Re: Distributed-Replicated with mismatch number of bricks

Amar Tumballi <atumball@xxxxxxxxxx> · Fri, 14 Sep 2018 17:49:18 +0530

On Wed 12 Sep, 2018, 4:42 PM Jose V. Carrión, <jocarbur@xxxxxxxxx> wrote:
Hi,
I would like to implement  a distributed-replicated architecture but 
some of my nodes have different number of bricks. My idea is to do a 
replica 2 through 6 nodes (all bricks with the same size).

My gluster architecture is:

Name node  |  Brick 1 | Brick 2
------------------------------------------

Node1          |      R1     |             |

Node2          |      R1     |             |

Node3          |      R2     |     R3    |
Node4          |      R2     |     R3    |
Node5          |      R4      |     R5   |
Node6          |      R4      |     R5   |
My questions are: 

1.  Would be this correct ?

gluster volume create vol0 replica 2 transport tcp \
    node1:/data/gluster/vol0/brick1 node2:/data/gluster/vol0/brick1 \
    node3:/data/gluster/vol0/brick1 node4:/data/gluster/vol0/brick1 \
    node3:/data/gluster/vol0/brick2 node4:/data/gluster/vol0/brick2 \
    node5:/data/gluster/vol0/brick1 node6:/data/gluster/vol0/brick1 \
    node5:/data/gluster/vol0/brick2 node6:/data/gluster/vol0/brick2 \
Looks correct.

2 If I need to add a new node in the future, I will be able to extend my distributed replication system ? Something like:
Name node  |  Brick 1 | Brick 2
------------------------------------------

Node1          |      R1     |             |

Node2          |      R2      |            |

Node3          |      R3      |     R4   |
Node4          |      R3      |     R4   |
Node5          |      R5      |     R6   |
Node6          |      R5      |     R6   |
Node7          |      R1      |     R2   |

gluster volume create vol0 replica 2 transport tcp \
    node1:/data/gluster/vol0/brick1 node7:/data/gluster/vol0/brick1 \
    node2:/data/gluster/vol0/brick1 node7:/data/gluster/vol0/brick2 \
    node3:/data/gluster/vol0/brick1 node4:/data/gluster/vol0/brick1 \
    node3:/data/gluster/vol0/brick2 node4:/data/gluster/vol0/brick2 \
    node5:/data/gluster/vol0/brick1 node6:/data/gluster/vol0/brick1 \
    node5:/data/gluster/vol0/brick2 node6:/data/gluster/vol0/brick2 \

This should be add-brick to  existing volume, and not another create volume 
3. My nodes will also mount volumes (gluster clients) so in order to optimize the read performance operations,  what is the best replication system:  distributed-replicated or only replicated.

In this case, if you have just 3 machines, for read perf, replica is better. If you have more nodes then to utilise all the storage you need distributed replicate.

Thanks in advance.
_______________________________________________

Gluster-users mailing list

Gluster-users@xxxxxxxxxxx

https://lists.gluster.org/mailman/listinfo/gluster-users
_______________________________________________
Gluster-users mailing list
Gluster-users@xxxxxxxxxxx
https://lists.gluster.org/mailman/listinfo/gluster-users