On Tue, Oct 06, 2015 at 15:35:05 +0530, Deepak C Shetty wrote:
On 10/06/2015 02:53 PM, Peter Krempa wrote:
On Tue, Oct 06, 2015 at 10:38:23 +0530, Deepak C Shetty wrote:
On 10/05/2015 07:33 PM, Peter Krempa wrote:
On Mon, Oct 05, 2015 at 18:59:54 +0530, Deepak C Shetty wrote:
On 10/05/2015 04:31 PM, Peter Krempa wrote:
On Mon, Oct 05, 2015 at 15:30:42 +0530, Prasanna Kumar Kalever wrote:
[...]
We just figured ....
1) qemu-img create -f qcow2 -b json:"{...}" gluster://hostname/volname/snapfile.qcow2 - works
Okay, you can use it as the backing store string ...
2) qemu-img create -f qcow2 -b json:"{.../base.img}" json:"{.../snapfile.qcow2}" - fails saying json: protocol unsupported
... but not as the image itself. That is workable, since you could emulate the behavior by re-trying the command with different single-server URIs.
This is probably bcos qemu-img utility doesn't have support for json: syntax and IIRC in the qemu thread it was said that we don't need to add this to the qemu-img utility as it adds complexity
So using #1 we are able to pre-create snapfiles that have backing-file: json string
[...]
gluster protocol* Erm, how are you attempting to test this if you don't even have libvirt with gluster support installed? Hmm, what makes u think we don't have libvirt with gluster support installed ? libvirtd is running from source with this patch included. So your configure options or build environment is wrong then. The hint is the error message you've posted.
error: internal error: missing storage backend for network files using gluster protocol
We configured it correctly now --with-gluster and --with-storage-gluster
That is a generic message from the storage driver that is used in case when the appropriate backend is not available.
As I've pointed out, it's the part of the storage driver that assists the qemu driver in regular operations like reading the image header or setting permissions on image files, not the actual pool/volume storage driver.
So just to understand this correctly, the storage driver is called as part of parsing the backing file and yes we do see that it fails when backing file is of json: syntax but works when its URI syntax.
Well, this is the thing I've pointed out in the original review.
This also means that the snapshot operation (of having the VM switch to the new active file) is done on qemu side and when it returns back to libvirt, libvirt is unable to parse as the backing file is json format, so fixing the parse URI to support parseJSON needs to be done.
Exactly, that's why I've requested it.
What happens when we don't use --reuse-external. IIUC the newfile is
If you don't use --reuse-external, then the file is pre-created by libvirt and the image metadata (including the backing store string) are filled by the qemu process once we start the "blockdev-snapshot-sync" operation in qemu.
created by libvirt so which part of libvirt needs to be fixed so ensure that libvirt creates the file using json: syntax and not URI syntax ? The above
No, this happens in qemu. Libvirt needs to be able to process that file on next start of the VM.
qemu-img snip shows that json is only supported as part of -b option only, so will that be a inhibitor for libvirt when creating new files ?
Not at this point, but we might want to use it in the future in that way.
[2]
Additionally this might be the reason you are not seeing any failures right now. The code that uses the storage backend queries whether the storage backend is actually supported before attempting to use it, so please make sure that your daemon is properly built with gluster support.
pkalever is now working with libvirtd having gluster support
(Shouldn't that be a pre-requisite for posting a patch like this?)
We will know more as we make more experiments... the whole gamut of network disk support in qemu/libvirt is confusing esp with the addn of json and URI syntax both :) So pls bear with our questions above :)
Peter