Gluster users - Oct 2012 - Making the mount server/point highly available

Hello,

I want to make sure that we can keep mounting gluster volumes even if one node
is down.
For this we would need something to keep the mount point highly available.
Am I safe to assume that putting port 24007 behind our load-balancer will do
this?
As far as I can see regular gluster traffic does not use this port and it looks
like it is used for the mount process.

Is this correct?
Can I safely do this or will bad things happen?

Thanks,
Robert van Leeuwen
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://supercolony.gluster.org/pipermail/gluster-users/attachments/20121016/d365f641/attachment.html>

On 10/16/12 01:26, Robert van Leeuwen wrote:
> Hello,
>
> I want to make sure that we can keep mounting gluster volumes even if
> one node is down.
> For this we would need something to keep the mount point highly available.
> Am I safe to assume that putting port 24007 behind our load-balancer
> will do this?
> As far as I can see regular gluster traffic does not use this port and
> it looks like it is used for the mount process.
>
> Is this correct?
> Can I safely do this or will bad things happen?
I use keepalived to keep a well known IP available for the volume info 
to be retrieved from.  Once the client has the volume info, it talks 
directly to the bricks.

Unless you are NFS mounting the volume...

-- 
Mr. Flibble
King of the Potato People

On 10/16/2012 05:57 AM, Robert Hajime Lanning wrote:
> On 10/16/12 01:26, Robert van Leeuwen wrote:
>> Hello,
>>
>> I want to make sure that we can keep mounting gluster volumes even if
>> one node is down.
>> For this we would need something to keep the mount point highly 
>> available.
>> Am I safe to assume that putting port 24007 behind our load-balancer
>> will do this?
>> As far as I can see regular gluster traffic does not use this port and
>> it looks like it is used for the mount process.
>>
>> Is this correct?
>> Can I safely do this or will bad things happen?
>
> I use keepalived to keep a well known IP available for the volume info 
> to be retrieved from.  Once the client has the volume info, it talks 
> directly to the bricks.
>
> Unless you are NFS mounting the volume...
>
Use round-robin DNS or the backupvolfile-server mount option.

>> Am I safe to assume that putting port 24007 behind our load-balancer
>> will do this?
>Use round-robin DNS or the backupvolfile-server mount option.
I understand there are different solutions but I would prefer a solution based
on a loadbalancer.
The loadbalancer option simply has the least drawbacks for our us if it works
reliably.

Thx,
Robert van Leeuwen

On 10/16/2012 06:32 AM, Robert van Leeuwen wrote:
>>> Am I safe to assume that putting port 24007 behind our
load-balancer
>>> will do this?
>>>
>> Use round-robin DNS or the backupvolfile-server mount option.
> I understand there are different solutions but I would prefer a solution
based on a loadbalancer.
> The loadbalancer option simply has the least drawbacks for our us if it
works reliably.
No, a loadbalancer will not work. The correct way is to use rrDNS or the 
backupvolfile-server mount option.