[ovirt-devel] Issues with the VM display type value(s)

[Einav Cohen]

> ----- Original Message -----
> From: "Frantisek Kobzik" <fkobzik at redhat.com>
> Sent: Thursday, February 19, 2015 10:24:03 AM
> 
> Hi,
> 
> sorry for late response, I was on something else. I'll explore that problem
> ASAP. I'll have to start working on the new graphics/display representation
> in restapi anyway so I'll try to solve this nasty behavior as well.
> 
> Cheers!
> Franta.

Thanks, Franta - much appreciated :)

> 
> 
> ----- Original Message -----
> From: "Einav Cohen" <ecohen at redhat.com>
> To: "Frantisek Kobzik" <fkobzik at redhat.com>, "Juan Hernández"
> <jhernand at redhat.com>
> Cc: devel at ovirt.org
> Sent: Thursday, February 19, 2015 2:47:21 PM
> Subject: Re: [ovirt-devel] Issues with the VM display type value(s)
> 
> > ----- Original Message -----
> > From: "Juan Hernández" <jhernand at redhat.com>
> > Sent: Thursday, February 19, 2015 5:44:42 AM
> > 
> > On 02/18/2015 11:10 PM, Einav Cohen wrote:
> > > Hi,
> > > 
> > > This is about the changes introduced by patch
> > > http://gerrit.ovirt.org/#/c/35281 (restapi: adjust api to new
> > > graphics representation).
> > > 
> > > There are a couple of issues with the VM's display type value:
> > > 
> > > (1) the "display -> type" value of the VM in the REST API seem to
> > > map to different properties based on the VM status.
> > > see http://i.imgur.com/Mn1RITW.png:
> > > When the VM is Up, the "display -> type" value in the REST API
> > > matches the 'Display' value in the GUI.
> > > But when the VM is Down, the "display -> type" value in the REST
> > > API matches the 'Graphics Protocol' value in the GUI and does not
> > > necessarily match the 'Display' value (e.g. in case we invoked
> > > 'Run Once' on the VM with a different Console value).
> > > 
> > > Is there a reason for the values in the REST API to not match the
> > > ones in the GUI? it's a bit confusing.
> > > Maybe need to have two properties in the REST API to match the two
> > > displayed values in the GUI: 'Display' (which, IIUC, is the 'run-
> > > time' value) and 'Graphics Protocol' (which, IIUC, is the
> > > 'configured'/persisted value)?
> > > I understand that there may be backward-compatibility issues with
> > > naming/terminology, but perhaps we can come up with a solution that
> > > will not break the backward compatibility?
> > > 
> > 
> > This behavior wasn't introduced by that patch, rather the patch was
> > carefully crafted to preserve it, for backwards compatibility.
> > 
> > My understanding is that the virt team plans to do larger changes to
> > this area of the RESTAPI to reflect the new separation of the "video
> > device" and "graphics device" concepts. I think that your concerns can
> > be handled in the context of those larger changes.
> 
> I opened the following BZ:
> Bug 1194287 - VM display type in REST API doesn't have 1:1 match with GUI
> [https://bugzilla.redhat.com/show_bug.cgi?id=1194287]
> 
> virt: If there are plans to refactor the entire VM-display area in the code,
> feel free to block this BZ on the relevant refactor BZ / close this BZ as
> duplicate / ...
> 
> > 
> > > (2) in patch http://gerrit.ovirt.org/#/c/35281, it seems that the
> > > REST API is calling a query *per VM* in order to re-populate the
> > > display-related values.
> > > This will potentially overload the engine, especially if we are
> > > planning to move the GUI to work over the REST API and request
> > > VMs every 5 seconds (by default) per client, as we are doing today.
> > > 
> > > Are there any plans to optimize this behavior (e.g. populate the
> > > correct display-related values on the db-side, rather than re-
> > > populating them on the REST-API side using a per-VM query)?
> > > 
> > 
> > Currently there are no plans to change this because the backend doesn't
> > have a mechanism to provide all the related display information using
> > only one query. I'd suggest to implement at least a query that takes
> > multiple VM identifiers and returns the graphics devices of those
> > multiple VMs. This would allow for a 1+1 queries scenario, instead of
> > 1+N. Once the backend provides this the RESTAPI can be changed to use
> > it. This is for the virt team to decide, I'd suggest you open a bug.
> 
> I opened the following bug:
> Bug 1194291 - REST API is calling a query _per VM_ in order to re-populate
> the display-related values -> potential engine overload
> [https://bugzilla.redhat.com/show_bug.cgi?id=1194291]
> 
> Thanks.
> 
> > 
> > > Many thanks in advance.
> > > 
> > > ----
> > > Regards,
> > > Einav
> > 
> > --
> > Dirección Comercial: C/Jose Bardasano Baos, 9, Edif. Gorbea 3, planta
> > 3ºD, 28016 Madrid, Spain
> > Inscrita en el Reg. Mercantil de Madrid – C.I.F. B82657941 - Red Hat S.L.
> >
>