[Gluster-users] Managing volumes is unreliable

[Vijay Bellur]

On 10/02/2013 06:05 PM, Jeff Darcy wrote:
> On 10/02/2013 06:57 AM, Heggland, Christian wrote:
>> Unsure if its recommended but you could try and do the following if
>> the set command stops adding / removing settings:
>>
>> 1. Stop the volume 2. Edit the config file for that volume on each
>> node ( /var/lib/glusterd/vols ) 3. Start the volume again.
>>
>> I've had to do this in order to update the access list since the set
>> command stopped working.
>
> Definitely not recommended.  Any changes you make that way are likely to
> be overwritten the next time the standard management mechanisms are used
> (including at system startup).
>
> There is apparently a bug in the glusterd connection code.  We need to
> concentrate on fixing that, and then see where things stand.


The problem here seems to be stemming from the privileged port check in 
glusterd for CLI commands. This check can be bypassed by setting

  option rpc-auth-allow-insecure on

in

/etc/glusterfs/glusterd.vol. Once this is done, restarting glusterd 
would be necessary.

With this option, CLI commands should run seamlessly when executed 
through a script on a single node.Note that enabling this option would 
let unprivileged ports connect to glusterd.

-Vijay