[Gluster-users] Gluster Failover Mount

[James]

On Fri, Jan 17, 2014 at 7:37 AM,  <Mike.Peters at opengi.co.uk> wrote:
>> -----Original Message-----
>> From: James [mailto:purpleidea at gmail.com]
>> Sent: 17 January 2014 11:58
>> To: Mike Peters
>> Cc: gluster-users at gluster.org
>> Subject: Re: [Gluster-users] Gluster Failover Mount
>>
>> On Fri, Jan 17, 2014 at 6:46 AM,  <Mike.Peters at opengi.co.uk> wrote:
>> > Hi,
>> >
>> > I am currently testing GlusterFS and am looking for some advice. My setup
>> uses the latest 3.4.2 packages from www.gluster.org  SLES11-SP3.
>> >
>> > I currently have a storage pool shared read-write across 2 gluster server
>> nodes. This seems to work fine. However, I would also like to mount this pool
>> on 4 further client machines running a legacy web application. Because of
>> some limitations in the application, I would like to be able to tell these
>> servers to mount the storage pool from one particular gluster server node,
>> but to fail over to the second node if and only the first node becomes
>> unavailable. I can mount the storage on the client nodes with both gluster
>> nodes specified or with only one node specified but cannot see a way in the
>> documentation of preferring one particular node and having the second
>> node configured as a fail over. Is this possible? What am I missing?
>>
>> You do realize that the initial connection is just for retrieving the volfiles, and
>> then all hosts are used, right? If so, carrying on:
>>
>> You can use VRRP and a VIP to specify which host to mount from. An
>> example of this is done in my Puppet-Gluster setup:
>> https://ttboj.wordpress.com/2014/01/08/automatically-deploying-glusterfs-
>> with-puppet-gluster-vagrant/
>>
>> You can specify ordering of the VIP with priority arguments to keepalived for
>> example.
>>
>> You can also specify more than one server on the mount command for
>> glusterfs. I forget the syntax for that, but it's easy to google.
>>
>> I hope this answers your questions!
>>
>> James
>
> Hi James,
>
> That definitely sounds feasible. I hadn't thought of doing it at that layer. We use ldirectord for load balancing the webservices so I'll give that a shot this afternoon.
No worries! It's always helpful to get fresh eyes on a problem sometimes.

>
> Thanks for your help.
My pleasure. Let me know how it goes!

>
> Mike

James