On July 20, 2018 2:45 pm, Paul Belanger wrote:
On Fri, Jul 20, 2018 at 06:21:48AM +0200, Matthieu Huin wrote:
Here's another reason to apply the Tech Preview model that I painfully
became aware of while working on my upstream patches: a lot of my work
in SF itself depends on my upstream patches being included in Zuul.
The tenant-scoped web API in Zuul implies work on cauth and sf-config
to support the JWT auth. If I wait for upstream to merge my patch, I
cannot start implementing these changes in cauth and sf-config since
our version of zuul won't have the needed feature.

Is there any way to circumvent that, or should we accept that a lot of
our dev will be slowed down by upstream?

Yup, can see how that is a blocker. I guess my question would be, if SF decided
to work to land features in upstream zuul master first, then wait to make
changes to sf-config / cauth only after a tagged release, what would that
looking?

That would be ideal, but without a release schedule and some sort of
blueprints/expected features, it's not realistic.

For example, the /tenant/job REST api[0] (to enable simple things like exposing
the location of a job) has been submitted 8 months ago already...

Similarly, I proposed an implementation[1] of the container build resource
right after we agreed on the design at Vancouver, more than 2 months
ago.

We don't know what will be in the next version of Zuul or
when it will be released.
SF developments are blocked without tech-previews.

Actually, as long as we are not releasing SF before a tagged version of
upstream zuul, depends-on here should work. And shouldn't block any development
of sf-config / cauth patches right?

Note that we are already waiting for tagged version upstream to release SF and
we often work on the upstream blockers to make the releases happen.

I think the key thing here, is we won't be able to land the new RPM on sf.io
(rdoproject / ansible tenants) due to upstream not releasing a tag, but nothing
saying we cannot do the testing in a staging environment first.

Technically, I do not think the tech previews need a staging environment as they
are fairly isolated from the zuul/nodepool logic.
For example, the extra nodepool drivers are optional.

Regards,
-Tristan

[0] https://review.openstack.org/527579
[1] https://review.openstack.org/570667

Attachment: pgpwK28XLgCoZ.pgp
Description: PGP signature

_______________________________________________
Softwarefactory-dev mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/softwarefactory-dev

Reply via email to