Hi, Can you share a few more details about the data source? Are you continuously ingesting files from a folder?
You are correct, that the parallelism should not affect the results, but there are a few things that can affect that: 1) non-determnistic keys 2) out-of-order data with inappropriate watermarks Note that watermark configuration for file ingests can be difficult and that you need to ensure that files are read in the "right" order. AFAIK, Flink's continuous file source uses the modification timestamp of files to determine the read order. Best, Fabian Am So., 25. Aug. 2019 um 19:32 Uhr schrieb Hanan Yehudai < hanan.yehu...@radcom.com>: > I have an issue with tumbling windows running in parallel. > > > > I run a Job on a set of CSV files. > > > > When the parallelism is set to 1. I get the proper results. > > While it runs in parallel. I get no output. > > Is it due to the fact the parallel streams take the MAX(watermark) from > all the parallel sources. > > And only one of the streams advances the watermark ? > > > > It seems wrong that the result is not deterministic and depends on the > parallel level. > > What am I doing wrong ? > > >
