On Thu, May 09, 2019 at 12:19:49PM +0900, Tristan Cacqueray wrote:
On Mon, May 06, 2019 at 14:57 Paul Belanger wrote:
On Mon, May 06, 2019 at 02:25:45PM -0400, Clark Boylan wrote:
On Mon, May 6, 2019, at 2:52 AM, Clint Byrum wrote:
Anyway, I think it may make sense to think hard as a community about why things do seem to sporadically sit for months, or land in a few days. It seems like there are some nice features sitting out there that aren't getting enough review traction.
For many of us the last month or so has been a time of vacation, opendev migration prep and execution, 9 days of summit + PTG, and now recovery. We've also fixed a security bug and the memory leak that ended up in that release. Finally, the test suite has been quite unhappy in recent days though many people stepped up to work on fixing those flaky tests (thank you!)
All that to say I think we've all been quite busy with items that have higher priorities in recent days/weeks. While I agree we should think hard about this an do what we can to fix it I worry that recent experiences are not representative.
I understand zuul.opendev.org is the flagship deployment of Zuul and it arguably takes priority over Zuul's backlog. Unfortunately, this lack of roadmap/backlog is not recent. Here are a few concret examples:
There is a runC driver for Nodepool that has been sitting in review for almost two years now: https://review.opendev.org/468753
The admin web API spec was proposed a year ago along with an implementation and a keycloack demo: https://review.opendev.org/562321
Lastly, the freeze-job stack was supposed to be reviewed after the multi-ansible feature; new changes are still being merged resulting in extra work to fix the rebase conflicts.
If using storyboard to track those features is not the right solution, what can we do to improve the contributor experience?
With zuulv3 development, I personally found it valuable to have a weekly IRC meeting to help discuss the current issue and if any action was needed. A while back we decided to move that to the ML, but I myself haven't really found that to be an improvement (cannot speak for others).
In the case of above, perhaps weekly emails to the ML to help track progress would help.
Most of my last 2 weeks for zuul was helping stabilize our testing, but currently that isn't part of my day to day job. Most of the items I am currently focused on in zuul, are making the github driver better[1], getting wheels published to pypi for dev releases[2] and nodepool features around network_cli connection for ansible (zuul too, this is dropping blacklist support for modules).
Of the things linked by TristanC, scale out of scheduler is likely something I'll be able to contribute too, however I don't think we've landed the spec yet, have we?
The spec is not approved yet, and the last review is from January 18th.
Oh, and zuul ansible 2.8 support[3], I'd love to get that done as close as ansible 2.8 release as possible. I think that is an important story for cross project collaboration.
[1] https://review.opendev.org/655204 [2] https://review.opendev.org/655474 [3] https://review.opendev.org/631933
Well, perhaps it is time to update the maintainer list? It seems like most of the maintainers only participate in the change they are going to use, resulting in new features staling indefinitely.
Regrads, -Tristan