[Gluster-devel] NetBSD portability (was Re: NetBSD tests not running to completion)

[Kaleb KEITHLEY]

On 01/07/2016 02:24 PM, Emmanuel Dreyfus wrote:
> Kaleb KEITHLEY <kkeithle at redhat.com> wrote:
> 
>> I was in the middle of composing a reply along pretty much the same
>> lines when I saw Jeff's reply land in my inbox.
>>
>> We are migrating from gnfs to NFS-Ganesha for NFS; NFS-Ganesha uses
>> GFAPI. If Samba isn't already using a GFAPI-based VFS, it will be soon.
>>
>> FUSE isn't the best answer, IMO, to the "universal storage" question.
> 
> I thought FUSE was the only way to acheive automatic failover on server
> failure. Am I wrong?
> 

GFAPI has exactly the same AFR fail-over mechanism as FUSE. We've
actually discussed — very theoretically — reimplementing the FUSE client
with GFAPI, but for now "if it ain't broke, don't fix it."

On Linux we use Pacemaker+Corosync to provide Highly Available NFS with
NFS-Ganesha. At present Samba still uses CTDB, but will also be able to
use Pacemaker+Corosync in a forthcoming release.

I haven't looked to see what sort of HA solutions exist in the *BSD
world. That might make for a good Google Summer of Code project.

--

Kaleb