[Gluster-devel] QEMU (and other libgfapi client?) crashes on add-brick / replace-brick

[Guido De Rosa]

Hello,

I'm using QEMU (1.5.1) compiled with GlusterFS  native support, i.e.
linked to libgfapi (v3.4.0) to bypass FUSE overhead and improve
performance.

All is explained in Bharata Rao's blog:

http://raobharata.wordpress.com/2012/10/29/qemu-glusterfs-native-integration/

And the problem I ecounter is explained in my comment:

http://raobharata.wordpress.com/2012/10/29/qemu-glusterfs-native-integration/#comment-520

As the author suggests, I'm here asking what's the current situation
about libgfapi, when a client application is running on a replicated
volume which undergoes add-brick or replace-brick operations. As per
my test with QEMU, it just crashes, anytime bricks are added or
replaced - which, as opposed, does never happen on FUSE mounts.

Thanks very much.

Guido