[ovirt-users] add fourth full gluster node and remove arbiter: ovirt 4.1 with hosted engine

[knarra]

On 07/03/2017 06:53 PM, yayo (j) wrote:
> Hi,
>
> And sorry for delay
>
> 2017-06-30 14:09 GMT+02:00 knarra <knarra at redhat.com 
> <mailto:knarra at redhat.com>>:
>
>     To add a fully replicated node  you need to reduce the replica
>     count to 2 and add new brick to the volume so that it becomes
>     replica 3. Reducing replica count by removing a brick from replica
>     / arbiter cannot be done from UI currently and this has to be done
>     using gluster CLI.
>      AFAIR, there was an issue where vm's were going to paused state
>     when reducing the replica count and increasing it to 3. Not sure
>     if this still holds good with the latest release.
>
>     Any specific reason why you want to move to full replication
>     instead of using an arbiter node ?
>
>
> We have a new server with the same hard disk size of other two node, 
> so, why not? Why join the cluster as an arbiter when we can have the 
> same disk capacity to add extra replication?
>
>
>>     and remove the arbiter node (Also a way to move the arbiter role
>>     to the new node, If needed)
>     To move arbiter role to a new node you can move the node to
>     maintenance , add  new node and replace  old brick with new brick.
>     You can follow the steps below to do that.
>
>       * Move the node to be replaced into Maintenance mode
>       * Prepare the replacement node
>       * Prepare bricks on that node.
>       * Create replacement brick directories
>       * Ensure the new directories are owned by the vdsm user and the
>         kvm group.
>       * # mkdir /rhgs/bricks/engine
>       * # chmod vdsm:kvm /rhgs/bricks/engine
>       * # mkdir /rhgs/bricks/data
>       * # chmod vdsm:kvm /rhgs/bricks/data
>       * Run the following command from one of the healthy cluster members:
>       * # gluster peer probe <new_node>
>       * add the new host to the cluster.
>       * Add new host address to gluster network
>       * Click Network Interfaces sub-tab.
>       * Click Set up Host Networks.
>       * Drag and drop the glusternw network onto the IP address of the
>         new host.
>       * Click OK
>       * Replace the old brick with the brick on the new host
>       * Click the Bricks sub-tab.
>       * Verify that brick heal completes successfully.
>       * In the Hosts tab, right-click on the old host and click Remove.
>       * Clean old host metadata
>       * # hosted-engine --clean-metadata --host-id=<old_host_id>
>         --force-clean
>
>
>
> I need this (reads: I need the arbiter role) if I reduce replica count 
> then I add the new node as full replica and increasing replica count 
> again to 3? (As you expained above)
>
Above steps hold good if you want to move the arbiter role to a new node.

If you want to move to full replica, reducing the replica count will 
work fine but increasing it again back to 3 might cause vm pause issues.




> Thank you


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/users/attachments/20170703/8be20e30/attachment.html>