[Gluster-users] Gluster NFS high-availibility

[Vijay Bellur]

On 01/09/2014 10:17 PM, Jia-Hao Chen wrote:
> Hi,
>
> I am doing a preliminary test of using Gluster NFS to replace our old
> NFS server. I created a replicate volume with 2 bricks and setup up CTDB
> to manage IP takeover.
>
> My test is very simple. Copy a large file to the nfs mount, and unplug
> power cord while the file is copying. After couples of seconds, the
> survival node take over IP address successfully and nfs client
> established a new connection to it. Everything seems to be fine. But the
> copy didn't continue due to an inode lock still granted to someone.
> After around 15~20 minutes, the lock was gone and copy continues.
>
> But after copy finished, the checksum of the file is wrong. I repeatedly
> test this case several times but the result are the same. Have anyone
> known how to setup a high-availability Gluster NFS server correctly? or
> can give me some hint about it?
>

This does not look right and I was unable to recreate the problem with 
simple tests in my setup. Can you please file a bug with more 
description at:

https://bugzilla.redhat.com/enter_bug.cgi?product=GlusterFS

Having a bug will help in tracking the issue better.

Thanks,
Vijay