[openstack-dev] [glance] proposed priorities for Mitaka

Doug Hellmann doug at doughellmann.com
Tue Sep 15 12:46:53 UTC 2015


Excerpts from Flavio Percoco's message of 2015-09-15 10:54:04 +0200:
> On 14/09/15 15:51 -0400, Doug Hellmann wrote:
> >Excerpts from Flavio Percoco's message of 2015-09-14 14:41:00 +0200:
> >> On 14/09/15 08:10 -0400, Doug Hellmann wrote:
> 
> >> This is definitely unfortunate. I believe a good step forward for this
> >> discussion would be to create a list of issues related to uploading
> >> images and see how those issues can be addressed. The result from that
> >> work might be that it's not recommended to make that endpoint public
> >> but again, without going through the issues, it'll be hard to
> >> understand how we can improve this situation. I expect most of this
> >> issues to have a security impact.
> >
> >A report like that would be good to have. Can someone on the Glance team
> >volunteer to put it together?
> 
> Here's an attempt from someone that uses clouds but doesn't run any:
> 
> - Image authenticity (we recently landed code that allows for having
>   signed images)
> - Quota management: Glance's quota management is very basic and it
>   allows for setting quota in a per-user level[1]
> - Bandwidth requirements to upload images
> - (add more here)

That seems like a good start. You could add a desire to optionally
take advantage of advanced object store services like Swift and
Ceph.

> >> The mistake here could be that the library should've been refactored
> >> *before* adopting it in Glance.
> >
> >The fact that there is disagreement over the intent of the library makes
> >me think the plan for creating it wasn't sufficiently circulated or
> >detailed.
> 
> There wasn't much disagreement when it was created. Some folks think
> the use-cases for the library don't exist anymore and some folks that
> participated in this effort are not part of OpenStack anymore.

OK. There is definite desire outside of the Glance team to have
*some* library for talking to the image store directly. The evidence
for that is the specs in nova related to a "seam" library, and the
requests by some Cinder driver authors to have something similar.
>From what I can tell, everyone else thought that's what glance-store
was going to be, but it's not quite what is needed.  It seems like
the use cases need to be revisited so the requirements can be
documented properly and then we can figure out what steps to take
with the existing code.

Doug



More information about the OpenStack-dev mailing list