On Tue, Feb 16, 2016 at 06:55:44PM +0100, Dennis Jacobfeuerborn wrote:

Hi, I've noticed that the capacity of the disk in one of my VMs is completely wrongly reported by the python api. The image was transfered using scp but the transfer was aborted and now the api show a strange value:

[root@virt83 ~]# du -h /var/lib/libvirt/images/test-disk1 706M /var/lib/libvirt/images/test-disk1 [root@virt83 ~]# ls -lh /var/lib/libvirt/images/test-disk1 -rw-r--r--. 1 root root 706M 3. Nov 20:03 /var/lib/libvirt/images/test-disk1 [root@virt83 ~]# qemu-img info /var/lib/libvirt/images/test-disk1 image: /var/lib/libvirt/images/test-disk1 file format: raw virtual size: 705M (739573760 bytes) disk size: 705M

Notice how that raw image is seen with a size of 705M by ls, du and qemu-img however when I get a blockinfo using libvirt-python using the blockInfo() call I get this:

{'source': '/var/lib/libvirt/images/test-disk1', 'target': 'vda'} [2423276348619671841L, 739577856L, 739577856L]

Obviously the first value (capacity) is way too large to be correct. Any ideas if this is a bug or if there is something else going on?

If it also happens with 'virsh domblkinfo', then the bug is not in the python bindings. Running libvirtd with debug logs might give you a hint where the value came from: http://libvirt.org/logging.html Also, the code paths are different for running and inactive domains - for inactive domains libvirtd figures out the data by looking at/into the file, for active domains it asks QEMU via the query-block{,stats} commands. Jan