[Gluster-users] Gluster configuration NUFA+AFR+distibuted

[Kirill V. Borodin]

Hello, Daniel!

I need only 2 uniq nodes to be affected with AFR (or replicate?), 4 
servers = 2+2 AFR nodes, if one of them down, ther will be no data-loss, 
one of you'r examples:

[..]
volume replicate1
type cluster/replicate
subvolumes remote1 remote2
end-volume

volume replicate2
type cluster/replicate
subvolumes remote3 remote4
end-volume

volume replicate3
..
volume replicateN
...

volume distribute
type cluster/distribute
subvolumes replicate1 replicate2 replicate3 replicate4 replicate5 
replicate6 .. replicateN
end-volume
[..]

This is distributed+replicate mechanism. How can I combine it with NUFA 
or change replicate mech. to AFR? And what difference betwen AFR and 
replicate? Will it work fine, or there is more easy way to have 
distibuted storage in more than 100 nodes where each two of them 
replicated?

Thanks for your answers.

Daniel Maher пишет:
> Kirill V. Borodin wrote:
>
>> Is it possible to combine 3 translators in single clinet/server-process:
>> NUFA (each server will be the client) + distributed (no file-space, 
>> all in hash) + AFR (network mirror, one server down, no users 
>> affected) ?
>
> If your intention with AFR is to ensure that each node (client == 
> server) has exactly the same data, then you only need to use AFR, and 
> not any of the other clustering translators.
>
> Perhaps you'd like to expand on your desired architecture further ?
>