 
> Right now the setup consists of 1 storage server with no replica serving
> gluster volumes to clients.
> 
> We need to have some sort of replication of it by adding a second server but
> the catch is this second server will have spinning disks while the current
> one has SSD disks.
> 
> I have read that in Gluster, the clients are the one running the replication
> of data by sending the same bytes to the gluster servers, so by using one
> gluster server with spinning disks the performance of clients will be as
> fast as a spinning disk speed even when the other server has SSD.
> For budget reasons we can't have SSDs in second storage server.
> 
> I kept reading and found the geo replication feature which makes the server
> do the replication of data instead of the clients, which is more likely my
> case but looks like there is no automatic failover mechanism of it and the
> administrator need to intervene to make the slave server a master one
> according to this document:
> https://access.redhat.com/documentation/en-US/Red_Hat_Storage/2.0/html/Administration_Guide/ch11s05.html
> 
> Given this scenario, I really need a piece of advice from the gluster users
> on how would be the best approach to have a replicated setup with SSD+HDD
> storage servers.
> 
> Thanks,
> 
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://lists.gluster.org/mailman/listinfo/gluster-users