[ 
https://issues.apache.org/jira/browse/KAFKA-16585?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17840907#comment-17840907
 ] 

Matthias J. Sax commented on KAFKA-16585:
-----------------------------------------

{quote}I can use the regular Processor, but as I understand it add some 
overhead comparing with FixedKeyProcessor
{quote}
Where did you get this? The Processor itself does not have overhead. – The only 
think that could happen downstream is, that a unnecessary repartition step 
could be inserted. We are tackling this via 
[https://cwiki.apache.org/confluence/display/KAFKA/KIP-759%3A+Unneeded+repartition+canceling]
{quote}{color:#172b4d}Really, I think FixedKeyProcessor do not need to be 
"ensure that the key is not changed". IMHO there is enough to have a key from 
the same partition. So, if you will provide the way to generate the 
{color}*FixedKeyRecord*{color:#172b4d} from any local store it will be 
enough.{color}
{quote}
{color:#172b4d}Well, technically yes, but there is no simply way to 
enforce/check this... We would need to serialize the provided key, pipe it 
through the Partitioner, and compare the computed partition. Or is there some 
other way to do this? – This would be quite expensive to do.{color}

{color:#172b4d}If you feel strong about all this, feel free to do a POC PR and 
write a KIP about it, and we can take it from there. I don't see a simple way 
to do it, and I believe that using a regular Processor is the right way to go 
(especially with KIP-759 on the horizon). {color}

> No way to forward message from punctuation method in the FixedKeyProcessor
> --------------------------------------------------------------------------
>
>                 Key: KAFKA-16585
>                 URL: https://issues.apache.org/jira/browse/KAFKA-16585
>             Project: Kafka
>          Issue Type: Bug
>          Components: streams
>    Affects Versions: 3.6.2
>            Reporter: Stanislav Spiridonov
>            Priority: Major
>
> The FixedKeyProcessorContext can forward only FixedKeyRecord. This class 
> doesn't have a public constructor and can be created based on existing 
> records. But such record usually is absent in the punctuation method.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to