[Gluster-users] VMDK file replication issue

[Keith Freedman]

At 09:11 PM 8/25/2008, Raghavendra G wrote:
>Hi Marco,
>
>comments are inlined.
>
>On Tue, Aug 12, 2008 at 12:57 PM, Marco Trevisan 
><<mailto:marco.trevisan at cardinis.com>marco.trevisan at cardinis.com> wrote:
>The backend filesystem is XFS. The command used to create it was 
>"mkfs.xfs" without any option.
>I later (Aug. 5th) changed the mount options in /etc/fstab from 
>"defaults" to "noatime,attr2".
>
>My problems started to happen on Aug. 6-7, so one of my suspictions 
>is that having set "noatime,attr2" on already existing data (used by 
>glusterfs) may have led to errors on the AFR side.
>
>
>I think  you are right here.  Since AFR depends on extended 
>attributes for file/directory synchronization, there is a 
>possibility of data corruption.

as I understand it switching from attr1 to attr2 isn't as easy as 
just specifying it on the mount point.
you should make your filesystem with attr=2 otherwise the behavior 
may not be what you expect.

you can possibly re-add the extended attributes and things will 
continue merrily along, but there's some debate apparently over 
whether or not adding the attributes to a live file system works as expected.

I want a vacation :(

Keith
I