kafka-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Boyang Chen <reluctanthero...@gmail.com>
Subject Re: Kafka Exactly-Once Semantics in .NET support
Date Wed, 01 Jul 2020 06:43:36 GMT
Thanks for the interest Saher. TBH, I don't know how the .NET confluent
library works, but I would assume it is based on the
https://github.com/edenhill/librdkafka. Maybe propose an issue there as
well to attract some expertes.

Boyang

On Tue, Jun 30, 2020 at 3:22 PM Saher Ahwal <saher.ahwal@gmail.com> wrote:

> Hi
>
> I am working on exactly-once semantics with Kafka and I have streaming
> scenario of read-process-write. I noticed the new exactly-once scalability
> design with correctness in case of partition reassignment here:
>
> https://docs.google.com/document/d/1LhzHGeX7_Lay4xvrEXxfciuDWATjpUXQhrEIkph9qRE/edit#
>
> It all looks great and I understand we need to rely on
> PendingTransatcionException and ProducerFencedException. However, we are
> using confluent library and .NET and I don't see any of these exceptions.
>
> How do I ensure correctness of fencing of zombie producer transactions when
> using .NET confluent library? What exceptions are retry-able and what are
> not? Can I call abortTransaction on any Kafka exception ? I don't find good
> examples in the documentation.
> Any pointers or answers are kindly appreciated.
>
> Thanks in advance
> Saher
>
>
>
> --
> *Saher Ahwal*
>
> *Massachusetts Institute of Technology '13, '14*
> *Department of Electrical Engineering and Computer Science*
> *saher@alum.mit.edu <saher@mit.edu> | 617 680 4877*
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message