Thanks Till What about this workaround. If I after the watermark assignment split the stream in elements that fits in the watermark (s1) and those that don’t (s2). The s1 I process with the table api with a window aggregate using watermark and s2 I handle with an unbounded non-windows aggregate with IdleStateRentionTime so state are removed when my devices are up to date again. I then merge the two outputs and continue. By doing this I handle 99% as standard and only keeping state for the late data.
Make sense? And would it work? Med venlig hilsen / Best regards Lasse Nedergaard > Den 24. apr. 2019 kl. 19.00 skrev Till Rohrmann <trohrm...@apache.org>: > > Hi Lasse, > > at the moment this is not supported out of the box by Flink. The community > thought about this feature but so far did not implement it. Unfortunately, > I'm also not aware of an easy workaround one could do in the user code space. > > Cheers, > Till > >> On Wed, Apr 24, 2019 at 3:26 PM Lasse Nedergaard <lassenederga...@gmail.com> >> wrote: >> Hi. >> >> We work with IoT data and we have cases where the IoT-device delay data >> transfer if it can't get network access. We would like to use table windows >> aggregate function over each device to calculate some statistics, but for >> windows aggregate functions to work we need to assign a watermark. This >> watermark is general for all devices. We can set allow latency, but we can't >> set it to months. >> So what we need is to have a watermark for each device (key by) so the >> window aggregate work on the timestamp delivered for the device and not the >> global watermark. >> Is that possible, or have anyone consider this feature? >> >> Best >> >> Lasse Nedergaard