[Gluster-users] 2 Node glusterfs quorum help

[Sachidananda URS]

Excuse my top posting:

Kaamesh,

There are two types of quorum implementations:

1)
cluster.server-quorum-type: server/none
cluster.server-quorum-ratio: (0-100%) default >50%

Whose behavior is to kill the glusterfsd daemons (server) if server-quorum-ratio is not met.

2) And we have quorum type:

cluster.quorum-type: fixed
cluster.quorum-count: 1

Whose behavior is to make the cluster read-only if the quorum is not met.
In your case if you don't want the cluster to go down (Without adding another node to the cluster),  
disable the cluster.server-quorum-type: server, and use the `cluster.quorum-type: fixed and 
cluster.quorum-count: 1' options.

In your volume options I see that you are using both the above methods to achieve quorum.

ref: http://www.gluster.org/community/documentation/index.php/Gluster_3.2:_Setting_Volume_Options

You can see some explanation in the above link.

-sac


----- Original Message -----
> From: "ML mail" <mlnospam at yahoo.com>
> To: gluster-users at gluster.org
> Sent: Monday, February 9, 2015 1:53:56 PM
> Subject: Re: [Gluster-users] 2 Node glusterfs quorum help
> 
> This seems to be a workaround, isn't there another proper way with the
> configuration of the volume to achieve this? I would not like to have to
> setup a third fake server just in order to avoid that.
> 
> 
> 
> On Monday, February 9, 2015 2:27 AM, Kaamesh Kamalaaharan
> <kaamesh at novocraft.com> wrote:
> 
> 
> It works! Thanks to craig's suggestion . i setup a third server without a
> brick and added it to the trusted pool. now it doesnt go down. thanks alot
> guys!
> 
> Thank You Kindly,
> Kaamesh
> Bioinformatician
> Novocraft Technologies Sdn Bhd
> C-23A-05, 3 Two Square, Section 19, 46300 Petaling Jaya
> Selangor Darul Ehsan
> Malaysia
> Mobile: +60176562635
> Ph: +60379600541
> Fax: +60379600540
> 
> On Mon, Feb 9, 2015 at 2:19 AM, < prmarino1 at gmail.com > wrote:
> 
> 
> 
> Quorum only appli‎es when you have 3 or more bricks replicating each other.
> In other words it doesn't mean any thing in a 2 node 2 brick cluster so it
> shouldn't be set.
> 
> In other words based on your settings it's acting correctly because it thinks
> that the online brick needs to have a minimum of one other brick it agrees
> with online.
> 
> Sent from my BlackBerry 10 smartphone.
> From: Kaamesh Kamalaaharan
> Sent: Sunday, February 8, 2015 05:50
> To: gluster-users at gluster.org
> Subject: [Gluster-users] 2 Node glusterfs quorum help
> 
> Hi guys. I have a 2 node replicated gluster setup with the quorum count set
> at 1 brick. By my understanding this means that the gluster will not go down
> when one brick is disconnected. This however proves false and when one brick
> is disconnected (i just pulled it off the network) the remaining brick goes
> down as well and i lose my mount points on the server.
> can anyone shed some light on whats wrong?
> 
> my gfs config options are as following
> 
> 
> Volume Name: gfsvolume
> Type: Replicate
> Volume ID: a29bd2fb-b1ef-4481-be10-c2f4faf4059b
> Status: Started
> Number of Bricks: 1 x 2 = 2
> Transport-type: tcp
> Bricks:
> Brick1: gfs1:/export/sda/brick
> Brick2: gfs2:/export/sda/brick
> Options Reconfigured:
> cluster.quorum-count: 1
> auth.allow: 172.*
> cluster.quorum-type: fixed
> performance.cache-size: 1914589184
> performance.cache-refresh-timeout: 60
> cluster.data-self-heal-algorithm: diff
> performance.write-behind-window-size: 4MB
> nfs.trusted-write: off
> nfs.addr-namelookup: off
> cluster.server-quorum-type: server
> performance.cache-max-file-size: 2MB
> network.frame-timeout: 90
> network.ping-timeout: 30
> performance.quick-read: off
> cluster.server-quorum-ratio: 50%
> 
> 
> Thank You Kindly,
> Kaamesh
> 
> 
> 
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://www.gluster.org/mailman/listinfo/gluster-users
> 
> 
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://www.gluster.org/mailman/listinfo/gluster-users
> 
> 
> 
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://www.gluster.org/mailman/listinfo/gluster-users