> > There's no need to reset the librados state, so connections to the > cluster can stick around. I'm a bit unclear on the bdrv_reopen_* > functions though - what is their intended use and semantics?
My motivation for implementing this basic reopen support is getting active block commit in qemu (driven by libvirt) to work for rbd. For example: - VM is running and using rbd device as primary disk - Create external local qcow2 snapshot with rbd as backing device - All writes end up in local qcow2 file - Do active block commit to get changes back to rbd after a while Without the patch the last part (active block commit) fails in qemu: Block format 'rbd' used by device '' does not support reopening files I think, in this case, block commit wants to switch the rbd device to RW to make sure it can write back the changes. Which is unnecessary in this case because it already is writeable. Unfortunately i don't think i can implement proper/full-blown reopen support (like gluster) for rbd - this would be way above my head ... Sebastian
