Regarding correctness, I think it makes sense to change default values and
possibly add a warning note when there's a known risk of wrong results.
Needless to say that we should try to fix as many issues as possible; we
still need volunteers to review open PRS.

Performances regressions are trickier but if we have the query plans (CBO +
full) along with logs (including task counters) for fast and slow execution
we may be able to understand what happens. Don't hesitate to create Jira
tickets with these information if available.

Last regarding 4.0.0 blockers, I don't think we need a special label. The
built-in and widely used priority "blocker" seems enough to capture the
importance and urgency of a ticket.
Since I am the release manager for the next release I will go over tickets
marked as blockers and reevaluate priorities if necessary.

Best,
Stamatis

On Thu, Mar 23, 2023, 10:27 AM Denys Kuzmenko <dkuzme...@apache.org> wrote:

> Thanks, Sungwoo for running the TPC-DS benchmark. Do we know if the same
> level of performance degradation was present in 4.0.0-alpha1?
>
> All: please use the `hive-4.0.0-must` label in a ticket if you think it's
> a show-stopper for the release.
>

Reply via email to