I agree that it is acceptable to break it (again). At the moment, we
already prohibit users from resuming 1.0 savepoints with 1.1 and so
far (~1 week since release) no one complained. I think that at this
point most users use it only with the same major version as they used
when triggering the save
Hi Aljoscha,
I'm not very deep into the state backend implementation. However, I
think a breaking change is unavoidable with the new key groups. The
only way that we achieve backwards-compatibility is to include a
translator from the old state format to the new one. As you already
mentioned, this
Hi,
I think this is a very important change for the the future of the system
that provides a much cleaner internal representation of the states.
You are right that this can in theory break programs written in 1.1 when
upgraded to 1.2 but I wonder if this will be actually a practical problem
in th
Hi,
Stefan and I are currently working on preparing our state infrastructure
for the introduction of key-grouped state. This is the parent issue for
key-grouped state https://issues.apache.org/jira/browse/FLINK-3755 while
this is the specific issue that we are currently working on
https://issues.ap
