[Gluster-users] The mysterious NUFA - help needed

[Ravishankar N]

On 10/15/2013 06:38 PM, Hafeez Bana wrote:
> Thanks Ravi, that solves it.
>
> Would it have helped if I had specified AAAHOST:nufa-r4  i.e. removed 
> the slash?
>
Actually it wouldn't. Nufa really applies to bricks.What enabling nufa 
does is if you happen to mount the gluster volume on one of the machines 
that contained a brick of that volume, then all writes from that mount 
point will go that brick, irrespective of which brick the file name 
hashes to.

Considering a distributed volume of 3 bricks on 3 machines,
Brick1: A_HOST:/brick
Brick2: B_HOST:/brick
Brick3: C_HOST:/brick

Without nufa, the files that you create from the mount point will hash 
to (i.e reside in)  one of the 3 bricks depending on the filename.
With nufa enabled, if you happen to mount the voume on B_HOST, then all 
files created from that mount point will always go to B_HOST:/brick 
irrespective of which brick the filename really hashes to.
> Asking because the documentation for nufa uses the words sub-volumes - 
> and as far as I understand a subvolume is a translator with a brick 
> underneath.
Well a subvolume is a brick after being processed by a translator 
(http://www.gluster.org/community/documentation/index.php/GlusterFS_Concepts).


>
> Thx again.
> hb
>
>
> On Tue, Oct 15, 2013 at 1:34 PM, Ravishankar N <ravishankar at redhat.com 
> <mailto:ravishankar at redhat.com>> wrote:
>
>     On 10/15/2013 02:37 PM, Hafeez Bana wrote:
>>     Hi All,
>>
>>     I've been playing with Gluster 3.4x - in particular the nufa
>>     configuration. All the setups are via the gluster CLI.
>>
>>     A simple nufa config across 2 machines with the subvolumes being
>>     bricks works without issue.
>>
>>     However anything more complex gives very strange behaviors. For
>>     instance the below config.
>>
>>     This volume is meant to have a replica set on each machine and
>>     then tied together into a nufa config.
>>
>>     I then mount nufa4 on BBBHOST and AAAHOST and simulate some
>>     writes. This works and the speeds seem to indicate the writes are
>>     happening locally.
>>
>>     However the strange thing is if you look into any of the brick
>>     dirs i.e. /DIR_ON_HOST/test004a/ - no files are present!
>>
>>     The volume continues to operate but I am baffled about
>>     1) Where the files are being stored
>     Hi hb,
>     The brick path present in the volume info for 'nufa4' tells you
>     where the files go. So
>
>     Volume Name: nufa4
>     .
>     .
>
>     .
>     Brick1: AAAHOST:/nufa-r4
>     Brick2: BBBHOST:/nufa-r4b
>
>     means the files are created in the root file system, in the
>     /nufa-r4 and /nufa-r4b directories of your hosts. I think you have
>     misunderstood 'AAAHOST:/nufa-r4' and 'BBBHOST:/nufa-r4b' as
>     referring to the 2 replica volumes that you have created. They are
>     actually the brick paths.
>
>
>
>>     2) Whether such a configuration is possible i.e  a nufa sitting
>>     on top of 2 replica systems (and in fact how nufa figures out
>>     which replica sit is on the local host).
>>
>
>     Nufa makes sense only on distributed volumes. Enabling nufa on a
>     plain replicate volume has no effect- the files will be created on
>     all bricks that constitute the replica.
>     Thanks,
>     Ravi
>>     Any insight would be helpful.
>>
>>     Thx.
>>     hb
>>
>>
>>     ----------------
>>     Volume Name: nufa-r4
>>     Type: Replicate
>>     Volume ID: d47ceecf-11be-49a3-8b27-9d1df61f588c
>>     Status: Started
>>     Number of Bricks: 1 x 2 = 2
>>     Transport-type: tcp
>>     Bricks:
>>     Brick1: AAAHOST:/DIR_ON_HOST_A/test004a
>>     Brick2: AAAHOST:/DIR_ON_HOST_A/test004b
>>
>>     Volume Name: nufa-r4b
>>     Type: Replicate
>>     Volume ID: 98bc1402-fbd9-4637-81c2-2f24029d51bc
>>     Status: Started
>>     Number of Bricks: 1 x 2 = 2
>>     Transport-type: tcp
>>     Bricks:
>>     Brick1: BBBHOST:/DIR_ON_HOST_B/test004a
>>     Brick2: BBBHOST:/DIR_ON_HOST_B/test004b
>>
>>     Volume Name: nufa4
>>     Type: Distribute
>>     Volume ID: 061a3359-ec92-4ab3-9305-bae783df0ec2
>>     Status: Started
>>     Number of Bricks: 2
>>     Transport-type: tcp
>>     Bricks:
>>     Brick1: AAAHOST:/nufa-r4
>>     Brick2: BBBHOST:/nufa-r4b
>>     Options Reconfigured:
>>     cluster.nufa: on
>>
>>
>>     _______________________________________________
>>     Gluster-users mailing list
>>     Gluster-users at gluster.org  <mailto:Gluster-users at gluster.org>
>>     http://supercolony.gluster.org/mailman/listinfo/gluster-users
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20131015/cecd8fcc/attachment.html>