https://github.com/apache/airflow/pull/68368 - PR for workflows that should monitor regressions in the future.
On Wed, Jun 10, 2026 at 7:07 PM Henry Chen <[email protected]> wrote: > Wow, that's super helpful! Thanks for doing this, Kaxil. > > Aaron Chen <[email protected]> 於 2026年6月10日週三 下午11:57寫道： > > > Amazing! This is a really useful feature! > > > > Thanks, > > Aaron > > > > On Wed, Jun 10, 2026 at 8:06 AM Rahul Vats <[email protected]> > wrote: > > > > > Nice one! > > > > > > Regards, > > > Rahul Vats > > > > > > > > > > > > On Wed, 10 Jun 2026 at 19:42, Jarek Potiuk <[email protected]> wrote: > > > > > > > Probably we should add some monitoring to see if length of running of > > our > > > > jobs is increasing above some recent trend - I have no idea when this > > > > started to happen, but with all things changing and runner queue we > are > > > > experiencing having such an automated warning on "main" builds would > > help > > > > us to catch things like this earlier. > > > > > > > > On Wed, Jun 10, 2026 at 3:50 PM Amogh Desai <[email protected]> > > > wrote: > > > > > > > > > 50% is an amazing improvement, thanks Kaxil! > > > > > > > > > > Thanks & Regards, > > > > > Amogh Desai > > > > > > > > > > > > > > > On Wed, Jun 10, 2026 at 7:19 PM Jarek Potiuk <[email protected]> > > wrote: > > > > > > > > > > > Nice! > > > > > > > > > > > > On Wed, Jun 10, 2026 at 3:02 PM Kaxil Naik <[email protected]> > > > > wrote: > > > > > > > > > > > > > Hi team, > > > > > > > > > > > > > > I just merged https://github.com/apache/airflow/pull/68261 > which > > > > > should > > > > > > > drop CI run time (where it runs all tests) by 50% from ~62 min > to > > > ~29 > > > > > > mins > > > > > > > in the case of MySQL. But same applies for Postgres & Sqlite. > > > > > > > > > > > > > > The motivation was that our MySQL tests were timing out on some > > PRs > > > > > > because > > > > > > > it took more than 60 mins. > > > > > > > > > > > > > > While there is still room for improvement and I'd like the time > > to > > > > > > decrease > > > > > > > further, we need to be careful not to unintentionally lose test > > > > > > isolation. > > > > > > > > > > > > > > *It doesn't apply for all PRs that are filtered by selective > > > checks - > > > > > > > example a Yandex provider change. So they might not see these > > > > > > improvements. > > > > > > > > > > > > > > Regards, > > > > > > > Kaxil > > > > > > > > > > > > > > > > > > > > > > > > > > > >
