[pve-devel] RFC : cloudinit : PATCH V8

[Wolfgang Bumiller]

On Tue, Aug 04, 2015 at 12:24:06PM +0200, Alexandre DERUMIER wrote:
> >>Regarding the error when the image already exists: In Qemu.pm only disks 
> >>that were previously marked as unused get deleted 
> >>($test_deallocate_drive), also this function ignores cdrom drives. 
> >>Maybe a media=cloudinit would be useful after all? 
> >>On one side it would give us an easy condition to allow deleting the 
> >>file when removing the image, on the other side it would also allow us 
> >>to easily pick out cloudinit images from the config without even having 
> >>to care about the image name. 
> >>What do you think? 
> yes, good idea.
> (I don't have too much time today to work on this)

Dietmar said matching the filename should be fine, too. I originally
didn't like the idea since we don't match filenames in qemu yet, but the
storage backend is strict enough about naming conventions anyway. And
currently I think only deletion and config-regeneration need to actually
do this. I'll just try it :-)

> ----- Mail original -----
> De: "Wolfgang Bumiller" <w.bumiller at proxmox.com>
> À: "aderumier" <aderumier at odiso.com>
> Cc: "pve-devel" <pve-devel at pve.proxmox.com>
> Envoyé: Mardi 4 Août 2015 10:41:58
> Objet: Re: [pve-devel] RFC : cloudinit : PATCH V8
> 
> Regarding the error when the image already exists: In Qemu.pm only disks 
> that were previously marked as unused get deleted 
> ($test_deallocate_drive), also this function ignores cdrom drives. 
> Maybe a media=cloudinit would be useful after all? 
> On one side it would give us an easy condition to allow deleting the 
> file when removing the image, on the other side it would also allow us 
> to easily pick out cloudinit images from the config without even having 
> to care about the image name. 
> What do you think? 
> 
> On Tue, Aug 04, 2015 at 10:03:06AM +0200, Alexandre DERUMIER wrote: 
> > >>I'd like to pick this up again. Have you made any more changes? 
> > 
> > No change since the last time. (I would like to work on it again too) 
> > 
> > >>Currently generating the cloudinit image is a one-shot mechanism (and 
> > >>also currently fails if the image already exists.), it ends up as a 
> > >>"normal" (persistent) image in a cd drive. 
> > >>I'm wondering if this is ideal, and how this would be represented in the 
> > >>UI. (Probably a simple «generate cloudinit» button?) 
> > 
> > I thinked to something like a dropdown list. 
> > 
> > cloudinit : -disable 
> > -ide3 (enable) 
> > -ide4 (enable) 
> > -sataX (enable) 
> > 
> > 
> > selecting a disk controller as value, generate and attach the config drive. 
> > (to match "qm set vmid -(ide|sata)x storeid:cloudinit" ) 
> > 
> > 
> > >>Do we still like to have a way to auto-regenerate it when changing 
> > >>network configuration? 
> > I think the main problem is that currently, we regenerate the config at each start. 
> > That's why I thinken to add the digest as property, in vmid.conf. 
> > 
> > >>Because for this the current situation is 
> > >>somewhat inconvenient as we'd have to match the existing drives for a 
> > >>disk named vm-${vmid}-cloudinit. 
> > 
> > It's not a problem is we need to simply parse the vm config file. 
> > But it can be slow if we need to scan the storage itself. (depend of the storage api) 
> > 
> > 
> > 
> > ----- Mail original ----- 
> > De: "Wolfgang Bumiller" <w.bumiller at proxmox.com> 
> > À: "aderumier" <aderumier at odiso.com> 
> > Cc: "pve-devel" <pve-devel at pve.proxmox.com> 
> > Envoyé: Mardi 4 Août 2015 08:45:06 
> > Objet: Re: [pve-devel] RFC : cloudinit : PATCH V8 
> > 
> > I'd like to pick this up again. Have you made any more changes? 
> > Currently generating the cloudinit image is a one-shot mechanism (and 
> > also currently fails if the image already exists.), it ends up as a 
> > "normal" (persistent) image in a cd drive. 
> > I'm wondering if this is ideal, and how this would be represented in the 
> > UI. (Probably a simple «generate cloudinit» button?) 
> > Do we still like to have a way to auto-regenerate it when changing 
> > network configuration? Because for this the current situation is 
> > somewhat inconvenient as we'd have to match the existing drives for a 
> > disk named vm-${vmid}-cloudinit. 
> > 
> > On Tue, Jul 07, 2015 at 09:14:32AM +0200, Alexandre Derumier wrote: 
>