☂️ Release v0.9.0 requirements

[mimowo]

We are targeting the release for the first week of Nov 2024, with a release candidate in the second half of Oct.

Must Haves

Beta Give feedback

1. Graduate LendingLimit to Beta #2909
   approved cncf-cla: yes kind/feature lgtm release-note size/L +6
   
2. Graduate MultiplePreemptions to Beta #2864
   approved cncf-cla: yes kind/feature lgtm release-note size/XS +6
   
3. Promote VisibilityOnDemand to Beta #2973
   kind/feature +1
   
4. Configurable mechanism for Resource Abstraction #2937
   kind/feature +1
   
5. [MultiKueue] Promote MultiKueue API and feature gate to Beta #3095
   kind/feature +1
   
6. MVP support / extension of support for serving workloads #2717
   kind/feature +1
   
7. Topology Aware Scheduling #2724
   kind/feature +1
   
8. Support for configurable ProvisioningRequest retries #1353
   kind/feature +1
   

Note that at #2717, we may need to drop LWS support.

Nice To Haves

Beta Give feedback

1. [MultiKueue] AdmissionCheck conditions are confusing #3094
   kind/bug +1
   
2. Introduce Job-agnostic API to declare the maximal execution time for a Job #3125
   kind/feature +1
   
3. Decide if we release kjobctl as part of Kueue #2778
   +0
4. Make LossLessDefaulter selective to only prevent dropping fields outside of schema #3174
   kind/feature +1
   
5. Expose Flavors in LocalQueue Status #3122
   kind/feature +1
   
6. MVP support for serving workloads running as LeaderWorkerSet #3232
   kind/feature +1
   
7. Support MPIJob managedBy feature for the MultiKueue #3257
   kind/feature +1
   

[mimowo]

/cc @tenzen-y @alculquicondor @mwielgus @mwysokin @dgrove-oss
PTAL and let me know if something is missing or needs to be re-prioritized

[mimowo]

/cc @trasc

[mimowo]

/cc @andrewsykim

[kannon92]

Could we maybe add an API promotion to this release or the next one? #768

I have been asked to find out how stable the Kueue API is and when can we expect stability.

(I think it should be fine but folks would like to know what kinds of guarantees we would have with using Kueue).

It seems that the project is pretty stable and has a lot of users. Can we discuss about stabilizing the core API?

[mimowo]

+1, thanks for raising this. Given we are just a month away from the planned 0.9 release it is unlikely to be included, as it would increase the risk of extending the release timeline. I would consider this as one of the major themes for the next release.

We will use this release to give a deprecation warning for some APIs we want to drop, like #2437, #2256 or #3094.

[tenzen-y]

I would like to move #768 after the HierarchyCohort and TAS graduate to beta since both features introduce significant changes to the Kueue, and we can imagine evolving the API based on user feedback on both features.

[tenzen-y]

Additionally, this is a lightweight nice to have: #3257
But, this has a significant UX-improving impact on the MultiKueue.

[mimowo]

I make the following changes:

• moved Topology Aware Scheduling #2724 and Support for configurable ProvisioningRequest retries #1353 to must haves (based on the user requests we get)
• moved [MultiKueue] AdmissionCheck conditions are confusing #3094 as nice to have (but still high priority)
• dropped Extend waitForPodsReady config to account for AdmissionChecks #3231 and Add retry mechanism for AdmissionChecks in Kueue #3258

Also added nice to have #3257. Cc @mszadkow