[Gluster-devel] combining AFR and cluster/unify

[Daniel van Ham Colchete]

On 3/14/07, Krishna Srinivas <krishna at zresearch.com> wrote:
>
> Pooya,
>
> Your client spec was wrong. For a 4 node cluster with 2 replicas of
> each file following will be the spec file: (similarly you can write
> for 20 nodes)
>
> ### CLIENT client.vol ####
> volume brick1
>   type protocol/client
>   option transport-type tcp/client
>   option remote-host 172.16.30.11
>   option remote-port 6996
>   option remote-subvolume brick
> end-volume
>
> volume brick1-afr
>   type protocol/client
>   option transport-type tcp/client
>   option remote-host 172.16.30.12
>   option remote-port 6996
>   option remote-subvolume brick-afr
> end-volume
>
> volume brick2
>   type protocol/client
>   option transport-type tcp/client
>   option remote-host 172.16.30.12
>   option remote-port 6996
>   option remote-subvolume brick
> end-volume
>
> volume brick2-afr
>   type protocol/client
>   option transport-type tcp/client
>   option remote-host 172.16.30.13
>   option remote-port 6996
>   option remote-subvolume brick-afr
> end-volume
>
> volume brick3
>   type protocol/client
>   option transport-type tcp/client
>   option remote-host 172.16.30.13
>   option remote-port 6996
>   option remote-subvolume brick
> end-volume
>
> volume brick3-afr
>   type protocol/client
>   option transport-type tcp/client
>   option remote-host 172.16.30.14
>   option remote-port 6996
>   option remote-subvolume brick-afr
> end-volume
>
> volume brick4
>   type protocol/client
>   option transport-type tcp/client
>   option remote-host 172.16.30.14
>   option remote-port 6996
>   option remote-subvolume brick
> end-volume
>
> volume brick4-afr
>   type protocol/client
>   option transport-type tcp/client
>   option remote-host 172.16.30.11
>   option remote-port 6996
>   option remote-subvolume brick-afr
> end-volume
>
> volume afr1
>   type protocol/client
>   subvolumes brick1 brick1-afr
>   option replicate *:2
> endvolume
>
> volume afr2
>   type protocol/client
>   subvolumes brick2 brick2-afr
>   option replicate *:2
> endvolume
>
> volume afr3
>   type protocol/client
>   subvolumes brick3 brick3-afr
>   option replicate *:2
> endvolume
>
> volume afr4
>   type protocol/client
>   subvolumes brick4 brick4-afr
>   option replicate *:2
> endvolume
>
> volume unify1
>   type cluster/unify
>   subvolumes afr1 afr2 afr3 afr4
> ...
> ..
> endvolume
>

I'm no gluster expert but I think this config will put each file pair in the
same server, doesn't it? Like, volume afr4 uses the brick4 and brick4-afr,
that happend to be on the same server, on it's subvolumes.

Shouldn't it be something like:

volume afr1
 type protocol/client
 subvolumes brick1 brick2-afr
 option replicate *:2
endvolume

volume afr2
 type protocol/client
 subvolumes brick2 brick1-afr
 option replicate *:2
endvolume

volume afr3
 type protocol/client
 subvolumes brick3 brick4-afr
 option replicate *:2
endvolume

volume afr4
 type protocol/client
 subvolumes brick4 brick3-afr
 option replicate *:2
endvolume

So that everyfile has a copy of itself on two diferent servers?

Best regards,
Daniel Colchete