does anyone know what the flink git tags mean. how...
# troubleshooting
f
does anyone know what the flink git tags mean. how does
release-1.15.2-rc2
differ from
release-1.15
? i've noticed that
rc2
does not contain the FlinkKinesisConsumer pyflink bindings whereas
release-1.15
does.
im on AWS KDA and I have a feeling it's using rc2 as I cannot resolve this kinesis consumer when i submit my job. i guess this means I need to manually build the python mappings for this? anyone using aws kda 1.15. also have any experience with this problem?
m
@Hong Teoh @Danny Cranmer ?
d
@Felix Angell
release-1.15.2-rc2
is the tag for
release-1.15.2
while
release-1.15
is the branch for release-1.15.x series(that’s, all release-1.15.x are created from branch release-1.15 and currently the latest release is
release-1.15.4
) Regarding to Kinesis Support for PyFlink, it is added since 1.16(see https://issues.apache.org/jira/browse/FLINK-21966 for more details).
h
Yes, Dian Fu seems correct. There are no python wrapper for Kinesis in Flink 1.15.2. https://github.com/apache/flink/blob/release-1.15.2/flink-python/pyflink/datastream/connectors.py If I’m not mistaken, a possible workaround is to use the Table API… 👀
f
dang ok. i think we are too far gone for the table api sadly
is there much risk in copying over the python wrappers from a later version?
d
@Felix Angell I think it should work.
👍 2
f
noice! i can give that a go and see how we get on 😄
ah. can i assume that FlinkKinesisConsumer exists in the KDA runtime for PyFlink 1.15? or do i need to bundle that jar dependency in myself
d
It is not bundled; you will need to bundle that
f
is it ok to use this or does this mean i need to get a jar from flink do your eckon: https://github.com/awslabs/amazon-kinesis-connector-flink
d
This fork is for 1.8/1.11 only. Please use the Apache version
👍 1
f
cool, cheers for the help all!
another KDA question relating to python backporting stuff that is related to this sorry! the WatermarkStrategy PyFlink API introduces 'source alignment' in version 1.16 despite being available as a feature in 1.15. is it reasonable/safe to backport this or would it be safer to roll our own alignment (somehow).
d
I do not think this is a KDA question. I am unsure on the answer, suggest repost as a general question.
f
ah that's a fair point! thanks