Gluster users - Feb 2017 - Always writeable distributed volume

Hi,

I am wondering if it is possible to create an always writeable distributed
volume.

Reading the documentation I can figure out how. So is it possible? 
If I understand the docs correctly. The DHT determines based on a hash of the
filename, which brick to place the file. And if you have two bricks and loose
one brick I can't create files DHT determines should be places on the failed
brick.

I have tried creating a distributed volume on two bricks/nodes. And as the
feared I can't write files DHT determines should be placed on a failed node.
I am well aware and can accept that I can't access or re-create files
already created on the failed node. But I would like to write new files.

Is there a setting/feature I can enable that allows me to create file on the
available/online bricks even if DHT determines that the file should be placed on
the unavailable/failed?

My use case is: I want to use a Gluster volume for temporary storage that is
always available, as in I can always mount and write to it . We have a lot of
media file that is transcoded on user request and there need at temporary
storage for this operation. All I need is a temporary, fast and always
accessible storage with no data security/replica.

Regards
Jesper

Short answer : I think you need to add an arbiter node, this way the cluster
keeps being writable when there is at least 2 nodes presents (eg 1 data node is
down). This solve the split brain case where only 2 nodes are involved in the
setup.

Cheers 

--
C?dric Lemarchand
> Le 1 f?vr. 2017 ? 13:18, Jesper Led Lauridsen TS Infra server <JLY at
dr.dk> a ?crit :
> 
> Hi,
> 
> I am wondering if it is possible to create an always writeable distributed
volume.
> 
> Reading the documentation I can figure out how. So is it possible? 
> If I understand the docs correctly. The DHT determines based on a hash of
the filename, which brick to place the file. And if you have two bricks and
loose one brick I can't create files DHT determines should be places on the
failed brick.
> 
> I have tried creating a distributed volume on two bricks/nodes. And as the
feared I can't write files DHT determines should be placed on a failed node.
I am well aware and can accept that I can't access or re-create files
already created on the failed node. But I would like to write new files.
> 
> Is there a setting/feature I can enable that allows me to create file on
the available/online bricks even if DHT determines that the file should be
placed on the unavailable/failed?
> 
> My use case is: I want to use a Gluster volume for temporary storage that
is always available, as in I can always mount and write to it . We have a lot of
media file that is transcoded on user request and there need at temporary
storage for this operation. All I need is a temporary, fast and always
accessible storage with no data security/replica.
> 
> Regards
> Jesper   
> 
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://lists.gluster.org/mailman/listinfo/gluster-users