Borislav Markov

This feature is quite important to many people. When people don't see it in argo, they just switch to a workaround.

Thanks for explanation. Why this cannot be addressed in this project by putting a call from time to time to check expired time windows and fire `getOutput()` respectively ?

And another question how to workaround this problem. If I generate a message towards my grouping vertex, does it have to be keyed, because I have 2 partitions at the...

I also made few more tests and I put in the pipeline same message again and again and it seems reducer still keep old data inside and doesn't fire `getOutput()`...

> In Numaflow the progression of time is data-driven rather than being based on the clock time. This means that the watermark progression or window closure is not dependent on...

> @borkox With [idle source watermark](https://github.com/numaproj/numaflow/issues/633), you should be able to fix this (Numaflow 1.1 release). The ideal generic solution will be through early triggers, which will be possible via...

> @borkox thank you for being an early adopter of Numaflow. As we work through upcoming releases, we love to learn about our community use cases and take feedback, wanted...

> @borkox thank you for being an early adopter of Numaflow. As we work through upcoming releases, we love to learn about our community use cases and take feedback, wanted...

So things around this watermark are getting more weird and weird, seems nothing works as expected around reducer. So I have 2 messages with number 32 which goes to transformer...