Re: [Gluster-devel] two disks on each node.

[Amar S. Tumballi]

you can do this instead,
====================
server spec: (write proper type option and other things)

volume brick1
  option directory /export # the '/' partition
end-volume

volume brick2
 option directory /glusterfs/export # disk reserved for glusterfs
end-volume

volume ns-local
 option directory /glusterfs/ns-local
end-volume

volume unify
 option scheduler rr # or alu
 option namespace ns-local
 subvolumes brick1 brick2
end-volume

volume ns
 option directory /glusterfs/ns
end-volume

volume server
 subvolumes ns unify
 # export 'unify' and 'ns'
 . . . . . .
end-volume
=================

client spec:

volume client-local
 remote-host localhost
 remote-subvolume unify
end-volume

volume client[1-n]
..
..
..
volume ns
 remote-host IP
 remote-subvolume ns
end-volume

volume unify
 option scheduler nufa
 option nufa.local-volume-name client-local
 option namespace ns
 subvolumes client-local client[1-n]
end-volume

volume writebehind

volume iocache
===========

Hope this works better and simple for you..

-amar

On Dec 13, 2007 10:14 PM, Albert Shih <address@hidden> wrote:

> Hi all.
>
> On my cluster (for computation) I've 2 disks. One for the / and software
> and another for glusterfs.
>
> But on the first disk I've a unused partition (because the disk size is
> 140Go, and Linux don't eat enought Mo ;-) ).
>
> Until today I've use only the second disk for glusterfs. But If I
> calculate
> the sum of the unused partition for all cluster I've 1.3To. It's lot of
> space.
>
> Well now I want use this place too. But I don't see how I can do that.
>
> Of course this two partitions have different size.
>
> Actually I use this kind of configuration :
>
> On node X
>
>        volume nodeX
>          type storage/posix
>          option directory /_glusterfs
>        end-volume
>        volume nodeY
>          type protocol/client
>          option transport-type tcp/client     # for TCP/IP transport
>          option remote-host ip_address_of_nodeY
>          option transport-timeout 30
>          option remote-subvolume brick
>        end-volume
>        volume node....
>
>
>        end-volume
>        etc...
>
>        volume unify
>          type cluster/unify
>          subvolumes node1....nodeN
>           option scheduler nufa
>           option nufa.local-volume-name nodeX
>           option nufa.limits.min-free-disk 10
>           option nufa.refresh-interval 1
>          option namespace ns
>        end-volume
>        volume work
>           type performance/write-behind
>           option aggregate-size 1MB
>           option flush-behind on
>           subvolumes unify
>        end-volume
>
> As you can see I want use nufa scheduler.
>
> How can I use nufa scheduler with two local disk ? Is that mean anything ?
> Maybe the solution is I add
>
>        volume nodeX-bis
>           type storage/posix
>           option directory my_second_partition
>        end-volume
>        volume nodexY-bis
>           type protocol/client
>        ...
>        end-volume
>
>        volume unify-bis
>          type cluster/unify
>          subvolumes node1-bis....nodeN-bis
>           option scheduler nufa
>           option nufa.local-volume-name nodeX-bis
>           option nufa.limits.min-free-disk 10
>           option nufa.refresh-interval 1
>          option namespace ns2
>        end-volume
>
>        volume bigunify
>           type cluster/unify
>           subvolumes unify unify-bis
>
>        ? what kind of scheduler ?
>
>           option namespace ns3
>        end-volume
>
>        volume work
>           type performance/write-behind
>           option aggregate-size 1MB
>           option flush-behind on
>           subvolumes bigunify
>        end-volume
>
> What's your opinion ?
>
>
> Regards.
>
> --
> Albert SHIH
> Observatoire de Paris Meudon
> SIO batiment 15
> Heure local/Local time:
> Jeu 13 déc 2007 17:28:26 CET
>
>
> _______________________________________________
> Gluster-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/gluster-devel
>



-- 
Amar Tumballi
Gluster/GlusterFS Hacker
[bulde on #gluster/irc.gnu.org]
http://www.zresearch.com - Commoditizing Supercomputing and Superstorage!