[Gluster-users] Xen - Backend or Frontend or Both?

[Raghavendra G]

Hi Nathan,

Comments inlined.

----- Original Message -----
From: "Nathan Stratton" <nathan at robotics.net>
To: gluster-users at gluster.org
Sent: Thursday, July 30, 2009 1:48:03 AM GMT +04:00 Abu Dhabi / Muscat
Subject: Re: [Gluster-users] Xen - Backend or Frontend or Both?


Sorry, with config:

On Wed, 29 Jul 2009, Nathan Stratton wrote:

> I have 6 boxes with a client config (see below) across 6 boxes. I am using 
> distribute across 3 replicate pairs. Since I am running xen I need to 
> disable-direct-io and that slows things down quite a bit. My thought was to 
> move the replicate / distribute to the backend server config so that self 
> heal can happen on faster backend rather then frontend client with 
> disable-direct-io.

For your configuration, I think it may not be of much help since only one child of replicate will be on the local disk. Among three replicate, one will have a local disk and remote disk as children and other two will have children which are remote. Also, If you are moving distribute and replicate to server/side, you should use High availability translator on client, otherwise, the cluster will become inaccessible if the server, client is speaking to goes offline. Considering the complexity, I think it may not be worth the effort.

>
> Does this make sense? I would think that it would speed up self-heal 
> operations quite a bit.
>
> Also, is it possible to use both? Could you replicate / distribute on the 
> server side AND on the client side using the same raw volumes?

I am not sure, how you can use distribute and replicate both on client and server side. Can you give an example configuration for what you are thinking?


volume brick0
  type protocol/client
  option transport-type ib-verbs/client
  option remote-host 172.16.0.200
  option remote-subvolume brick
end-volume

volume mirror0
  type protocol/client
  option transport-type ib-verbs/client
  option remote-host 172.16.0.201
  option remote-subvolume brick
end-volume

volume brick1
  type protocol/client
  option transport-type ib-verbs/client
  option remote-host 172.16.0.202
  option remote-subvolume brick
end-volume

volume mirror1
  type protocol/client
  option transport-type ib-verbs/client
  option remote-host 172.16.0.203
  option remote-subvolume brick
end-volume

volume brick2
  type protocol/client
  option transport-type ib-verbs/client
  option remote-host 172.16.0.204
  option remote-subvolume brick
end-volume

volume mirror2
  type protocol/client
  option transport-type ib-verbs/client
  option remote-host 172.16.0.205
  option remote-subvolume brick
end-volume

volume block0
  type cluster/replicate
  subvolumes brick0 mirror0
end-volume

volume block1
  type cluster/replicate
  subvolumes brick1 mirror1
end-volume

volume block2
  type cluster/replicate
  subvolumes brick2 mirror2
end-volume

volume unify
  type cluster/distribute
  subvolumes block0 block1 block2
end-volume

regards,
Raghavendra

><>
Nathan Stratton                                CTO, BlinkMind, Inc.
nathan at robotics.net                         nathan at blinkmind.com
http://www.robotics.net                        http://www.blinkmind.com
_______________________________________________
Gluster-users mailing list
Gluster-users at gluster.org
http://gluster.org/cgi-bin/mailman/listinfo/gluster-users