On Sat, Jan 17, 2026 at 04:06:52PM +0000, Arthur Diniz wrote:
> > 3. Rather than a cron job on a fixed schedule, consider how we could
> >     incorporate content updates into the publication pipeline to reduce to
> >     refresh latency.  The pipelines are defined at
> >     https://salsa.debian.org/cloud-admin-team/debian-cloud-images-daily and
> >     https://salsa.debian.org/cloud-admin-team/debian-cloud-images-release
> >     Additional executables should be added to the main debian-cloud-images
> >     project, which is shared by the -daily and -release projects.
> 
> Would be nice having the pipeline publishing to cloud-image-finder but I
> also like to have a way to back populate the database using cloud.d.o in
> case something fail in the pipeline.
> 
> So the optimal solution would be both, having the pipeline pushing and
> cronjob making sure nothing is missing.

Agreed.

> Since image-finder is now using OpenSearch, we could have a stage in both
> pipelines (daily, release) that could do a simple curl command posting the
> image metadata payload to OpenSearch.
> 
> I will try to work on that over the next few weeks.

Thanks.

> > 4. In the Search filters, you mis-spelled UUID as UID. That had me
> >     totally confused for a minute.
> 
> This is not misspell is just the same key stored in the metadata JSON
> published to cloud.d.o.
> 
> You will be able to see uid in 
> https://cloud.debian.org/cdimage/cloud/trixie/20260112-2355/debian-13-ec2-amd64-20260112-2355.json
> 

"UID" will typically be interpreted as User ID, so we shouldn't use it
in other contexts without more explanation.  Note also that the "uid"
field in the json files is not an image ID.  It's more of a job ID.
You'll note that the json file you linked has multiple 'uid' fields: one
for the image build and one for the upload.  It's probably not usually
the thing people are going to be keying on when they search using the
image finder.  We don't publish that identifier anywhere, so the only
way anybody would ever find it is by looking at the json file, and if
you have the json file you already know where the images are.

So my recommendation would be to not show that search field, at least in
the normal UI.  Maybe hide it behind an "advanced" switch or something,
or get rid of it altogether.  And maybe clarify what the search is
actually looking for.  Most people looking for cloud images won't be at
all familiar with those identifiers.

noah

Reply via email to